Molecular mechanisms of resistance to PI3K inhibitors Pau Castel Morales ADVERTIMENT. La consulta d’aquesta tesi queda condicionada a l’acceptació de les següents condicions d'ús: La difusió d’aquesta tesi per mitjà del servei TDX (www.tdx.cat) i a través del Dipòsit Digital de la UB (diposit.ub.edu) ha estat autoritzada pels titulars dels drets de propietat intel·lectual únicament per a usos privats emmarcats en activitats d’investigació i docència. No s’autoritza la seva reproducció amb finalitats de lucre ni la seva difusió i posada a disposició des d’un lloc aliè al servei TDX ni al Dipòsit Digital de la UB. No s’autoritza la presentació del seu contingut en una finestra o marc aliè a TDX o al Dipòsit Digital de la UB (framing). Aquesta reserva de drets afecta tant al resum de presentació de la tesi com als seus continguts. En la utilització o cita de parts de la tesi és obligat indicar el nom de la persona autora. ADVERTENCIA. La consulta de esta tesis queda condicionada a la aceptación de las siguientes condiciones de uso: La difusión de esta tesis por medio del servicio TDR (www.tdx.cat) y a través del Repositorio Digital de la UB (diposit.ub.edu) ha sido autorizada por los titulares de los derechos de propiedad intelectual únicamente para usos privados enmarcados en actividades de investigación y docencia. No se autoriza su reproducción con finalidades de lucro ni su difusión y puesta a disposición desde un sitio ajeno al servicio TDR o al Repositorio Digital de la UB. No se autoriza la presentación de su contenido en una ventana o marco ajeno a TDR o al Repositorio Digital de la UB (framing). Esta reserva de derechos afecta tanto al resumen de presentación de la tesis como a sus contenidos. En la utilización o cita de partes de la tesis es obligado indicar el nombre de la persona autora. WARNING. On having consulted this thesis you’re accepting the following use conditions: Spreading this thesis by the TDX (www.tdx.cat) service and by the UB Digital Repository (diposit.ub.edu) has been authorized by the titular of the intellectual property rights only for private uses placed in investigation and teaching activities. Reproduction with lucrative aims is not authorized nor its spreading and availability from a site foreign to the TDX service or to the UB Digital Repository. Introducing its content in a window or frame foreign to the TDX service or to the UB Digital Repository is not authorized (framing). Those rights affect to the presentation summary of the thesis as well as to its contents. In the using or citation of parts of the thesis it’s obliged to indicate the name of the author. Facultat de Farmàcia Universitat de Barcelona Programa de Doctorat en Biomedicina Molecular mechanisms of resistance to PI3K inhibitors Memòria presentada per Pau Castel Morales per optar al títol de Doctor per la Universitat de Barcelona Pau Castel Morales Directors: Dr. Josep Baselga Dr. Maurizio Scaltriti Tutor: Dr. Albert Tauler Barcelona, 2016 2   3 ACKNOWLEGMENTS 4 5 Over the four years of hard work on this thesis, many people have played a special role and have helped me accomplish this important academic and personal goal. Without all their support and confidence in me, this work would not have been possible, and this section has been dedicated to thank them for their contribution. First, I would like to thank my directors Josep and Mauri, who allowed me to perform all these exciting projects and experiments in their lab. All the experiences in this environment have given me a good understanding of the cancer field and have contributed to my critical thinking and scientific maturity. Una especial dedicació a la meva família, al meu pare, mare i germà, que sempre han estat un exemple de dedicació i esforç. Per haver-me ensenyat a aprendre de tot, a lluitar per allò que m’agrada, però sobretot a ser bona persona. Per tot el temps que no hem estat junts durant aquests anys, aquesta tesis va per vosaltres. A l’Alba, per ser el millor moment del dia. Per estar sempre que cal, per estimar- me, per fer-me feliç, per aguantar-me, per tot. Però sobretot per tot el que ens queda per fer, per els nous plans i per tot una vida per davant. T’estimo. Als meus quatre avis, us enyoro i se que estareu orgullosos de mi. A la Patricia i a tots els tiets i cosins que esteu escampats per tot el món. A la família de l’Alba. I would also like to thank my lab mates for all the hours that we have spent together, all the fruitful scientific discussions, all the “eureka” and “failure” moments while performing experiments, the afterhours craziness, and the exciting ideas. From MGH to MSK, Haley, Natasha, Ana, Javi, Eneda, Guotai, Pedram, Moshe, Zach, Joanne, Carmen, Neel, and all the students. Thank you for making the “Baselgaquarium” such a special place to work. I am grateful to all the collaborators, scientists, and clinicians that have helped and taught me so much during this process, with stimulating projects and ideas. També agraeixo a tots els amics i amigues que durant aquests anys han estat al meu costat, m’han donat suport, i han estat simplement bons amics. Finally, thank you to the patients, families, and mice for their contribution to science and cancer research. 6 7 TABLE OF CONTENTS List of Figures Page 9 List of Tables Page 10 Abbreviations Page 11 Summary Page 13 Resum Page 15 Introduction Page 17 1. Personalized medicine Page 19 2. Drug development and clinical trials Page 23 3. Targeted therapies Page 28 4. Mechanisms of resistance to targeted therapies Page 29 5. PI3K pathway Page 38 5.1. Structural and biochemical determinants Page 38 5.2. Signaling downstream of Class IA PI3K Page 41 6. Protein kinases and regulation of AGC kinases Page 46 7. Genetic alterations in the PI3K pathway Page 54 8. PI3K inhibitors Page 58 9. Other inhibitors targeting the PI3K/AKT/mTOR pathway Page 64 9.1. mTORC1 inhibitors Page 64 9.2. AKT inhibitors Page 65 9.3. mTOR catalytic inhibitors Page 66 9.4. Dual PI3K-mTOR inhibitors Page 67 Objectives Page 71 Results Page 75 8 Summary of results Page 207 Discussion Page 219 Conclusions Page 237 References Page 241 9 LIST OF FIGURES Figure 1: The concept of personalized medicine Page 21 Figure 2: Drug development: from the lab to the approval Page 23 Figure 3: Common mechanisms of resistance to targeted therapies Page 30 Figure 4: Structural overview of T790 residue in EGFR Page 32 Figure 5: Evolution of tumor mutations upon therapy Page 37 Figure 6: General overview of PI3K regulation Page 39 Figure 7: Structural features and regulation of PI3K Page 41 Figure 8: PI3K and AKT downstream signaling Page 45 Figure 9: Phosphoryl transfer reaction of kinases Page 47 Figure 10: Conserved regulatory motifs in AGC kinases Page 48 Figure 11: Mechanisms of AGC kinase activation Page 50 Figure 12: PDK1 inhibitor GSK2334470 chemical structure Page 52 Figure 13: Conserved mTORC2 signaling in lower species Page 53 Figure 14: Overview PIK3CA mutations in breast cancer Page 56 Figure 15: Structural features of PI3K inhibitors Page 61 Figure 16: Clinical response to a PI3Kα inhibitor Page 62 Figure 17: Phase Ib clinical trial of fulvestrant with GDC0032 Page 222 10 LIST OF TABLES Table 1 Page 35 Table 2 Page 54 Table 3 Page 63 Table 4 Page 69 Table 5 Page 69 11 ABBREVIATIONS bp Base pairs BrdU Bromo-deoxy-Uridine ChIP Chromatin Immunoprecipitation CNS Central Nervous System CNV Copy Number Variation Da Dalton DMEM Dulbecco’s Modified Eagle Medium DMSO Dimethyl Sulfoxide DNA Deoxyribonucleic Acid EGF Epidermal Growth Factor EGFR Epidermal Growth Factor Receptor ELISA Enzyme Linked Immunosorbent Assay ER Estrogen Receptor FACS Fluorescence Activated Cell Sorting FBS Fetal Bovine Serum FDA Food and Drug administration FFPE Formalin-Fixed Paraffin-Embedded FISH Fluorescence In Situ Hybridization FOXO Forkhead O transcription factors GEMM Genetically Engineered Mouse Models H&E Hematoxylin & Eosin HER2/3/4 Human Epidermal growth factor Receptor 2/3/4 HDACs Histone Deacetylases HRP Horseradish Peroxidase HUVEC Human Umbilical Vein Cord Ig Immunoglobulin IHC Immunohistochemistry IP Immunoprecipitation KO Knock Out kDa kilo-Dalton LB Lysogeny Broth MAPK Mitogen-Activated Protein Kinase MEF Mouse Embryonic Fibroblasts miRNA micro RNA MRI Magnetic Resonance Image mRNA messenger RNA mTOR mammalian Target of Rapamycin NF1 Neurofibromatosis Factor 1 NF-κB Nuclear Factor Kappa-light-chain-enhancer of activated B cells NGS Next Generation Sequencing PBS Phosphate Buffered Saline PCR Polymerase Chain Reaction PD Pharmacodynamics PFS Progression Free Survival PI3K Phosphatidylinositol 3-Kinase PK Pharmacokinetics 12 PKC Protein Kinase C PR Progesterone Receptor qRT-PCR Quantitative Reverse Transcription PCR RB Retinoblastoma RNA Ribonucleic Acid RPMI Roswell Park Memorial Institute Medium RSK Ribosomal S6 Kinase RTK Receptor-associated Tyrosine Kinases SD Standard Deviation SEM Standard Error of the Mean sgRNA single guide RNA shRNA short hairpin RNA siRNA small interference RNA SNP Single Nucleotide Polymorphism STAT Signal Transducer and Activator of Transcription S6 Ribosomal protein S6 S6K Ribosomal protein S6 kinase TBS Tris-buffered Saline TCGA The Cancer Genome Atlas TE Tris-EDTA TF Transcription Factor TNBC Triple Negative Breast Cancer TUNEL Terminal deoxynucleotidyl transferase dUTP nick end labeling UTR Untranslated Region WB Western blot WES Whole Exome Sequencing WGS Whole Genome Sequencing WT Wild Type 13 SUMMARY The development of high-throughput sequencing technologies has prompted the evaluation of large cohorts of different tumor types. The results from these comprehensive genomic studies have revealed the most commonly mutated genes across human cancers, providing further understanding of the pathogenesis, molecular classification, and therapeutic strategies for this disease. PIK3CA, the gene encoding the PI3Kα isoform, is among the most frequently mutated genes in breast, head and neck, colorectal, and lung cancer, among others. Activating mutations in PIK3CA promote hyperactivation of the PI3K/AKT pathway, leading to increased proliferation, cell growth, survival, and metabolism. Current efforts are aimed to develop PI3K inhibitors as an effective therapy for PIK3CA mutated cancers and, despite promising clinical responses, the emergence of drug resistance is a clear limitation. In this doctoral thesis, we have explored these mechanisms of resistance in order to provide a better understanding of tumor evolution upon therapy, define subpopulations of patients that are likely to respond to PI3K inhibitors, and provide novel pharmacological combinations to overcome therapy refractoriness. The loss of the tumor suppressor PTEN was found to play an important role in the resistance of PI3Kα inhibitors in preclinical models and patients, mainly by reactivating the PI3K/AKT pathway as a result of an increased dependency on the PI3Kβ isoform. Our work also demonstrated the notion of tumor evolution and phenotypic convergent evolution in response to therapeutic pressure. Moreover, we have established that intrinsic resistance to PI3Kα inhibitors occurs as a result of incomplete inhibition of the mammalian target of rapamycin complex 1 (mTORC1), a downstream effector of the PI3K/AKT pathway. PI3Kα inhibitor-resistant cells could be sensitized through the blockade of phosphoinositide-dependent kinase (PDK1), a constitutively active kinase, using genetic or pharmacologic inhibition. Further experiments showed that the downstream effector of PDK1 is the serum and glucocorticoid-induced kinase 1 (SGK1), which promotes cell survival through the phosphorylation of key proteins such as FOXO3 and TSC2. Accordingly, the resistant phenotype could be also reverted by inhibiting SGK1, a novel pharmacological approach that has revealed interesting roles of this kinase in tumor biology. 14 Genetically engineered mouse models represent reliable tools for investigating the etiology, biology, and progression of human diseases, as well as for exploring novel therapeutic approaches. By serendipity, we discovered the role of PIK3CA mutations in the genesis of venous malformations, an aberration of normal venous development that currently lacks effective treatments. Our mouse models recapitulated the histopathologic features of the disease and provided an experimental platform to test novel pharmacological approaches. PI3Kα inhibitors were effective at reducing the morbidity and mortality of mice carrying venous malformations. The results from this thesis highlight the importance of defining the molecular determinants of sensitivity and resistance to PI3K inhibitors, a therapy that will most likely benefit PIK3CA mutant patients. 15 RESUM El desenvolupament de noves tècniques de seqüenciació massiva ha fomentat l’estudi d’un gran nombre de mostres de diversos tipus tumorals. Els resultats d’aquests estudis genòmics exhaustius ha revelat els gens que es troben mutats en major prevalença, contribuint a una millor comprensió dels processos de patogènesis, classificació molecular i estratègies terapèutiques per a aquesta malaltia. PIK3CA, el gen que codifica per a la isoforma PI3Kα, es troba entre els gens mes freqüentment mutats en el carcinoma de mama, cap i coll, colorectal, pulmó, entre d’altres. Les mutacions activadores a PIK3CA promouen la hiperactivació de la via de senyalització de PI3K/AKT, donant lloc a un increment en la proliferació, la supervivència, i el metabolisme de les cèl·lules tumorals. Els esforços actuals es centren en el desenvolupament d’inhibidors de l’enzim PI3K com a una possible teràpia efectiva en tumors que presenten mutacions a PIK3CA. Tot i que els assajos clínics inicials son prometedors, l’emergència de resistència a aquestes teràpies és una clara limitació. En aquesta tesis doctoral s’han explorat els possibles mecanismes de resistència per intentar entendre com els tumors evolucionen enfront d’aquest fàrmacs, poder definir les subpoblacions de pacients que respondran als inhibidors de PI3K i proporcionar noves combinacions farmacològiques per combatre el fenomen de la resistència. Hem demostrat que la pèrdua del supressor tumoral PTEN juga un paper important en la resistència als inhibidors de PI3Kα, tant en models preclínics com en pacients, mitjançant la reactivació de la via de PI3K/AKT que és resultat d’un increment en la dependència de la isoforma PI3Kβ. El nostre treball també ha evidenciat la noció d’evolució tumoral i ha demostrat el concepte d’evolució convergent fenotípica en resposta a la pressió terapèutica. També s’ha demostrat que la resistència intrínseca als inhibidors de PI3Kα es pot donar com a resultat d’una inhibició incompleta del complex 1 de mTOR (mTORC1), un efector clau de la via de PI3K/AKT. Cèl·lules resistents a inhibidors de PI3Kα es van poder sensibilitzar amb el bloqueig genètic o farmacològic de PDK1, una quinasa constitutivament activa. Experiments addicionals van poder demostrar que l’efector molecular de PDK1 era la quinasa SGK1, la qual promou la supervivència cel·lular a través de la fosforilació de proteïnes clau com FOXO3 i TSC2. El fenotip resistent es va 16 poder revertir mitjançant la inhibició farmacològica d’aquesta proteïna, una aproximació terapèutica que ha revelat un rol interessant en la biologia tumoral. Els models murins modificats genèticament representen una eina segura per a l’estudi de la etiologia, biologia i progressió de malalties humanes, així com per explorar noves aproximacions terapèutiques. Com a resultat d’un descobriment imprevist, també hem pogut revelar el rol de les mutacions de PIK3CA en la formació de malformacions venoses, una aberració del desenvolupament normal de les venes que actualment no tenen un tractament específic. El nostre model animal de malformació venosa recapitula les característiques histopatològigues de la malaltia i proporciona una plataforma experimental única per a l’estudi de noves teràpies. En aquests models animals, els inhibidors de PI3Kα han demostrat ser efectius en la reducció de la morbiditat de les malformacions venoses. Els resultats d’aquesta tesis emfatitzen la importància de definir els determinants moleculars de sensibilitat i resistència a inhibidors de PI3K, una teràpia de la que segurament es beneficiaran els pacients amb mutacions a PIK3CA si s’administren adequadament. 17 INTRODUCTION 18 19 1. Personalized medicine Over the last decades, the model of personalized medicine has emerged in the field of oncology and other specialties of medicine. This concept refers to the idea in which each patient is characterized by a specific molecular alteration and using precise therapy to target this molecular aberration (Haber et al., 2011). This is especially important in the case of cancer; although tumors can share histological or anatomical features, the mutated genes that drive tumor growth are sometimes different. In the oncology field, personalized medicine can be illustrated by the discovery of different RTKs involved in cancers, such as EGFR and HER2 (Gharwan and Groninger, 2015). Specific antibodies targeting these membrane receptors (Cetuximab for EGFR and Trastuzumab for HER2) were found to be active agents only in patients harboring overexpression of these proteins and prolonged survival when compared to the conventional treatment. Today, these therapies are part of the standard-of-care treatment for breast, colorectal, lung, head and neck, and gastric cancers. Additional examples of personalized medicine include the treatment with targeted therapies such as Rituximab (anti-CD20), Fulvestrant (anti-estrogen receptor), Erlotinib and Gefitinib (EGFR small molecule inhibitors), and many others (Vivanco, 2014). Therefore, it is logical to design therapies for each of the genomic aberrations found in tumors rather than treating all patients with drugs that are cytotoxic, thereby decreasing toxicity and increasing efficacy. An important methodology that has greatly improved over the last years is next generation sequencing (NGS), which allows interrogation of the cancer’s genetic code. The development of novel sequencing technologies has catapulted the use of this technique for the diagnosis and molecular classification of cancers and also for the study of tumor genomic behavior (heterogeneity, evolution, dynamism, and other concepts that were more commonly used in population genetics) (Gagan and Van Allen, 2015). The cost of sequencing genomes and initial input material required for sequencing has decreased dramatically during the last decade. Consequently, small samples of DNA purified from needle biopsies can be used to perform whole exome (WES) or whole genome sequencing (WGS), despite the fact that DNA may not be of optimal quality (Shen et al., 2015). 20 Other genomic platforms have also been developed with clinical purposes such as the targeted-sequencing panels for cancer-related genes. In these panels, only genes with known functions or with available therapies are included, allowing for a time- and cost- effective and high-coverage sequencing result with clinical implications. This approach is currently being undertaken by several hospitals and cancer centers for its importance in personalized medicine (Cheng et al., 2015). However, all these advances in the technology of DNA sequencing cannot be translated into a therapeutic strategy unless the field of medicinal chemistry advances accordingly. Fortunately, the concept of targeting oncogenes has evolved rapidly in the chemistry field since the approval of the first kinase small-molecule inhibitor imatinib (Muller, 2009). The concept of creating a molecule that can compete with endogenous ATP in order to target a kinase remained elusive for years. Now in the kinase field, we can not only compete with ATP but also exploit structural susceptibilities such as the allosteric pockets created when kinases are in the inactive conformation (DFG-out) (Muller et al., 2015). Several strategies have been developed in order to improve the potency and the selectivity of compounds that are aimed to target oncogenes that drive cancer growth and survival. Recently, a novel concept that also seemed improbable has emerged: targeting small protein GTPases (Ostrem et al., 2013). This is especially important, since one of the most common oncogenes in human cancer is RAS, a small GTPase that activates the MAPK signaling pathway (Pylayeva-Gupta et al., 2011). In summary, the field of medicinal chemistry has developed probes that can be used to target many oncogenes found in human cancers using strategies that take advantage of allosteric sites, protein-protein interactions, alkylating moieties for irreversible inhibitors, mutation- specific sites, and many others (Zhao et al., 2014). Once the fields of DNA sequencing and medicinal chemistry have evolved to this point, one might envision a model of medicine that involves the molecular interrogation of the disease, the identification of a suitable target, and the selection of a targeted therapy (Figure 1). However, it is important to consider the nature of cancer as an evolving disease, implying that treatments suitable for today might not be appropriate for tomorrow (Haber et al., 2011; Schmitt et al., 2015). In this context, there is increasing interest in the development of non-invasive techniques that allow for the interrogation of the tumor’s genome such as liquid biopsies (Crowley et 21 al., 2013). Liquid biopsies refer to the use of blood (or other fluids that are in contact with the tumor, i.e. cerebrospinal fluid or urine) as a matrix to determine the somatic mutations present in the tumor. This can be achieved by the selection of tumor “debris” present in the fluid, mainly circulating tumor cells (CTCs) that result from tumor cell detachment and circulating tumor DNA (ctDNA) resulting from tumor cell apoptosis or necrosis (Garcia-Olmo et al., 2010). The isolation of CTCs can be achieved by different biological or physicochemical means such as the positive immune-affinity to epithelial markers (i.e. EpCAM), density, size, and others depending on the histological features of the tumor (Yu et al., 2011). On the other hand, the isolation of ctDNA is relatively easier, using kits containing columns that enrich for nucleic acids. Together with the increased sensitivity of the sequencing techniques, the liquid biopsy method appears to be a suitable approach to follow-up the life cycle of the tumor (Arnedos et al., 2015). Importantly, there are several reports that indicate that ctDNA could be more accurate than imaging in terms of recurrence and can also identify subclonal resistant populations that might arise upon treatment (Diehl et al., 2008). Figure 1. The concept of personalized medicine Upon diagnosis, tumor biopsy or resection is characterized by different molecular profiling techniques. DNA sequencing is the most advanced technique for molecular profiling, and allows very low input material including circulating tumor DNA (ctDNA). Other techniques are based on expression platforms, proteomic signatures, or others. At the same time, when available, tumors that engraft in immunocompromised mice can serve as an excellent platform for experimental therapy testing. Data from in vivo experiments and molecular profiling leads to a suitable molecular target or cancer driver that can be pharmacologically inhibited. Finally, patients treated with personalized medicine are required to be monitored clinically and, if resistant clones emerge, a new cycle of tumor characterization will be required to identify novel pharmacological targets. Source: modified from (Arnedos et al., 2015) 22 Finally, another notable technology is the patient-derived xenograft (PDX). Using this approach, tumors harvested during surgery or biopsies are engrafted into immunocompromised mice (DeRose et al., 2011). This presents a potential limitation, as not all tumors have the ability to engraft; however, if the procedure is successful the tumor can be expanded into different mice. These tumors tend to recapitulate the morphological and genomic features of the original tumors, probably because they are not disaggregated like primary cultures (Aparicio et al., 2015). PDX can be characterized using DNA sequencing and can also be treated experimentally with drugs aimed to target specific signaling pathways, ultimately selecting the best agent for clinical treatment (Eirew et al., 2015). This approach is especially useful for analyzing resistance mechanisms, since mice can be used to test combination therapies or characterize the emerging resistant clones (Hidalgo et al., 2014). In summary, personalized medicine covers different scientific specialties that are aimed at ascertaining the right medicine for each patient, using evidence-based therapy (Arnedos et al., 2015). For example, a patient diagnosed with a breast cancer will be screened for mutations using NGS, the tumor will be transplanted in mice, and the results will yield an activating mutation in PIK3CA. The patient will now be treated with standard-of-care therapy and a PI3K inhibitor that better works in the patient's specific tumor type. Liquid biopsies in conjunction with imaging will be performed periodically to assess the response to therapy. If successful, the patient will be monitored until cure; if not, sequencing might reveal novel genetic drivers that can be pharmacologically targeted. It is also important to keep in mind that not all the mechanisms of resistance are caused by genetic mechanisms. For this reason, techniques that measure RNA and protein expression might be useful. 23 2. Drug development and clinical trials The development of a drug is a complex process that generally begins with the discovery of the biological target and results in the commercialization of the drug for therapeutic purposes (Hughes et al., 2011). This process requires the validation of the new therapeutic agent in preclinical models, but more importantly, the clinical development in clinical trials (Owens et al., 2015). In cancer research, the development of high-throughput DNA sequencing has led to the identification of many targets that can be pharmacologically inhibited (Haber et al., 2011). Together with the advances in screening platforms and molecular biology techniques, a large list of molecular targets has been validated as possible cancer therapies and the development of compounds specifically targeting these oncoproteins is currently undergoing. Briefly, after the preclinical testing and validation of the target, the clinical study is designed and comprises 3 phases of development (Munos, 2009). If all these are positive, the drug will receive approval form the regulatory agencies, allowing for commercialization of the new medicine (Figure 2). Figure 2. Drug development: from the lab to the approval Compounds that are synthesized with a therapeutic purpose are studied in preclinical settings using in vitro and in vivo models of the disease. Next, 3 phases of clinical development are required in order to prove that the compound has better activity than the current standard-of- care therapy. Regulatory agencies (FDA in US and EMA in Europe) are essential for the final review and approval of the drug. After approval, Phase IV or surveillance phase is required to identify adverse effects that were not originally described. NDA: New Drug Application. Source: (Dickson and Gagnon, 2004) 24 For the purpose of this thesis, I will focus on the clinical development of compounds in the field of oncology. After the in vitro and in vivo preclinical studies there are three clinical phases of development: Phase I: These studies, also called first-in-human trials, are aimed at determining the safety, tolerability, pharmacodynamics (PD), and initial pharmacokinetics (PK) of the new drug. In oncology, the population recruited for these studies are cancer patients, in contrast with other fields in which Phase I clinical trials are carried out in healthy volunteers (Rubin and Gilliland, 2012). Because safety and preliminary PD/PK are the most important information that will be obtained from Phase I trials, patients are not required to have the same type of cancer, and hence patients with different tumor types can be enrolled, unless is specified. In these studies a reduced number of patients is required (14- 40) and investigators will determine the dose at which the new drug has to be administered based on several factors: pharmacodynamics (ability of the drug to reach the affected tissue, i.e. tumor), maximum tolerated dose (MTD) as a result of a dose-escalation schedule, dose-limiting toxicities (DLTs), pharmacokinetics, and other factors that might influence the bioavailability of the new compound. Phase I trials are often carried out in subjects with metastatic cancers that have exhausted the different available lines of therapy. Thus, these are heavily pre- treated patients that have limited treatment options, and sometimes the durability of the experimental treatment is limited (Owens et al., 2015). There is increasing popularity in determining responses in Phase I clinical trials, especially in the case of targeted therapies, aiming to speed up a process that sometimes can take many years until approval (Ivy et al., 2010). To further observe responses in early phases of clinical development, selection criteria are becoming more important. In the case of PI3K inhibitors, the phase I clinical trial of BYL719 was the first PI3K inhibitor to select only patients carrying PIK3CA mutations (Juric and Baselga, 2012). This approach of patient selection has been critical for the clinical development of Vemurafenib, which was initially administered in BRAF mutant melanoma (Chapman et al., 2011). Importantly, with the discovery of actionable targets of cancer (drivers) that can be pharmacologically inhibited, trials in histologically independent tumor types have been proposed (Hyman et al., 2015). In these trials, also known as basket trials, the enrollment of patients is 25 done mainly by the mutation status of an oncogene rather than the histological origin of the tumor. Another concept that is becoming very important in Phase I clinical trials as a result of the use of signal transduction inhibitors is pharmacodynamics. Tissue biopsies or imaging techniques are necessary to establish if the drug is specifically targeting the oncogenic pathway. In the case of PI3K inhibitors, some PD markers that have been used include tissue stainings (pAKT, pPRAS40, pS6, p4EBP1), imaging techniques (Fludeoxyglucose 18FDG positron emission tomography), and plasma biomarkers (glucose levels and insulin) (Dienstmann et al., 2014; Rodon et al., 2013). Phase Ib: These trials are a slight variation of the Phase I clinical trials and are undertaken to assess the safety, dose, and PK/PD of a drug in combination with another agent. Normally, one of the two agents has already been approved or the dose has been previously determined clinically, while the other compound requires investigation. These trials are especially important in cases in which pharmacokinetic interaction might exist or one drug potentiates the effect of the other (Ivy et al., 2010). Phase II: Phase II studies are aimed at determining whether an agent has antitumor activity and estimates the response rate in a defined patient population. In general, phase II clinical trials only enroll enough patients to ensure the detection of antitumoral activity, limiting the number of patients to less than 100. These trials are disease-oriented (for instance, only patients with metastatic Her2+ breast cancer) (Hughes et al., 2011). This is important because every tumor type has a different prognostic factor and response to therapy, therefore a mixed population would compromise the study. Therefore, patient stratification is key in this process. In this line, the development of biomarkers that can predict the response or the resistance to an experimental agent will be useful to define the patient population. Preferentially, Phase II trials are performed in a randomized manner, compared to a control arm, and the end points are usually response rate, time to progression (TTP), and progression free survival (PFS). Moreover, these trials confirm the safety results from Phase I and can be used to improve the understanding of the PK profile (population PK). 26 Phase III: This phase consists of a large clinical study that enrolls multiple patients with a specific disease (hundreds to thousands). Ideally these are run as multicentric studies, randomized, and double-blinded in order to achieve the highest clinical evidence. The main goal of this study is to determine whether the new therapy is better than the current standard-of-care treatment. Because these trials are required for the approval of a new compound or for novel use by regulatory agencies, they are sometimes referred to as pivotal studies. The most common end-points for these studies are PFS and overall survival (OS). Phase III trials can be carried out together with Phase II trials in order to accelerate the process of therapy approval. In these trials, two stages will be designated in which initial efficacy is assessed in a subset of patients and then the second stage will be used to recruit additional patients. These trials are especially useful in the cases where therapies show remarkable activity in the Phase I clinical trial. When the results in this phase are positive, regulatory agencies generally approve the use of the new compound for clinical use and the compound is commercialized and used as a new standard of care treatment for a particular disease/ tumor type (Dickson and Gagnon, 2004). All these Phases discussed above are needed to bring a molecule from the laboratory to the final approval, in a process that might take almost a decade and cost on average one billion US dollars (Dickson and Gagnon, 2004). Therefore, strategies aimed at improving the response ratio are very useful. Among these, we can include screening of patients for known biomarkers of response or resistance and the appropriate PD measurements along the trial. For instance, a concept that is becoming popular in the discovery of novel biomarkers is the exceptional or outlier responders. This notion consists of the identification of patients that have an uncommon clinical response to the therapy (either extremely positive or negative) with subsequent genotyping in the appropriate lesions (Brannon and Sawyers, 2013). These n=1 cases are intended to identify the molecular mechanism by which the tumor remains sensitive or resistant to the therapy. An example of sensitivity was described for the Rapamycin analogue Everolimus, an mTORC1 inhibitor, in a clinical trial for bladder cancer (Iyer et al., 2012). In this phase II clinical trial, a patient exhibited a complete response for a period of over 2 years. Whole genome sequencing of the lesion revealed a frameshift mutation in 27 TSC1, a negative regulator of the mTORC1 complex. Mutations in the TSC complex are associated with response to Rapamycin, and the findings were validated in a larger cohort of patients that revealed that TSC1 mutation is a biomarker of response to the therapy. Conversely, there are also examples in which exceptional responders’ lesions can uncover mechanisms of resistance. In the lesions that have exhibited a remarkable clinical response, unfortunately the emergence of resistant clones is sometimes a reality (Arnedos et al., 2015). Those lesions can be obtained by biopsy or rapid autopsy procedures and once sequenced, can reveal novel biomarkers of resistance that have appeared when compared to the pre-treatment biopsies or surgical resections (Rubin et al., 2000). 28 3. Targeted therapies As discussed before, targeted therapies represent a key part of personalized medicine (Arnedos et al., 2015). The concept of targeted therapy was initially used in the field of infectious disease, where the design of drugs that selectively kill bacteria but not host cells was developed after the pioneering discovery of the “magic bullet” Salvarsan (Kaufmann, 2008). During the 1940s, the use of chemotherapy began to gain popularity after nitrogen mustards, agents used in chemical warfare, were revealed to have anti-proliferative activity in tumor cells from blood malignancies (DeVita and Chu, 2008). From these seminal observations, a variety of novel compounds were developed and new combinations were assayed, leading to the present scenario of multiple chemotherapy agents with different mechanism of action and activity. These agents can be divided depending on their mechanism of action into alkylating agents (e.g. cyclophosphamide), antimetabolites (e.g. methotrexate), anti-microtubule polymerization (e.g. vincristine), topoisomerase inhibitors (e.g. irinotecan), and cytotoxic antibiotics (e.g. doxorubicin) (Verrill, 2009). Although chemotherapy has prolonged the life of a large number of patients and is currently the standard-of-care for many tumors, the pharmacologic use of these compounds has several caveats, being systemic toxicity the most important one (Nguyen et al., 2014). Chemotherapy does not selectively kill cancer cells, thus normal cells with high proliferative capacities (such as intestinal epithelia or bone marrow derived cells) can be equally affected by these therapies, producing profound immunosuppression of the patient on some occasions (Chabner and Roberts, 2005). Because tumors are defined by oncogenic signals that drive the growth and malignant characteristics, therapies that are aimed at targeting this "Achilles’ heel" are needed. This opened a novel therapeutic perspective aimed at specifically targeting the aberrant signaling (Gharwan and Groninger, 2015). Moreover, the combination of these therapies with systemic treatments such as chemotherapy and radiotherapy are predicted to have a major impact on decreasing tumor burden without affecting healthy cells, particularly when administered in the neoadjuvant setting. 29 4. Mechanisms of resistance to targeted therapies Despite the benefit from targeted therapies, the best-known challenge for the use of these therapies is the appearance of drug resistance. In this section, I will describe the general mechanisms that are known to limit the antitumor effect of the targeted therapies. Resistance to targeted therapies can be a result of different processes during the pharmacokinetic and pharmacodynamic (PK/PD) course of a drug (Gottesman, 2002). While most of the studies focus on the tumor-specific mechanisms, we have to keep in mind that pharmacological mechanisms of resistance are also a possibility (Baguley, 2010). In fact, during the PK process, there are several steps that might result in the reduction of drug efficacy, such as decreased absorption, increased metabolism, unfavorable compartmentalization of the drug, or increased elimination (Calcagno, 2011; Luqmani, 2005). For instance, a patient experiencing pathologies associated with gastric pH acidification/alkalization can modify the absorption of drugs that are susceptible to ionization as a result of their pKa (Wojtkowiak et al., 2011). In terms of metabolism, factors such as the inter-individual polymorphism in the CYP genes or enzyme induction may result in increased metabolism of the drug (Rodriguez-Antona and Ingelman- Sundberg, 2006). Drugs that are very liposoluble or lacking the ability to cross the brain blood barrier (BBB) account for examples in which the drug is compartmentalized and cannot properly reach the desired target. A good example of this is the poor activity of Gefitinib in central nervous system (CNS) disease due to limited BBB penetration (Agarwal et al., 2010). Therefore, it is mandatory that robust PD biomarkers are available for targeted therapies, so clinicians can ensure the appropriate targeting of the molecular alteration and avoid these pharmacological mechanisms of resistance (Figure 3). 30 Tumor-specific or pharmacodynamic resistance can be the result of several mechanisms that include increased efflux, drug inactivation, target modification, adaptive responses, and impaired cell death. Here, I will summarize the different mechanisms known to drive resistance to targeted therapies. Drug efflux: All the cells in the body have transporters in the plasma membrane that are required for the translocation of small molecules from/to the cytoplasm. Cancer cells have been found to upregulate some of these transporters in order to increase the efflux of drugs, leading to a decreased intracellular concentration of the drug and, as a consequence, the resistant phenotype. Transporters associated with resistance were widely studied in the field of chemotherapy drug resistance, where this mechanism plays an important role. However, the role of this mechanism of resistance in targeted therapies is less evident, with minimal supporting evidence, such as the role of ABCG2 (Borst and Elferink, 2002). Interestingly, some studies have shown that tyrosine kinase inhibitors have the ability to inhibit multidrug transporters, maybe due to the Figure 3. Common mechanisms of resistance to targeted therapies Most targeted therapies are designed to be administered orally. Similar to other drugs, when administered in the organism, they undergo several steps and processes termed ADME. The distribution of the molecule to the cancer cells is required to reach the appropriate molecular target. This process results in cytotoxicity of the cancer cells and the clinical activity of the molecule. However, all of these steps from the absorption to the elimination can be altered, leading to resistance to the therapeutic agent. Source: (Holohan et al., 2013) 31 nature of the ATP mimetic structural characteristic (Gottesman et al., 2002). One could speculate that kinase inhibitors have some inhibitory activity on the ATP pocket of these pumps, explaining why these mechanisms of resistance are not as important in targeted therapies as in hydrophobic molecules such as chemotherapy agents (Kartal-Yandim et al., 2015). Nevertheless, it is still important to acknowledge that some of these transporters are associated with resistance in cancer therapy, including the multiple drug resistant (MDR) and ATP-binding cassette (ABC) family of transporters — the most important molecules involved in this efflux mechanism (Holohan et al., 2013). Drug inactivation: This mechanism of resistance is also extensively studied in the field of chemotherapy and metabolism, since cancer cells have the ability to express metabolic enzymes that are required to alter the chemical structure of the drug. In general terms, metabolic reactions are a method by which the organism increases the polarity of small molecules in order to solubilize them for excretion (Baguley, 2010). Among these reactions, we can find enzymes that have the ability to breakdown chemical groups (esterases, amidases), reduce or oxidize groups, and conjugate chemical groups (acetylation, sulfation, glucuronidation, and glutathione conjugation). These mechanisms are very specific for each class of drug because they depend on the chemical structure. For example, irinotecan, a topoisomerase I inhibitor, can be inactivated by UDP glucuronosyltransferase 1 (UGT1A1) (Belanger et al., 2010). Remarkably, the expression of this metabolic enzyme is regulated by DNA promoter methylation. Target modification: This mechanism of resistance was previously described in the field of infectious disease and there are also some examples in the case of chemotherapeutic agents. However, the most straightforward cases were described for targeted therapies that are currently standard-of-care. For example, in EGFR-driven lung cancer, the presence of the drug Gefinitib selects for resistant clones that harbor the gatekeeper mutation T790M (Kobayashi et al., 2005; Shih et al., 2005) (Figure 4). This mutation, as shown in the image below, confers resistance by at least two mechanisms. First, the alteration of the ATP pocket by the methionine residue, a bulky substitute, will occupy the space where 32 the drug interacts. Second, this mutation increases the affinity between ATP and EGFR, decreasing the ability of this drug to compete with the endogenous concentrations of ATP. Similar cases have been described for KIT, ALK, and other RTKs with acquired resistance to their equivalent inhibitors. To further stress this concept of target modification, a recent report has shown a very peculiar case leading to resistance to ALK inhibitors in lung cancer (Shaw et al., 2016). A patient treated with the ALK inhibitor Crizotinib developed the ALK mutation C1156Y, which confers resistance to these agents. The patient was then treated with lorlatinib, an ALK inhibitor that is not affected by this mutation, and the disease improved significantly. However, after months of therapy, the patient developed resistance to lorlatinib via the ALK mutation L1198F that sterically interferes with drug binding. Surprisingly, this mutation renders sensitivity to crizotinib, re-sensitizing the tumor to the previous therapy and improving the clinical parameters when the patient was treated again with crizotinib. Another case that is sometimes observed as a mechanism of resistance to targeted therapies involving the modification of the target is the increase or decrease of the expression of the target. There are some reports indicating that the amplification or downregulation of the target is observed upon Figure 4. Structural overview of T790 residue in EGFR The residue Thr790 is shown in red, and the EGFR inhibitor is shown in yellow. In WT EGFR, the Thr residue is responsible for the hydrogen bond with the inhibitor, but in the mutation T790M, the bulky aa methionine occupies the ATP pocket where the inhibitor sits. This steric impediment does not allow the drug to interact with EGFR, leading to resistance to the inhibitor. 33 treatment with some inhibitors. For example, our group has reported that some patients that are sensitive to a PI3Kα inhibitor exhibit a copy number loss of PIK3CA, the gene encoding PI3Kα, upon therapy progression (Juric et al., 2015). Adaptive responses: This group contains all the mechanisms of resistance that are as a result of increased oncogenic signaling not associated with the original oncogenic driver. In other words, tumors that become refractory by adaptive responses exhibit increased dependency in other mutations, signaling pathways, or other molecules. This is better exemplified in the case of EGFR targeting, where it was found that the amplification of MET, another RTK, can sustain the mechanisms of oncogenicity independently of EGFR (Engelman et al., 2007). The amplification of MET induces the heterodimerization with Her3 and the downstream signaling mediated by PI3K and the MAPK pathway. In the research field of drug resistance, this is probably the area that yields the most surprising discoveries, since mechanisms of resistance can be virtually unlimited (Dalton, 1990; Gottesman, 2002). Often, the pathway involved in the resistance is the same, probably as a result of oncogenic addiction. For instance, in the MAPK pathway it has been shown that multiple feedback mechanism can lead to the reactivation of the same pathway independently of the target inhibition (Lito et al., 2013). There are at least two major mechanisms of resistance for RAF inhibitors classified between dimerization-dependent and –independent. The first includes the splice variant of BRAF (Poulikakos et al., 2011), the acquired mutation in NRAS (Q61) (Nazarian et al., 2010), and CRAF overexpression (Montagut et al., 2008). In the second other mechanisms such as COT overexpression or MEK mutations can be found (Johannessen et al., 2010). In general, all these mechanisms are aimed to reactivate ERK signaling, a process which is very likely vital for the survival of cancer cells (Lito et al., 2013). Similar to these cases is the mutations observed in PIK3CA in Her2-positive tumors that drive resistance to anti-Her2 therapies, a finding that has been clinically validated (Majewski et al., 2015). In other cases, the pathway involved in the resistance is completely unrelated inducing a change in the dependency of the tumor. For example, our group has recently reported evidences suggesting that ER positive tumors that are treated with PI3Kα inhibitors can develop resistance by up-regulating the expression of 34 ER and its target genes, leading to a more luminal-like phenotype (Bosch et al., 2015). In these cases, combinations with inhibitors that target these “horizontal” mechanisms of resistance are expected to improve tumor response. Impaired cell death: This is a broad mechanism of resistance that not only affects targeted therapies but any other therapy aimed to induce cell death in the tumor. It is a result of genetic or transcriptional modifications in the apoptosis machinery, leading to increased resistance to this biological process. It has been widely studied in the field of apoptosis, where therapeutic antibodies that activate death receptors are being investigated. In these cases some alterations such as the reduction of Caspase 8 expression or the amplification of cFLIP (an antiapoptotic molecule) are observed (Holohan et al., 2013). It is also becoming apparent that resistance to EGFR therapies can be a result of deregulated apoptosis response. For instance, amplifications in BCL2 or BCL-xL, antiapoptotic proteins, are found in tumors resistant to these therapies (Faber et al., 2012). In fact, combination of these agents with BCL inhibitors induces increased sensitivity and better clinical responses. In the following Table (Holohan et al., 2013), there is a summary of well-accepted mechanisms of resistance to targeted therapies that are FDA-approved. 35 Classically, and based on the observations in infectious diseases, the pharmacodynamic mechanisms of resistance have always been classified in two groups: intrinsic or acquired. Intrinsic resistance refers to the resilient phenotype of tumor cells upon treatment, indicating that once a drug is administered cancer cells still have the ability to grow. Intrinsic resistance is common, especially in the case of targeted therapies, and can be also considered as an insensitive phenotype (Gottesman, 2002). Many times, the lack of the molecular target is the major mechanism of resistance to certain therapies. For instance, tumors from Her2-positive breast cancer are intrinsically resistance to the EGFR inhibitor cetuximab for the only reason that cancer cells do not express the Table 1. The common mechanisms of resistance to approved targeted therapies are shown. Many mechanisms can coexist including mutations and upregulation of parallel signaling pathways. Source: (Holohan et al. 2013). 36 receptor or simply do not rely on the activity of this molecule for their growth. However, there are occasions in which tumor cells are clearly driven by an oncogene yet the therapy against this molecular aberration is not successful. For example, in Her2- positive breast cancers there are a subset of patients that are intrinsically resistant to trastuzumab, lapatinib, or both (Rexer and Arteaga, 2012). In some occasions pharmacodynamic measurements reveal that the target is correctly inhibited and yet the tumor remains insensitive. In this scenario, we can think in a Her2-positive tumor that is treated with trastuzumab. The biopsy post-treatment reveals that the downstream signaling is completely abrogated (i.e. phospho-Her2, phospho-ERK, phospho-AKT, and others) but there is no clinical response. In general, in these cases the underlying mechanism of resistance can be explained by the presence of bypass tracks or secondary drivers. Acquired resistance represents the opposite situation. In this case, tumors are sensitive to the therapeutic agent administered, but after some time of drug exposure the tumor becomes resistant to the therapy (Gottesman, 2002). For instance, a Her2-positive tumor is treated with trastuzumab and the patient exhibits a remarkable response, with a decrease in the tumor size and an improvement in the clinical parameters. However, several months after the treatment, the tumor recurs and grows again. Now, this tumor is fully resistant to the therapy. There are evidences suggesting that at least two different scenarios can exist or co-exist in terms of acquired resistance. On one hand, that the tumor cells engage adaptive responses to the therapy (increase drug efflux, drug transformation, decrease dependency on the target). This is a dynamic process involving the active role of cancer cells and can be explained by the plasticity of tumor cells and the presence of DNA replication errors as a result of the high proliferation rate (Holohan et al., 2013). On the other hand, that pre-existing resistant clones are found in the tumor bulk and therapeutic pressure only selects for these resistant populations (Figure 5). As an example, if a clone formed by 100 cells is present in a tumor at a low percentage (<0.001%) and it does not respond to Trastuzumab, but all the other surrounding cells respond, upon Trastuzumab treatment all the tumor bulk will decrease leaving only this small fraction of cells that could repopulate the tumor (Schmitt et al., 2015). This can be explained by the heterogeneity of tumors that are composed by multiple clones with different mutations (Zahreddine and Borden, 2013). 37 Although these two different models of acquired resistance seem to be contradicted hypotheses, it is very likely that both mechanisms co-exist in tumors undergoing acquired resistance. Figure 5. Evolution of tumor mutations upon therapy (Above) Heterogeneity is a feature of human cancers. Within a tumor, several clones are affected by different mutations as a result of DNA replication errors. The number of tumor cells increases exponentially until the tumor exhibits clinical symptoms. At that point therapy is administered, but clonal populations that represent a small fraction of the tumor and that are resistant to the therapy will now grow in the presence of therapeutic pressure. (Below) The percentage of mutant cells is shown. The current techniques for the detection of these mutations might be limited by their detection threshold. Therefore, it is required to increase the sensitivity of DNA sequencing tools. Source: (Schmitt et al., 2015) 38 5. PI3K pathway 5.1. Structural and biochemical determinants There are three classes of PI3K (Class I to III) depending on their structural and biochemical properties (Figure 6). Regarding their biochemical activity, Class I PI3Ks have the ability to phosphorylate the lipid substrate phosphatidylinositol (4, 5) biphosphate (PIP2) at the 3ʹ-hydroxyl group, giving rise to the second messenger phosphatidylinositol (3, 4, 5) triphosphate (PIP3) (Fruman and Rommel, 2014; Thorpe et al., 2015). In terms o structural regulation, Class I PI3K is further divided into the two subclasses IA and IB. Class IA consists of heterodimers that contain a catalytic and a regulatory subunit. There are three known genes encoding the different catalytic subunits of the PI3K heterodimer, namely PIK3CA, PIK3CB, and PIK3CD, which encode for the p110α, β, and δ isoforms, respectively. On the other hand, there are five different regulatory isoforms that associate with any of the p110 catalytic subunit. These regulatory subunits are encoded by the genes PIK3R1 (p85α, and splicing variants p55α and p50α), PIK3R2 (p85β), and PIK3R3 (p55γ). Class IB PI3Ks also form heterodimers between the catalytic subunit encoded by PIK3CG (p110γ), and any of the regulatory subunits encoded by PIK3R5 (p101) and PIK3R6 (p87). The functions of the regulatory subunits of PI3K are: stabilization of the catalytic subunit, inhibition of the basal kinase activity, and they also engage the activation of the catalytic subunit downstream of phosphorylated tyrosine motifs as a results of the interaction with their SH2 domains (Fruman and Rommel, 2014). Class IA isoforms p110α and p110β are virtually expressed in all the cells of the organisms, while the expression of the p110δ and p110γ isoforms seems to be restricted to leukocytes (Hawkins and Stephens, 2014). Class II PI3Ks are not very well studied. There are three genes encoding the class II isoforms, PIK3C2A (PI3K-C2α), PIK3C2B (PI3K-C2β), PIK3C2G (PI3K-C2γ). In the case of class II PI3Ks, the lipid kinase activity does not require a regulatory subunit so these 39 act as monomers. The physiological substrate of these kinases is phosphatidylinositol (PI) giving rise to phosphatidylinositol 3-phosphate (PI3P) (Vanhaesebroeck et al., 2010). Class III PI3K consists only of one member, VPS34, which is encoded by the PIK3C3 gene. VPS34 interacts with the regulatory subunit VPS15, encoded by PIK3R4, in order to generate a functional dimer. Similar to class II PI3Ks, VPS34 phosphorylates phosphatidylinositol (PI) and produces phosphatidylinositol 3-phosphate (PI3P) (Vanhaesebroeck et al., 2010). Figure 6. General overview of PI3K regulation (Left) Substrates for the different class of PI3Ks are shown. (Right) Representative view of the domains present in the catalytic and regulatory subunits of the different class PI3K. p85-BD: p85 binding domain; RBD: Ras-binding domain; P: proline-rich region; SH3: Src homology domain 3; SH2: Src homology domain 2; iSH2: inter SH2 domain; Gβγ BD: Gβ/γ protein binding domain; PX: Phox homology domain; HEAT: Huntingtin, elongation factor 3 (EF3), protein phosphatase 2A (PP2A), and yeast kinase TOR1 domain. Source: (Thorpe et al., 2015) 40 While the role of Class I PI3Ks is widely studied and these are known to play a key role in the transduction of extracellular receptors, including RTK and GPCR, the role of Class II and III PI3Ks is less characterized. However, it is now clear that these enzymes are required for endosome biology, especially VPS34, whose pharmacological inhibition impairs PI3P-mediated signaling (Bago et al., 2014). For the purpose of this work, I will mainly focus in the biology of Class IA PI3K, which are involved in cancer. The catalytic subunit of Class I PI3Ks, p110, contains five different recognizable conserved domains that are responsible for the structural regulation of the heterodimer (Engelman et al., 2006). At the C-terminal there is a kinase domain that confers the lipid kinase activity to the enzyme. Similar to many protein kinases, the kinase domain of PI3K is organized in a N- and C-lobe. A helical domain can be also found, although the function is not fully understood. Other domains found are the C2 domain, which is responsible for the binding to membranes, the Ras-binding domain (RBD), responsible for the interaction with this GTPase, and the adaptor-binding domain (ABD), that is necessary and sufficient to mediate protein-protein interactions with the intermediate SH2 (iSH2) domain of the regulatory subunit p85 (Costa and Engelman, 2014). In the case of p85, other domains can be recognized such as the N-terminal SH3 domain, the breakpoint cluster region (BCR) homology domain, which is flanked by proline-rich regions, and the two different SH2 domains, nSH2 and cSH2, that are separated by a coiled-coil region termed the inter-SH2 domain (iSH2) (Vanhaesebroeck et al., 2010). In general, all these structural domains of p85 are known to be important for the binding to phosphorylated tyrosine motifs present in activated RTK and for the inhibition of the catalytic subunit. In the basal state, p85 would interact and inhibit the catalytic subunit, but upon phospho-tyrosine stimulation the SH2 domains interact with this motif and release the inhibitory interaction towards p110 (Falasca and Maffucci, 2012). X-ray crystallography and hydrogen-deuterium exchange mass spectrometry have shown that there are several inhibitory interfaces between p85 and p110 (Figure 7). For instance, p85 nSH2 domain creates inhibitory interfaces with the C2, helical, and C-lobe kinase domains of p110, the p85 iSH2 domain with the p110 C2 domain, and the p85 cSH2 domain with the p110 C-lobe kinase domain. This last interaction is, in fact, an 41 isoform-specific regulatory mechanism, since only the p110β and p110δ isoforms exhibit this contact (Burke and Williams, 2013). The crystal structure of PI3Kα in complex with the regulatory subunit p85α was initially solved in 2007 (Huang et al., 2007). 5.2. Signaling downstream of Class IA PI3K Once mitogenic stimuli induce activation of the PI3K enzyme, the readily available substrate PIP2 is phosphorylated at the 3’-OH position. PIP3, which is an uncommon lipid in the plasma membrane, engages the signaling of downstream pathways. Phosphatidylinositol (3, 4) biphosphate (PI(3,4)P2), which is a PIP3 metabolite generated as a result of the 5’-PO4-3 dephosphorylation by the phosphatases SHIP and INPP5, can also induce downstream signaling of the pathway (Thorpe et al., 2015). Figure 7. Structural features and regulation of PI3K (A) Contact interactions found between the regulatory subunit of PI3K, p85, and the catalytic subunits of Class I PI3Ks. (B) Overview of the PI3K structure. Domains are indicated. (C) Contact interactions found between the regulatory subunit of PI3K, p84 and p101, and the catalytic subunits of Class IB PI3K. Source: (Burke and Williams, 2013) 42 In general terms, when the PI3K enzyme is active, the generated PIP3 or PI(3,4)P2 lipids serve as a scaffold for several proteins that contain phosphoinositide recognition domains. The interaction between these domains and the lipid induces conformational changes or membrane translocation that are required for the downstream activity of the enzyme (Vanhaesebroeck and Alessi, 2000). In cancer, the most studied PI3K effector is AKT, which is responsible for the activation of several downstream cellular effectors that regulate cell cycle, proliferation, apoptosis, metabolism, and translation. The key events downstream of PI3K can be summarized in: Phosphoinositide recognition: Several proteins have the ability to recognize phosphoinositides in the cell. These lipids can be found in the plasma membrane, but also in the intracellular compartments such as the lysosome, endosome, or other organelles. Structurally, this interaction can be explained by the presence of highly conserved motifs termed phospholipid-binding domains. In mammals, there are approximately twelve different domains known to interact with phospholipids: PH, PKC C2, C1, PX, FYVE, Discoidin C2, GRAM, F-BAR, Annexin, Gla, N-BAR, and ENTH/ANTH domains (Lemmon, 2008). For the purpose of this dissertation, I will only focus on the PH domain for its ability to recognize mainly PIP2, PIP3, and PI(3,4)P2. Pleckstrin-homology (PH) domains refer to a region of approximately 100 aa that is found in Pleckstrin, a PKC substrate, and homology regions are also found in other protein with the ability to interact with phosphoinositides. This domain has the ability to selectively recognize the second messenger lipids among many others present in the membrane. The structural basis for this recognition has been widely studied using the PH domain of Phospholipase C (PLCδ) as a model. The headgroup of the phosphoinositide can interact with a PH domain surface composed by the 7- stranded β-sandwich structure and the β1–β2 loop between the first two β- strands. This generates a pocket that contains the sequence motif KXn(K/R)XR, where K is a Lysine, R an Arginine, and X any aa, and interacts with the phosphate group (Lemmon, 2008). Among the most studied proteins containing a PH domain relevant for the PI3K pathway we can find the effectors Bruton’s Tyrosine Kinase (BTK), PLCδ, Phosphoinositide-dependent kinase (PDK1), AKT, and other adaptor proteins, GTPase activating proteins (GAP), and Guanine Nucleotide exchange proteins (GEF) (Falkenburger et al., 2010). 43 AKT: Activation of AKT, a kinase that belongs to the AGC family of the human kinome, is considered a key output of the PI3K pathway, due to the large number of substrates interacting with this Ser/Thr kinase (Arencibia et al., 2013). There are three isoforms of AKT (1-3) encoded by their respective genes AKT1, AKT2 and AKT3. After remaining elusive for several years, there is now consensus on the mechanism of activation of this kinase. Upon generation of the second messenger PIP3 in the membrane by PI3K, both phosphoinositide- dependent kinase-1 (PDK1, a constitutively active kinase that contains a highly affine PH domain) and AKT, which also contains a PH domain, are recruited to the membrane (Alessi et al., 1996). The proximity of these kinases allows PDK1 to phosphorylate AKT at the residue Thr308 of the activation loop (T-loop) (Alessi and Cohen, 1998). Subsequently, AKT is phosphorylated at the residue Ser473 of the hydrophobic motif by the rapamycin-insensitive mTOR complex 2 (mTORC2) (Sarbassov et al., 2005). This phosphorylation if thought to fully activate the kinase activity of AKT. However, several reports suggest that the phosphorylation at the T-loop may be enough to engage AKT activity in selected substrates. Activated AKT, in turn, phosphorylates several substrates involved in apoptosis and cell cycle regulation. For instance, AKT is able to phosphorylate and inhibit BAD, a pro-apoptotic member of the BCL-2 family, and Caspase 9, two main regulators of the mitochondrial apoptotic pathway. It also inhibits p21CIP1 and p27KIP, directly related with the inhibition of cell cycle progression (Vanhaesebroeck et al., 2010). Moreover, AKT can also inhibit the forkhead transcription factors FOXO1, 3, 4 and 6, involved in the transcriptional regulation of several genes including the pro-apoptotic CD95L, BCL2L11 (BIM), BBC3 (PUMA), and the cell cycle inhibitors CDKN2A (p21CIP) and CDKN2B (p27KIP) (Webb and Brunet, 2014). In addition to these effectors, AKT can phosphorylate PRAS40 and TSC2, two negative regulators of mTORC1 activity. These phosphorylations inhibit the activity of PRAS40 and TSC2, leading to an increased mTORC1 signaling (Huang and Manning, 2009; Sancak et al., 2007). Mammalian target of Rapamycin (mTOR): mTOR, a 289 kDa serine- threonine protein kinase, is the core of this pathway and acts as a master integrator (Laplante and Sabatini, 2012). It senses and responds to 44 environmental cues such as nutrient availability, stress, and mitogens to regulate anabolic reactions through a highly orchestrated and complex mechanism. mTOR was originally identified in the early 1990s as a mutated protein that can confer resistance to the growth inhibitory effects of rapamycin in yeast. The first evidence that mTOR can form two distinct complexes came also from work in yeast, in which two evolutionary conserved proteins TOR1 and TOR2 form complexes with distinct roles in cell growth control (Guertin and Sabatini, 2007, 2009). mTORC1 which contains mTOR, Raptor, mLST8, and PRAS40 has been considered a master regulator of cell growth and metabolism that signals mainly though 4E-binding protein (4EBP1) and 40S ribosomal protein S6 kinase (S6K), which are both important in the physiological control of translation. mTORC1 promotes protein synthesis by phosphorylating 4EBP1 which in turn prevents 4EBP1 binding to the eukaryotic initiation factor 4E (eIF4E), enabling eIF4E to initiate cap-dependent translation. On the other hand, activation of S6K1 by mTORC1 leads to an increase in mRNA biogenesis and cap-dependent translation. mTORC1 has also been demonstrated to activate RNA Pol I transcription and thus rRNA synthesis through a process involving the protein phosphatase 2A (PP2A) and the transcription initiation factor IA (TIF-IA) (Laplante and Sabatini, 2009). Soon after the discovery of mTORC1, it was identified a second nutrient-sensitive but rapamycin-insensitive complex, mTORC2, which consists of six different proteins, several of those in common with mTORC1. mTORC2 is formed by mTOR, Deptor, and mLST8 but, instead of raptor, it contains three other proteins: rapamycin-insensitive companion of mTOR (Rictor) (Sarbassov et al., 2004), mammalian stress-activated protein kinase interacting protein (mSIN1) (Frias et al., 2006; Jacinto et al., 2006), and protein observed with Rictor-1 (Protor-1) (Pearce et al., 2007). It is known that mTOR2 activity responds to growth factors, but how mTORC2 is regulated and the exact molecular function of most of its interacting proteins still remains elusive (Alessi et al., 2009). Functionally, mTORC2 regulates organization of the actin cytoskeleton through the phosphorylation of protein kinase Cα and also activates AKT and SGK1 through phosphorylation at the hydrophobic motif Ser473 and Ser422, respectively (Garcia-Martinez and Alessi, 2008; Sarbassov et al., 2005). Although other kinases such as DNA-PK and ATM, have been suggested to phosphorylate AKT at S473, Rictor, mSin1, and mLST8 knockout 45 mice have shown that intact mTORC2 complex is required for maximal phosphorylation and activation of AKT in mouse embryonic cells (Guertin et al., 2006). Figure 8. PI3K and AKT downstream signaling Upon PI3K activation, the presence of PIP3 leads to increased downstream signaling that is mediated by the interaction of PH-containing proteins. Among them, PDK1 and AKT are well studied. These translocate to the membrane, interact, and PDK1 phosphorylates AKT, which is subsequently phosphorylated by mTORC2. Now, active AKT will phosphorylate a variety of downstream targets that regulate cellular events such as cell growth, proliferation, survival, metabolism and autophagy. Source: (Vanhaesebroeck et al., 2012). 46 6. Protein kinases and regulation of AGC kinases Although PI3K is a lipid kinase, the propagation of the biochemical signals occurs mainly through the action of protein kinases, which have the ability to phosphorylate effector proteins. Protein kinases are highly conserved across species and essential players in signal transduction. The deregulation of several kinases has been linked to multiple diseases, including cancer, overgrowth syndromes, and diabetes, among others. The human kinome is classified in eight groups based on their phylogenic relationship: AGC (Containing PKA, PKG, PKC families), CAMK (Calcium/calmodulin-dependent protein kinase), CK1 (Casein kinase 1) CMGC (Containing CDK, MAPK, GSK3, CLK families), STE (Homologs of yeast Sterile 7, Sterile 11, Sterile 20 kinases), TK (Tyrosine kinase), and TKL (Tyrosine kinase-like). An additional group includes atypical families of kinases, containing for instance PI3K-related kinases such as mTOR, DNA-PK, ATM, ATR, and others (Endicott et al., 2012). Protein kinases are enzymes that catalyze the transfer of a phosphoryl group from the γ- phosphate of ATP to the hydroxyl group of serine, threonine, or tyrosine of a protein substrate according to the following reaction: Substrate-OH + ATP-4Mg+2 à Substrate-O-PO3-2 + ADP-3Mg+2 + H+ Structurally, protein kinases contain in general two lobes in their kinase domain. The N- terminal lobe is composed of a β-sheet and a single α-helix, known as the αC helix, while the C-terminal lobe is larger and is composed mainly of α-helix. These two lobes create a cleft, where the active site resides. The C-terminal lobe also contains the activation loop between two highly conserved segments: DFG and APE motifs. The Asp (D) residue form the DFG motif chelates an Mg+2 ion to orientate ATP in the active conformation (Endicott et al., 2012). In the inactive conformation, the interaction between Asp and the Mg+2 ion is disrupted and the Phe (F) from the DFG motif is turned towards the ATP site. Mechanistically, the catalysis of the γ-phosphate of ATP requires a precise molecular mechanism (Figure 9). First, the –OH group of the substrate is aligned in a manner in 47 which the lone of pair electrons are pointed towards the γ-phosphate of ATP and carry a nucleophilic attack. This creates a metaphosphate intermediate in which the negative charge on the γ-phosphate is compensated by the Mg+2 and the neighbor lysine, K168 in the case of PKA. During the reaction, the pKa of the hydroxyl group decreases below the pKa of the catalytic aspartate allowing the transfer of a proton from the hydroxyl group to the aspartate. The figure below describes the mechanism of catalysis exemplified in PKA, a widely studies protein kinase. Although all the protein kinases have the ability to transfer γ-phosphate of ATP to the hydroxyl groups of Ser, Thr, or Tyr, not all of them have the ability to recognize the same substrates. This is in part due to the structural and electrostatical properties of the kinase and the substrate. In general, kinases require mechanisms of activation in order to increase their catalytic activity. In some cases, when kinases are constitutively active their mechanism of activation relies mainly in the transcriptional regulation or the limited substrate availability. On the other hand, many kinases require upstream phosphorylation in activation segments that are needed for a structural rearrangement leading to an active conformation. The AGC family of kinases is a highly conserved family of Ser/Thr protein kinases that are highly involved in the regulation of cell growth and survival, and play key roles downstream of mitogenic signals including PI3K and RAS. Among the members of this Figure 9. Phosphoryl transfer reaction of kinases In green, the hydroxyl group of the substrate is shown. In red, the three different phospho groups of ATP are shown and in pink the two Mg+2 ions. Source: (Endicott et al., 2012) 48 family, we find AKT, S6K, RSK, SGK, PKC, and other kinases that are important in cellular stress signaling (Pearce et al., 2010). Due to the relevance of this family of kinases for the work performed in this thesis, I will discuss the different mechanisms of activation known for AGC kinases in the context of PI3K signaling. AGC kinases contain two conserved motifs in which its phosphorylation is required for the optimal enzymatic: hydrophobic motif and activation loop (Arencibia et al., 2013) (Figure 10). Some of these sites are indicated below: Although these two highly conserved regions seem to dictate most of the activity of these kinases, there are other regulatory domains that can be important for the kinase activity in vivo. For instance, AKT contains a PH domain that interacts with PIP3 and is required for the subsequent phosphorylations at these motifs. Analogously, SGK3 contains a PX domain, which binds to PI(3)P but not PIP3, allowing the localization on the endosomes (Bago et al., 2014). This localization is also required for the kinase activity of the enzyme, since mutations that abrogate the binding to PI(3)P fail to activate SGK3 activity. The phosphorylation at the hydrophobic motif is mediated by different kinases including mTORC1 (S6K), mTORC2 (AKT, SGK, and PKC), ERK (RSK and MSK), and others, Figure 10. Conserved regulatory motifs in AGC kinases The sequence for the activation loop and hydrophobic motifs for the different AGC kinases are shown. In red, the phosphorylation residue is highlighted and in blue the phosphomimetic residue found in some AGC kinases is shown. Source: (Storz and Toker, 2002). 49 whereas the phosphorylation at the activation loop seems to be restricted to PDK1 (Cohen et al., 1997). PDK1 is a constitutively active Ser/Ther kinase that phosphorylates the activation loop of at least 23 AGC kinases using the consensus substrate motif (S/T)FCGT (Arencibia et al., 2013; Pearce et al., 2010). Although PDK1 is constitutively active and found thorough the cell, only phosphorylates its substrates when these have been primed appropriately. For most of the AGC kinases, this priming mechanism results from the phosphorylation at the hydrophobic motif that creates a docking platform for PDK1 to interact with (Biondi et al., 2000; Biondi et al., 2001; Collins et al., 2003). In other cases, the priming mechanism is mainly based on the localization. As mentioned before AKT and SGK3 are two examples of kinases that require subcellular mobilization for the appropriate phosphorylation by PDK1 at the activation loop and hence optimal kinase activity. In the presence of PIP3 in the membrane or PI(3)P in the endosome, AKT or SGK3 will translocate respectively. In that location these kinases interact with the hydrophobic motif kinase and PDK1 activates them (Arencibia et al., 2013). The mechanism by which PDK1 phosphorylates AKT was elucidated with the discovery of its PH domain, which interacts to PIP3 with more affinity than the AKT PH domain (Currie et al., 1999). On the other hand, the mechanism by which PDK1 phosphorylates kinases that are primed at the hydrophobic motif was later proposed based on the results from a yeast two-hybrid screening that reveled that PDK1 interacts with a region of the protein kinase C-related kinase-2 (PRK2), termed the PDK1 interacting fragment (PIF) (Balendran et al., 2000). Several evidences revealed that this fragment was conserved in several AGC kinases and was acting as a docking motif. Later, structural observations obtained from the PDK1 high-resolution crystal structure revealed the presence of a pocket that was consistent with the previously described PIF-binding pocket (Biondi et al., 2000). Upon phosphorylation at the hydrophobic motif, PDK1 interacts with the AGC kinase using the PIF-binding pocket and phosphorylates the activation loop to fully activate the kinase (Pearce et al., 2010). Mutations in this PIF- binding pocket (mainly at residue L155) abrogate the interaction with the phospho- peptide and fail to activate the AGC kinases regulated by this mechanism (Collins et al., 2003). 50 An example for this activation mechanism is S6K, the main substrate of mTORC1. Upon mitogenic stimulation, S6K is phosphorylated by mTORC1 at the hydrophobic motif T389. This primes the kinase for PDK1, interacting by means of the PIF-binding pocket, facilitating the phosphorylation at the activation loop (Alessi et al., 1998). It is believed that upon activation loop phosphorylation, AGC kinases are more stable and the phosphorylations are less exposed to phosphatases. Similar is the case for RSK, although in this case the hydrophobic motif phosphorylation is carried out by ERK1/2 instead of mTORC1. Again, this phosphorylation allows the docking to PDK1 inducing the maximal activity upon phosphorylation (Figure 11). Importantly, genetic studies using PDK1 knockout mice have revealed that these animals die at embryonic day E9.5 due to severe abnormalities in their somites, forebrain, and neural crest (Lawlor et al., 2002). However, mice containing hypomorphic alleles that only express 10% of PDK1 are enough to activate downstream AGC kinases Figure 11. Mechanisms of AGC kinase activation Three different mechanisms of AGC kinase activation are shown. Activation of AKT (PKB) requires the plasma membrane translocation that is mediated by the high-affinity interaction between the PH domain and the lipid messenger PIP3. In the case of S6K, hydrophobic motif phosphorylation is primed by mTORC1-dependent phosphorylation. For RSK, RAS-dependent signaling is required for its activity, since ERK1/2 kinase is responsible for the phosphorylation at the hydrophobic motif. Source: (Collins et al., 2005). 51 at the same level as wild-type mice (Bayascas et al., 2005; Lawlor et al., 2002). This indicates that robust inhibition of PDK1 is required in order to obtain strong phenotypes. PDK1 knockout mice also exhibit decreased cell size, probably as a result of the AKT inhibition, a result in line with the phenotypes reported for the PH domain mutant (K465E) knock-in mice, which fails to activate AKT in the presence of Insulin (McManus et al., 2004). Since many AGC kinases are involved in cancer and PDK1 knockout mice do not show activation of any of these kinases, one might speculate that PDK1 is a suitable target strategy for cancer therapy (Vanhaesebroeck and Alessi, 2000). Instead, pharmacologic inhibition of PDK1 exhibits a marked reduction of some AGC kinases activity, but not AKT (Najafov et al., 2011; Peifer and Alessi, 2008). This is a result of a mechanism of resistance to PDK1 inhibitors, which can be explained by the fact that AKT can also be activated by taking advantage of the high affinity PH domain and become phosphorylated by mTORC2. Once mTORC2 phosphorylates the hydrophobic motif of AKT, the phosphorylation can also serve as a mechanism to improve affinity to PDK1 using the PIF-binding pocket (Najafov et al., 2012). Therefore, in order to pharmacologically inhibit PDK1 and downstream AKT in cancer cells, it is required that very high doses of drug is used, limiting the application of these compounds in the clinical setting. This can be partially solved by using a combinatorial approach with an mTOR kinase inhibitor, which would avoid the mechanism of resistance, or similarly by reducing the PIP3 levels in the membrane using PI3K inhibitors. These observations are also consistent with the observations made with the hypomorphic mice, in which full reduction of PDK1 levels are required to observe a phenotype on downstream AKT (Lawlor et al., 2002). Further work is required to better understand PDK1 inhibitors and design treatment protocols that yield clinical responses in patients. Although several PDK1 inhibitors are currently available, only GSK2334470 (GlaxoSmithKline) appears to have significant selectivity and potency towards PDK1 (Figure 12). The IC50 for this compound in vitro is 10 nM and it does not inhibit the activity of related AGC kinases such as AKT, S6K, RSK, and SGK, or lipid kinases of the PI3K family and mTOR (Najafov et al., 2011). Treatment of cells with GSK2334470 decreases the phosphorylation at the activation loop of many AGC kinases and, as a consequence, inhibits the activity of these kinases. On the other hand, as discussed above, AKT activation loop phosphorylation is more 52 resistant to PDK1 inhibition and AKT remains active at concentration in which the other AGC kinases are inhibited. Currently, in cells there are not known genetic determinants of sensitivity or response to the compound. Figure 12. PDK1 inhibitor GSK2334470 chemical structure Another AGC kinase that is relevant for the purpose of this thesis is the serum and glucocorticoid-induced kinase (SGK). The SGK family is composed of three members SGK1-3 and was discovered as a kinase that displayed increased activity upon corticosteroid treatment. These Ser/Thr kinases are in fact very similar to the AKT kinase domain, sharing about 60% of homology. Different regulatory domains are present in these kinases, for instance they lack the PH domain present in all the isoforms of AKT. However, the mechanism of activation of SGKs is similar to AKT, since they require both mTORC2 and PDK1 as the hydrophobic motif and activation loop kinases, respectively. In fact these upstream regulators of SGK1 are highly conserved and the orthologue for these molecules have similar functions in lower species. Interestingly, in the budding and fission yeast (Saccharomyces cerevisiae and Schizosaccharomyces pombe) the SGK1 orthologues Ypk2 and Gad8 are key proteins responsible to maintain several biological processes (Kamada et al., 2005; Matsuo et al., 2003). In these organisms there are no AKT orthologues, probably because they do not have the ability to synthesize PIP3 (Figure 12). 53 Figure 13. Conserved mTORC2 signaling in lower species mTORC2 complex from different species is shown. Although these complexes are slightly different in terms of members, the main downstream substrate is the SGK1 orthologue. Similarly, PDK1 is also highly conserved and both kinases regulate the phosphorylation events of SGK1. Source: modified from (Cybulski and Hall, 2009). 54 7. Genetic alterations in the PI3K pathway During the last decades, since the publication of the first complete human genome, the technology for high throughput DNA sequencing has improved substantially. Several platforms have been developed to efficiently decipher the genetic code for virtually any sample (Shen et al., 2015). These platforms have different technologies for the sample preparation and sequencing and as a consequence its pros and cons. There are currently four widely used platforms for next generation sequencing (NGS) that include: Illumina (Solexa), Roche 454, Ion torrent: Proton / PGM, and SOLiD sequencing. Because the technologies that these platforms use are vey different we have summarized the major characteristics of each one in the following table: Table 2. The main characteristics for the different high throughput DNA sequencing techniques are detailed. Source: (Metzker, 2010). 55 As a result of this increasing efficiency in the DNA sequencing and the reduction in the cost of the technology, several consortiums have been founded in order to sequence the genomes of many disease and species. A very relevant example for the field of the oncology is The Cancer Genome Atlas (TCGA), a consortium that is funded by the National Institute of Health (NIH) and is focused on providing comprehensive genomic analysis of the main histological tumor types. So far, the TCGA has been able to sequence a large number of cases and reported the genomic landscape of the following malignancies: Breast (Breast Ductal Carcinoma and Breast Lobular Carcinoma) Central Nervous System (Glioblastoma Multiforme and Lower Grade Glioma) Endocrine (Adrenocortical Carcinoma, Papillary Thyroid Carcinoma, and Paraganglioma and Pheochromocytoma) Gastrointestinal (Cholangiocarcinoma, Colorectal Adenocarcinoma, Liver Hepatocellular Carcinoma, Pancreatic Ductal Adenocarcinoma, Stomach-Esophageal Cancer) Gynecologic (Cervical Cancer, Ovarian Serous Cystadenocarcinoma, Uterine Carcinosarcoma, Uterine Corpus Endometrial Carcinoma) Head and Neck (Head and Neck Squamous Cell Carcinoma, Uveal Melanoma) Hematologic (Acute Myeloid Leukemia and Thymoma) Skin (Cutaneous Melanoma) Soft Tissue (Sarcoma) Thoracic (Lung Adenocarcinoma, Lung Squamous Cell Carcinoma, Mesothelioma) Urologic (Chromophobe Renal Cell Carcinoma, Clear Cell Kidney Carcinoma, Papillary Kidney Carcinoma, Prostate Adenocarcinoma, Testicular Germ Cell Cancer, and Urothelial Bladder Carcinoma) In all these specimens the TCGA has performed and extensive analysis of the molecular signatures that includes whole genome/exome sequencing, RNA sequencing (gene expression and miRNA), methylation status, and proteomic reverse phase protein arrays (RPPA). The results can be found at the TCGA website (http://cancergenome.nih.gov), published in high impact journals, and can be easily visualized in the cBioPortal, a website hosted by MSKCC that includes user-friendly interphases to visualize and 56 analyze the TCGA and other high throughput sequencing studies related to cancer (http://www.cbioportal.org). The results from the large sequencing cohorts have revealed the presence of multiple genetic alterations that are present in the PI3K pathway (Figure 14). Now, we have also been able to identify the subset of tumors in which some mutations are more or less prevalent and we understand the main oncogenic drivers for each tumor type. Mutations in the PIK3CA gene are found in a large proportion of tumors, especially in breast cancers. 40% of hormone positive and 30% of Her2-positive breast cancers harbor mutations in PIK3CA. Among the mutations found in this gene, there are two identifiable hot spots in the helical and kinase domain that are indicated below: PIK3CA mutations are frequent in breast cancer TCGA cBioPortal Kinase Domain ≈ 42% Helical Domain ≈ 30% Gabelli et al. 2010 Curr Top Micro and Immunology Figure 14: Overview PIK3CA mutations in breast cancer The percentage of PIK3CA mutations (green) and amplifications (red) in breast cancer are shown. The different mutations across the domains of PI3Kα are indicated. Note the presence of the two main hot-spots in the helical and kinase domain, containing more than 70% of the total mutations in the PIK3CA gene. The structural effects of PIK3CA mutations are shown, while the helical mutants disrupt the inhibitory effect of p85, the mutations in the kinase domain impinge the catalytic loop into the membrane. Source: cBioPortal at MSKCC 57 a) Helical mutations: Generally affecting the E542 or E545 residues with substitutions to the aminoacid lysine. The PIK3CA E542 and E545K/Q mutations result in constitutive activation of PI3Kα enzymatic activity (Huang et al., 2008). This aberrant activation of PI3Kα promotes downstream activation of AKT and other PIP3 dependent effectors, eventually leading to increased cell survival, proliferation, and motility. In vitro experiments have demonstrated that this PIK3CA mutation results in increased anchorage independent colony formation and augmented tumor growth in vivo (Zhao and Vogt, 2008a, b). Helical domain PIK3CA mutations E542K and E545K account for approximately 10% and 20% of PIK3CA mutations in breast cancer respectively, occurring in approximately 3-5% and 2-6% of all breast cancers. Structurally, the change in the charge promotes repulsion with the inhibitory domains of p85 leading to an activation of the catalytic subunit (Huang et al., 2008). b) Kinase domain: Although the most common mutation found in this domain is the H1047R variant, other mutations have also been reported, including H1047L/Q/Y. This mutation constitutively activates the catalytic subunit of PI3K by increasing the exposure of the catalytic loop into the plasma membrane where its substrate, PIP2, is found. In mammary epithelial cells this mutation has been found to be oncogenic in nature, promoting the transformation (anchorage-independent growth and tumorigenicity in nude mice). In vitro experiments in chicken embryo fibroblasts also identified the variant H1047L as an activating mutation that results in constitutive activation of downstream AKT signaling in a similar fashion as the H1047R mutation (Huang et al., 2007; Zhao and Vogt, 2008a, b). Kinase domain PIK3CA mutations at amino acid H1047 account for approximately 40-60% of PIK3CA mutations in breast cancer, occurring in approximately 20% of all breast cancers. 58 8. PI3K inhibitors Because PI3K is a key integrator of signal transduction, mutations have been found in PIK3CA in a large proportion of tumors, and many RTKs are amplified in cancer, PI3K has emerged as a suitable target for cancer therapy. There are currently ~200 clinical trials testing the activity, safety, and efficacy of different PI3K inhibitors (www.clinicaltrials.gov) alone or in combination with other treatments. The first inhibitors of PI3K, isolated almost two decades ago, were not specific. They also targeted mTOR, DNA-PK and other phosphatidylinositol 3-kinase-related kinases (PIKK) due to the high similarity in the active site of these enzymes. Because PI3K uses ATP as a phosphate donor to phosphorylate the substrate PIP2, most inhibitors work as competitive ATP mimetics. The first known inhibitor of PI3K, wortmannin, is a natural furanosteroid derivative isolated from the fungus Penicillium wortmanii, which acts as a potent and irreversible pan-PI3K inhibitor and has an IC50 ≈ 4.2 nM (Figure 15) (Norman et al., 1996). Unfortunately, this inhibitor also targets other members of the PIKK family and MAPK. In fact, many studies in the field of PI3K have been carried out using this inhibitor as a tool compound, thus results from these experiments should be interpreted with caution. Similarly, viridin, a natural steroid derived from Trichoderma viridae has been shown to inhibit PI3K in a potent and irreversible manner with an IC50 ≈ 5 nM (Wipf et al., 2004). Using the structure of the furanosteroid wortmannin as a pharmacophore, multiple drugs have been developed including PX-866 (sonolisib) and WAY-176. Although these compounds are potent when assessed by their ability to inhibit phosphorylated AKT (Ser473), their specificity properties are not ideal, indicating that the reduction of AKT Ser473 phosphorylation might be a result of mTORC2 inhibition. However, phase I and II clinical trials have been pursued for PX-866 (Ihle et al., 2004). Another natural product that was shown to inhibit PI3K is the natural flavonoid quercetin. Quercetin is a natural flavonol found in many vegetables such as Capparis spinosa and Levisticum officinale. Using this molecule as a model, LY294002 (Eli Lily) was synthesized (Figure 15). This drug has been used as a tool for many years in the PI3K field; however, it shows low potency towards PI3K (IC50 ≈ 1.4 µM) and co-inhibits mTOR and DNA-PK (Vlahos et al., 1994). Recently, it has also been shown that 59 LY294002 can inhibit BET bromodomains (Dittmann et al., 2014). More importantly, LY294002 has been used as a chemical probe to create the beta-specific inhibitors TGX-115, TGX-126, TGX-221, and TGX-286 that were initially designed as anti- thrombotic agents due to the involvement of PI3Kβ in the homeostasis (Knight et al., 2004). These compounds were the first described isoform-specific PI3K inhibitors, and although the initial use was intended to block the coagulation process, evidence suggests that these compounds could be useful in the inhibition of PI3Kβ activation that is mediated by the activity of GPCRs or the loss of PTEN. 1, 4-morpholino-2-phenylquinazolines were also found to inhibit PI3K in a chemical screening. Importantly, some derivatives have the ability to selectively inhibit PI3Kα with an IC50 of 1.3 µM. This structure was also used as a common pharmacophore backbone to create novel PI3K inhibitors. For example, PI-103 is a potent (IC50 ≈ 3.6 nM) but not very selective inhibitor of PI3Kα and β (Hayakawa et al., 2007; Hayakawa et al., 2006). In fact, this compound also targets mTOR, making very challenging the interpretation of each signaling node into the contribution of any phenotype studied. Probably, the most clinically relevant drug derived from morpholino quinazolines is the thienopyrimidine GDC-0941 (Genentech), which is a pan-inhibitor (inhibits all the isoforms of PI3K) potent against PI3Kα (IC50 ≈ 3 nM) but less towards PI3Kβ (IC50 ≈ 33 nM) (Figure 15) (Folkes et al., 2008). Nevertheless, the concentrations used in cells probably inhibit both isoforms. This molecule has shown good pharmacokinetic properties in animals and humans and is currently in Phase II for metastatic ER+ breast cancer and non-small cell lung cancer (clinicaltrials.gov). A simple search of this compound in the portal clinicaltrials.gov shows 17 different clinical trials in which GDC- 0941 has been used as a single agent or in combination for different malignancies. The results from the Phase I clinical trial were recently published and GDC-0941 was found to be well tolerated and safe. The most common toxicities were maculopapular rash, fatigue, and nausea. Hyperglycemia is often found in patients treated with PI3K inhibitors, as PI3K is a key signaling molecule in the insulin pathway. These mild to moderate adverse events are in general manageable with supportive medication (Rodon et al., 2013). A similar structural pharmacophore is Imidazo[4,5-c]quinolone, which served as a template for the synthesis of BEZ-235 (Novartis), a compound with inhibitory properties 60 towards Class IA PI3K (IC50 ≈ 4, 75, and 7 nM respectively) and mTOR (IC50 ≈ 21 nM). About a dozen clinical trials have been conducted using BEZ235 and in general, toxicities have been a major limitation for this drug. This is likely due to the fact that BEZ235 is a potent mTOR inhibitor (Maira et al., 2008). NVP-BKM120 (Novartis) is another pan-PI3K inhibitor currently in late clinical development (Figure 15). It was originally derived from the 2-morpholino-6- aminopyridyl-pyrimidine scaffold, based on the structure of PI3Kγ (Bendell et al., 2012; Geuna et al., 2015). This dimorpholino pyrimidine derivative shows approximate equipotency towards all the Class IA isoforms of PI3K (IC50 PI3Kα ≈ 52 nM; PI3Kβ ≈ 166 nM; PI3Kδ ≈ 116 nM) and is currently in multiple Phase II and III trials for breast, prostate, endometrial, and lung cancer, among others (clinicaltrials.gov). Recently, results from the neoadjuvant phase II trial assessing efficacy of BKM-120 in combination with anti-Her2 therapy showed promising results that were somewhat diminished by the toxicities observed. Toxicities observed are a result of on-target effects, which is a significant limitation regarding the use of pan-PI3K inhibitors in combination with other therapies. To date, there are several pan-PI3K inhibitors that have been reported, most of them as a result of extensive medicinal chemistry to improve non-specific scaffolds described before, or using rational design when the structure of PI3K in complex with an inhibitor was available. Among these, SAR245408 (Sanofi-Aventis), GSK1059615 (GlaxoSmithKline), CH5132799 (Chugai), and BAY 80-6946 (Bayer) are promising molecules currently under clinical investigation (Rodon et al., 2013). With the elucidation of PI3K crystal structure in 1999 and the co-crystal structure of PI3Kα with the iSH2 domain of p85 in 2007, the design of PI3K inhibitors has improved, yielding compounds of higher potency and selectivity. For example, the development of PI3K isoform specific inhibitors has been important therapeutically. Both preclinical and clinical studies have shown that PIK3CA mutations are associated with response to PI3Kα inhibition (Fritsch et al., 2014). Taking into account that PIK3CA mutations are frequent in numerous cancers, it has been hypothesized that the use of alpha isoform- specific PI3K inhibitors might yield better clinical results than pan-PI3K inhibitors. Moreover, it is assumed that toxicities associated with inhibition of the other isoforms can be reduced. This safer profile would also allow the use of higher doses of the drug without unexpected off-site effects. 61 The first reported PI3Kα specific inhibitor was the 2-aminothiazole derivative NVP- BYL719 that inhibits the alpha isoform with an IC50 ≈ 5 nM (Figure 15) (Furet et al., 2013). It also shows similar potency towards the mutant versions H1047R and E545K. This compound is currently in several Phase II clinical phases for breast, head and neck, and gastrointestinal tumors (clinicaltrials.gov). The data from Phase I clinical trials suggest that BYL719 is able to reduce PI3K signaling in vivo, as established by pharmacodynamic studies of the tumors. Regarding toxicities, similar adverse effects have been observed as compared to pan-PI3K inhibitors, including decreased hyperglycemia, likely due to the lack of PI3Kβ inhibition. Similarly, INK1117 (also known as MLN-1117) has been reported as another specific alpha PI3K inhibitor and is Figure 15. Structural features of PI3K inhibitors Representative structures from different PI3K inhibitors are shown. Wortmannin and LY294002 are precursor molecules that were used as compound tools for several years. These have also been used as scaffold to create more specific and potent PI3K inhibitors. Pan-PI3K inhibitors GDC-0941 and BKM-120 are probably the most clinically experienced PI3K inhibitors with more that a dozen of trials as single agents or in combination. BYL719 and CAL101 are two isoform- specific PI3K inhibitors that target the alpha and delta isoform, respectively. CAL101 has been the first PI3K inhibitor to be approved by the FDA and BYL719 has shown remarkable clinical results in breast cancer. 62 currently under Phase I investigation. The results of this trial showed safety and tolerability of the compound and remarkable stability of the drug in plasma (Thorpe et al., 2015). Pharmacodynamic studies in skin biopsies showed a decrease in phosphorylated 4EBP1 and S6 riboprotein, two different surrogates of pathway activity. However, it is not clear whether the use of skin biopsies as a pharmacodynamic measure is sufficient to assess the activity of targeted therapy in clinical development. Recently, a new molecule named GDC-0032 has been reported as a PI3Kβ sparing inhibitor, indicating that its chemical design has been focused on reducing the PI3Kβ activity as much as possible but keeping the high potency towards PI3Kα isoform. This agent is a highly potent and selective inhibitor of PI3Kα and PI3Kδ (IC50 ≈ 0.3 nM and 0.12nM, respectively) showing 3-fold more potency towards the oncogenic mutants H1047R and E545K, an interesting property taking into account that these compounds are aimed to be used in tumors harboring these oncogenic mutations. GDC-0032 is currently in Phase I clinical trials and has shown remarkable results, leading to the decision to move forward to a Phase III clinical trial enrolling breast cancer patients with estrogen-positive disease and PIK3CA mutation (Figure 16). Figure 16: Clinical response to a PI3Kα inhibitor The results from the Phase I clinical trial of GDC-0032 in monotherapy are shown. The majority of responses observed during the trial are in breast cancer and/or patients with tumors harboring mutations in PIK3CA. Source: (Juric et al, unpublished results) 63 Although the contribution of the delta isoform of PI3K in solid tumors is not significant, it is worth to mention the PI3Kδ inhibitor CAL-101, a phenylquinazolin derivative that selectively inhibits the delta isoform of PI3K with an IC50 ≈ 70 nM, compared to the other isoforms (IC50 >1 µM) (Lannutti et al., 2011) (Figure 15). Recently, the FDA approved the use of CAL-101 (Idelalisib) for the treatment of relapsed CLL (chronic lymphocytic leukemia), SLL (small lymphocytic lymphoma) and FL (follicular lymphoma). This is the first PI3K inhibitor that has been approved by a regulatory agency and the clinical experience will certainly reveal novel insights into the pharmacology of these compounds. Name Target Company Clinical development Comments BYL719 PI3Kα Novartis Phase II INK1117 PI3Kα Millenium Pharmaceuticals Phase I GDC0032 PI3Kα Genentech Phase III A66 PI3Kα University of Auckland Pre-clinical Not very potent drug. CNIO-PI3Ki PI3Kα CNIO (Spanish Institute Oncology) Pre-clinical BKM-120 PI3K Novartis Phase III Higher doses of this compound exhibit antimicrotubule effects in cells. LY294002 PI3K, DNA-PK, others Lilly Preclinical Poor selectivity and solubility. Not very potent either. SF1126 PI3K/mTOR Semafore Phase I Prodrug for LY294002 PWT-458 PI3K Wyeth/Pfizer Preclinical PX-866 PI3K Oncothyreon Phase I Irreversible PI3K inhibitor XL-147 PI3K Exelixis (Sanofi-Aventis) Phase I GSK1059615 PI3K GlaxoSmithKline Phase I ZSTK474 PI3K Zenyaku Kogyo Preclinical CH5132799 PI3K Chugai Phase I GDC-0941 PI3K Genentech Phase II PF-4989216 PI3K Pfizer Pre-clinical GNE-317 PI3K Genentech Pre-clinical Developed as a brain-permeant PI3K inhibitor. BAY 80-6946 PI3K Bayer Phase I Intravenously administered BGT226 PI3K/mTOR Novartis Phase I Development discontinued. BEZ235 PI3K/mTOR Novartis Phase II Discontinued. Poor pharmacokinetics. SAR254409 PI3K/mTOR Sanofi Phase I Wortmannin PI3K/mTOR Preclinical Irreversible inhibitor. Not specific. GDC-0980 PI3K/mTOR Genentech Phase I PI-103 PI3K/mTOR Piramed Pharma (Roche) Pre-clinical Poor pharmacokinets. Not selective. Discontinued. PF-05212384 PI3K/mTOR Pfizer Phase I Intravenous. PF-04691502 PI3K/mTOR Pfizer Phase I SB2343 PI3K/mTOR Singapore Biostar Pre-clinical GSK2126458 PI3K/mTOR GlaxoSmithKline Phase I PKI-402 PI3K/mTOR Wyeth/Pfizer Pre-clinical PI3Kα isoform specific inhibitors pan PI3K inhibitors Dual PI3K/mTOR inhibitors Table 3. Inhibitors that represent a therapeutic approach for PIK3CA-mutant tumors are described. Note that the list of isoform specific inhibitors is reduced, although preliminary clinical trials suggest that inhibition of this isoform could result in increased clinical benefit. Source: modified from (Rodon et al., 2013) 64 9. Other inhibitors targeting the PI3K/AKT/mTOR pathway 9.1. mTORC1 inhibitors Much of the knowledge about mTORC1 comes from early studies that elucidated the mechanism of action of rapamycin, an allosteric inhibitor that directly binds to mTORC1 and inhibits downstream phosphorylation of its substrates. Specifically, rapamycin binds to the immunophilin FKB12 to generate a highly specific complex that binds to the FKB (FKBP12-rapamycin-binding) domain of mTOR (Laplante and Sabatini, 2012; Sabatini, 2006). In contrast to its effects on mTORC1, rapamycin cannot interact with mTORC2. For this reason, mTORC1 has been known as rapamycin-sensitive complex and mTORC2 as rapamycin-insesitive complex. This paradigm, however, may not be entirely accurate, as it has been described that prolonged treatment with rapamycin in some cells inhibits mTORC2 activity (Sarbassov et al., 2006). This drop in mTORC2 activity seems sufficient to inhibit AKT signaling. Thus, rapamycin is considered to be a universal inhibitor of mTORC1 and seems to be a cell-type specific inhibitor of mTORC2 on a rapamycin dose-dependent manner. The pharmacological properties of rapamycin are not optimal for clinical development and several rapamycin analogues, termed rapalogues, have been developed. These analogues, such as temsirolimus, everolimus, and ridaforolimus, exhibit higher solubility and better pharmacokinetic properties than rapamycin (Guertin and Sabatini, 2009). Temsirolimus has been approved by the FDA for treating renal cell carcinoma based from a phase III study that demonstrated improved overall survival among patients with metastatic renal-cell carcinoma when compared to interferon alpha or both drugs combined. In Europe, temsirolimus has also been approved for use in mantle cell lymphoma where it significantly improved PFS and objective response rate compared with investigator’s choice therapy in patients with relapsed or refractory disease in a phase III study. Temsirolimus was also used in combination with letrozole in patients with metastatic breast cancer in a phase III study. However, the combination of drugs did not show benefit over letrozole (aromatase inhibitor) alone. Everolimus achieved regulatory approval for use in pancreatic neuroendocrine tumors based on a phase III trial where it prolonged PFS when compared to best supportive care (Yao et al., 2011). A further phase III study demonstrated superiority for everolimus 65 over placebo in patients with metastatic renal cell carcinoma (mRCC) progressing on vascular endothelial growth factor receptor-tyrosine kinase inhibitors, leading to approval for this indication (Motzer et al., 2010). Everolimus has also been approved by the FDA for use in combination with anti-estrogen therapy in hormone-receptor positive HER2- negative breast cancer (Baselga et al., 2012). Ridaforolimus, formerly known as deferolimus, has no current approved indications. All three agents are currently under investigation across many clinical trials (Dienstmann et al., 2014). Despite a plethora of preclinical data on rapamycin and its analogs, these molecules have not shown universal anti-tumor activity in clinical trials. An important limitation of the rapalogs is the paradoxical increase in the AKT activity, which can have therapeutic contraindications (O'Reilly et al., 2006). This is only observed in cells in which rapalogs only have the ability to inhibit mTORC1 but not mTORC2. S6K (one of the key substrates of mTORC1) phosphorylates the ribosomal S6 protein, which in turn phosphorylates and inhibits IRS1, the adaptor protein linking the IGF-1 receptor and PI3K. This effect leads to a reduction of input into the PI3K pathway coming from the stimulation of the insulin/IGF-1 receptors. The inhibition of mTORC1 releases the S6K- IRS1-PI3K feedback inhibitory loop and results in increased AKT activity. Thus additional targeting of other key members of the pathway would be required to overcome the effects of this feedback. 9.2. AKT inhibitors The discovery of perifosine was probably one of the first AKT inhibitors, characterized by the alkylphospholipid structure. Structurally, perifosine resembles phospholipids, leading to the disruption of the normal PIP3 signaling upon growth factor stimulation. Perifosine has been shown to cause dose-dependent inhibition of AKT phosphorylation and activity in pre-clinical models of PI3K-induced tumors. A myristoylated form of AKT is able to bypass the effect of perifosine, indicating that this molecule probably acts by competing with endogenous AKT to translocate into the plasma membrane (Richardson et al., 2012). Other inhibitors have been discovered by their ability to compete with ATP and inhibiting the kinase activity of the protein. Because AKT is highly homologous across the three isoforms, most of these inhibitors have the ability to inhibit them all. 66 AKT catalytic inhibitors that are actually under clinical development include AZD5363 and GDC-0068 (Rodon et al., 2013). Pre-clinical models suggest that sensitivity to AKT- specific inhibitors is dependent upon activation of the PI3K/AKT pathway, however initial clinical trials with these compounds were disappointing. In fact, the few responses observed in the clinical trial of AZD5363 exhibited the AKT1 activating mutation E17K. The E17K mutation has been characterized structurally and biochemically, and has the ability to induce a stronger plasma membrane translocation as a result of increased exposure of the pleckstrin homology domain (Carpten et al., 2007). In cell lines, these inhibitors show activity in models of this mutation, and this has been the rational for the design of basket trials that select patient population based on the status of AKT. Another inhibitor that has been extensively studied is the allosteric inhibitor MK2206. This inhibitor has the ability to “lock” AKT in the inactive conformation, hampering the conformational change required for the full activation. In fact, the mutant W80A decreases the inhibitory effects of MK2206 (Hirai et al., 2010). The three AKT inhibitors mentioned before, MK-2206, AZD5363, and GDC-0068, have shown activity in cell lines containing PIK3CA mutations or PTEN loss. Interestingly, AKT1 E17K mutant cell lines are resistant to MK2206 consistent with the mechanism of action of the drug (Dienstmann et al., 2011). Early-phase trials with suggest that also PTEN loss could be a good indicator of sensitivity to AKT inhibitors, although there is also clinical evidence suggesting that these inhibitors could work in patients with PIK3CA-mutant tumors. 9.3. mTOR catalytic inhibitors The findings that mTORC2 has a direct role in the activation of AKT, and the limited clinical antitumor activity of the rapalogs possibly through the consequences of feedback loops, has led to the development of ATP-competitive inhibitors of mTOR kinase that potently inhibit both mTORC1 and mTORC2 complexes (Guertin and Sabatini, 2005). Interestingly, these compounds have shown to inhibit mTORC1 more potently than rapalogs. For instance, the mTOR kinase inhibitor AZD8055 inhibits 4EBP1 phosphorylation more effectively than rapamycin and also effectively inhibits mTORC2 and AKT S473 phosphorylation (Chresta et al., 2010). However, the inhibition of AKT signaling seems transient as inhibition by AZD8055 causes activation of RTKs, which in 67 turn induces PI3K signaling and reactivates AKT activity and signaling. Combined inhibition of mTOR kinase and RTKs fully abolishes AKT signaling resulting in tumor regression (Rodrik-Outmezguine et al., 2011). INK128 is another mTORC1/2 inhibitor in which in vitro and in vivo data has demonstrated successful inhibition of mTORC1 (pS6 and p4EBP1) and mTORC2 (pAKT at S473) (Hsieh et al., 2012). Interestingly, this agent has also shown marked activity in cell lines that are resistant to rapamycin and pan-PI3K inhibitors. A phase I clinical trial testing the activity of this molecule is currently undergoing in patients with solid tumors. Other compounds that aim to inhibit both mTOR complexes and potentially have a more profound antitumor activity in comparison to rapalogs are AZD2014, CC-223, and OSI- 027 (Dienstmann et al., 2014). 9.4. Dual PI3K-mTOR inhibitors A set of inhibitors that have the ability to inhibit both mTOR and PI3K are available as a result of the overlapping homology of the kinase domains of these two PI3K-related kinase family (PIKK) members. Several mTOR/PI3K dual inhibitors, namely SF1126, BEZ235, XL765, GDC-0980, PF- 04691502, PKI-587, GSK2126458, BGT226, and PWT3359, have shown promising activity in preclinical models and are in early stage clinical trials (Rodon et al., 2013). For instance, BEZ235 has been reported to inhibit tumor growth in many preclinical models such as prostate, breast, pancreatic, renal cancers, multiple myeloma and sarcomas. BEZ235 potently inhibited several rapamycin-resistant colon cancer cell lines at low concentration. Some other findings among the list of preclinical data about this agent include breast cancer cell lines with HER2 amplification and/or PIK3CA mutations to be highly sensitive to BEZ235. These and other findings have allowed BEZ235 to enter Phase I/II clinical trials in patients with advanced solid tumors including breast cancer. Unfortunately, the physicochemical properties of this compound have complicated the clinical development of BEZ235, requiring different formulations to improve the pharmacology of the drug. XL765 is another pan-class I PI3K/mTOR inhibitor that is administered in a phase I dose-escalation study of patients with solid tumors. Tumors treated with this compound have shown decreased phosphorylation of several components of the PI3K/mTOR pathway including AKT and reduced tumor tissue proliferation. 68 Another dual PI3K/mTOR inhibitor, GDC-0980 has demonstrated broad preclinical activity in breast, ovarian, lung, and prostate cancer models. It has been shown to be active against tumor cells bearing mutations in PI3K, PTEN, or KRAS. It is currently in phase I clinical development. Furthermore, recent work from our laboratory has demonstrated that mTORC1 inhibition is required for sensitivity to PI3K isoform p110α inhibitors in PIK3CA-mutant breast cancer. Our study demonstrated that breast cancer cell lines and patient tumors resistant to the PI3K inhibitor BYL719 have active mTOR signaling. Sustained mTORC1 activation may limit the effects of BYL719 against tumor growth. This hypothesis was confirmed by adding RAD00 (everolimus) to BYL719 insensitive cells resulting in reversal of resistance both in vitro and in vivo (Elkabets et al., 2013). Additionally, work from our laboratory has shown that progressive loss of PTEN leads to clinical resistance to the PI3Kα inhibitor BYL719. Specifically, loss of PTEN in BYL719-sensitive cell lines and in patients that progress to this agent leads to resistance to PI3Kα inhibition (Juric et al., 2015). Since PTEN deficient preclinical models have shown to mostly rely on the p110β subunit of the PI3K holoenzyme, the inhibition of both p110α and p110β isoforms of PI3K re-sensitized the cells to BYL719, in a similar fashion as it would be expected for dual PI3K/mTOR inhibition. As mechanisms of resistance to PI3K inhibitors might emerge or in some cases tumors are intrinsically resistant to these compounds, there is increasing evidence suggesting that these therapies will be required to be combines with other inhibitors. In the field of breast cancer, combination with anti-estrogen therapy seems to be an exciting approach for ER positive tumors, especially taking into account that in this subtype of up to 40% of the tumors harbor PIK3CA mutations (Juric and Baselga, 2012). Other combinations that may be beneficial include RTK antibodies or inhibitors in cases where these receptors are amplified, such as Her-2 or EGFR positive tumors. Current work is required to understand which mechanisms of resistance are associated to these compounds and which drug combinations will be the most appropriate to increase the clinical benefit. 69 In the following table, the drug combinations that are currently being tested with pan- PI3K inhibitors in the clinic are shown: In the following table, the drug combinations that are currently being tested with isoform- specific PI3K inhibitors in the clinic are shown: Genomic Alterations in the Pathway: PIK3CA and PTEN PIK3CA, the gene encoding p110a, is the most commonly mutated gene among the components of the PI3K pathway.45 Although there are 3 main mutation hotspots within the helical (E545 and E542) and kinase (H1047) domains,46 other muta- tions can be found across the whole gene. The E542K and E545K mutations located in the helical domain induce an impor- tant electrostatic switch in these residues by reversing the ionic charge. This modifies the interaction of the N-terminal SH2 domain of the regulatory subunit p85 with the helical and kinase domain of p110a, comparable to binding at phospho-Y resi- dues.47,48 In the case of the H1047R mutation within the kinase domain, it has been established that this alteration increases the kinase activity by inducing a new orientation of the C-terminal loop toward the plasma membrane, in which the active site has better access to its substrate PIP2. 49,50 Other mutations include the R39 (R39C and R39H) and the R88Q mutations that occur in the ABD domain. These muta- tions are reported to disrupt the interaction between the ABD and kinase domains of p110a, thus augmenting its activity. In addition, 2 mutations within the C2 domain, N345L and E453Q, have been proposed to alter the interaction between this domain and the iSH2 domain of the regulatory subunit, again increasing the kinase activity of p110a.47 Another well-studied alteration that can lead to hyperactiva- tion of the pathway is loss of function of the tumor suppressor PTEN.51 Low PTEN phosphatase activity results in increased levels of PIP3 and consequent activation of downstream PI3K effectors such as AKT and mTORC1 (Fig. 2). PTEN is a 55- kDa enzyme containing, among other structural motifs, a phos- phatase domain that controls the catalytic activity of the enzyme and a C2 domain that is responsible for lipid binding.10 Somatic mutations of PTEN are found throughout the entire gene, although there is a slightly higher frequency at the R130 residue. Somatic PTEN mutations are relatively frequent in endometrial carcinoma and glioblastoma, and PTEN copy num- ber loss is common in prostate, breast, and ovary cancer and glioblastoma.52,53 PI3K inhibitors There are currently approximately 200 clinical trials testing the activity, safety, and efficacy of different PI3K inhibitors (www.clinicaltrial.gov). The first inhibitors, isolated almost 2 decades ago, were not specific for PI3K and commonly inhibited other kinases, especially phosphatidylinositol 3-kinase-related kinases (PIKK) such as mTOR and DNA-PK, which contain structurally similar active sites.54 Because PI3K uses ATP as a phosphate donor to phosphorylate the substrate PIP2, most inhibitors work as competitive ATP mim- etics.55 The first known inhibitor of PI3K, wortmannin, is a fura- nosteroid derivative isolated from the fungus Penicillium wortmanii, which acts as a potent and irreversible pan-PI3K inhib- itor (IC50 ! 4.2 nM).56 Unfortunately, this inhibitor also targets other members of the PIKK family and MAPK. Similarly, viridin, a natural steroid derived from Trichoderma viridae, has been shown to inhibit PI3K in a potent and irreversible manner (IC50 !5 nM).57 Using the structure of the furanosteroid wortmannin as a pharmacophore, multiple drugs have been developed includ- ing PX-866,58 which is currently in Phase I/II trials. Another natu- ral product that was shown to inhibit PI3K is the flavonoid quercetin,55 which was used as a model in the synthesis of LY294002. Although LY294002 has been used for many years as a tool in the field of PI3K research, it actually has relatively low potency toward PI3K (IC50 ! 1.4 mM) and co-inhibits mTOR and DNA-PK.59 LY294002 has also been used as a model to cre- ate the b-specific inhibitors TGX-115, TGX-126, TGX-221, and TGX-286, which were initially designed as antithrombotic agents because of the involvement of PI3Kb in hemostasis.60 Another molecule, phenylquinazoline, inhibits PI3Ka with an IC50 of 1.3 mM, and is also a common pharmacophore used as a back- bone to create novel PI3K inhibitors such as PI-103, a potent (IC50 ! 3.6 nM) but not very selective inhibitor of PI3Ka and b.61,62 The most clinically promising drug derived from a mor- pholino quinazoline is probably the thienopyrimidine GDC- 0941, which is considered a pan PI3K inhibitor although it prefer- entially targets PI3Ka (IC50 ! 3 nM).63 This molecule has shown good pharmacokinetic properties in animals and humans and is currently in Phase II trials for metastatic estrogen receptor (ER)- positive breast cancer and non-small cell lung cancer. Table 1. Combinations with pan-PI3K inhibitors under clinical development Drug Target Company Phase Combinations CH5132799 PI3K Chugai I CLR457 PI3K Novartis I GDC-0941 PI3K Genentech I/II Cisplatin, paclitaxel, fulvestrant, erlotinib, bevacizumab, GDC0973. NVP-BKM120 PI3K Novartis II Radiotherapy, Chemotherapy (paclitaxel, irinotecan, carboplatin, gemcitabine, etc), RTKi (Gefitinib, Erlotinib, Cetuximab, lapatinib, rituximab), endocrine therapy (fulvestrant, abiraterone) imatinib, bevacizumab, olaparib, CDKi (LEE011), MEKi (MEK162, GSK1120212), everolimus, LDE225, INC280, BRAFi (encorafenib, vemurafinib). PX-866 PI3K Oncothyreon I/II Docetaxel, cetuximab, vemurafenib. SF1126 PI3K Semafore I XL147 PI3K Exelixis I/II Paclitaxel, carboplatin, letrozole, trastuzumab, erlotinib, XL647. ZSTK474 PI3K Zenyaku Kogyo I/II e963447-4 Volume 1 Issue 3Molecular & Cellular Oncology Do wn loa de d b y [ M em ori al Slo an K ett eri ng L ibr ary ] a t 0 7:1 5 1 2 J an ua ry 20 15 phosphorylation.82 However, the inhibition of AKT signaling seems to be transient, as inhibition by AZD8055 causes activation of RTKs, which in turn induce PI3K signaling and reactivate AKT activity and signaling.82 Combined inhibition of mTOR kinase and RTKs fully abolishes AKT signaling resulting in tumor regression.83 INK128 is another mTORC1/2 inhibitor for which in vitro and in vivo data have demonstrated successful inhibition of mTORC1 (S6K and 4EBP1) and mTORC2 (AKT at S473).84 Interestingly, this agent has also shown marked activity in cell lines that are resistant to rapamycin and pan-PI3K inhibitors.71 A Phase I trial testing the activity of this molecule is ongoing in patients with solid tumors. Other compounds that aim to inhibit both mTOR complexes and potentially have a more profound antitu- mor activity than rapalogs are AZD2014, CC-223, and OSI-027. Combinations of PI3K Inhibitors with RTK Inhibitors HER2-targeted therapies have produced clinical improve- ments in patient survival, both in the adjuvant and metastatic set- ting. However, despite initial responses the emergence of resistance occurs in the vast majority of patients with metastatic disease. Activation of the PI3K pathway, either by loss of PTEN or by the presence of PIK3CA mutations, is perhaps the most widely accepted mechanism of resistance to anti-HER2 ther- apy.84-88 Activating mutations in PIK3CA occur in approxi- mately 25–30% of HER2-amplified breast cancers (TCGA) and may be important for response to HER2-targeted therapies. For example, introduction of a PIK3CAH1047R activating mutation in HER2-driven mammary tumors in MMTV/neu transgenic mice accelerates tumor onset and progression, and generates resis- tance to anti-HER2 therapy.89 This evidence paved the way to investigate the efficacy of therapeutic strategies combining anti- HER2 agents with PI3K inhibitors. In preclinical models, this strategy proved to be synergistic in cells or tumors resistant to anti-HER2 therapy.84,86,90–92 In a clinical setting, trastuzumab in combination with everolimus showed promising antitumor activity in heavily pretreated patients who had progressed to tras- tuzumab-based therapy.93-95 Further, in a recent study testing the effect of the PI3K inhibitor NVP-BKM120 in a similar pop- ulation, researchers reported clinical responses in patients with tumors possessing an activated PI3K pathway.96 Additional trials testing other PI3K inhibitors in combination with anti-HER2 agents are currently ongoing. Interestingly, a recent study highlighted the need to design tri- ple inhibition strategies to effectively delay and/or avoid tumor relapse, especially in patients with large tumor burdens.97 Although toxicity remains a limiting factor when translating these combinatorial therapeutic strategies to patients, concomi- tant treatment of HER2-driven cancers with synergistic combina- tions targeting the PI3K/AKT pathway and additional driver events might render cancer cells vulnerable to efficient eradica- tion and lead to curative regimes for patients. Treatment with PI3K/AKT inhibitors leads to upregulation/ activation of RTKs such as EGFR, HER3, and IGFR1 that, in turn, can limit the antitumor effects of these therapies by increas- ing PI3K signaling or triggering the activation of other compen- satory pathways.2,4,82,98 This provides the rationale to test combinations of PI3K inhibitors with agents that block the activ- ity of RTKs. Due to space limitations we are unable to list the large (and increasing) number of preclinical studies conducted in this regard. Likewise, many clinical trials testing the efficacy of PI3K inhibitors and anti-RTK agents (including EGFR, HER2, HER3, FGFR, IGFR, and CD20) have been launched. Details are available at www.clinicaltrial.gov and are summarized in Table 1 and Table 2 . Combinations of PI3K Inhibitors with Endocrine Therapy Up to 75% of breast tumors are hormone receptor (HR)-posi- tive, expressing estrogen receptor (ER), progesterone receptor (PR), or both. These nuclear receptors are both targets and pre- dictors of response to anti-estrogen therapy such as selective estrogen receptor modulators (SERMs), aromatase inhibitors (AIs), or selective estrogen receptor degraders (SERDs). Despite the fact that HR-positive breast tumors generally have a favorable prognosis and effective therapies exist, up to one-third of women diagnosed with this type of cancer will relapse after 5 years of adjuvant therapy with tamoxifen99 and up to 20% of women treated with adjuvant AIs will undergo recurrence 10 years after the initial diagnosis.100 Several mechanisms of resistance to hor- mone therapy have been proposed101; however, a common theme emerging from many of these studies is the importance of the activation of growth factor receptors in endocrine resistance. Given that PI3K is the most frequently altered pathway in ERC breast tumors, deregulation of the PI3K pathway through Table 2. Combinations with isoform-specific PI3K inhibitors under clinical development Drug Target Company Phase Combinations CAL-101 p110d Gilead Sciences II/III/IV Chemotherapy (bendamustine), ofatumumab, rituximab, Bcl-2 inhibitor (GDC-0199). GDC-0032 p110a Novartis I/II Chemotherapy (docetaxel, paclitaxel), letrozole, fulvestrant. GSK2636771 p110b GlaxoSmithKline I IPI-145 p110d/g Infinity I/II/III Bendamustine, ofatu u a , rituximab. MLN1117 p110a Millennium I/Ib mTORi (MLN0128) NVP-BYL719 p110a Novartis I/II Chemotherapy (paclitaxel, gemcitabine, capecitabine), everolimus, RTKi (cetuximab, LJM716, AMG479, BGJ398), letrozole, T-DM1, encorafenib, MEK162, Hsp90i (AUY922), CDKi (LEE011), PIMi (LGH447). e963447-6 Volume 1 Issue 3Molecular & Cellular Oncology Do wn loa de d b y [ M em ori al Slo an K ett eri ng L ibr ary ] a t 0 7:1 5 1 2 J an ua ry 20 15 Table 4. List of pan-PI3K inhibitors that are currently being tested in the clinical setting in combination with other chemotherapeutic or targeted agents. Table 5. List of isoform specific PI3 inhibitor that are curre tl being tested in the clinical setting i combination with other chemotherapeutic or argeted ag nts. 70 71 OBJECTIVES 72 73 The main objectives of this thesis are to better understand the mechanisms of resistance to PI3K inhibitors in order to identify novel biomarkers of resistance and propose new drug combinations to overcome the resistant phenotype. These objectives are detailed as follows: 1) Characterize the mechanisms of sustained mTORC1 activation upon PI3K inhibition in BYL719 resistant cell lines - Perform a kinome and phosphatome RNAi screening to identify kinases or phosphatases involved in the mechanism of resistance and sustained mTORC1 activation in PI3Kα-resistant cell lines - Develop tools for the study of the kinase responsible for the resistant phenotype in cell lines and patient samples - Therapeutic strategies to overcome this resistance 2) Define the role of secondary AKT-like kinases as mechanisms of resistance to PI3K inhibitors - Use of biopsies and rapid autopsies as a model of acquired resistance to PI3Kα inhibitors in selected patients - Genomic characterization of the lesions known to be sensitive, intrinsically resistant, or acquired resistant to PI3Kα inhibitors - Generation of patient-derived xenografts to study the mechanism of resistant and the pharmacological intervention 3) Development of a mouse model as a pre-clinical tool to study the role of these kinases in vivo - Develop a new mouse model driven by the expression of oncogenic PIK3CA H1047R mutation - Test the efficacy of PI3K inhibitors - Study the role of PI3K kinase in the process of oncogenicity 74 75 RESULTS 76 77 Article 1: Convergent loss of PTEN leads to clinical resistance to a PI(3)Kα inhibitor Journal: Nature Impact Factor: 41.456 Article 2: PDK1-SGK1 signaling sustains AKT-independent mTORC1 activation and confers resistance to PI3Kα inhibition Journal: Cancer Cell (under review) Impact Factor: 23.523 Article 3: Somatic PIK3CA mutations as a driver of sporadic venous malformations Journal: Science Translational Medicine Impact Factor: 15.843 Other articles: Article 4: Antagonism of EGFR and HER3 Enhances the Response to Inhibitors of the PI3K-Akt Pathway in Triple-Negative Breast Cancer Journal: Science Signaling Impact Factor: 6.279 Article 5: PI3K inhibition results in enhanced estrogen receptor function and dependence in hormone receptor-positive breast cancer Journal: Science Translational Medicine Impact Factor: 15.843 Article 6: The tumor suppressor PTEN and the PDK1 kinase regulate formation of the columnar neural epithelium Journal: eLife Impact Factor: 9.322 78   79 ARTICLE 1 Title: Convergent loss of PTEN leads to clinical resistance to a PI(3)Kα inhibitor Journal: Nature Impact Factor: 41.456                                                                              80                                                                                               81   LETTER doi:10.1038/nature13948 Convergent loss of PTEN leads to clinical resistance to a PI(3)Ka inhibitor Dejan Juric1*, Pau Castel2*, Malachi Griffith3,4,5, Obi L. Griffith4,5,6, Helen H. Won2,7, Haley Ellis2, Saya H. Ebbesen8, Benjamin J. Ainscough5, Avinash Ramu5, Gopa Iyer2,9, Ronak H. Shah2, Tiffany Huynh1, Mari Mino-Kenudson1, Dennis Sgroi1, Steven Isakoff1, Ashraf Thabet1, Leila Elamine1,DavidB. Solit2,9, ScottW.Lowe8,10, CorneliaQuadt11,Malte Peters11, AdnanDerti12, Robert Schegel12, Alan Huang12, Elaine R. Mardis3,4,5,6, Michael F. Berger2,7, Jose´ Baselga2,13 & Maurizio Scaltriti2 Broad and deep tumour genome sequencing has shed new light on tumourheterogeneityandprovided important insights into theevolu- tion of metastases arising from different clones1,2. There is an addi- tional layerof complexity, in that tumour evolutionmaybe influenced by selective pressureprovidedby therapy, in a similar fashion to that occurring in infectious diseases. Here we studied tumour genomic evolution inapatient (indexpatient)withmetastaticbreast cancerbear- ing an activating PIK3CA (phosphatidylinositol-4,5-bisphosphate 3-kinase, catalytic subunit alpha,PI(3)Ka)mutation.Thepatientwas treatedwith thePI(3)Ka inhibitorBYL719,which achieved a lasting clinical response, but the patient eventually became resistant to this drug (emergence of lung metastases) and died shortly thereafter. A rapid autopsywas performed andmaterial from a total of 14metas- tatic sites was collected and sequenced. All metastatic lesions, when compared to thepre-treatment tumour, hada copy loss ofPTEN(phos- phatase and tensin homolog) and those lesions that became refract- ory toBYL719hadadditional anddifferentPTENgenetic alterations, resulting in the loss of PTENexpression.Toput these results in con- text, we examined six other patients also treatedwithBYL719.Acquired bi-allelic loss of PTENwas found in one of these patients,whereas in two others PIK3CAmutations present in the primary tumour were no longer detected at the time of progression. To characterize our findings functionally, we examined the effects of PTEN knockdown in several preclinicalmodels (both in cell lines intrinsically sensitive to BYL719 and in PTEN-null xenografts derived from our index patient), which we found resulted in resistance to BYL719, whereas simultaneousPI(3)Kp110b blockade reverted this resistance pheno- type.Weconclude thatparallel genetic evolutionof separatemetastatic sites with different PTEN genomic alterations leads to a convergent PTEN-null phenotype resistant to PI(3)Ka inhibition. We are currently engaged in testing the antitumour activity of a novel PI(3)Ka inhibitor, BYL719, in patients with tumours harbouring acti- vating PI(3)Ka mutations3. The PI(3)K pathway is essential for cell growth, proliferation, survival, andmetabolism4,5. ThePI(3)K family of enzymes isdivided into threemain classes (I to III),with class I being the most often implicated in human cancer6. Class IA PI(3)K is a hetero- dimer composed of a catalytic subunit (p110a, b or d) and a regulatory subunit7,8. PIK3CA, the gene encoding p110a, is mutated in up to 40% of oestrogen receptor (ER) and/orHER2 positive breast tumours9,10. In our ongoing phase I clinical study of BYL719,we have observed clinical responses inbreast, head andneck andother tumours3, providing proof of principle that PI(3)Ka targeting is active against tumours harbour- ing PIK3CA mutation. Wepresent the caseof a60-year-oldbreast cancerpatient (indexpatient) diagnosedwith invasiveductal carcinomawhounderwent surgery followed by adjuvant treatmentwith chemotherapy and the aromatase inhibitor examestane. Four years later, the patient developed bone metastases and started therapywith the ER antagonist fulvestrant, achieving stable disease.After 18months on therapy, her disease progressed in the liver, bone and lymph nodes. The archival tissue of the primary tumour was subjected to PCR-based genetic analysis11 and a hot spot mutation in PIK3CA (E542K) was detected. This finding led to the patient’s enrol- ment in a phase I clinical trial testing the tolerability and antitumour activity of BYL719 (NCT01219699). The patient rapidly achieved a con- firmedpartial response according to theRECIST1.0 criteria12 that lasted 9.5months (Table 1 and ExtendedData Fig. 1). At that point, while the tumour remained stable inmultiple sites including a peri-aortic lymph node location, progressionoccurred in the lungs (Fig. 1) and consequently *These authors contributed equally to this work. 1MassachusettsGeneralHospital CancerCenter, 55 Fruit Street, Boston,Massachusetts 02114,USA. 2HumanOncology andPathogenesisProgram (HOPP),Memorial SloanKetteringCancerCenter, 1275 York Avenue, Box 20, New York, New York 10065, USA. 3Department of Genetics, Washington University School of Medicine, 4566 Scott Avenue, St Louis, Missouri 63110, USA. 4Siteman Cancer Center, Washington University School of Medicine, 660 South Euclid Avenue, St Louis, Missouri 63110, USA. 5The Genome Institute, WashingtonUniversity School of Medicine, 4444 Forest Park Avenue, St Louis, Missouri 63108, USA. 6Department of Medicine, Washington University School of Medicine, 660 South Euclid Avenue, St Louis, Missouri 63110, USA. 7Department of Pathology, Memorial Sloan Kettering Cancer Center, 1275 York Avenue, Box 20, New York, New York 10065, USA. 8Cancer Biology and Genetics Program, Memorial Sloan Kettering Cancer Center, 1275 York Avenue, Box 20, New York, New York 10065, USA. 9Division of Genitourinary Oncology, Department of Medicine, Memorial Sloan Kettering Cancer Center, 1275 York Avenue, Box 20, New York, New York 10065, USA. 10Howard Hughes Medical Institute, Cancer Biology and Genetics Program, Memorial Sloan Kettering Cancer Center, 1275 York Avenue, Box 20, New York, New York 10065, USA. 11Novartis Pharma AG, Forum 1, Novartis Campus, CH-4056 Basel, Switzerland. 12Oncology Translational Medicine, Novartis Institutes for BioMedical Research, Cambridge, Massachusetts 02139, USA. 13Breast Medicine Service, Department of Medicine, Memorial Sloan Kettering Cancer Center, 1275 York Avenue, Box 20, New York, New York 10065, USA. Cycle 8 Cycle 10 Cycle 8 Cycle 10 Responding New lesions Figure 1 | Clinical response of index patient treated with BYL719. CT scans showing stable (responding) peri-aortic lymph node metastasis (yellow circles, left column) and the appearance of new lung metastatic lesions (yellow circles, right column) after the completion of the tenth cycle of BYL719 therapy. Arrow, pleural effusion. 2 4 0 | N A T U R E | V O L 5 1 8 | 1 2 F E B R U A R Y 2 0 1 5 Macmillan Publishers Limited. All rights reserved©2015    82   PIK3CA (E542K) PIK3CA (D752G) PSTK (S315G) ESR1 (Y537N) ATM (E1787K) CRY2 (G466R) BRCA2 (L971S) PIK3CA (E542K) PIK3CA (D752G) PSTK (S315G) ESR1 (Y537N) ATM (E1787K) CRY2 (G466R) BRCA2 (L971S) 43.8% VAF 42.0% 41.4% 33.3% 26.7% 22.8% 20.4% 39.0% VAF 35.1% 0.0% 19.0% 0.0% 0.0% 27.3% 31.4% VAF 21.8% 0.0% 0.0% 4.8% 4.0% 3.7% 0.0% 0.0% 9.2% PTEN (del 339fs) PTEN (del 339fs) Tumour and metastases Metastases only Lymph metastasis only Lung metastasis only Presence of each mutation b a Primary tumour Lung metastasis c Breast tumour d RESPONDING PTEN expression NEW LESION PTEN loss PIK3CA E542K PIK3CA D725G PTEN WT Primary tumour Lung metastasis Lymph metastasis Chromosome 10 Chromosome 10 p1 5.3 p1 5.2 p1 5.1 p1 4 p1 3 p1 2.3 3 p1 2.3 2 p1 2.3 1 p1 2.2 p1 2.1 p1 1.2 3 p1 1.2 2 p1 1.2 1 p1 1.1 q1 1.1 q1 1.2 1 q1 1.2 2 q1 1.2 3 q2 1.1 q2 1.2 q2 1.3 q2 2.1 q2 2.2 q2 2.3 q2 3.1 q2 3.2 q2 3.3 1 q2 3.3 2 q2 3.3 3 q2 4.1 q2 4.2 q2 4.3 1 q2 4.3 2 q2 4.3 3 q2 5.1 q2 5.2 q2 5.3 q2 6.1 1 q2 6.1 2 q2 6.1 3 q2 6.2 q2 6.3 100 2 × 107 4 × 107 6 × 107 8 × 107 108 1.2 × 108 1.4 × 108 100 2 × 107 4 × 107 6 × 107 8 × 107 108 1.2 × 108 1.4 × 108 –2 –1 0 1 C N V di ffe re nc e Loss (Windows = 13,554) Loss < –0.5 (n = 1,892) Loss < –0.75 (n = 2) Loss < –0.9 (n = 0) HMM Loss Segs p1 5.3 p1 5.2 p1 5.1 p1 4 p1 3 p1 2.3 3 p1 2.3 2 p1 2.3 1 p1 2.2 p1 2.1 p1 1.2 3 p1 1.2 2 p1 1.2 1 p1 1.1 q1 1.1 q1 1.2 1 q1 1.2 2 q1 1.2 3 q2 1.1 q2 1.2 q2 1.3 q2 2.1 q2 2.2 q2 2.3 q2 3.1 q2 3.2 q2 3.3 1 q2 3.3 2 q2 3.3 3 q2 4.1 q2 4.2 q2 4.3 1 q2 4.3 2 q2 4.3 3 q2 5.1 q2 5.2 q2 5.3 q2 6.1 1 q2 6.1 2 q2 6.1 3 q2 6.2 q2 6.3 –3 –2 –1 0 1 C N V di ffe re nc e Loss (Windows = 13,554) Loss < –0.5 (n = 1,892) Loss < –0.75 (n = 2) Loss < –0.9 (n = 0) HMM Loss Segs Peri-aortic lymph node metastasis Figure 2 | Loss of PTEN upon BYL719 resistance. a, Circos plots fromWGS analysis of primary tumour (before BYL719 treatment) and a lung metastasis appearing after the tenth cycle of BYL719 therapy. b, Copy number variation of chromosome 10. c, WES of the peri-aortic lymph node showing durable stable disease during BYL719 therapy compared to both primary tumour and the progressing lung lesion. The diagram shows the variant allele fractions (VAF) of the listed gene mutations in the three lesions. The estimated tumour purities are 44% for the breast primary tumour, 50% for the lung metastasis, and 59% for the lymph node metastasis. d, PTEN IHC of primary tumour, peri-aortic lymph node, and lung metastasis. Images were taken from Servier Medical Art (licensed under a Creative Commons Attribution 3.0 Unported License). LETTER RESEARCH 1 2 F E B R U A R Y 2 0 1 5 | V O L 5 1 8 | N A T U R E | 2 4 1 Macmillan Publishers Limited. All rights reserved©2015   83   therapywith BYL719was discontinued. The clinical status of the patient deteriorated rapidly and she died two months after termination of the BYL719 treatment. A rapid autopsy was performed three hours after death and a total of 14 metastases with tumour cells present were iden- tified and collected for sequencing (Extended Data Table 1). In order to proceed systematically to identify possible genetic deter- minants of acquired resistance to PI(3)Ka inhibition, we took a three- step approach. First, we examined both the primary tumour (before BYL719 treatment) and the new lung metastasis by whole genome se- quencing (WGS).Althoughboth samples sharedmany somatic genetic aberrations (Fig. 2a andExtendedData Fig. 2),PTEN copynumber loss was detected only in the lungmetastasis (Fig. 2b). Second, we analysed by whole exome sequencing (WES) the primary tumour, lung meta- stasis, and the peri-aortic lesion that remained stable (responding) at the timeof progression toBYL719 therapy (Fig. 2c). This analysis revealed that both peri-aortic and lung lesions harbouredmutations inPIK3CA, ESR1 and BRCA2, and single copy loss of PTEN. Importantly, in addi- tion to the PTEN copy number loss, we identified a PTEN del339FS (frameshift) mutation only in the lung metastasis (Fig. 2c). By immu- nohistochemistry (IHC),weobserved thatPTENprotein expressionwas lost in the lungmetastasis but was present in both the primary tumour and peri-aortic lesion (Fig. 2d). Third, to confirm and expand our findings, we sequenced the prim- ary tumour and all the metastatic lesions to.500-fold coverage using a custom targeted deep-sequencing assay, IMPACT13,14 (Methods). A numberofmutationswere sharedby theprimary tumour and themetas- tatic sites, whereas others were observed only in all or in selected meta- static lesions (Fig. 3a and Supplementary Table). We confirmed that the PIK3CA E542Kmutation in the primary tumour was conserved in the metastatic samples and detected the presence of another PIK3CA mutation (D725G).Moreover,we found increasedcopynumberofFGFR1 and EI4EBP1 in all tumour samples, consistent with the relatively fre- quent 8p11-12 amplificationdescribed inbreast cancer15,16.ESR1Y537N and BRCA2 L971S alterations were present in all themetastatic lesions but not in the primary tumour.We speculate that theESR1Y537Nmu- tation, reported to promote ligand-independent ER activation13, was selected upon anti-oestrogen therapy received by the patient before BYL719 treatment. Central to ourwork, allmetastatic lesions appeared toharbour a single copy loss of PTEN (Extended Data Fig. 3). Furthermore, we found that 10 of the 14 metastatic lesions harboured additional genomic altera- tions within PTEN. The spectrum of PTEN alterations was heterogen- eous across the 10 samples and included a splice sitemutation at K342, a frameshift indel at P339 (confirming theWES result), and 4 different exon-level deletions (Fig. 3a and Extended Data Fig. 4). All 10 speci- menswith either secondaryPTENmutations or copy number losswere confirmed negative for PTEN staining by IHC, whereas the four speci- mens that retained a PTEN wild-type allele were positive for PTEN protein expression (ExtendedData Fig. 5). In addition, for those lesions M 01 M 02 M 08 M 15 M 03 M 09 M 04 M 05 M 07 M 12 M 13 M 06 M 10 M 11 JAK1 PIK3CA FGFR1 EIF4EBP1 ESR1 BRCA2 ATM PTEN BRCA1 MEN1 M 01 M 02 M 08 M 15 M 03 M 09 M 04 M 05 M 07 M 12 M 13 M 06 M 10 M 11 Primary M04 M01 M02 M08 M15 M11 M06 M03 JAK1 T416fs PIK3CA E542K PIK3CA D725G FGFR1 amp EIF4EBP1 amp ESR1 Y537N BRCA2 L971S PTEN single copy loss PTEN deletion (exons 6-7) PTEN K342_splice BRCA1 G1062S PTEN deletion (exons 1–9) M10 ATM E1787K PTEN S339fs M09 PTEN deletion (exons 1–2) PTEN deletion (exons 5–9) M05 M07 M12 M13 MEN1 R234C a b PTEN (deletion exons 1–2) PTEN (deletion exons 5–9) PTEN (deletion exons 6–7) PTEN (deletion exons 1–9) PTEN (p.K342_splice) PTEN (pS339fs) P rim ar y P rim ar y Figure 3 | Loss of PTEN by different genetic alterations. a, Heatmap of the non-silent genetic alterations across the primary tumour and the 14 metastases (M) collected during the autopsy of the index patient. Gene mutations are depicted in green, gene amplifications in red, and gene copy number loss in blue. b, Dendrogram showing the proposed phylogenetic evolution of the metastases in the index patient. Shaded circles represent metastatic lesions with bi-allelic loss of PTEN and lack of PTEN expression by IHC. Table 1 | Patient information Patient Site PIK3CA baseline Dose (mg) Response (RECIST) DOT (days) PTEN Post-T PIK3CA Post-T Index Breast E542K 400 PR (252.4%) 285 Loss E542 1 Breast H1047R 400 SD* 181 Unch. H1047R 2 Breast H1047R 400 SD (226.3%) 424 Loss H1047R 3 Breast H1047R 400 SD (228.5%) 179 Unch. WT 4 Breast H1047L 400 SD (211.3%) 504 Unch. H1047L 5 Breast H1047R 400 SD (224.9%) 110 Unch. H1047R 6 Salivary E545K 400 SD (217%) 112 Unch. WT RECIST, Response Evaluation Criteria In Solid Tumors; DOT, duration of treatment; PR, partial response; SD, stable disease; Unch., unchanged; WT, wild-type; Amp., amplified; Post-T, post-treatment. *With interval decrease in left breast and left chest wall skin thickening. RESEARCH LETTER 2 4 2 | N A T U R E | V O L 5 1 8 | 1 2 F E B R U A R Y 2 0 1 5 Macmillan Publishers Limited. All rights reserved©2015    84   thatwere visualized byCT (computerized tomography) scan, therewas a tight correlation between progression of disease and loss of PTEN expression. The peri-aortic lesion (M02) thatwas responding at the time of disease progression still contained one PTEN wild-type allele and protein expression. Conversely, the lung lesions (M04, M06, M09 and M11) and liver lesion (M12) with documented progression to therapy had bi-allelic PTEN alteration and lack of expression. In an effort to integrate the genomic data from our patient, we con- structed a dendrogram mapping the phylogenetic evolution of the dis- ease. Our findings suggest that all the lesions were derived from the PTENwild-type primary tumour, and that there was a progressive and parallel loss of PTENunder BYL719 selective pressure (Fig. 3b). Of note, the two-month duration between progression to BYL719 and autopsy needs to be considered as well. In order to expandourobservations,we analysedpaired samples (pre- treatment and at progression) from six additional patients enrolled in the same study at our institution (Table 1). Targeted sequencing iden- tified homozygous loss of PTEN in a post-treatment sample of a breast cancerpatientwhodeveloped resistance toBYL719after initially experi- encing a durable response to therapy (Table 1).We also confirmed lack of PTEN expression by IHC in the post-treatment sample (Extended Data Fig. 6). We found no detectable PIK3CA mutations in the post- treatment samples of two patients (Table 1). Given that the presence of PIK3CAmutations drove sensitivity toBYL719 inour cell line screens17, positive selection of clones bearing wild-type alleles of PIK3CA may explain the emergence of resistance to BYL719 in these two additional cases. These results may be an indication that in some cases PIK3CA mutations are not early founder/truncal events but branched subclonal drivers that are cleared from the tumours under the selective pressure of PI(3)Ka inhibition. In any case, the fact that loss of PTENexpression and emergence of PIK3CA wild-type clones are mutually exclusive in our patient samples indicates that both events may be important in opposing the therapeutic efficacy of BYL719. No other alterations with an obvious connection to BYL719 resist- ance were found in the responding cases, with the exception of a mu- tation (E1490*) and an in-frame deletion in MAP3K1 in one of the three patients for whom neither PTEN nor PIK3CA status changed during BYL719 treatment. Further characterization is needed to deter- mine whether these mutations lead to increased MEK and ERK signal- ling and limit the effects of PI(3)K inhibition. PTEN encodes for a phosphatase that regulates the activity of PI(3)K by limiting the accumulationof phosphatidylinositol-3,4,5-trisphosphate (PtdIns(3,4,5)P3 or PIP3), a required mediator to initiate the PI(3)K/ AKT/mTOR signalling cascade18. In the absence of PTEN, cancer cells become dependent mostly on the activity of the p110b isoform of PI(3)K (PI(3)Kb) topropagate signalling throughdownstreampathway a b 0 5 10 15 20 25 30 35 40 0 100 200 300 400 500 600 Days of treatment P er ce nt ag e ch an ge in tu m ou r v ol um e Vehicle BYL719 BKM120 pA K T (S 47 3) pS 6 (S 24 0/ 4) Vehicle BYL719 BKM120 c pS 6 (S 24 0/ 4) pA K T (S 47 3) Vehicle BYL719 AZD6482 Combination d e 0 5 10 15 20 25 30 0 100 200 300 400 500 Days of treatment P er ce nt ag e ch an ge in tu m ou r v ol um e Vehicle BYL719 AZD6482 Combination f –6 –4 –2 0 2 4 0 20 40 60 80 100 log2 [AZD6482 + BYL719 (μM)] Vi ab ili ty (% ) shRenilla shPTEN no. 2 T47D g 100 μm 100 μm –6 –4 –2 0 2 4 0 20 40 60 80 100 120 log2 [BYL719 (μM)] Vi ab ili ty (% ) shRenilla shPTEN no. 1 shPTEN no. 2 T47D –6 –4 –2 0 2 4 0 20 40 60 80 100 120 log2 [BKM120 (μM)] Vi ab ili ty (% ) shRenilla shPTEN no. 1 shPTEN no. 2 T47D * * BYL719: pPRAS40 (T246) pGSK3β (S9) pS6 (S240/4) Actin PTEN no. 1 PTEN no. 2 DOX – + – + – + – + – DOX – pAKT (T308) PTEN T47D (PIK3CAH1047R) pAKT (S473) Figure 4 | Loss of PTEN expression and sensitivity to PI(3)Ka and PI(3)Kb blockade. a, Western blot showing PTEN knockdown by two independent shRNAs and its effects on the PI(3)K/AKT/mTOR pathway. b, Cell viability assay in T47D cells with inducible PTEN knockdown (shPTEN no. 1 and no. 2) or PTENexpressing controls (shRenilla) treatedwith increasing concentrations of BYL719 or BKM120. Error bars, s.e.m. c, Antitumour activity of either BYL719 (25mg kg21 daily) or BKM120 (25mgkg21 daily) in PDXs subcutaneously grown in nude mice (n5 6 (vehicle) and n5 8 (treatments)). Error bars, s.e.m. d, Representative immunostaining for phosphorylated AKT (pAKT) andphosphorylated S6 (pS6) in PDXs treated as shown.Tumourswere collected at the end of the experiment of c, 2 h after the last dosage. Scale bar, 100mm. e, Cell viability assay in T47D cells with PTEN expression or PTEN knockdown treated with increasing concentrations of the combination of BYL719 and AZD6482. Error bars, s.e.m. f, Antitumour activity of either BYL719 (25mg kg21 daily) or the combination of BYL719 and AZD6482 (25mgkg21 daily) in PDXs subcutaneously grown in nude mice (n5 6 (vehicle) and n5 8 (treatments)). Error bars, s.e.m. g, Representative immunostaining for pAKT (S473) and pS6 (S240/244) in PDXs treated as shown. Tumours were collected at the end of the experiment of f, 2 h after the last dosage. Scale bar, 100mm. *P, 0.05. LETTER RESEARCH 1 2 F E B R U A R Y 2 0 1 5 | V O L 5 1 8 | N A T U R E | 2 4 3 Macmillan Publishers Limited. All rights reserved©2015   85   effectors19,20. Therefore, we hypothesized that progressive decrease or loss of PTEN expression in the presence of PI(3)Ka inhibition might restore PI(3)K/AKT signalling through PI(3)Kb activity. To test our hypothesis, we established cell lines expressing either doxycycline- inducibleor constitutive shorthairpin (sh)RNAagainstPTENmessenger RNAusingT47DandMCF7 cells, known to be intrinsically sensitive to BYL71921. As expected, induction of PTENdownregulation led to acti- vation of AKT and the downstream effectors PRAS40, GSK3b and S6 in bothT47D(Fig. 4a) andMCF7 (data not shown) cells under basal con- ditions. PTENdownregulationmarkedly limited the effects of BYL719, both at the signalling and cell viability level. On the other hand, PTEN knockdown did not result in resistance to the pan-PI(3)K inhibitor BKM120, which blocks all the PI(3)K p110 isoforms (Fig. 4b and Ex- tendedData Fig. 7a). Similar effectswere observed in another BYL719- sensitive cell line (MDA-MB-453) with constitutivePTEN knockdown (Extended Data Fig. 7b). Fromourpatient’snon-respondingPTEN-null lungmetastatic lesion, we were able to establish xenografts in nude mice. Consistent with the in vitrodata, this patient-derivedxenograft (PDX)was resistant toBYL719 treatment but sensitive to BKM120 (Fig. 4c). The degree of inhibition of phospho-AKT and phospho-S6was also higher with BKM120 (Fig. 4d and Extended Data Fig. 7c and d). These results were complemented by the combination of BYL719 and the PI(3)Kb inhibitor AZD6482. Upon PTEN knockdown, only combined PI(3)Ka and b blockade was capable of reverting the resistant phenotype (Fig. 4e andExtendedData Fig. 8a). Similarly, theBYL719-resistantPDXwas insensitive toAZD6482 alone but responded to the combination of both compounds (Fig. 4f). Profound inhibition ofAKT and S6phosphorylationwas achieved only upon treatment with BYL719 in combination with AZD6482 (Fig. 4g and Extended Data Fig. 8b and c). Taken together, these data indicate that inhibition of the PI(3)Kb isoform is required to achieve antitu- mour activity in cells/tumours that lost PTEN expression and become resistant to BYL719. We have reported a case of parallel genetic evolution under selective therapeutic pressure leading to a progressive loss of PTEN expression and consequent gain of dependency on the PI(3)Kb isoform. Parallel evolution under selective pressure has been described in conditions where treatments are highly efficacious, such as in HIV22. Our case highlights that this tumour, despite its heterogeneity, was dependent on PI(3)K signalling, probably as a result of the presence of the same activating PIK3CA mutation in all the tumour sites. Upon continued suppression of PI(3)Ka, diverse genomic alterations emerged, leading to PTEN loss as an alternative mechanism of PI(3)K activation. More- over, our study emphasizes the importance of tumour interrogation upon progression to therapy and the dynamic nature of tumour gen- omes under selective therapeutic pressure. Online ContentMethods, along with any additional Extended Data display items andSourceData, are available in theonline versionof thepaper; referencesunique to these sections appear only in the online paper. Received 29 April; accepted 7 October 2014. Published online 17 November 2014. 1. Gerlinger, M. et al. Intratumor heterogeneity and branched evolution revealed by multiregion sequencing. N. Engl. J. Med. 366, 883–892 (2012). 2. Swanton, C. Intratumor heterogeneity: evolution through space and time. Cancer Res. 72, 4875–4882 (2012). 3. Juric, D. BYL719, a next generation PI3K alpha specific inhibitor: preliminary safety, PK, and efficacy results from the first-in-human study. Cancer Res. 72, Abstr. CT-01 (2012). 4. Engelman, J. A., Luo, J. & Cantley, L. C. The evolution of phosphatidylinositol 3-kinases as regulators of growth andmetabolism.Nature Rev. Genet. 7, 606–619 (2006). 5. Cantley, L. C. The phosphoinositide 3-kinase pathway. Science 296, 1655–1657 (2002). 6. Engelman, J. A. Targeting PI3K signalling in cancer: opportunities, challenges and limitations. Nature Rev. Cancer 9, 550–562 (2009). 7. Brachmann, S. M., Ueki, K., Engelman, J. A., Kahn, R. C. & Cantley, L. C. Phosphoinositide 3-kinase catalytic subunit deletion and regulatory subunit deletion have opposite effects on insulin sensitivity in mice.Mol. Cell. Biol. 25, 1596–1607 (2005). 8. Zhao, L.&Vogt, P.K.Helical domainandkinasedomainmutations inp110alphaof phosphatidylinositol 3-kinase induce gain of function by different mechanisms. Proc. Natl Acad. Sci. USA 105, 2652–2657 (2008). 9. Miller, T. W., Rexer, B. N., Garrett, J. T. & Arteaga, C. L. Mutations in the phosphatidylinositol 3-kinasepathway: role in tumor progressionand therapeutic implications in breast cancer. Breast Cancer Res. 13, 224–235 (2011). 10. Samuels, Y. et al. High frequency of mutations of the PIK3CA gene in human cancers. Science 304, 554 (2004). 11. Sequist, L. V. et al. Implementing multiplexed genotyping of non-small-cell lung cancers into routine clinical practice. Ann. Oncol. 22, 2616–2624 (2011). 12. Therasse, P. et al. New guidelines to evaluate the response to treatment in solid tumors. European Organization for Research and Treatment of Cancer, National Cancer Institute of the United States, National Cancer Institute of Canada. J. Natl. Cancer Inst. 92, 205–216 (2000). 13. Toy, W. et al. ESR1 ligand-binding domainmutations in hormone-resistant breast cancer. Nature Genet. 45, 1439–1445 (2013). 14. Won, H. H., Scott, S. N., Brannon, A. R., Shah, R. H. &Berger,M. F. Detecting somatic genetic alterations in tumor specimens by exon capture and massively parallel sequencing. J. Visual. Expts e50710 (2013). 15. Yang, Z. Q., Liu, G., Bollig-Fischer, A., Giroux, C. N. & Ethier, S. P. Transforming properties of 8p11-12 amplified genes in human breast cancer. Cancer Res. 70, 8487–8497 (2010). 16. Karlsson, E. et al. High-resolution genomic analysis of the 11q13 amplicon in breast cancers identifies synergy with 8p12 amplification, involving the mTOR targets S6K2 and 4EBP1. Genes Chromosom. Cancer 50, 775–787 (2011). 17. Fritsch, C. et al. Characterization of the novel and specific PI3Ka inhibitor NVP- BYL719 and development of the patient stratification strategy for clinical trials. Mol. Cancer Ther. 13, 1117–1129 (2014). 18. Stambolic, V. et al. Negative regulation of PKB/Akt-dependent cell survival by the tumor suppressor PTEN. Cell 95, 29–39 (1998). 19. Jia, S. et al. Essential roles of PI(3)K-p110b in cell growth, metabolism and tumorigenesis. Nature 454, 776–779 (2008). 20. Edgar, K. A. et al. Isoform-specific phosphoinositide 3-kinase inhibitors exert distinct effects in solid tumors. Cancer Res. 70, 1164–1172 (2010). 21. Elkabets, M. et al.mTORC1 inhibition is required for sensitivity to PI3K p110a inhibitors in PIK3CA-mutant breast cancer. Sci. Transl. Med. 5, 196ra199 (2013). 22. Lemey, P. et al.Molecular footprint of drug-selective pressure in a human immunodeficiency virus transmission chain. J. Virol. 79, 11981–11989 (2005). Supplementary Information is available in the online version of the paper. AcknowledgementsWe thank members of the MSKCC Diagnostic Molecular Pathology Laboratory and theMSKMaria-Jose´e andHenry Kravis Center forMolecular Oncology for assistance with sequencing. We thank M. Asher and U. Bhanot from the MSKCC Pathology Core for assistance with tissue staining. This work was funded bya ‘‘StandUp toCancer’’DreamTeamTranslationalResearchGrant, aProgramof the Entertainment Industry Foundation (SU2C-AACR-DT0209), the Breast Cancer Research Foundation, the Geoffrey Beene Cancer Research Center, the Starr Cancer Consortium and an MMHCC grant (CA105388). D.J. is also funded by a National Institutes of Health Training Grant (T32 CA-71345-15) and by philanthropic support from Stephen and Kathleen Chubb. Author Contributions D.J., P.C., M.F.B., J.B. and M.S. conceived the project, designed and analysed the experiments, and wrote the manuscript. M.G., O.L.G., B.J.A., A.R. and E.R.M. performed and analysed theWGS andWESdata. T.H.,M.M.-K., D.S., S.I., A.T., L.E., C.Q., M.P., A.D., R.B. and A.H. collected and analysed patients’ samples. P.C., H.E., S.H.E. and S.W.L. performed and supervised the laboratory experiments. H.H.W., G.I., R.H.S., D.B.S. and M.F.B. performed and supervised the IMPACT sequencing and analysis. Author Information DNA sequences have been deposited in the European Genome-phenomeArchive with accession number EGAS00001000991. Reprints and permissions information is available at www.nature.com/reprints. The authors declare competing financial interests: details are available in the online version of the paper. Readers are welcome to comment on the online version of the paper. Correspondence and requests for materials should be addressed to M.S. (scaltrim@mskcc.org), J.B. (baselgaj@mskcc.org) or M.F.B. (bergerm1@mskcc.org). RESEARCH LETTER 2 4 4 | N A T U R E | V O L 5 1 8 | 1 2 F E B R U A R Y 2 0 1 5 Macmillan Publishers Limited. All rights reserved©2015    86   METHODS PIK3CA mutant cell lines MCF7 (E545K) and T47D (H1047R) (ATCC) were transduced with the retroviral TRMPV vector. Doxycycline (Sigma) was used to temporally activate the expression of amicroRNA 30-embedded shRNA targeting Renilla luciferase (control) or PTENmRNA. The hairpin sequences used were as follows. Renilla luciferase. CTCGAGAAGGTATATTGCTGTTGACAGTGAGCGCAG GAATTATAATGCTTATCTATAGTGAAGCCACAGATGTATAGATAAGCA TTATAATTCCTATGCCTACTGCCTCGGAATTC PTEN no. 1. CTCGAGAAGGTATATTGCTGTTGACAGTGAGCGACCAGC TAAAGGTGAAGATATATAGTGAAGCCACAGATGTATATATCTTCACCT TTAGCTGGCTGCCTACTGCCTCGGAATTC PTENno. 2.CTCGAGAAGGTATATTGCTGTTGACAGTGAGCGCCCAGAT GTTAGTGACAATGAATAGTGAAGCCACAGATGTATTCATTGTCACTAA CATCTGGTTGCCTACTGCCTCGGAATTC Cell viability was assessed using the tetrazolium-basedMTTassay after 6 days of treatment. All cell lines resulted negative formycoplasma contamination.Western blotting was carried out using previously describedmethods21. All the in vitro experi- ments were performed in triplicate. Patient-derived xenografts and IHC. Animals were maintained and treated in accordance with Institutional Guidelines of Memorial Sloan Kettering Cancer Center (Protocol number 12-10-019). Tumours were implanted subcutaneously in six-week-old female athymicNU/NUnudemice.Once the tumours reached a volume of,200mm3, it was expanded in multiple mice which were then randomized to the following treatments: BYL719, BKM120 (a pan-class I PI(3)K inhibitor), or AZD6482 (a PI(3)Kb inhibitor), each administered orally at 25mgkg21 once a day. After treatment, mice were euthanized and tumours were harvested and pro- cured for IHC analysis. IHCwas performed on a Ventana Discovery XT processor platformusing standard protocols and the following antibodies: pAKT (S473) (D9E), Cell Signaling Technology, #4060, dilution 1:70. pS6 (S240/4) (D68F8)XP, Cell Signaling Technology, #5364, dilution 1:500. PTEN (138G6), Cell Signaling Tech- nology, #9559, dilution 1:30. All the in vivo experiments were run with at least n5 8 for each treatment arm. Two-way t-test was performed using GraphPad Prism (GraphPad Software). Error bars represent the s.e.m. *P, 0.05. Whole genome and whole exome sequencing. For whole genome (WG) and whole exome (WE) sequencing, DNAwas derived from the primary tumour, lung metastasis, and peri-aortic lymph nodemetastasis. DNA from the spleen was used as a normal control. WG libraries were produced as previously described23 and sequencedusing the IlluminaHiSeq2500platformaspaired-end100basepair reads, producing ,30-fold (primary tumour, spleen normal)-50-fold (lung metastasis) coverage for WG sequencing. By hybrid capture (Nimblegen version 3.0) of the lymph node and lung metastases, primary tumour and spleen normal, we gener- ated,100-fold coverage for WE sequencing. All patients provided written informed consent for the genetic research studies performed in accordancewithprotocols approved byDana Farber/HarvardCancer Center Institutional Review Board. Autopsy in the index patient was performed within the first three hours post mortem. Targeted exome sequencing (IMPACT).DNAderived from the primary tumour, 14 metastases, and matched normal spleen tissue was further subjected to deep- coverage targeted sequencing of key cancer-associated genes. Our assay, termed IMPACT (Integrated Mutation Profiling of Actionable Cancer Targets), involves hybridization of barcoded libraries to custom oligonucleotides (Nimblegen SeqCap) designed to capture all protein-coding exons and select introns of 279 commonly implicated oncogenes, tumour suppressor genes, andmembers of pathways deemed actionable by targeted therapies14. The captured pool was subsequently sequenced on an Illumina HiSeq 2500 as paired-end 75-base pair reads, producing 513-fold coverageper tumour. Sequencedatawere analysed to identify three classes of somatic alterations: single-nucleotide variants, small insertions/deletions (indels), and copy number alterations. Barcoded sequence libraries were prepared using 250ng genomic DNA (Kapa Biosystems) and combined in a single equimolar pool. Sequence data were demul- tiplexed using CASAVA, and reads were aligned to the reference human geno- me(hg19)usingBWAandpostprocessedusing theGenomeAnalysisToolkit (GATK) according to GATK best practices24,25. MuTect andGATKwere used to call single-nucleotide variants and small indels, respectively26. Exon-level copy number gains and losses were inferred from the ratio inTumour:Normal sequence coverage for each target region, following a loss- normalization to adjust for the dependency of coverage on GC content27. Statistical analysis. Two-way t-tests were performed using GraphPad Prism (GraphPad Software). Error bars represent the s.e.m., P values are indicated as *P, 0.05. All cellular experiments were repeated at least three times. All the in vivo experiments were run with at least 6–8 tumours for each treatment arm. Sample sizewas chosen to detect a difference inmeans of 20%with a power of 90%. Animalswere randomized in groupswith similar average in tumour size. Investigators were blinded when assessing the outcome of the in vivo experiments. For the cell viability graphs, nonlinear regressionwas applied to the experimental data sets. Curves were compared using the extra-sum-of-squares F test using a5 0.05. Hypothesis was rejected when nonlinear models were not nested within each other and was considered statistically significant. 23. Li, S. et al. Endocrine-therapy-resistant ESR1 variants revealed by genomic characterization of breast-cancer-derived xenografts. Cell Rep. 4, 1116–1130 (2013). 24. Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760 (2009). 25. DePristo, M. A. et al. A framework for variation discovery and genotyping using next-generation DNA sequencing data. Nature Genet. 43, 491–498 (2011). 26. Cibulskis, K. et al. Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples. Nature Biotechnol. 31, 213–219 (2013). 27. Wagle, N. et al. High-throughput detection of actionable genomic alterations in clinical tumor samples by targeted, massively parallel sequencing. Cancer Discov. 2, 82–93 (2012). LETTER RESEARCH Macmillan Publishers Limited. All rights reserved©2015   87 ARTICLE 2 Title: PDK1-SGK1 signaling sustains AKT-independent mTORC1 activation and confers resistance to PI3Kα inhibition Journal: Cancer Cell (under review) Impact Factor: 23.523    88   89 PDK1-SGK1 signaling sustains AKT-independent mTORC1 activation and confers resistance to PI3Kα inhibition Pau Castel1, Haley Ellis1, Ruzica Bago2, Eneda Toska1, Srinivasaraghavan Kannan3, F. Javier Carmona1, Pedram Razavi1, Chandra S. Verma3,4,5, Maura Dickler6, Sarat Chandarlapaty1,6, Edi Brogi7, Dario R. Alessi2, José Baselga1,6, Maurizio Scaltriti1. 1Human Oncology & Pathogenesis Program (HOPP), Memorial Sloan Kettering Cancer Center, 1275 York Avenue, Box 20, New York, NY 10065 2MRC Protein Phosphorylation and Ubiquitylation Unit, College of Life Sciences, University of Dundee, Dow Street, Dundee DD1 5EH, Scotland 3Bioinformatics Institute (A*STAR), 30 Biopolis Street, #07-01 Matrix, Singapore 138671 4School of Biological Sciences, Nanyang Technological University, 60 Nanyang Drive, Singapore 637551 5Department of Biological Sciences, National University of Singapore, 14 Science Drive 4, Singapore 117543 6Department of Medicine, Memorial Sloan Kettering Cancer Center, 1275 York Avenue, Box 20, New York, NY 10065 7Department of Pathology, Memorial Sloan Kettering Cancer Center, 1275 York Avenue, Box 20, New York, NY 10065, USA Corresponding Authors: Maurizio Scaltriti, PhD Human Oncology & Pathogenesis Program (HOPP) Memorial Sloan Kettering Cancer Center 1275 York Avenue, Box 20 New York, NY 10065 Tel: 646-888-3519 Fax: 646-422-0247 E-mail: scaltrim@mskcc.org José Baselga, MD, PhD Department of Medicine Memorial Sloan Kettering Cancer Center 1275 York Avenue - Suite M2015 New York, NY 10065 Phone: 212 639-8000 Fax: 212 794-3182 E-mail: baselgaj@mskcc.org    90 Summary PIK3CA, the gene encoding the alpha isoform of PI3K (PI3Kα), is frequently mutated and oncogenic in breast cancer. PI3Kα inhibitors are in clinical development and despite promising early clinical activity, primary and acquired resistance is frequent among patients. We have previously reported that residual downstream mTORC1 activity upon treatment with PI3Kα inhibitors drives resistance to these agents. However, the underlying mechanisms that mediate this phenotype are not fully understood. Here we show that in cancer cells resistant to PI3Kα inhibition, PDK1 blockade restores sensitivity to these therapies. SGK1, which is activated by PDK1, contributes to the maintenance of residual mTORC1 activity and cell survival through direct phosphorylation and inhibition of TSC2. Targeting either PDK1 or SGK1 prevents mTORC1 activation and restores the antitumoral effects of PI3Kα inhibition in resistant cancer cells.   91 Introduction The phosphoinositide 3-kinase (PI3K) pathway integrates many extracellular stimuli and upon activation, triggers the phosphorylation of key downstream effectors such as AKT and the mammalian Target of Rapamycin Complex 1 and 2 (mTORC1 and 2). This signaling cascade is essential for regulating cell size, proliferation, survival, and metabolism (Engelman, 2009; Thorpe et al., 2015). Activation of PI3K results in the synthesis of the second messenger phosphatidylinositol-(3,4,5)-triphosphate (PIP3) at the plasma membrane, which in turn promotes the recruitment of the pleckstrin homology (PH) domain-containing proteins PDK1 and AKT (Currie et al., 1999; Pearce et al., 2010). At the plasma membrane, the constitutively active kinase PDK1 phosphorylates AKT at the activation loop (T308) (Alessi et al., 1997) and a second phosphorylation in the hydrophobic motif (S473) is then carried out by mTORC2 to fully activate AKT (Sarbassov et al., 2005). Once active, AKT phosphorylates a variety of substrates including antiapoptotic and cell cycle-related proteins, as well as transcription factors (Manning and Cantley, 2007). Moreover, as a result of PI3K signaling, AKT activates downstream mTORC1 through the phosphorylation of the negative regulators TSC2 and PRAS40 (Inoki et al., 2002; Manning et al., 2002; Potter et al., 2002; Sancak et al., 2007). Activating mutations in PIK3CA, the gene that encodes for the α isoform of the p110 catalytic subunit of PI3K (PI3Kα), loss of function of phosphatase and tensin homolog (PTEN), the phosphatase that modulates levels of PIP2, and overexpression of membrane-bound receptor tyrosine kinases result in hyperactivation of the PI3K/AKT/mTOR pathway (Engelman, 2009). These events are common in breast cancer and provide the rationale for the development of inhibitors targeting the different nodes of the PI3K pathway (Fruman and Rommel, 2014). PI3Kα specific inhibitors are currently showing promising results in patients with tumors bearing activating mutations in PIK3CA (Juric et al., 2013; Juric et al., 2012). However, despite these encouraging results, some tumors treated with these agents remain insensitive. Understanding the molecular mechanisms by which these tumors bypass the pharmacological inactivation of PI3Kα is crucial for the identification of patients that are more likely to respond to these inhibitors and for devising therapeutic strategies to prevent or revert the emergence of resistance. We have previously reported that the activation status of mTORC1 upon PI3Kα blockade is a determinant of drug sensitivity in PIK3CA-mutant tumors. Despite full inhibition of    92 PI3K/AKT, the presence of residual mTORC1 activity is sufficient to weaken the antitumor activity of PI3Kα inhibition. Resistant tumors are sensitized by co-treatment with the mTORC1 allosteric inhibitor everolimus, underscoring the causative role of mTORC1 in limiting the effects of PI3Kα blockade (Elkabets et al., 2013). In this work, we have elucidated the molecular pathway that allows mTORC1 to retain activity in the presence of PI3K and AKT inactivation. These results uncover new aspects of the biology of PI3K signaling upon pharmacological inhibition and offer novel therapeutic approaches for the clinical setting. Results PDK1 inhibition sensitizes resistant cells to BYL719. Aiming to identify possible kinases or phosphatases responsible for the AKT- independent sustained mTORC1 activity in cells resistant to PI3Kα inhibition, we performed an arrayed siRNA screening using a library targeting the genes encoding for the 710 kinases and 298 phosphatases of the human genome. We measured cell viability and S6 ribosomal protein (S6) phosphorylation, a bona fide read-out of mTORC1 activity, in the presence of BY719, a PI3Kα-specific inhibitor. The screen design is shown in Figure S1A. Three different siRNAs targeting each gene of the kinome/phosphatome and negative (scrambled siRNA) and positive (siPLK1) controls were transfected in JIMT1 and HCC1954 cell lines, which are both PIK3CA- mutant and insensitive to BYL719. After treatment with BYL719 over 6 days, cell viability was quantified using both Alamar Blue and nuclear count. We found that knockdown of 37 genes in HCC1954 and 35 genes in JIMT1 sensitized cells to BYL719 (Figure S1B; Table S1). Among these genes, five were found to be shared in both cell lines: MTOR, PDPK1, PIK3CA, PPP1R12A, and PAPL (Figure S1B). These findings were validated with a second targeted screening using the two most active siRNAs against these five genes, interrogating for both cell viability and phosphorylation of S6. With this more stringent approach, we found that only knockdowns of MTOR and PDPK1, the genes encoding mTOR and PDK1 respectively, were capable of sensitizing cells to BYL719 and reducing S6 phosphorylation (S240/4) in the presence of PI3Kα inhibition (Figure S1C). While the ablation of MTOR confirmed our previous data (Elkabets et al., 2013), the contribution of PDK1 in maintaining the resistant phenotype was a new finding. PDK1 is a kinase that belongs to the Containing PKA, PKG, and PKC (AGC) kinase   93 family, a phylogenetically related family of 60 serine-threonine kinases that includes some well-known members such as AKT, PKC, RSK, SGK, and S6K (Pearce et al., 2010). To confirm that PDK1 limits the sensitivity to PI3Kα inhibition by maintaining mTORC1 activity upon pharmacological PI3K inhibition, we generated HCC1954 and JIMT1 cell lines stably expressing a PDPK1 short hairpin RNA (shRNA). We observed that PDK1 knockdown is sufficient to decrease cell viability to BYL719 treatment (Figure 1A; Figure S2A). As previously described, treatment with BYL719 alone reduced AKT phosphorylation (S473 and T308) but not downstream mTORC1 targets (Elkabets et al., 2013). In contrast, the combination of PDK1 knockdown with BYL719 decreased the phosphorylation of the mTORC1 downstream targets p70 S6 Kinase (S6K) and translation initiation factor 4E-binding protein (4EBP1), as well as phosphorylated S6 at both S240/4 and S235/6 sites (Figure 1B; Figure S2B). Moreover, by means of m7GTP pulldowns we found that the combination of BYL719 and PDK1 knockdown decreased cap-dependent translation, a downstream cellular process directly regulated by mTORC1 (Silvera et al., 2010). In PDK1 knockdown cells, inhibition of PI3K induced an increased binding of 4EBP1 to the cap m7GpppN mRNA analogue m7GTP, to a similar extent as the mTOR inhibitor AZD8055 (Figure S2C). On the contrary, we observed a reduction of the eukaryotic initiation factors (eIF) eIF4G and eIF4A, components of the eIF4F cap-initiation complex. As expected, eIF4E remained unchanged. In long-term treatments, the combination of BYL719 and PDK1 knockdown induced PARP cleavage (Figure 2C) and increased caspase 3/7 activity (Figure 2D), surrogate markers of apoptotic activity. Next, we sought to test the in vivo antitumor activity of BYL719 in HCC1954 xenografts expressing either shGFP control or shPDK1. Pharmacological inhibition of PI3Kα resulted in a modest delay in tumor growth in shGFP xenografts but was sufficient to induce durable tumor shrinkage in tumors with ablated PDK1 (Figure 2E). Pharmacodynamic analyses of the tumors collected at the end of the experiment showed that BYL719 treatment effectively suppressed AKT phosphorylation (S473) in both shGFP and shPDK1 tumors, whereas S6 and 4EBP1 phosphorylation was inhibited only in shPDK1 xenografts (Figure 2F, S2D). To translate our observations into a therapeutically exploitable strategy, we tested the activity of BYL719 in combination with GSK2334470, a highly selective PDK1 inhibitor. Of note, a previous study showed that GSK2334470 did not exhibit activity towards any PI3K isoform or any other AGC kinases when it was screened against a panel of over 140 protein and lipid kinases (Najafov et al., 2011).    94 We determined the appropriate dose of GSK2334470 by analyzing the phosphorylation of the direct PDK1 target RSK2 (S227) and cell viability upon incubation with increasing concentrations of the PDK1 inhibitor. At 1 µM, pRSK2 (S227) was greatly reduced and no significant changes on cell viability were observed (Figure S2E, F). Despite the minimal effect on cell viability when used as a single agent, treatment with GSK2334470 was sufficient to sensitize both HCC1954 and JIMT1 cells and the triple- negative BYL719-resistant breast cancer cell line BT20 to PI3Kα inhibition (Figure 1G; Figure S2G, H). Again, we observed that BYL719 effectively suppressed AKT activity but not mTORC1 signaling and only the combination of BYL719 and GSK2334470 resulted in the inhibition of mTORC1 (Fig 1H; Figure S2I, J). Phosphorylation of RSK2 (S227) was inhibited when cells are treated with GSK2334470 alone or in combination, indicating a good degree of inhibition of this enzyme in vitro. Paradoxically, PDK1 inhibition did not decrease the phosphorylation of AKT at the activation loop (T308), an observation in line with previous reports that is a result of a compensatory mechanism involving PIP3 and mTORC2 (Najafov et al., 2012). Analysis of cap-dependent translation complex formation revealed an increase in 4EBP1 and a decrease in eIF4G and eIF4A in m7GTP pull downs when both drugs were combined, consistent with a robust mTORC1 inhibition (Figure S2K). Consistent with the knockdown experiments, the combination of BYL719 and GSK2334470 induced apoptosis in HCC1954 cells when measured by PARP cleavage (Figure 1I) and caspase 3/7 activity (Figure 1J). Next, we expanded our results in vivo by treating HCC1954 xenografts with BYL719, GSK2334470, or the combination of both agents. Although some antitumor activity was observed with single agent treatment, only the combination of both compounds induced durable tumor shrinkage (Figure 1K). We measured the levels of pAKT (S473), pS6 (S240/4), and p4EBP1 (T37/46) in the tumors at the end of the experiment and observed that while BYL719 monotherapy was sufficient to suppress pAKT, pS6 and p4EBP1 were only inhibited when both agents were used in combination (Figure 1L, S2L). Consistent results were obtained with JIMT1 xenografts, although this cell line does not exhibit apoptosis upon drug combination but induces cell cycle arrest instead (Figure S2M-O). Taken together, these results indicate that PDK1 inhibition sensitizes cells to PI3Kα blockade via suppression of mTORC1.   95 The PIF-binding pocket of PDK1 is required for sustained mTORC1 activation upon PI3Kα inhibition The activation of AGC kinases requires phosphorylation at two highly conserved regulatory motifs termed the hydrophobic motif (HM), at the C-terminal region, and the activation loop, in the catalytic domain. Extracellular stimulation triggers several kinases to prime AGC kinases for activation through phosphorylation at the HM. PDK1, which acts as a master regulator of this family of kinases, scaffolds at the phosphorylated HM of AGC kinases using the so-called PIF (PDK1-Interacting Fragment) binding pocket. This interaction enables the phosphorylation of the activation loop, thereby fully activating their kinase activity. However, this is not the case for AKT, which does not require the PIF binding pocket of PDK1 but needs its PH domain instead in order to physically interact with PDK1 at the plasma membrane in a PIP3- dependent manner (Alessi et al., 1997; Arencibia et al., 2013; Biondi et al., 2001; Collins et al., 2003; McManus et al., 2004). In order to explore the PDK1 regulatory mechanism required to sustain mTORC1 activity upon PI3Kα inhibition, we used the HCT116 parental and PDPK1-null (PDPK1-/-) isogenic model (Ericson et al., 2010). HCT116 cells harbor the H1047R PIK3CA-activating mutation, and the addition of BYL719 decreases AKT phosphorylation independently of genetic manipulation (Figure S2P). In parental cell lines, the addition of BYL719 did not decrease mTORC1 signaling, mimicking the phenotype observed in BYL719-resistant breast cancer cell lines. However, in PDPK1-/- cells, the addition of BYL719 inhibited mTORC1, consistent with our previous experiments with either shPDK1 or PDK1 inhibitor (Figure S2P). We reconstituted HCT116 PDPK1-/- cells with different PDK1 mutants and tested the contribution of each regulatory mechanism of PDK1 to mTORC1 activation. We included wild type (WT), kinase inactive K111N (KD), PIP3-binding deficient K546E (KE), and PIF pocket-deficient L155E (LE) mutants (Figure S2Q). Reconstitution of PDK1 WT, but not the kinase inactive mutant KD, restored mTORC1 activation in the presence of BYL719. The PH domain mutant KE, which is unable to bind PIP3, was also able to restore the phenotype. On the other hand, the PIF binding pocket mutant (LE) was unable to rescue mTORC1 signaling (Figure S2R). This set of experiments suggests that the maintenance of mTORC1 activity requires both kinase activity and the PIF binding pocket of PDK1 but it is PIP3-, and consequently, AKT- independent.    96 Combined suppression of PI3Kα and PDK1 activates FOXO-dependent transcription We next investigated whether mTORC1 suppression upon inhibition of both PI3Kα and PDK1 in BYL719-resistant cell lines was accompanied by specific transcriptional changes that would reveal a mechanistic explanation for the observed superior activity. We performed gene expression analysis in HCC1954 and JIMT1 cells treated with BYL719, GSK2334470, or the combination of both. While the differences in gene expression upon single agent treatment were modest, the combination of both drugs induced marked changes in the transcriptomic profiles when compared to the DMSO- treated control cells (Figure 2A, Figure S3A). Gene set enrichment analysis (GSEA) using this gene expression data showed enrichment of FOXO3 transcription factor targets in both HCC1954 and JIMT1 cells (Figure 2B; Figure S3B). Individual genes described to be positively (CCNG2, BCL6, IRS2) or negatively (CCND1) regulated by FOXO3 (Webb and Brunet, 2014) were confirmed to be induced or repressed, respectively, upon dual PI3Kα and PDK1 blockade (Figure 3C). These results were further validated by performing quantitative RT-PCR to measure the relative mRNA expression levels of four well-described FOXO3 targets: ERBB3, TNFSF10, BCL6, and IRS2 (Webb and Brunet, 2014) following different treatments. Significant increases in the mRNA levels of these genes only occurred when cells were treated with the combination of BYL719 and GSK2334470 (Figure 2D; Figure S3C). Upon growth factor stimulation (e.g. Insulin or oncogenic PIK3CA mutations), FOXO transcription factors are phosphorylated at several residues, including T32 (FOXO3A), a well-known AKT phosphorylation site, and inhibited due to the interaction with the 14-3-3 proteins that prevents FOXO nuclear shuttling and gene transcription (Webb and Brunet, 2014). Inhibition of these mitogenic signals induces a rapid de-phosphorylation and nuclear translocation of FOXOs that allows expression of downstream target genes involved in apoptosis and/or cell cycle arrest (Webb and Brunet, 2014). In our cells, we found that treatment with both BYL719 and GSK234470, but not single agent, resulted in strong nuclear localization of FOXO3 (Figure 2E; Figure S3D). This was consistent with a decreased phosphorylation of this transcription factor at the residue T32 (Figure 2F). Moreover, using a FOXO-luciferase reporter system, we observed that only the combination of BYL719 with GSK2334470 stimulated endogenous FOXO transcriptional activity (Figure 2G). An increased occupancy of FOXO3A at the promoters of two well- known FOXO targets, IRS2 and TNFSF10, was confirmed by Chromatin   97 Immunoprecipitation (ChIP) only when HCC1954 and JIMT1 cells were treated with BYL719 in combination with GSK234470 (Figure 2H; Figure S3E). These results suggest that dual PI3Kα and PDK1 inhibition induces a FOXO-dependent transcriptional activity in BYL719-resistant cells. SGK1 is upregulated in BYL719-resistant cell lines AKT has been shown to phosphorylate FOXO1/3 at T24 and T32 residues, respectively, causing its dissociation from 14-3-3 proteins and consequent nuclear shuttling (Brunet et al., 1999). However, we observed that despite full inhibition of AKT by PI3Kα inhibition, FOXO3 was not efficiently primed to migrate to the nucleus and exert its transcriptional activity in cells resistant to BYL719 (Figure 2 and S3). Moreover, we had shown that the PH domain of PDK1 is not required for sustained mTORC1 signaling upon PI3Kα inhibition (Figure S2S). Since PDK1 requires downstream AGC kinases as molecular effectors (Pearce et al., 2010), we reasoned that in BYL719-resistant cells, a downstream AGC kinase dependent on the PDK1 catalytic activity and docking with PIF binding pocket regulates both FOXO1/3 phosphorylation and mTORC1 activity, independently of AKT. Serum and glucocorticoid-induced kinase (SGK) is a family of AGC serine/threonine kinases that comprises three members (SGK1, SGK2, and SGK3) highly homologous to AKT, sharing 55% identity in the kinase domain (Kobayashi and Cohen, 1999). SGK1 activation is mediated by mTORC2-dependent phosphorylation at the HM (S422) and subsequent PDK1 phosphorylation at the activation loop (T256) in a PIF binding pocket- dependent manner (Garcia-Martinez and Alessi, 2008; Pearce et al., 2010). Earlier reports have demonstrated that SGK1 is able to directly phosphorylate FOXO1 at residues T32 and S315 (Brunet et al., 2001). Furthermore, SGK1 has been correlated with resistance to AKT inhibition (Sommer et al., 2013). Therefore, we hypothesized that SGK1 plays a critical role downstream of PDK1 in sustaining mTORC1 activity and inducing resistance to PI3Kα inhibition in our models. We analyzed the basal mRNA expression of 27 breast cancer cell lines, including cell lines previously characterized as sensitive or resistant to BYL719, (Elkabets et al., 2013) and found that resistant cell lines had significantly higher levels of SGK1 mRNA compared to sensitive cells (Figure 3A; Figure S4A). This held true when only breast cancer cells harboring PIK3CA-activating mutations, which are known to be sensitive to PI3Kα inhibition (Elkabets et al., 2013), were considered in the analysis (Figure 3B and    98 C). The mRNA levels of SGK2 and SGK3 were similar between sensitive and resistant cell lines (Figure S4B, C), although JIMT1 cells also express high levels of SGK2. The ratio of phosphorylated N-Myc Downstream Regulated 1 (NDRG1) (T346), a specific substrate of SGK1 (Murray et al., 2004), versus total NDRG1 was also higher in BYL719-resistant cells (Figure 3C and S4D). We then aimed to investigate the variability of SGK1 expression in breast tumors. Analysis of SGK1 mRNA levels in the TCGA breast cancer patient cohort revealed that about 10% of breast tumors express high levels of SGK1 mRNA, independently of the PIK3CA status (Fisher’s exact test, p=0.4) (Cerami et al., 2012; Ciriello et al., 2015). In our hands, none of the commercially available antibodies against SGK1 that we have tested for IHC provided reliable results. Thus, we analyzed expression of the downstream target pNDRG1 (T346) as the read- out of SGK1 activity in 273 breast invasive carcinomas, comprised of 138 triple-negative breast cancer (TNBC), 68 ER/PR receptor-positive, and 67 HER2-positive breast cancer patients. High expression of pNDRG1 (see Methods) was found in TNBC (21%) and HER2-positive tumors (12%) (Figure 3D), demonstrating that SGK1 activity is upregulated in a subset of breast cancer patients and in line with the SGK1 expression in the TCGA cohort. To explore the correlation between pNDRG1 and clinical outcome to PI3Kα inhibition, we secured tumor samples from 12 patients with hormone receptor-positive breast cancer and harboring PIK3CA mutations that are currently participating in a clinical trial evaluating the activity of BYL719 in combination with an aromatase inhibitor being conducted at our institution (NCT01870505) (see Methods). Two of these tumors were positive for pNDRG1 expression while the remaining ten tumors had no detectable staining. The two patients with tumors positive for pNDRG1 did not respond to therapy and their disease progressed rapidly (Figure 3E, F). On the contrary, in the pNDRG1 negative group, we observed two partial responses and six disease stabilizations by RECIST criteria (Therasse et al., 2000), in addition to a longer time to disease progression when compared to the pNDRG1 positive tumors (Figure 3E, F). These results in this small sample size are suggestive of a role of SGK1 in mediating intrinsic resistance to PI3Kα inhibitors. Because SGK1 levels are differentially expressed across breast cancer patients and cell lines, we sought to investigate the mechanism underlying this transcriptional variability. We analyzed the promoter of SGK1 and realized that in the region comprised between - 56bp and 391bp of the transcription start site (TSS) there are 12 CpG islands that are   99 susceptible for epigenetic methylation. We used bisulphite sequencing in 8 different sensitive and resistant cell lines and found that in three of these CpG islands DNA methylation was differentially present (Figure S4E). We confirmed our results quantitatively using direct pyrosequencing in a cohort of 11 cell lines (8 sensitives and 3 resistant). Sensitive cell lines, which have low levels of SGK1 mRNA, exhibited high levels of promoter methylation (Mean CpG1=65%, CpG2=67%, and CpG3=40%), while resistant cell lines displayed low levels of SGK1 promoter methylation (Mean CpG1=11%, CpG2=13%, and CpG3=16%) (Figure S4F). Importantly, the degree of promoter methylation inversely correlated with SGK1 mRNA levels in these cells (Figure S4G). In order to address if these epigenetic marks have a functional effect in the transcriptional activity of SGK1 gene, we performed ChIP-qPCR assays using antibodies against RNA polymerase II (Pol II), an enzyme essential for transcription. In the resistant cell lines HCC1954 and JIMT1, where SGK1 levels are high and promoter methylation is low, we found high occupancy of Pol II and phosphorylated Pol II (S5), which indicates successful transcription initiation, demonstrating that SGK1 transcription is active (Figure S4H). On the other hand, sensitive cell lines MDA-MB-453 and T47D, with low levels of SGK1 and high promoter methylation, we found low occupancy of both Pol II and phosphorylated Pol II (S5) in the SGK1 promoter (Figure S4H). Furthermore, we treated sensitive cell lines with the DNA demethylating agent 5-Aza-2’-deoxycytidine and the histone deacetylase inhibitor panobinostat to revert the promoter methylation, which resulted in increased mRNA levels of SGK1 in the four cell lines tested (Figure S4I). Our results indicate that the differential expression of SGK1 is mediated, at least in part, by epigenetic DNA methylation of the promoter region. Despite the correlation observed in vivo between pNDRG1 levels and SGK1 expression (Murray et al., 2004), in cultured cell lines AKT can also phosphorylate NDRG1 in the absence of SGK1. As a matter of fact, cells that do not express SGK1 but have high AKT activity exhibit NDRG1 phosphorylation, which is sensitive to AKT inhibitors (Sommer et al., 2013). In support of the role of AKT in NDRG1 phosphorylation, we observed that cancer cells sensitive to BYL719 displayed decreased NDRG1 phosphorylation at T346 when treated with BYL719 (Figure 3G). In contrast, resistant cell lines treated with BYL719 maintain NDRG1 phosphorylation, underscoring the role of SGK1 in this setting (Figure 3G). Central to our work, only the combination of BYL719 and GSK2334470 was able to decrease the phosphorylation of NDRG1 in BYL719-    100 resistant cell lines, confirming that the combination of both drugs is required in order to effectively inhibit both SGK1 and AKT activity (Figure 3H). Next, we aimed to characterize the upstream events that regulate SGK1 activity upon PI3Kα and PDK1 inhibition. When we immunoprecipitated endogenous SGK1, we found that BYL719 treatment was not sufficient to completely abolish the kinase activity of SGK1, in contrast to GSK2334470 (Figure 3H). On the other hand, immunoprecipitation of endogenous AKT revealed that while BYL719 treatment completely abrogated AKT kinase activity, this is not the case when cells are treated with GSK2334470, as previously observed (Najafov et al., 2012). This indicates that the combination of PI3Kα and PDK1 inhibitors can simultaneously block the activity of endogenous AKT and SGK1, and single agent treatment is not enough to induce tumor regression as a result of a signaling compensation between AKT and SGK1. To further demonstrate that AKT and SGK1 compensate for each other in phosphorylating downstream shared targets in resistant cells, we took advantage of the known different sensitivity to the inhibition of their HM phosphorylation. While mTORC2- mediated phosphorylation at the HM is indispensable for SGK1 kinase activity (Kobayashi and Cohen, 1999), several reports indicate that AKT remains active in the absence of HM phosphorylation, as phosphorylation at the activation loop (T308) is sufficient to partially activate the kinase (Guertin et al., 2006; Jacinto et al., 2006; Rodrik- Outmezguine et al., 2011). Thus, we treated HCC1954 cells with the mTOR catalytic inhibitor AZD8055, which targets both mTORC1 and mTORC2 and completely inhibits SGK1 but not AKT activity. Increasing concentrations of AZD8055 did not reduce the levels of the substrates pFOXO3 (T32) and pNDRG1 (T346), confirming that mTORC2 inhibition is not sufficient to abolish AKT activity in these cells (Figure S4J). Consistent with previous results, GSK2334470 alone is not capable of inhibiting AKT activity (Figure S4D and Figure 3H), explained by the fact that AKT is efficiently activated by PDK1 in a PIF binding pocket-dependent manner upon 1 µM GSK2334470 treatment (Najafov et al., 2012). Therefore, in the presence of mTOR inhibition, the phosphorylation at the HM is abrogated and this mechanism of AKT activation is no longer supported. Consistently, the addition of GSK2334470 to resistant cells treated with the mTOR inhibitor AZD8055 resulted in a marked decrease in the phosphorylation of both FOXO3 and NDRG1. The combination of AZD8055 and GSK2334470 also   101 resulted in decreased cell viability in both HCC1954 and JIMT1 cells, phenocopying the effects observed by dual PDK1 and PI3Kα inhibition (Figure S4K). To rule out the possibility that this effect was not mediated by mTORC1 inhibition, we knocked down the mTORC2-specific component RICTOR in JIMT1 cells. We observed that RICTOR knockdown decreased the phosphorylation of NDRG1 (T346) only in the presence of PDK1 inhibition (Figure S4L). These experiments demonstrate that both PI3K and PDK1 activity has to be suppressed in order to inhibit downstream AKT and SGK1 phosphorylation and activity in our resistant models. SGK1 mediates resistance to BYL719 We next assessed the contribution of SGK1 in mediating resistance to PI3Kα inhibition. The overexpression of a constitutively active form of SGK1 in MDA-MB-361 cells, which are highly sensitive to PI3Kα inhibition, was sufficient to increase cell viability in the presence of BYL719 (Figure 4A). In parental cells, PI3Kα inhibition decreased both AKT phosphorylation and mTORC1 signaling, as assessed by S6K, S6, and 4EBP1 phosphorylation. On the contrary, cells overexpressing SGK1 maintained mTORC1 signaling in the presence of the BYL719 (Figure 4B). Genetic inactivation of SGK1 has been shown to be challenging in cell lines with high levels of SGK1 (Sommer et al., 2013). Pharmacologically, the few SGK inhibitors currently available have low activity in cellular models. Thus, in order to overcome this problem, we characterized a recently described SGK inhibitor (SGK1-inh) that was discovered using 3D ligand-based virtual screening (Halland et al., 2015). SGK1-inh exhibited an IC50 of 4.8 nM at 10 µM ATP using recombinant SGK1 kinase assay (Figure 4C), with appreciable activity also towards SGK2 and SGK3 (IC50 of 2.8 nM and 590 nM, respectively) (Figure 4D). The specificity of this compound was tested in vitro at a concentration of 1 µM (200X higher than the SGK1 inhibitory dose) against a panel of 140 human kinases. SGK1-inh showed remarkable selectivity towards SGK1 (Figure S5A). Importantly, no activity against AKT1, PDK1, PKC, or RSK was detected at this concentration. However, we found that at this high concentration the activity of S6K was also inhibited, probably due to the high similarity of their catalytic site. Because S6K is a key downstream substrate of mTORC1, we further characterized the activity of SGK1- inh towards S6K. Recombinant in vitro kinase assay of S6K demonstrated an IC50 of 33 nM; seven times higher that than SGK1 IC50 (Figure S5B). At the cellular level, we    102 performed S6K kinase assay in 293T cells overexpressing constitutively active S6K (ΔCT T389E) treated with increasing concentrations of SGK1-inh and we found that the cellular IC50 is ~20 µM (Figure S5C). Next, we further characterized the activity of SGK1-inh towards endogenous S6K in a more physiological model. We used two fibroblast cell lines that lack TSC2 and are derived from TSC2 KO mice and a Lymphangioleiomyomatosis patient, respectively. Increasing concentrations of SGK1-inh up to 30 µM were not able to reduce S6K signaling in these cellular models, as assed by the phosphorylation of the downstream S6K targets pS6 (S235/6), pS6 (S240/4), and pmTOR (S2448) (Figure S5D). This suggests that in cellular models, SGK1-inh does not have activity towards S6K at concentrations below 20-30 µM. We then excluded any potential inhibition of mTORC1 by SGK1-inh by testing this compound against mTOR in a recombinant kinase assay using 4EBP1 as a substrate (IC50 of >5000 nM, Figure S5E). The chemical structure and preliminary characterization of SGK1-inh suggested that this compound acts as an ATP-competitive inhibitor. To experimentally validate these observations, we performed an ATP-competition assay. We found that addition of increasing concentrations of ATP at 100 µM, 250 µM, 350 µM, and 500 µM decreased the potency of SGK1-inh in a dose-dependent manner (Figure S5F). Our computational analyses suggest that SGK1-inh is a type II kinase inhibitor as it binds preferentially to the inactive conformation of the kinase (Figure 4E). Consistently, docking models using the active conformation of SGK1 show that the sulfonamide moiety with the terminal hydrophobic ring points out from the pocket towards the solvent (Figure S5G), rendering the bound state unstable. In contrast, in the inactive conformation, several hydrophobic residues mediate interactions with SGK1-inh within the allosteric DFG-out pocket (mainly V149, L159, V154, and V160; Figure 4F). The pyrazolo(3,4-b)pyrazine head portion of SGK1-inh interacts with the key residues D177 and I179, similar to the interactions of the adenine moiety of ATP (Figure 4F; Figure S5H). The energetics of SGK1-SGK1-inh binding is more favorable than SGK1-ATP, as assessed by binding free energy calculations (Figure S5I). The electrostatic components of these interactions are similar between ATP and SGK1-inh and the majority of the binding energy arises from more favorable packing (van der Waals interactions) made between SGK1-inh and the kinase (Figure S5I). Next, we analyzed the energetic contribution of each individual residue of SGK1 by decomposing the binding free energies. Most of the favorable interactions that take place between SGK1 and SGK1-inh are with amino acids found within the SGK1   103 active site (Figure S5J, upper panel). In silico alanine scanning of the key residues resulted in substantial loss of binding free energies of hydrophobic residues and K127, confirming the importance of these amino acids in the protein-ligand interactions (Figure S5J, lower panel). Given our in vitro and modeling results, we moved forward to cell-based experiments. In our models, we estimated that the appropriate concentration of SGK1-inh to fully inhibit endogenous SGK1 is 10 µM, based on the ability to inhibit NDRG1 phosphorylation in the presence of BYL719 (Figure 4G). This relatively high concentration required to inhibit endogenous SGK1 may be explained by the fact that these sulfonamide derivatives exhibit poor permeability (133 × 10-7 cm/s in Caco2 cells permeability assays) (Halland et al., 2015). Treatment of HCC1954 and JIMT1 cells with the combination of BYL719 and SGK1-inh not only abrogated pNDRG1 (T346) but also mTORC1 signaling (Figure 4I; Figure S5K). Using m7GTP pull downs we also found that the combination of PI3Kα and SGK1 inhibitors induces a decreased cap-dependent translation as seen by the increased 4EBP1 and decreased eIF4A and eIF4G binding to the m7GTP beads (Figure S5L). This translated to superior inhibition of cell viability of BYL719-resistant cell lines treated with the combination of BYL719 and SGK1-inh (Figure 4J; Figure S5M, N). Our results support that in cells expressing high levels of SGK1, both AKT and SGK kinases need to be inhibited simultaneously in order to block mTORC1 and proliferation. We then assessed the potential antitumor activity of SGK1-inh in HCC1954-derived xenografts treated with BYL719, SGK1-inh, and the combination of both agents. We observed that only the combination of BYL719 and SGK1-inh reduced tumor burden in this model (Figure 4L). Once harvested, these tumors exhibited decreased phosphorylation in S6, 4EBP1, and NDRG1 when animals where treated with the combination of BYL719 and SGK1-inh (Figure 4L; Figure S5O). These results show that targeting SGK1 pharmacologically is feasible, and demonstrate that dual inhibition of AKT and SGK1 is required to achieve full suppression of mTORC1. SGK1 interacts with and phosphorylates TSC2 Due to its similarity with AKT, we reasoned that SGK1 could modulate mTORC1 activity by interacting with a component of the TSC/RHEB/mTORC1 axis. Immunoprecipitation of recombinant Flag-tagged TSC1, TSC2, RHEB, and mTOR in 293T cells revealed that SGK1 physically interacts with both mTOR and TSC2 proteins (Figure 5A, Figure S6A).    104 While the interaction between SGK1 and mTOR has previously been described, as mTORC2 is responsible for the HM phosphorylation of SGK1 (Garcia-Martinez and Alessi, 2008), to our knowledge this is the first report showing an interaction between SGK1 and TSC2. This result was corroborated in a cell-live context by performing fluorescence resonance energy transfer (FRET) experiments using EGFP-tagged TSC2 and EYFP-tagged SGK1 in HeLa cells. Excitation of the donor (EGFP) molecule led to emission from the acceptor molecule (EYFP), demonstrating the in vivo direct interaction of these proteins, as assessed by FRET efficiency calculations (Figure 5B). We further confirmed the novel interaction between SGK1 and TSC2 by endogenous co- immunoprecipitation in JIMT1 cells (Figure 5C). Moreover, in order to determine the proportion of endogenous SGK1 that is associated with the TSC complex, we undertook sucrose gradient experiments in JIMT1 lysates. The TSC complex fractionated at high- density fractions (fraction 5), as assessed by the presence of the three components TSC1, TSC2, and TBC1D7 (Figure S6B), as previously shown (Dibble et al., 2012). Although most of SGK1 fractionated at low molarity fractions, approximately 20% of the kinase eluted at similar fractions as the TSC complex. Considering SGK1 as a monomer (or maybe a dimer (Zhao et al., 2007)), only the association with a larger complex such as the TSC complex can explain the elution at these high sucrose gradients. Co-immunoprecipitation assays using five different fragments of TSC2 demonstrated that SGK1 binds to the N-terminal region found between amino acids 1-608 (Figure 5D). This region of TSC2 contains a Leucine Zipper (LZ) domain important for protein-protein interactions and is also required for the interaction with TSC1 (Li et al., 2004). In our immunoprecipitation assays, TSC2 truncation mutants that lack this region of the protein are unable to bind SGK1 (Figure 5D). SGK1 has high similarity to AKT in the kinase domain and thus shares many substrates that contain the AGC-kinase consensus motif RXRXX(S/T), where R is Arginine, X is any amino acid, and (S/T) is a phosphorylatable Serine or Threonine (Alessi et al., 2009). The use of a degenerated phospho-specific motif antibody allows detection of these phosphosites and has previously been shown to be a reliable surrogate for phospho-TSC2 detection (Manning et al., 2002). When we analyzed the TSC2 protein sequence searching for identifiable RXRXX(S/T) motifs, we found seven putative sites of phosphorylation: S939, S981, T993, S1130, S1132, T1462, and S1798. All these sites were conserved across lower species, including mouse, rat, cattle, chicken, frog, and zebra fish (Figure S6C). To systematically test the ability of SGK1 to phosphorylate   105 these residues, we established an in vitro kinase assay using recombinant active SGK1 and TSC2 as a substrate, immunoprecipitated from 293T cells expressing Flag-TSC2. In order to deplete endogenous phosphorylation of TSC2, we pre-treated 293T cells with the AKT inhibitor MK2206. The addition of recombinant SGK1 kinase increased the phosphorylation of the RXRXX(S/T) sites of TSC2, even when cells were pre-treated with AKT inhibitor (Figure 5E). Using mass spectrometry to identify the phosphorylation status of the aforementioned residues in our in vitro kinase assay, we found increased phosphorylation in all these sites, except at T993 (Figure 5F). Mutation of these six phosphorylatable sites into the non-phosphorylatable amino acid alanine (TSC2 6A) completely abrogated the ability of SGK1 to phosphorylate TSC2 in vitro (Figure 5G). SGK1 and AKT share the capability to phosphorylate at least five of these six sites in TSC2. It is well accepted that phosphorylation of these sites increase downstream RHEB-GTP loading and mTORC1 signaling, as a result of a translocation from the lysosome to the cytoplasm (Inoki et al., 2003a; Menon et al., 2014). The phosphorylation and inhibition of TSC2 is phenocopied by the loss of expression of the protein itself, as demonstrated by the induction of mTORC1 activity and consequent resistance to BYL719 in the T47D PI3Kα inhibitor-sensitive cell line depleted of TSC2 (Figure S6D, E). To confirm that our biochemical findings are consistent with the proposed mechanism of resistance to BYL719, we treated HCC1954 and JIMT1 cells with BYL719, GSK2334470, SGK1-inh and the combination of these agents and found that the phosphorylation of endogenous TSC2 decreases only upon dual PI3Kα and PDK1 or SGK1 suppression (Figure 5H). These results demonstrate that SGK1 can sustain mTORC1 activity in BYL719-resistant cells by phosphorylating and inhibiting the mTORC1 negative regulator TSC2. Besides AKT and SGK1, other kinases have been found to phosphorylate TSC2, probably because this GAP acts as an integrator of upstream kinases involved in the regulation of mTORC1 (Li et al., 2004) (Figure S6F). Therefore, we asked in the resistant cell line HCC1954 if other kinases might be involved in the phosphorylation of TSC2 and sustained activation of mTORC1. Extracellular Signal-regulated Kinase (ERK), and the downstream AGC kinase RSK phosphorylate TSC2, activating downstream mTORC1 (Ma et al., 2005; Roux et al., 2004). However, we did not detect changes in TSC2 phosphorylation at S939 or mTORC1 downstream signaling when HCC1954 cells were treated with the MEK inhibitors PD0325901 and GSK1120212 and downstream ERK and RSK was fully inhibited (Figure S6G). AMP-dependent protein    106 kinase (AMPK), which is activated in conditions of energy stress, phosphorylates TSC2 at S1345 and induces the inhibition of mTORC1 (Inoki et al., 2003b). Treatment of HCC1954 cells with the stress-inducing agent 2-deoxyglucose (2DG) and the AMPK inductor A769662 is not able to rescue the sustained phosphorylation of S6 in this resistant model (Figure S6H). In line with the AMPK regulation of mTORC1 signaling, GSK3 kinase has also been reported to phosphorylate TSC2 using the AMPK-specific site S1345 as a priming event, in a process downstream of WNT signaling (Inoki et al., 2006). However, incubation of HCC1954 cells with the recombinant WNT antagonist DKK-1 did not rescue either the sustained S6 phosphorylation, probably in part because these cells have low WNT signaling, as assessed by the immunofluorescence staining of β-catenin (Figure S6J). Altogether, these results suggest that in our resistant models SGK1 is the main kinase involved in the phosphorylation of TSC2 and sustained mTORC1 activation. Discussion In this work, we show that inhibition of the constitutively active kinase PDK1 overcomes resistance to PI3Kα inhibitors in PIK3CA-mutant breast cancer cells insensitive to PI3Kα inhibition. We discovered that in a setting of PI3Kα inhibition and thus low levels of PIP3 and full suppression of AKT, SGK1 contributes to the maintenance of residual mTORC1 activity and cell survival through direct phosphorylation and inhibition of TSC2. Suppression of either PDK1 or SGK1 prevents mTORC1 activation and sensitizes resistant cells to PI3Kα blockade, underscoring the causative role of this signaling route in inducing the resistance phenotype (Figure 6). Summarizing our current knowledge, resistance to PI3Kα inhibitors in PIK3CA-mutant malignancies may occur either as a result of PI3K-dependent or -independent mechanisms. An example of PI3K-dependent acquired resistance mechanism has recently been shown by the observation that loss of PTEN results in activation of PI3K p110β and therefore forfeiting PI3Kα signaling (Juric et al., 2015). Similarly, reactivation of PI3K p110β signaling has also been revealed to be a mechanism of adaptive resistance in PI3Kα-driven cells (Costa et al., 2015). In terms of PI3K-independent mechanisms, we now propose that mTORC1 sustained activity is mediated by PDK1- SGK1 signaling. In this context, AKT activity would be dispensable for cell survival. This is in accordance with previous reports showing that AKT activity is not always required to   107 transduce the downstream PI3K signaling (Gasser et al., 2014; Vasudevan et al., 2009). The role of SGK1 in mediating mTORC1 activation upon PI3Kα inhibition can be explained by the differential regulation of AKT and SGK1 upon pharmacological stress. Although both kinases share the same upstream regulators, mTORC2 and PDK1, AKT contains a PH domain that is required for the PI3K-dependent plasma membrane translocation and subsequent activation. In contrast, SGK1 does not require plasma membrane localization, which could explain, at least in part, why it remains active in the absence of PIP3. Moreover, in our resistant cell lines treated with PI3Kα inhibitor, we observe a substantial but incomplete decrease in SGK1 activity. This can be partially explained by the fact that PIP3 controls mTORC2 (Gan et al., 2011), although other PIP3- independent pools of mTORC2 might be responsible for residual SGK1 activity. While PDK1 is a constitutively active kinase and can be present in both the cytoplasm and the membrane (upon PIP3 synthesis), the subcellular localization of mTORC2 is ambiguous (Cybulski and Hall, 2009). Therefore, it is plausible that different pools of mTORC2 can be found in the membrane, cytoplasm, and other organelles. Pharmacological inhibition of PDK1 has been reported to have a profound effect on the activity of several AGC kinases such as RSK, S6K, PKC, and SGK (Najafov et al., 2011). However, in order to achieve the same inhibitory effects on AKT, higher doses of PDK1 inhibitors are required. As a matter of fact, in the presence of 1 µM GSK2334470, AKT can be efficiently activated by PDK1 through different PIP3-dependent or PIF binding pocket-dependent mechanisms (Najafov et al., 2012). In our experiments using endogenous immunoprecipitated SGK1 and AKT, we show that this is indeed the case. In the presence of BYL719, SGK1 but not AKT remains active; conversely, upon GSK2334470 treatment, SGK1 but not AKT is fully inhibited. Single activity of any of these kinases seems to be sufficient to propagate downstream pro-survival signaling through mTORC1 activation and FOXO3 repression. This is also confirmed by the fact that the combination of both agents efficiently inhibits FOXO3 and mTORC1, eliciting a powerful anti-tumor effect. In this setting, rather than inhibition of AKT, NDRG1 phosphorylation (substrate of both AKT and SGK1) should be used as readout of pathway inhibition (Kobayashi et al., 1999). The role of SGK1 in regulating signaling downstream of mTORC2 is intriguing but not entirely unexpected. From the evolutionary point of view, the SGK1 orthologue in Drosophila melanogaster is the dAkt gene, which shares the same similarity with human SGK1 (63%) and AKT1 (67%) genes. Thus, it is tempting to speculate that due to the    108 high overlap of their substrates, Drosophila dAkt plays the role of both AKT and SGK1. In fact, Ypk2 and Gad8, the SGK1 orthologues in budding and fission yeast, respectively, are the main TORC2 downstream effectors (Cybulski and Hall, 2009; Kamada et al., 2005; Matsuo et al., 2003). Similarly, in the worm Caenorhabditis elegans, sgk-1 appears to be essential for TORC2 signaling, lifespan, and growth (Jones et al., 2009; Soukas et al., 2009). Some evidence also suggests that SGK3 may play an important role in the oncogenicity of PI3K-driven cells (Gasser et al., 2014; Vasudevan et al., 2009). Our results indicate that this is not the case in intrinsic resistance to PI3K pathway inhibitors since low levels of SGK1, but not SGK2 or SGK3, are correlated with sensitivity to BYL719 (Figure 4) and AKT inhibitors (Sommer et al., 2013). Moreover, SGK3 is a direct downstream target of class III phosphoinositide 3-kinase and is indirectly inactivated upon class I PI3K inhibitor treatment (Bago et al., 2014), so it is an unlikely mediator of intrinsic resistance. Despite this, in the future it would be interesting to study whether the different members of the SGK family induce resistance to PI3K inhibitors in acquired resistance models. In summary, our findings show that SGK1 mediates resistance to PI3Kα inhibitors through the activation of mTORC1, which can be reverted by PDK1 blockade. This study highlights the importance of understanding the underlying mechanisms of protein kinase regulation in order to uncover critical nodes for pharmacological intervention and improve the therapeutic options for oncogene-driven cancers. Experimental Procedures RNAi screening The synthetic lethal RNAi screening was carried out at the High-Throughput Screening Core Facility of MSKCC. The kinome and phosphatome Ambion Silencer Select v4.0 libraries were purchased from Life Technologies and contain 2130 unique siRNAs targeting each of the 710 human kinase genes and 894 unique siRNAs targeting each of the 298 human phosphatase genes. Diluted siRNA were transferred into assay plates at a final concentration of 50 nM. As a reference, we used Silencer Select Negative Control #1 siRNA (4390843) as a negative control and PLK1 siRNA (s449) as the positive control.   109 JIMT1 and HCC1954 cells were seeded and were reverse transfected using Dharmafect-1 at 0.05 µL/well. Next, cells were treated with DMSO or BYL719 1 µM and 7 days after transfection, cell viability was assessed using Alamar blue and Nuclei Count using Hoechst staining and quantified using LEADseeker (GE Healthcare) and INCA2000 (GE Healthcare), respectively. For the hit nomination, the BDA method was used as previously described (Bhinder and Djaballah, 2012). Briefly, this method comprises 5 steps to analyze and score active siRNA duplexes and genes: (1) active duplex identification, (2) active gene identification, (3) off-target effects filtering, (4) re-scoring, and (5) biological classifications. To identify modulators of BYL719 resistance, active genes were nominated from the active siRNA duplexes using a hit rate per gene (H score) of ≥60. H score is defined as follows: H  score =   number  of  active  siRNA  duplextotal  number  of  siRNA  duplexes  ×  100 Using this approach, 5 genes were identified and the two most active duplexes of each gene were purchased and screened for cell viability and pS6 staining in the presence of BYL719 1 µM. siRNA were from Ambion: PIK3CA (s10520, s10522), MTOR (s602, s603), PDPK1 (s10274, s10275), PAPL (s52890, s52892), and PP1R12A (s935, s937). Confirmation screening was carried out as described above. For pS6 (S240/4) staining, cells were reverse-transfected and after 72 hours, they were treated for 4 hours with BYL719 at 1 µM. Next, cells were fixed with 4% Paraformaldehyde in PBS and stained using pS6 (S240/4) antibody from Cell Signaling (2215), followed by Alexa Fluor 488 secondary antibody. Fluorescence was quantified using INCA2000 (GE Healthcare). Final nomination was performed using the H score described above and genes that sensitized cells to BYL719 and decreased pS6 (S240/4) were selected. Plasmids and site-directed mutagenesis The Myc-tagged constructs pCCL-PDK1 WT, KD (K111N), K465E, and L155E were a gift from Dr. Primo and Dr. Gagliardi (University of Turin). pLPCX-HA-SGK1(Δ60) was obtained from Dr. Conzen (The University of Chicago) and was used as a template to subclone the cDNA and generate pdEYFP-SGK1(Δ60) and pLenti7.3-V5- SGK1(Δ60,S422D). The kinase-inactive K127A and constitutively active S422D mutant were generated using PCR-based site-directed mutagenesis.    110 Plasmids expressing Flag-tagged mTOR (26603), TSC1 (8995), TSC2 (8996), and RHEB (15888) were obtained from Addgene. pcDNA3-Flag-TSC2 WT and 5A (S939A, S981A, S1130A, S1132A, T1462A) were a gift from Dr. Manning and were used as a template for the generation of pcDNA3-Flag-TSC2 6A (5A, S1798A) and pcDNA3-Flag- TSC2 6E, respectively. Plasmids encoding for the TSC2 truncation mutants were provided by Dr. Xiong (University of North Carolina at Chapel Hill) and pEGFP-TSC2 was from Dr. Krymskaya (University of Pennsylvania). PDPK1 targeting shRNA pLKO- based vector used in this study was TRCN0000039782, although other clones were also tested. RICTOR-targeting shRNA plasmid was from Addgene (1853). All constructs were validated by Sanger sequencing. Cells and lentiviral production All cell lines were obtained from ATCC except for JIMT1 (AddexBio), used at low passages, and maintained at 37°C in a 5% CO2 atmosphere in the recommended culture media. HCT116 PDPK1-/- and +/+ cells were a gift from Dr. Mills (MD Anderson) and were originally generated by Dr. Vogelstein’s laboratory (Johns Hopkins University) (Ericson et al., 2010). For lentiviral production, 293T cells were seeded in 10-cm plates, transfected with pCMV-VSVG, pCMV-dR8.2, and the plasmid of interest using FuGene HD (Promega). Viruses were collected 72h post-transfection, filtered through a 0.45 µm filter (Millipore), and recipient cells were infected twice using viral supernatants supplemented with 8 µg/µL of polybrene (Sigma). Transduced cells were selected using puromycin (2 µg/mL) or Fluorescence Activated Cell Sorting (FACS) for the pCCL and pLenti7.3 vectors, which contain EGFP as a selectable marker. Reagents, cell viability and apoptosis BYL719 and MK2206 were obtained from the Stand Up to Cancer (SU2C) pharmacy. GSK2334470 and Staurosporine were purchased at Selleckchem. SGK1-inh was a gift from M. Nazare and N. Halland. All drugs were dissolved in DMSO for in vitro experiments. Cell viability was measures using the MTT assay. Briefly, 5000 cells were seeded in 96 well plates, treated for 6 days, and assayed using 0.25% MTT (Sigma) and 50 mM sodium succinate (Sigma) solutions for 3 hours. Formazan crystals were dissolved with DMSO and absorbance was measured at 570 nm of wavelength.   111 For Caspase 3/7 activity, the Caspase-Glo® 3/7 Assay kit from Promega was used following manufacturer’s instructions. The caspase inhibitor zVAD-fmk was used to inhibit apoptosis in cells and was also obtained from Promega. Immunoblot, immunoprecipitation, and kinase assay For western blot analysis, proteins were extracted in RIPA buffer supplemented with protease and phosphatase inhibitors (Roche). Protein lysates were separated using SDS-PAGE gels and transferred to a PVDF membrane. Then, membranes were probed using specific antibodies. PDK1, pAKT (S473), pAKT (T308), pS6K (T389), pS6 (S240/4), pS6 (235/6), p4EBP1 (S65), PARP, Actin, pRSK (S227), cleaved Caspase 3, pFOXO1/3 (T24/T32), SGK1, SGK2, SGK3, pNDRG1 (T346), NDRG1, Flag, HA, and phospho-RXRXX(S/T) were from Cell Signaling Technology (CST). For SGK1 HM and activation loop phosphorylation detection, we used pS6K (T389) antibody (9205) and pPKC (pan) (γT514) (9379) from CST, as previously reported (Garcia-Martinez and Alessi, 2008). The SGK1 and AKT antibodies for endogenous immunoprecipitation were raised in sheep by the Division of Signal Transduction Therapy (DSTT) at the University of Dundee and affinity-purified against the indicated antigens: anti-AKT1 (S695B, third bleed; raised against residues 466–480 of human Akt1: RPHFPQFSYSASGTA), anti- SGK1 antibody (S062D, third bleed, raised against recombinant SGK1 protein (DU35257). For immunoprecipitation assays, 293T cells were transiently transfected with appropriate plasmids and 24h post-transfection, cells were washed in cold PBS and lysed using NP- 40 buffer (150 mM NaCl, 10 mM Tris pH 8, 1% NP-40, 10% glycerol). Lysates were rotated at 4°C for 4 hours with EZview™ Red ANTI-FLAG® M2 or ANTI-HA agarose beads (Sigma) and washed three times using NP-40 buffer. For in vitro kinase assay, immunoprecipitated Flag-TSC2 was used as a substrate in a reaction with recombinant His-SGK1 (Δ60) (MRC-PPU Reagents) and ATP (Signalchem) in kinase assay buffer containing 25 mM MOPS pH 7.2, 12.5 mM β-glycerophosphate, 25 mM MgCl2, 5 mM EGTA, 2 mM EDTA and 0.25 mM DTT at 30oC for 30 minutes. In vitro kinase activity of endogenous SGK1 and AKT was assayed by measuring [γ-32P] ATP incorporation into Crosstide substrate peptide [GRPRTSSFAEGKK]. SGK1 and AKT were immunoprecipitated from HCC1954 cell line 4 hours after treatment. Immunoprecipitates were washed once with lysis buffer containing 500 mM NaCl, once with lysis buffer, and twice with Buffer A (50 mM Tris pH7.5, 0.1mM EGTA). Reactions were carried out in 40    112 µL total volume containing 0.1 mM [γ-32P] ATP (400-1000 cpm/pmol), 10 mM magnesium acetate, and 30 µM Crosstide peptide. Reactions were terminated by adding 10 µL 0.1 mM EDTA. 40 µL of the reaction mix was spotted on P81 paper, which was immediately immersed into 50 mM ortophosphoric acid and washed several times. Papers were rinsed in acetone and air dried. Radioactivity was quantified by Cerenkov counting. One unit of enzyme activity was defined as the amount of enzyme that catalyzes incorporation of 1 nmol of [γ-32P] ATP into the substrate over one minute. Mass spectrometry Kinase assay reactions were performed in biological triplicates and resolved using SDS- polyacrylamide gel electrophoresis, stained with SimplyBlue SafeStain (Life Technologies, Thermo Fisher Scientific), and the band corresponding to Flag-TSC2 was excised and digested with trypsin as described by (Shevchenko et al., 2006). The tryptic peptides were resuspended in buffer A containing 3% formic acid and analyzed by microcapillary liquid chromatography with tandem mass spectrometry using a NanoAcquity LC (Waters) with a 100 µm-inner-diameter x  10 cm-length C18 column (1.7µm BEH130, Waters) configured with a 180 µm x 2 cm trap column coupled to a Q- Exactive mass spectrometer (Thermo Fisher Scientific) scanning 380-1800 m/z at 70,000 resolution with AGC set at 3 x 106. Peptides were eluted with a linear gradient of 2-30% acetonitrile (0.1% formic acid) in water over 90 min at a flow rate of 300 nL/min. Key parameters for the data dependent MS were top 10 DDA, AGC 5e4, and ms/ms resolution of 17,000. Data were analyzed using MaxQuant (Max Planck Institute of Biochemistry, Germany; version 1.5.1.0) at default settings with a few modifications. The default was used for first search tolerance and main search tolerance: 20 ppm and 6 ppm, respectively. MaxQuant was set up to search the reference human proteome database. Maxquant performed the search using trypsin digestion with up to 2 missed cleavages. Peptide, Site and Protein FDR were all set to 1% with a minimum of 1 peptide needed for identification but 2 peptides needed to calculate a protein ratio. LFQ quantitation was confirmed by manual integration of the MS1 data for the phosphorylation sites of interest. Raw data as well as original MaxQuant results files can be provided upon request. Microarray, qPCR, and ChIP-qPCR   113 RNA was isolated from cells using the QIAGEN RNeasy kit. For microarray analysis, biotinylated cRNA was prepared according to the standard Illumina protocol. After fragmentation, cRNA was hybridized with Illumina GX HT12 Human Array. Slides were washed and stained in the Illumina instrument following manufacturer’s protocol. Slides were scanned using Illumina Bead Array Reader. Data were analyzed using GenomeStudio software. No normalization and background correction are performed first, then quantile normalization and background correction are done. For mRNA expression analysis, cDNA was prepared using the Bio-Rad cDNA synthesis kit. cDNA was amplified by quantitative PCR using SYBR Select Master Mix (Applied Biosystems) in the ViiA 7 Real-Time PCR system. All reactions were carried out in triplicate. Primers used for mRNA expression were: ERBB3: 5’-CTGATCACCGGCCTCAAT; Rv-GGAAGACATTGAGCTTCTCTGG IRS2: Fw-TTCTTGTCCCACCACTTGAA; Rv-CTGACATGTGACATCCTGGTG TNFSF10: 5’-CCTCAGAGAGTAGCAGCTCACA; Rv-CAGAGCCTTTTCATTCTTGGA BCL6: Fw-CTGCAGATGGAGCATGTTGT; Rv-TCTTCACGAGGAGGCTTGAT Actin: Fw-CGTCTTCCCCTCCATCGT; Rv-GAAGGTGTGGTGCCAGATTT ChIP assays were performed as described previously (Toska et al., 2012). Briefly, cells were treated with 1% formaldehyde for 15 min at room temperature and quenched with ice-cold 125 nM glycine for 5 min. Lysed cells were sonicated on ice to yield 200-800 bp DNA fragments. Chromatin was incubated overnight at 4°C with 2 µg of anti-FOXO3A antibody (Santa Cruz Biotechnology; sc-11351) or nonspecific IgG. Immunocomplexes were precipitated by incubation overnight with protein G-conjugated beads. Immunoprecipitates were washed and crosslinks were reversed by heating to 65°C for 6 hours and then treated with proteinase K for 1 hour at 55°C. Chromatin was purified using QiaQuick PCR clean-up columns. ChIP primers used in this study were: Control: Fw-CAGCTCAGTGCTGTTGGTGG; Rv-ACCATCCAACCCTGGAGATC IRS2: Fw-GACAATCAAAGTCCTTCCCAAA; Rv-CCTTTTGACCTGTGCTGTTGT TNFSF10, Fw-AAAGAAAATCCCTCCCCTCTT; Rv-CACTCACCTCAAGCCCATTT Animal studies and IHC Animals were maintained and treated in accordance with Institutional Guidelines of Memorial Sloan Kettering Cancer Center (Protocol number 12-10-019). 5 x 106 cells in 1:1 PBS/Matrigel (Corning) were injected subcutaneously into six-week-old female athymic Foxn1nu nude mice. When tumors reached a volume of ~150mm3, mice were    114 randomized, treated, and tumors were measured twice a week during a month. At least 10 tumors per group were used in all the studies. Treatments were as follows: BYL719 (25 mg × kg-1 in 0.5% carboxymethylcellulose (Sigma), daily p.o.); GSK2334470 (100 mg × kg-1 in 10% of 1:1 Kolliphor® EL/EtOH (Sigma), three times/week, i.p.); SGK1-inh (50 mg × kg-1 in 40% of 3:1 Glycofurol/Kolliphor® RH 40 mixture (Sigma) in 0.9% saline, daily p.o.). Tumors were harvested at the end of the experiment three hours after the last dosage, fixed in 4% formaldehyde in PBS, and paraffin-embedded. IHC was performed on a Ventana Discovery XT processor platform using standard protocols and the following antibodies from Cell Signaling Technology: pAKT(S473) (4060),1:70; pS6 (S240/4) (5364),1:500; pNDRG1 (T346) (5482), 1:200. Primary staining was followed by 60 minutes incubation with biotinylated goat anti-rabbit IgG (Vector labs) 1:200. Blocker D, Streptavidin- HRP and DAB detection kit (Ventana Medical Systems) were used according to the manufacturer instructions. Docking and molecular dynamics simulations The structure of SGK1 kinase is only available in its inactive form, with missing structural information such as the coordinates of the αC helix. We constructed the 3D structures of SGK1 kinase both in its active and inactive forms using comparative modeling methods based on homology. The templates used were the available crystal structure of SGK1 kinase in the inactive state (PDB: 2R5T) (Zhao et al., 2007), high-resolution crystal structure of the kinase domain of AKT (55% homology) in its active (PDB: 1O6K) (Yang et al., 2002a) and inactive (PDB: 1GZN) (Yang et al., 2002b) states. The program Modeller (version 9.12) (Sali and Blundell, 1993) was used for the generation of homology models. Several models were generated and the models with the best physicochemical properties were further refined using all atom molecular dynamics (MD) simulations. The 3D structures of SGK1-inh and ATP were built using the Maestro module and minimized using the Macromodel module, employing the OPLS-2005 force field, in the program Schrodinger 9.0. The minimized SGK1 inhibitor and ATP were docked into the binding pockets of SGK1 kinase models with Glide (Friesner et al., 2004) using standard docking protocols (Kannan et al., 2015). Refinement of the docked models of SGK1- inhibitor and SGK1-ATP complexes were carried out using MD simulations under the Sander module of the program Amber14. The partial charges and force field parameters for SGK1 inhibitor and ATP were generated using the Antechamber module in Amber.   115 All atom versions of the Amber 03 force field (ff03) (Duan et al., 2003) and the general Amber force field (GAFF) (Wang et al., 2004) were used for the protein and the inhibitors respectively. All the simulations were carried out at 300 K using standard protocols (Kannan et al., 2015). Three independent MD simulations (assigning different initial velocities) were carried out on each equilibrated SGK1-ATP and SGK1-inhibitor structure for 100 ns each, with conformations saved every 10 ps. Simulation trajectories were visualized using VMD (Humphrey et al., 1996) and figures were generated using Pymol. The binding free energies (enthalpic components), energy decompositions (to identify “hot spot” residues) and computational alanine scans (of the “hot spot” residues) were calculated using the MMPBSA (Molecular Mechanics Poisson–Boltzmann Surface Area) methodology (Kannan et al., 2015). FRET For FRET experiments, HeLa cells were seeded in chambered coverglass and transfected with 0.5 µg of EGFP Donor plasmid, 0.5 µg of EYFP Acceptor plasmid, or both constructs. 16h post-transfection cells were imaged with a Leica TCS SP8 microscope using the established parameters for Donor (Ex: 458 nm laser at 15%; Em: 466-501 nm) and Acceptor (Ex: 528 nm laser at 3%; Em: 555-600 nm). FRET efficiency was calculated using the following equation as described in (van Rheenen et al., 2004): 𝐸FRET = !"#$  –  !"#$  ×  !  –  !"#$    ×   !  –  !  ×  !!"#$    ×   !!  !  ×  ! Where FRET, EGFP and EYFP refers to the FRET, Donor and Acceptor channels respectively. The corrections factors were α=0.01; β=0.37; γ=0.31; δ=0.02, where α corrects for acceptor cross-excitation crosstalk (α=Donor/Acceptor), β corrects for donor crosstalk (β=FRET/Donor), γ corrects for acceptor cross-excitation (γ=FRET/Acceptor), and δ corrects for FRET crosstalk (δ=Donor/FRET). Mock-transfected cells were used to calculate the background threshold level (background intensity mean + 4 Standard deviation). TMA and patients Formalin-fixed paraffin-embedded (FFPE) tissue blocks from primary invasive breast carcinomas were used to construct the TMA reported in this study. A certified pathologist    116 (E.B.) microscopically examined hematoxylin and eosin-stained sections of all the tumors and selected representative areas, excluding foci of ductal carcinoma in situ and tumor necrosis. All carcinomas were represented in the TMAs in triplicate 0.6-mm cores. An Automatic Tissue MicroArrayer (ATA-27, Beecher Instruments Inc) was used to construct TMAs from a total of 273 breast invasive carcinomas. This comprised clinically and pathologically confirmed triple-negative breast cancer patients (138), ER/PR receptor-positive breast cancer patients (68), and HER2-positive cancer patients (67). Tumor were considered ER/PR receptor-positive if >10% of neoplastic cells showed nuclear positivity. Cases with HER-2 staining intensity of 3+ were considered positive, whereas those with 2+ staining intensity of HER2 were further evaluated by ERBB2 FISH using the PathVision HER2 probe Kit (Abbott Laboratories), and scored as positive if the HER2/Cep17 ratio was 2.2 or greater. 5-µm thick TMA sections were stained for pNDRG1 (T346) following the protocol described above. Based on the observed staining across the different samples, cases were scored as High expression when pNDRG1 staining intensity of 2+ was found >20% of the neoplastic cells. Intermediate staining represented tumors that had 1+ staining intensity in >10% of the neoplastic cells. For the study of patients treated with the PI3Kα inhibitor BYL719, pre-treatment FFPE blocks from patients enrolled in the clinical trial NCT01870505 conducted at MSKCC were used for IHC as described above. For the selection of patients, PIK3CA and other tumor genomic drivers were analyzed using MSK-IMPACT (Cheng et al., 2015). Patients that did not exhibit toxicity during the trial, harbored hot-spot mutations in PIK3CA, and did not harbor mutations in PTEN or KRAS (known to cause resistance to PI3Kα inhibitors) were selected for the biomarker study. The MSKCC Institutional Review Board approved the study. Accession numbers The microarray data has been deposited in the Gene Expression Omnibus database under accession number GSE69189. Author contributions P.C., D.R.A., J.B., and M.S. designed the research. P.C. and H.E. performed the in vitro and in vivo experiments. R.B. performed endogenous kinase assays. E.A.T. performed ChIP experiments. S.K. and C.S.V. performed the docking and MS. F.J.C. performed expression analysis. E.B. scored and quantified IHC experiments. P.C. and M.S.   117 prepared the figures. P.C., M.S., and J.B. wrote the manuscript. Acknowledgments We thank all the investigators that have contributed to this work by providing key reagents. Specially, we are grateful to M. Nazare (Leibniz-Institute for Molecular Pharmacology) and N. Halland (Sanofi-Aventis) who generously provided the SGK1 inhibitor characterized in this work. We also would like to thank H. Djaballah, D. Shum, and B. Bhinder from the High-Throughput Drug Screening Facility for their help in the performance of the siRNA screening, R. Soni and R. Hendrickson from the Proteomics and Microchemistry Core for the help with mass spectrometry, V. Boyko from the Molecular Cytology Core for the help with FRET, C. Jones from the Bioinformatic Core for the help with microarray analysis, and M. Asher form the Pathology Core for the help with IHC. This work has been supported by the Molecular Cytology Core Facility-Core grant P30 CA0087748, the R01CA190642-01A1 (to J. Baselga), the Breast Cancer Research Foundation (to J. Baselga), the Geoffrey Beene Cancer Research Center (to M. Scaltriti), the A*CRC (A*STAR) Singapore for computing resources and the BMSI (A*STAR) Singapore for funding. D. Alessi and R. Bago are supported by the Medical Research Council, and the pharmaceutical companies supporting the Division of Signal Transduction Therapy Unit (AstraZeneca, Boehringer-Ingelheim, GlaxoSmithKline, Merck, Janssen Pharmaceutica and Pfizer). FJ. Carmona holds a fellowship from the Terri Brodeur Foundation. References Alessi, D.R., James, S.R., Downes, C.P., Holmes, A.B., Gaffney, P.R., Reese, C.B., and Cohen, P. (1997). Characterization of a 3-phosphoinositide-dependent protein kinase which phosphorylates and activates protein kinase Balpha. Current biology : CB 7, 261- 269. Alessi, D.R., Pearce, L.R., and Garcia-Martinez, J.M. (2009). New insights into mTOR signaling: mTORC2 and beyond. Science signaling 2, pe27. Arencibia, J.M., Pastor-Flores, D., Bauer, A.F., Schulze, J.O., and Biondi, R.M. (2013). AGC protein kinases: from structural mechanism of regulation to allosteric drug development for the treatment of human diseases. Biochimica et biophysica acta 1834, 1302-1321. Bago, R., Malik, N., Munson, M.J., Prescott, A.R., Davies, P., Sommer, E., Shpiro, N., Ward, R., Cross, D., Ganley, I.G., et al. (2014). Characterization of VPS34-IN1, a selective inhibitor of Vps34, reveals that the phosphatidylinositol 3-phosphate-binding    118 SGK3 protein kinase is a downstream target of class III phosphoinositide 3-kinase. The Biochemical journal 463, 413-427. Bhinder, B., and Djaballah, H. (2012). A simple method for analyzing actives in random RNAi screens: introducing the "H Score" for hit nomination & gene prioritization. Combinatorial chemistry & high throughput screening 15, 686-704. Biondi, R.M., Kieloch, A., Currie, R.A., Deak, M., and Alessi, D.R. (2001). The PIF- binding pocket in PDK1 is essential for activation of S6K and SGK, but not PKB. The EMBO journal 20, 4380-4390. Brunet, A., Bonni, A., Zigmond, M.J., Lin, M.Z., Juo, P., Hu, L.S., Anderson, M.J., Arden, K.C., Blenis, J., and Greenberg, M.E. (1999). Akt promotes cell survival by phosphorylating and inhibiting a Forkhead transcription factor. Cell 96, 857-868. Brunet, A., Park, J., Tran, H., Hu, L.S., Hemmings, B.A., and Greenberg, M.E. (2001). Protein kinase SGK mediates survival signals by phosphorylating the forkhead transcription factor FKHRL1 (FOXO3a). Molecular and cellular biology 21, 952-965. Cerami, E., Gao, J., Dogrusoz, U., Gross, B.E., Sumer, S.O., Aksoy, B.A., Jacobsen, A., Byrne, C.J., Heuer, M.L., Larsson, E., et al. (2012). The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer discovery 2, 401-404. Cheng, D.T., Mitchell, T.N., Zehir, A., Shah, R.H., Benayed, R., Syed, A., Chandramohan, R., Liu, Z.Y., Won, H.H., Scott, S.N., et al. (2015). Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT): A Hybridization Capture-Based Next-Generation Sequencing Clinical Assay for Solid Tumor Molecular Oncology. The Journal of molecular diagnostics : JMD 17, 251-264. Ciriello, G., Gatza, M.L., Beck, A.H., Wilkerson, M.D., Rhie, S.K., Pastore, A., Zhang, H., McLellan, M., Yau, C., Kandoth, C., et al. (2015). Comprehensive Molecular Portraits of Invasive Lobular Breast Cancer. Cell 163, 506-519. Collins, B.J., Deak, M., Arthur, J.S., Armit, L.J., and Alessi, D.R. (2003). In vivo role of the PIF-binding docking site of PDK1 defined by knock-in mutation. The EMBO journal 22, 4202-4211. Costa, C., Ebi, H., Martini, M., Beausoleil, S.A., Faber, A.C., Jakubik, C.T., Huang, A., Wang, Y., Nishtala, M., Hall, B., et al. (2015). Measurement of PIP3 levels reveals an unexpected role for p110beta in early adaptive responses to p110alpha-specific inhibitors in luminal breast cancer. Cancer cell 27, 97-108. Currie, R.A., Walker, K.S., Gray, A., Deak, M., Casamayor, A., Downes, C.P., Cohen, P., Alessi, D.R., and Lucocq, J. (1999). Role of phosphatidylinositol 3,4,5-trisphosphate in regulating the activity and localization of 3-phosphoinositide-dependent protein kinase-1. The Biochemical journal 337 ( Pt 3), 575-583. Cybulski, N., and Hall, M.N. (2009). TOR complex 2: a signaling pathway of its own. Trends in biochemical sciences 34, 620-627. Dibble, C.C., Elis, W., Menon, S., Qin, W., Klekota, J., Asara, J.M., Finan, P.M., Kwiatkowski, D.J., Murphy, L.O., and Manning, B.D. (2012). TBC1D7 is a third subunit of the TSC1-TSC2 complex upstream of mTORC1. Molecular cell 47, 535-546. Duan, Y., Wu, C., Chowdhury, S., Lee, M.C., Xiong, G., Zhang, W., Yang, R., Cieplak, P., Luo, R., Lee, T., et al. (2003). A point-charge force field for molecular mechanics   119 simulations of proteins based on condensed-phase quantum mechanical calculations. Journal of computational chemistry 24, 1999-2012. Elkabets, M., Vora, S., Juric, D., Morse, N., Mino-Kenudson, M., Muranen, T., Tao, J., Campos, A.B., Rodon, J., Ibrahim, Y.H., et al. (2013). mTORC1 inhibition is required for sensitivity to PI3K p110alpha inhibitors in PIK3CA-mutant breast cancer. Science translational medicine 5, 196ra199. Engelman, J.A. (2009). Targeting PI3K signalling in cancer: opportunities, challenges and limitations. Nature reviews Cancer 9, 550-562. Ericson, K., Gan, C., Cheong, I., Rago, C., Samuels, Y., Velculescu, V.E., Kinzler, K.W., Huso, D.L., Vogelstein, B., and Papadopoulos, N. (2010). Genetic inactivation of AKT1, AKT2, and PDPK1 in human colorectal cancer cells clarifies their roles in tumor growth regulation. Proceedings of the National Academy of Sciences of the United States of America 107, 2598-2603. Friesner, R.A., Banks, J.L., Murphy, R.B., Halgren, T.A., Klicic, J.J., Mainz, D.T., Repasky, M.P., Knoll, E.H., Shelley, M., Perry, J.K., et al. (2004). Glide: a new approach for rapid, accurate docking and scoring. 1. Method and assessment of docking accuracy. Journal of medicinal chemistry 47, 1739-1749. Fruman, D.A., and Rommel, C. (2014). PI3K and cancer: lessons, challenges and opportunities. Nature reviews Drug discovery 13, 140-156. Gan, X., Wang, J., Su, B., and Wu, D. (2011). Evidence for direct activation of mTORC2 kinase activity by phosphatidylinositol 3,4,5-trisphosphate. The Journal of biological chemistry 286, 10998-11002. Garcia-Martinez, J.M., and Alessi, D.R. (2008). mTOR complex 2 (mTORC2) controls hydrophobic motif phosphorylation and activation of serum- and glucocorticoid-induced protein kinase 1 (SGK1). The Biochemical journal 416, 375-385. Gasser, J.A., Inuzuka, H., Lau, A.W., Wei, W., Beroukhim, R., and Toker, A. (2014). SGK3 mediates INPP4B-dependent PI3K signaling in breast cancer. Molecular cell 56, 595-607. Guertin, D.A., Stevens, D.M., Thoreen, C.C., Burds, A.A., Kalaany, N.Y., Moffat, J., Brown, M., Fitzgerald, K.J., and Sabatini, D.M. (2006). Ablation in mice of the mTORC components raptor, rictor, or mLST8 reveals that mTORC2 is required for signaling to Akt-FOXO and PKCalpha, but not S6K1. Developmental cell 11, 859-871. Halland, N., Schmidt, F., Weiss, T., Saas, J., Li, Z., Czech, J., Dreyer, M., Hofmeister, A., Mertsch, K., Dietz, U., et al. (2015). Discovery of N-[4-(1H-Pyrazolo[3,4-b]pyrazin-6- yl)-phenyl]-sulfonamides as Highly Active and Selective SGK1 Inhibitors. ACS medicinal chemistry letters 6, 73-78. Humphrey, W., Dalke, A., and Schulten, K. (1996). VMD: visual molecular dynamics. Journal of molecular graphics 14, 33-38, 27-38. Inoki, K., Li, Y., Xu, T., and Guan, K.L. (2003a). Rheb GTPase is a direct target of TSC2 GAP activity and regulates mTOR signaling. Genes & development 17, 1829-1834. Inoki, K., Li, Y., Zhu, T., Wu, J., and Guan, K.L. (2002). TSC2 is phosphorylated and inhibited by Akt and suppresses mTOR signalling. Nature cell biology 4, 648-657. Inoki, K., Ouyang, H., Zhu, T., Lindvall, C., Wang, Y., Zhang, X., Yang, Q., Bennett, C., Harada, Y., Stankunas, K., et al. (2006). TSC2 integrates Wnt and energy signals via a    120 coordinated phosphorylation by AMPK and GSK3 to regulate cell growth. Cell 126, 955- 968. Inoki, K., Zhu, T., and Guan, K.L. (2003b). TSC2 mediates cellular energy response to control cell growth and survival. Cell 115, 577-590. Jacinto, E., Facchinetti, V., Liu, D., Soto, N., Wei, S., Jung, S.Y., Huang, Q., Qin, J., and Su, B. (2006). SIN1/MIP1 maintains rictor-mTOR complex integrity and regulates Akt phosphorylation and substrate specificity. Cell 127, 125-137. Juric, D., Castel, P., Griffith, M., Griffith, O.L., Won, H.H., Ellis, H., Ebbesen, S.H., Ainscough, B.J., Ramu, A., Iyer, G., et al. (2015). Convergent loss of PTEN leads to clinical resistance to a PI(3)Kalpha inhibitor. Nature 518, 240-244. Juric, D., Krop, I., Ramanathan, R.K., Xiao, J., Sanabria, S., Wilson, T.R., Choi, Y., Parmar, H., Hsu, J., Baselga, J., et al. (2013). Abstract LB-64: GDC-0032, a beta isoform-sparing PI3K inhibitor: Results of a first-in-human phase Ia dose escalation study. Cancer Res 73, LB-64-. Juric, D., Rodon, J., Gonzalez-Angulo, A.M., Burris, H.A., Bendell, J., Berlin, J.D., Middleton, M.R., Bootle, D., Boehm, M., Schmitt, A., et al. (2012). Abstract CT-01: BYL719, a next generation PI3K alpha specific inhibitor: Preliminary safety, PK, and efficacy results from the first-in-human study. Cancer Res 72, CT-01-. Kannan, S., Poulsen, A., Yang, H.Y., Ho, M., Ang, S.H., Eldwin, T.S., Jeyaraj, D.A., Chennamaneni, L.R., Liu, B., Hill, J., et al. (2015). Probing the binding mechanism of Mnk inhibitors by docking and molecular dynamics simulations. Biochemistry 54, 32-46. Kobayashi, T., and Cohen, P. (1999). Activation of serum- and glucocorticoid-regulated protein kinase by agonists that activate phosphatidylinositide 3-kinase is mediated by 3- phosphoinositide-dependent protein kinase-1 (PDK1) and PDK2. The Biochemical journal 339 ( Pt 2), 319-328. Kobayashi, T., Deak, M., Morrice, N., and Cohen, P. (1999). Characterization of the structure and regulation of two novel isoforms of serum- and glucocorticoid-induced protein kinase. The Biochemical journal 344 Pt 1, 189-197. Li, Y., Corradetti, M.N., Inoki, K., and Guan, K.L. (2004). TSC2: filling the GAP in the mTOR signaling pathway. Trends in biochemical sciences 29, 32-38. Ma, L., Chen, Z., Erdjument-Bromage, H., Tempst, P., and Pandolfi, P.P. (2005). Phosphorylation and functional inactivation of TSC2 by Erk implications for tuberous sclerosis and cancer pathogenesis. Cell 121, 179-193. Manning, B.D., and Cantley, L.C. (2007). AKT/PKB signaling: navigating downstream. Cell 129, 1261-1274. Manning, B.D., Tee, A.R., Logsdon, M.N., Blenis, J., and Cantley, L.C. (2002). Identification of the tuberous sclerosis complex-2 tumor suppressor gene product tuberin as a target of the phosphoinositide 3-kinase/akt pathway. Molecular cell 10, 151-162. McManus, E.J., Collins, B.J., Ashby, P.R., Prescott, A.R., Murray-Tait, V., Armit, L.J., Arthur, J.S., and Alessi, D.R. (2004). The in vivo role of PtdIns(3,4,5)P3 binding to PDK1 PH domain defined by knockin mutation. The EMBO journal 23, 2071-2082. Menon, S., Dibble, C.C., Talbott, G., Hoxhaj, G., Valvezan, A.J., Takahashi, H., Cantley, L.C., and Manning, B.D. (2014). Spatial control of the TSC complex integrates insulin and nutrient regulation of mTORC1 at the lysosome. Cell 156, 771-785.   121 Murray, J.T., Campbell, D.G., Morrice, N., Auld, G.C., Shpiro, N., Marquez, R., Peggie, M., Bain, J., Bloomberg, G.B., Grahammer, F., et al. (2004). Exploitation of KESTREL to identify NDRG family members as physiological substrates for SGK1 and GSK3. The Biochemical journal 384, 477-488. Najafov, A., Shpiro, N., and Alessi, D.R. (2012). Akt is efficiently activated by PIF- pocket- and PtdIns(3,4,5)P3-dependent mechanisms leading to resistance to PDK1 inhibitors. The Biochemical journal 448, 285-295. Najafov, A., Sommer, E.M., Axten, J.M., Deyoung, M.P., and Alessi, D.R. (2011). Characterization of GSK2334470, a novel and highly specific inhibitor of PDK1. The Biochemical journal 433, 357-369. Pearce, L.R., Komander, D., and Alessi, D.R. (2010). The nuts and bolts of AGC protein kinases. Nature reviews Molecular cell biology 11, 9-22. Potter, C.J., Pedraza, L.G., and Xu, T. (2002). Akt regulates growth by directly phosphorylating Tsc2. Nature cell biology 4, 658-665. Rodrik-Outmezguine, V.S., Chandarlapaty, S., Pagano, N.C., Poulikakos, P.I., Scaltriti, M., Moskatel, E., Baselga, J., Guichard, S., and Rosen, N. (2011). mTOR kinase inhibition causes feedback-dependent biphasic regulation of AKT signaling. Cancer discovery 1, 248-259. Roux, P.P., Ballif, B.A., Anjum, R., Gygi, S.P., and Blenis, J. (2004). Tumor-promoting phorbol esters and activated Ras inactivate the tuberous sclerosis tumor suppressor complex via p90 ribosomal S6 kinase. Proceedings of the National Academy of Sciences of the United States of America 101, 13489-13494. Sali, A., and Blundell, T.L. (1993). Comparative protein modelling by satisfaction of spatial restraints. Journal of molecular biology 234, 779-815. Sancak, Y., Thoreen, C.C., Peterson, T.R., Lindquist, R.A., Kang, S.A., Spooner, E., Carr, S.A., and Sabatini, D.M. (2007). PRAS40 is an insulin-regulated inhibitor of the mTORC1 protein kinase. Molecular cell 25, 903-915. Sarbassov, D.D., Guertin, D.A., Ali, S.M., and Sabatini, D.M. (2005). Phosphorylation and regulation of Akt/PKB by the rictor-mTOR complex. Science 307, 1098-1101. Shevchenko, A., Tomas, H., Havlis, J., Olsen, J.V., and Mann, M. (2006). In-gel digestion for mass spectrometric characterization of proteins and proteomes. Nature protocols 1, 2856-2860. Silvera, D., Formenti, S.C., and Schneider, R.J. (2010). Translational control in cancer. Nature reviews Cancer 10, 254-266. Sommer, E.M., Dry, H., Cross, D., Guichard, S., Davies, B.R., and Alessi, D.R. (2013). Elevated SGK1 predicts resistance of breast cancer cells to Akt inhibitors. The Biochemical journal 452, 499-508. Therasse, P., Arbuck, S.G., Eisenhauer, E.A., Wanders, J., Kaplan, R.S., Rubinstein, L., Verweij, J., Van Glabbeke, M., van Oosterom, A.T., Christian, M.C., et al. (2000). New guidelines to evaluate the response to treatment in solid tumors. European Organization for Research and Treatment of Cancer, National Cancer Institute of the United States, National Cancer Institute of Canada. Journal of the National Cancer Institute 92, 205- 216.    122 Thorpe, L.M., Yuzugullu, H., and Zhao, J.J. (2015). PI3K in cancer: divergent roles of isoforms, modes of activation and therapeutic targeting. Nature reviews Cancer 15, 7-24. Toska, E., Campbell, H.A., Shandilya, J., Goodfellow, S.J., Shore, P., Medler, K.F., and Roberts, S.G. (2012). Repression of transcription by WT1-BASP1 requires the myristoylation of BASP1 and the PIP2-dependent recruitment of histone deacetylase. Cell reports 2, 462-469. van Rheenen, J., Langeslag, M., and Jalink, K. (2004). Correcting confocal acquisition to optimize imaging of fluorescence resonance energy transfer by sensitized emission. Biophysical journal 86, 2517-2529. Vasudevan, K.M., Barbie, D.A., Davies, M.A., Rabinovsky, R., McNear, C.J., Kim, J.J., Hennessy, B.T., Tseng, H., Pochanard, P., Kim, S.Y., et al. (2009). AKT-independent signaling downstream of oncogenic PIK3CA mutations in human cancer. Cancer cell 16, 21-32. Wang, J., Wolf, R.M., Caldwell, J.W., Kollman, P.A., and Case, D.A. (2004). Development and testing of a general amber force field. Journal of computational chemistry 25, 1157-1174. Webb, A.E., and Brunet, A. (2014). FOXO transcription factors: key regulators of cellular quality control. Trends in biochemical sciences 39, 159-169. Yang, J., Cron, P., Good, V.M., Thompson, V., Hemmings, B.A., and Barford, D. (2002a). Crystal structure of an activated Akt/protein kinase B ternary complex with GSK3-peptide and AMP-PNP. Nature structural biology 9, 940-944. Yang, J., Cron, P., Thompson, V., Good, V.M., Hess, D., Hemmings, B.A., and Barford, D. (2002b). Molecular mechanism for the regulation of protein kinase B/Akt by hydrophobic motif phosphorylation. Molecular cell 9, 1227-1240. Zhao, B., Lehr, R., Smallwood, A.M., Ho, T.F., Maley, K., Randall, T., Head, M.S., Koretke, K.K., and Schnackenberg, C.G. (2007). Crystal structure of the kinase domain of serum and glucocorticoid-regulated kinase 1 in complex with AMP PNP. Protein science : a publication of the Protein Society 16, 2761-2769.                                   123                                           HCC1954 -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 20 40 60 80 100 120 BYL719 (Log2 µM) Vi ab ilit y ( % ) DMSO GSK2334470 HCC1954 -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 20 40 60 80 100 120 BYL719 (Log2 µM) Vi ab ilit y ( % ) shGFP shPDK1 Vehicle Vehicle BYL719 BYL719 pA K T (S 47 3) pS 6 (S 24 0/ 4) shGFP shPDK1 pAKT (T308) pAKT (S473) PDK1 pS6K (T389) pS6 (S235) pS6 (S240) p4EBP1 (S65) HCC1954 BYL719: - + - + shGFP shPDK1 Actin A B G S K 23 34 47 0 pAKT (T308) pAKT (S473) pS6K (T389) pS6 (S235) pS6 (S240) p4EBP1 (S65) pRSK2 (S227) Actin D M S O B Y L7 19 C O M B O HCC1954 C D F K H Castel et al. Figure 1 BYL719: - + - + shGFP shPDK1 PARP Actin E G I J L VEHICLE BYL719 GSK2334470 COMBO pA K T (S 47 3) pS 6 (S 24 0/ 4) D M S O B Y L7 19 G S K 23 34 47 0 C O M B O PARP Actin 0 5 10 15 20 25 30 0 200 400 600 800 1000 Days of treatment Tu m or v ol um e (% C ha ng e) Vehicle BYL719 Vehicle BYL719 p=0.011 shGFP shPDK1 D M S O B Y L7 19 G S K 23 34 47 0 C O M B O D M S O B Y L7 19 G S K 23 34 47 0 C O M B O S ta ur os po rin e0 200 400 600 800 1000 C as pa se 3 /7 a ct iv ity (R LU ) zVAD-fmk p<0.0001 + + + + S ta ur os po rin e0 100 200 300 400 500 C as pa se 3 /7 a ct iv ity (R LU ) zVAD-fmk BYL719: sh G FP sh G FP sh PD K1 sh PD K1 p=0.0009 0 5 10 15 20 25 30 0 100 200 300 400 500 600 700 800 900 1000 Days of treatment Tu m or v ol um e (% C ha ng e) Vehicle BYL719 (50mg/kg, daily) GSK2334470 (100 mg/kg, 3x/week) Combination p=0.01947    124 Figure 1. PDK1 inhibition sensitizes resistant cells to BYL719 A) Dose-response curves from HCC1954 cells transduced with shGFP and shPDK1 and treated with BYL719 for 6 days. B) Western blot comparing HCC1954 cells transduced with shGFP and shPDK1 and treated with BYL719 (1 µM) for 4 hours. C) Western blot of PARP comparing HCC1954 cells transduced with shGFP and shPDK1 and treated with BYL719 (1 µM) for 24 hours. D) Caspase 3/7 DEVDase activity of HCC1954 shGFP and shPDK1 cells treated with BYL719 (1 µM) for 12 hours in the presence or absence of caspase inhibitor zVAD-fmk (20 µM). Staurosporine is used as a positive control (1 µM; 4 hours). P-value is calculated using Student’s t-test. E) HCC1954 shGFP and shPDK1 in vivo xenograft treated with Vehicle or BYL719 (n=10/arm). P-value is calculated using Student’s t-test. F) IHC analysis of tumors from E collected at the end of the experiment after 4h of the last treatment. G) Dose-response curves from HCC1954 cells treated with BYL719 in the presence or absence of GSK2334470 (1 µM) over 6 days. H) Western blot comparing HCC1954 cells treated with BYL719 (1 µM), GSK2334470 (1 µM), or the combination of both agents for 4 hours. I) Western blot of PARP in HCC1954 cells treated with BYL719 (1 µM), GSK2334470 (1 µM), or the combination of both agents for 24 hours. J) Caspase 3/7 DEVDase activity of lysates from HCC1954 cells treated with BYL719 (1 µM), GSK2334470 (1 µM), or the combination of both agents for 12 hours in the presence or absence of caspase inhibitor zVAD-fmk (20 µM). Staurosporine is used as a positive control (1 µM, 4 hours). P-value is calculated using Student’s t-test. K) HCC1954 in vivo xenograft treated with Vehicle, BYL719 (25 mg × kg-1), GSK2334470 (100 mg × kg-1), or the combination of both agents (n=10/arm). P-value is calculated using Student’s t-test. L) IHC analysis of tumors from K collected at the end of the experiment after 4h of the last treatment.   125                                         CCNG2 CDKN2B BCL6 IRS2 ERBB3 TNFSF10 DUSP5 CCND1 D M S O B Y L7 19 G S K 23 34 47 0 C O M B O D M S O B Y L7 19 G S K 23 34 47 0 B Y L+ G S K pFOXO1/3 (T24/T32) HCC1954 D M S O B Y L7 19 G S K 23 34 47 0 C O M B O HCC1954 JIMT1 JIMT1 FOXO3 A B C D E DMSO BYL719 GSK2334470 BYL+GSK FOXO3 DAPI F G H Castel et al. Figure 2 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 E nr ic hm en t S co re (E S ) FOXO3 signature BYL+GSK DMSO Rank in ordered dataset NES: 2.1 p value ≤ 10-4 FDR ≤ 10-4 D M S O B Y L7 19 G S K 23 34 47 0 C O M BO 0 1 2 3 4 R el at iv e m R N A le ve ls ERBB3 D M S O B Y L7 19 G S K 23 34 47 0 C O M BO 0 2 4 6 8 R el at iv e m R N A le ve ls TNFSF10 D M S O B Y L7 19 G S K 23 34 47 0 C O M BO 0 1 2 3 4 R el at iv e m R N A le ve ls BCL6 D M S O B Y L7 19 G S K 23 34 47 0 C O M BO 0 2 4 6 8 R el at iv e m R N A le ve ls IRS2 D M S O B Y L7 19 G S K 23 34 47 0 B Y L+ G S K 0 500 1000 1500 2000 FO XO a ct iv ity (R LU ) FOXO reporter p=0.0008 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Fo ld E nr ic hm en t IRS2 promoter DMSO BYL719 GSK2334470 COMBO IgG FOXO3A p=0.0020 ChIP: 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Fo ld E nr ic hm en t TNFSF10 promoter DMSO BYL719 GSK2334470 COMBO IgG FOXO3A p=0.0030 ChIP:    126 Figure 2. FOXO activation upon PDK1 and PI3Kα inhibition A) Heat map showing changes in the top differentially expressed genes in both HCC1954 and JIMT1 cells treated with DMSO, BYL719 (1 µM), GSK2334470 (1 µM), or the combination of both agents for 4 hours. Gene expression up-regulation is indicated in red, while green represents gene expression down-regulation. B) Enrichment plot for the FOXO3 signature in HCC1954 cells using GSEA. NES: Normalized Enrichment Score. C) Heat map from the microarray results showing changes in previously described FOXO3 targets. The mRNA levels are an average of the differential expression of both HCC1954 and JIMT1 cells. D) ERBB3, TNFSF10, BCL6, and IRS2 mRNA expression in HCC1954 cells treated with DMSO, BYL719 (1 µM), GSK2334470 (1 µM), or the combination of both agents for 4 hours. E) Representative images of FOXO3 immunofluorescence (green) in HCC1954 cells treated with DMSO, BYL719 (1 µM), GSK2334470 (1 µM), or the combination of both agents for 4 hours. Nuclei are shown in blue (DAPI). F) Western blot analysis of FOXO1/3 phosphorylation (T24/T32) in HCC1954 and JIMT1 cells treated with DMSO, BYL719 (1 µM), GSK2334470 (1 µM), or the combination of both agents for 4 hours. G) Luciferase reporter assay in HCC1954 cells stably transduced with the FOXO consensus motif reporter construct treated as indicated for 12 hours. RLU (Relative luciferase units). P-value is calculated using Student’s t-test. H) ChIP-qPCR assay of FOXO3A for TNFSF10A and IRS2 promoters in HCC1954 cells treated as indicated in F. P-value is calculated using Student’s t-test.                                         127             T4 7D H C C 20 2 M D A M B 45 3 E FM 19 K P L1 H C C 19 54 JI M T1 B T2 0 C A L1 48 C A L5 10 10 20 30 40 50 130 140 SG K1 m R N A ex pr es si on Sensitive Resistant SGK1 pNDRG1 (T346) SGK3 SGK1 (long exp) Actin NDRG1 SGK2 JI M T1 C A L5 1 B T2 0 H C C 19 54 C A L1 48 T4 7D B T4 74 M D A -M B -3 61 M D A -M B -4 53 Sensitive Resistant A B C Sensitive Resistant 0 2 4 6 8 10 12 14 SG K1 m R N A ex pr es si on p< 0.0001 High Intermediate Low pN D R G 1 100µm Hig Inter iate Negative pN D R G 1 (T 34 6) D N = 67 High Intermediate Negative HER2+ N = 68 High Intermediate Negative ER/PR N = 138 High Intermediate Negative TNBC N = 138 High Intermediate Negative TNBC E pNDRG1 IHC Median PFS (days) POD POSITIVE (N=2) 53.5 2/2 NEGATIVE (N=10) 140.0 2/10 p=0.0072 F T47D MDA-MB-453 BT474 JIMT1 HCC1954 BT20 pNDRG1 BYL719: - + NDRG1 MCF7 S en si tiv e R es is ta nt - + G H G S K 23 34 47 0 JIMT1 HCC1954 pNDRG1 D M S O B Y L7 19 C O M B O BT20 I D M S O B Y L7 19 G S K 23 34 47 0 B Y L+ G S K 0 1 2 3 4 5 6 AK T ac tiv ity (m U /m g ly sa te ) D M S O B Y L7 19 G S K 23 34 47 0 B Y L+ G S K 0.00 0.02 0.04 0.06 0.08 0.10 SG K1 a ct iv ity (m U /m g ly sa te ) Castel et al. Figure 3 -80 -60 -40 -20 0 20 40 60 80 C ha ng e in tu m or s iz e fro m b as el in e (% ) pNDRG1 Low pNDRG1 High    128   Figure 3. SGK1 upregulation in BYL719-resistant cell lines A) SGK1 mRNA levels in breast cancer cell lines sensitive or resistant to BYL719 (n=27). P-value is calculated using Student’s t-test. B) SGK1 mRNA levels in a panel of PIK3CA-mutant breast cancer cell lines sensitive or resistant to BYL719. C) Western blot analysis of SGK1, SGK2, SGK3, and phosphorylated NDRG1 in a panel of PIK3CA- mutant breast cancer cell lines sensitive or resistant to BYL719. D) Representative images of phosphorylated NDRG1 (T346) IHC in breast cancer patients and quantification of the stainings observed in a cohort of 273 breast cancer patients including Triple negative breast Cancer (TNBC), Hormone positive (ER/PR) and Her2 (HER2+) tumors. E) Table depicting the median number of days of progression free survival (PFS) and best response according to RECIST criteria in pNDRG1 (T346) positive and negative patient samples. P value was calculated using Log-rank (Mantel- Cox) test. POD: Progression of disease; SD: Stable disease; PR: Partial response. F) Waterfall plot of the change in tumor size for the patients included in this study. In blue, pNDRG1 IHC high tumors are shown. Dotted lines represent the RECIST criteria, which establish stabilization of disease at +20%/-30% of tumor volume change. G) Western blot for NDRG1 and phosphorylated NDRG1 (T346) in BYL719-sensitive and -resistant breast cancer cell lines treated with BYL719 (1 µM) for 4 hours. H) Western blot of phosphorylated NDRG1 (T346) in the indicated cell lines treated with DMSO, BYL719 (1 µM), GSK2334470 (1 µM), or the combination of both agents for 4 hours. I) Endogenous kinase activity assay for SGK1 and AKT in HCC1954 cells treated with DMSO, BYL719 (1 µM), GSK2334470 (1 µM), or the combination of both agents for 4 hours.                                 129                                           F241 D240 G242 -1 0 1 2 3 4 0 20 40 60 80 100 120 log [SGK1-inh], nM Ac tiv ity (% ) SGK2 (IC50≈ 2.8 nM) SGK3 (IC50≈ 590 nM) -1 0 1 2 3 4 0 20 40 60 80 100 120 log [SGK1-inh], nM Ac tiv ity (% ) IC50 ≈ 4.8 nM SGK1 kinase assayMDA-MB-361 -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 20 40 60 80 100 120 BYL719 (Log2 µM) Vi ab ilit y (% ) LacZ SGK1 pAKT(S473) pS6K(T389) pS6 (S235) p4EBP1(S65) BYL719: - + - + LacZ SGK1 MDA-MB-361 Actin SGK1 A B C D E D177 I179 D240 K127 F 1 2 3 4 5 0 20 40 60 80 100 120 log [SGK1-inh], nM R es id ua l p N D R G 1 (% ) DMSO BYL719 G H HCC1954 -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 20 40 60 80 100 120 140 BYL719 (Log2 µM) Vi ab ilit y (% ) DMSO SGK1-inh I 0 5 10 15 20 25 30 0 200 400 600 800 1000 Days of treatment Tu m or v ol um e (% C ha ng e) Vehicle BYL719 SGK1inh BYL719+SGK1inh p=0.020 J D M S O B Y L7 19 S G K 1- in h B Y L+ S G K 1- in h pAKT (S473) pNDRG1 (T346) pS6 (S240/4) pS6 (S235/6) p4EBP1 (S65) Actin pS6K (T389) HCC1954 K VEHICLE BYL719 SGK1-inh COMBO pA K T (S 47 3) pS 6 (S 24 0/ 4) pN D R G 1 (T 34 6) Castel et al. Figure 4    130 Figure 4. A novel SGK1 inhibitor sensitizes resistant cells to BYL719 A) Dose-response curves from MDA-MB-361 cells transduced with pLenti7.3-LacZ or pLenti7.3-SGK1 (Δ60, S422D) constructs and treated with increasing concentrations of BYL719 for 6 days. B) Western blot analysis of LacZ and SGK1 transduced MDA-MB- 361 cells treated with BYL719 (1 µM) for 4 hours. C) Chemical structure of SGK1-inh and in vitro SGK1 kinase activity assay in the presence of increasing concentrations of SGK1-inh. IC50 is indicated below. 10 µM ATP is used (below SGK1 Km). D) In vitro kinase assay for SGK2 and SGK3, respectively, in the presence of increasing concentrations of SGK1-inh. IC50 is indicated above. 10 µM ATP is used (below SGK2 and SGK3 Km). E) Docking overview of SGK1-inh in the DFG-out conformation of SGK1. The hinge region is colored in red, the DFG motif in green, the ‘allosteric’ hydrophobic cavity that results from the DFG flip is colored in grey, and the rest of the kinase in colored in orange. The DFG motif aminoacids are indicated (D240, F241, G242). F) Detailed residues that mediate the interaction between SGK1-inh and the inactive conformation of SGK1. Hydrogen bonds are shown as purple dotted lines. G) Western blot quantification of NDRG1 phosphorylation (T346) in HCC1954 cells treated with increasing concentrations of SGK1-inh for 4 hours in the absence or presence of BYL719 (1 µM). H) Western blot analysis of HCC1954 cells treated with BYL719 (1 µM), SGK1-inh (10 µM), or the combination of both agents for 4 hours. I) Dose-response curves from HCC1954 cells treated with BYL719 for 6 days in the absence or presence of SGK1-inh (2 µM). J) HCC1954 in vivo xenograft treated with Vehicle, BYL719 (25 mg × kg-1), SGK1-inh (50 mg × kg-1), or the combination of both agents (n=10/arm). P- value is calculated using Student’s t-test. K) IHC analysis of tumors from K collected at the end of the experiment 4 hours after the last dosage. Scale bar represents 100 µm.   131 GAP CC CC LZ FL 1- 608 1-1200 1-1320 1100-1320 837-1765 GAP CC CC LZ S939 S981 S1130/2 T1462 S1798 RARSTS ! RCRSIS ! RDRVRS! RPRGYT ! RKRLIS ! RXRXXS/T ! RVRSMS! S939 S981 AGC motif S1130 S1132 T1462 S1798 0.5 1 2 4 8 16 32 64 128 256 512 S1798 T1462 S1130/2 S1130 S981 S939 Fold Change SGK1 Mock Flag Flag HA Flag Flag HA IP Flag WCL Fl ag -m TO R Fl ag -T S C 1 Fl ag -T S C 2 Fl ag -R he b Fl ag -E V HA-SGK1 (Δ60) A B C Castel et al. Figure 5 pRXRXX(S/T) Flag pAKT(S473) IP: Flag WCL His-SGK1: - + - + MK2206: Flag-TSC2 - - + + In vitro kinase assay D E Flag Flag pRXRXX(S/T) IP: Flag WCL Flag-TSC2: WT 6A His-SGK1: - + - + In vitro kinase assay G F GAP CC CC LZ EGFP% EYFP% KD 0.5 0 EGFP EYFP FRET FRET Efficiency Flag-SGK1: Flag HA HA-TSC2 FL 1- 60 8 1- 12 00 1- 13 20 11 00 -1 32 0 83 7- 17 65 E V E V - + + + + + + + IP: HA * TSC2 W C L B ea ds Ig G S G K 1 IP JIMT1 SGK1 H pTSC2 (S939) TSC2 D M S O D M S O S G K 1- in h S G K 1- in h G S K 23 34 47 0 G S K 23 34 47 0 DMSO BYL719 JI M T1 pTSC2 (S939) TSC2 H C C 19 54 Ba ck go run d EV FR ET 0 5 10 15 20 25 30 FR E T pi xe ls /R O I (> 5 0% e ffi ci en cy ) p<0.0001    132 Figure 5. SGK1 interacts with and phosphorylates TSC2 A) Flag co-immunoprecipitation assay in 293T cells transfected with the indicated plasmids. B) Representative efficiency images from the FRET experiment performed in HeLa cells 16 hours after transfection with the constructs indicated in the upper scheme. Quantification of FRET efficiency dots is indicated. P-value is calculated using Student’s t-test. C) Endogenous co-immunoprecipitation between SGK1 and TSC2 in JIMT1 cells. 10 mg of lysates were immunoprecipitated with the sheep antibody described in Methods (DU35257). D) (Upper panel) Schematic cartoon of the different truncation mutants used in co-immunoprecipitation assays. Domains are indicated: LZ (leucine zipper); CC (coiled coil); GAP (GTP-ase activation protein) (Lower panel) Co- immunoprecipitation assay in 293T cells between Flag-SGK1 and different truncation mutants of HA-TSC2. Asterisk indicates an unspecific band for the heavy chain of IgG. E) In vitro kinase assay using recombinant His-SGK1 and immunoprecipitated Flag- TSC2 from 293T cells as a substrate. MK2206 (2 µM, 1 hour). F) Schematic view and amino acid sequence of the predicted SGK1 phosphorylation sites in TSC2. Domains are indicated: LZ (leucine zipper); CC (coiled coil); GAP (GTP-ase activation protein). (Left panel) Quantification of the phosphorylated sited identified using LC-MS/MS in the absence or presence of recombinant SGK1 are indicated. G) In vitro kinase assay using recombinant His-SGK1 and immunoprecipitated and dephosphorylated Flag-TSC2 WT or 6A (S939A, S981A, S1130A, S1132A, T1462, S1798) as a substrate. H) Western blot of phosphorylated TSC2 (S939) in HCC1954 and JIMT1 cells treated with DMSO, BYL719 (1 µM), GSK2334470 (1 µM), SGK1-inh (10 µM), or the combination of both agents for 4 hours.                             133                                           Castel et al. Figure 6    134 Figure 6. Proposed model of PI3Kα resistance in SGK1 expressing cells PIK3CA-mutant breast tumors depend on the PI3K pathway, which mainly signals through AKT. AKT phosphorylates and inhibits FOXO3 and TSC2, promoting mTOC1 activity and tumor progression (left panel). In the presence of PI3Kα inhibitors, PIP3 levels in the plasma membrane are negligible and AKT cannot be activated. SGK1 expressing cells become resistant to PI3Kα inhibitors, as SGK1 is not fully inhibited in the presence of these therapies, supporting FOXO3 and TSC2 phosphorylation, which promotes mTORC1 activity and tumor progression (middle panel). When SGK1 expressing cells are treated with PI3Kα and PDK1 inhibitors, both AKT and SGK1 are inhibited, inducing tumor regression as a result of FOXO3 activation and mTORC1 inhibition.                                                             135   Figure S1. Relative to Figure 1 A) Overview of the large siRNA screening carried out in this work using libraries against the human kinome and phosphatome. B) Venn diagram indicating the number of genes found to sensitize to BYL719 treatment (1µM) in each individual cell line and in common. The table contains the gene name and NCBI mRNA accession number targeted by the siRNA found to sensitize both cell lines to BYL719. C) Quantification of pS6 (S240/4) staining in the siCTR, siMTOR, and siPDPK1 transfected cells in the presence of DMSO or BYL719 (1µM) in JIMT1 cells 4 hours after treatment. Quantification of the green fluorescence from the whole well is indicated as a fold change of the control untreated cells.             Castel et al. Figure S1 HCC1954 JIMT1 DMSO BYL719 Reverse Transfection Cell count and nomination 6 days 37 35 HCC1954 JIMT1 5 siC TR siM TO R# 1 siM TO R# 2 siP DP K1# 1 siP DP K1# 2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 pS 6 (S 24 0/ 4) st ain ing Fo ld ch an ge DMSO BYL719 A B C PAPL NM_001004318 MTOR NM_004958 PDPK1 NM_002613 PIK3CA NM_006218 PPP1R12A NM_002480    136             PDPK1+/+ PDPK1-/- 0 0.5 1 2 BYL719 (µM): 0 0.5 1 2 HCT116 pAKT (S473) pS6K (T389) pS6 (S235) pS6 (S240) Actin pRSK2 (S227) eIF4E eIF4A eIF4G 4EBP1 WCL eIF4E eIF4A eIF4G 4EBP1 IP: m7GTP S ep ha ro se D M S O B Y L7 19 G S K 23 34 47 0 C O M B O A ZD 80 55 HCC1954 pAKT(S473) pS6K (T389) pS6 (S235/6) pS6 (S240/4) Actin JIMT1 D M S O B Y L7 19 G S K 23 34 47 0 C O M B O p4EBP1 (S65) 4EBP1 eIF4G eIF4A eIF4E 4EBP1 eIF4G eIF4A eIF4E WCL IP: m7GTP HCC1954 S ep ha ro se D M S O D M S O B Y L7 19 B Y L7 19 A ZD 80 55 GFP PDK1 shRNA JIMT1 -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 20 40 60 80 100 120 BYL719 (Log2 µM) Vi ab ilit y ( % ) shGFP shPDK1 A B D M S O D M S O B Y L7 19 B Y L7 19 shGFP shPDK1 pS6K (T389) pAKT (T308) pS6 (S240/4) Actin pS6 (S235/6) JIMT1 PDK1 C D D M S O B Y L7 19 D M S O B Y L7 19 A ZD 80 55 0 1 2 3 4 5 Fo ld c ha ng e 4EBP1 GFP PDK1shRNA: D M S O B Y L7 19 D M S O B Y L7 19 A ZD 80 55 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Fo ld c ha ng e eIF4G GFP PDK1shRNA: D M S O B Y L7 19 D M S O B Y L7 19 A ZD 80 55 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Fo ld c ha ng e eIF4A GFP PDK1shRNA: D M S O B Y L7 19 D M S O B Y L7 19 A ZD 80 55 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Fo ld c ha ng e eIF4E GFP PDK1shRNA: p4 E B P 1 (T 37 /4 6) VEHICLE VEHICLE BYL719 BYL719 shGFP shPDK1 E pAKT (T308) pAKT (S473) pS6K (T389) pS6 (S235) pS6 (S240) p4EBP1 (S65) D M S O B Y L7 19 G S K 23 34 47 0 C O M B O BT20 Actin J D M S O B Y L7 19 G S K 23 34 47 0 C om bo 0 10 20 30 40 50 S ph as e (% ) JIMT1 p< 0. 00 05 0 5 10 15 20 25 30 0 200 400 600 800 1000 Days of treatment Tu m or v ol um e (% C ha ng e) Vehicle BYL719 GSK2334470 Combo p=0.0072 G H COMBO VEHICLE BYL719 GSK2334470 pA K T (S 47 3) pS 6 (S 24 0/ 4) I pRSK2 (S227) RSK2 Actin GSK2334470 (µM): 0 0.1 0.3 1 3 10 HCC1954 pRSK2 (S227) RSK2 Actin JIMT1 GSK2334470 (µM): 0 0.1 0.3 1 3 10 F HCC1954 -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 20 40 60 80 100 120 GSK2334470 (Log2 µM) Vi ab ilit y ( % ) JIMT1 -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 20 40 60 80 100 120 GSK2334470 (Log2 µM) Vi ab ilit y ( % ) JIMT1 -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 20 40 60 80 100 120 BYL719 (Log2 µM) Vi ab ilit y ( % ) DMSO GSK2334470 BT20 -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 20 40 60 80 100 120 BYL719 (Log2 µM) Vi ab ilit y ( % ) DMSO GSK2334470 K D M S O B Y L7 19 G S K 23 34 47 0 C O M B O A ZD 80 55 0 1 2 3 4 5 Fo ld c ha ng e 4EBP1 D M S O B Y L7 19 G S K 23 34 47 0 C O M B O A ZD 80 55 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Fo ld c ha ng e eIF4G D M S O B Y L7 19 G S K 23 34 47 0 C O M B O A ZD 80 55 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Fo ld c ha ng e eIF4A D M S O B Y L7 19 G S K 23 34 47 0 C O M B O A ZD 80 55 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Fo ld c ha ng e eIF4E L M N O p4 E B P 1 (T 37 /4 6) VEHICLE BYL719 GSK2334470 Combination P Q PH domain PIF pocket PDK1 (WT) PDK1 (KE) PDK1 (LE) AGC kinase AGC kinase PIP3 PIP3 PIP3 AGC kinase KD = K127N = kinase inactive KE = K465E = PIP3-binding inactive LE = L155E = PIF-binding inactive BYL719: PDK1 pS6 (S235/6) pAKT (S473) pS6K (T389) HCT116 PDPK1-/- EV Myc-PDK1 - + - + - + - + - + WT KD KE LE pS6 (S240/4) pS6K (T229) pRSK2 (S227) R Castel et al. Figure S2   137     Figure S2. Relative to Figure 1 A) Dose-response curves from JIMT1 shGFP and shPDK1 cells treated with BYL719 for 6 days. B) Western blot comparing JIMT1 shGFP and shPDK1 cells treated with BYL719 (1µM) for 4h. C) m7GTP pull down assay for HCC1954 shGFP and shPDK1 cells treated with BYL719 (1µM) for 4h. Quantification of the m7GTP-precipitated proteins is indicated in fold change. AZD8055 is used as a control at 1µM. D) p4EBP1 (T37/46) IHC from the tumors harvested from Figure 1E. E) Western blot analysis of phosphorylated RSK2 (S227) in HCC1954 and JIMT1 cells treated with increasing concentrations of GSK2334470 for 8h. F) Dose-response curves from HCC1954 and JIMT1 resistant cell lines treated with GSK2334470 for 6 days. G) Dose-response curves from JIMT1 cells treated with BYL719 in the presence or absence of GSK2334470 (1µM) during 6 days. H) Same as G using BT20 TNBC cells. I) Western blot comparing JIMT1 cells treated with BYL719 (1µM), GSK2334470 (1µM), or the combination of both agents for 4h. J) Same as I using BT20 TNBC cells. K) m7GTP pull down assay for HCC1954 cells treated with BYL719 (1µM), GSK2334470 (1µM), or the combination of both agents for 4h. Quantification of the m7GTP-precipitated proteins is indicated in fold change. AZD8055 is used as a control at 1µM. L) p4EBP1 (T37/46) IHC from the tumors harvested from Figure 1K. M) S-phase quantification in JIMT1 cells treated with BYL719 (1µM), GSK2334470 (1µM), or the combination of both agents for 24h and stained with Propidium iodide for cell cycle analysis. N) JIMT1 in vivo xenograft treated with Vehicle, BYL719, GSK2334470, or the combination of both agents (n=10/arm). P-value is calculated using Student’s t-test. O) IHC analysis of tumors from N collected at the end of the experiment 4 hours after the last dosage. P) Western blot analysis of HCT116 PDPK1+/+ and PDPK1-/- isogenic cell lines treated with increasing concentrations of BYL719 for 4 hours. Q) Schematic representation of the effects of the PIP3-binding and PIF-binding pocket deficient mutants used in R. Red circles indicate phosphate groups and green circles indicate hydrophilic-charged aminoacid E. Arrows indicate electric charge repulsion. R) Western blot of HCT116 PDPK1-/- cells transfected with different pCCL-PDK1 mutants. EV (empty vector), WT (wild type), KD (kinase death; K111N), KE (PIP3-binding deficient; K465E), LE (PIF-binding pocket deficient; L155E). Cells were treated with BYL719 (1µM) for 4 hours before collection.      138     Figure S3. Relative to Figure 2 A) Differentially expressed genes in JIMT1 (left) and HCC1954 (right) cells treated with BYL719 (1µM), GSK2334470 (1µM), or the combination of both agents for 4h. Gene expression up-regulation is indicated in red, while blue represents gene expression down-regulation. B) Enrichment plot for the FOXO3 signature of GSEA in JIMT1 cells. NES: Normalized Enrichment Score. C) ERBB3, TNFSF10, BCL6, and IRS2 mRNA expression in JIMT1 cells treated with DMSO, BYL719 (1µM), GSK2334470 (1µM), or the combination of both agents for 4h. D) FOXO3A immunofluorescence (green) in JIMT1 cells treated with DMSO, BYL719 (1µM), GSK2334470 (1µM), or the combination of both agents for 4h. Nuclei are shown in blue (DAPI). E) ChIP-qPCR assay of FOXO3A for TNFSF10A and IRS2 promoters in JIMT1 cells treated as indicated in C. P- value is calculated using Student’s t-test.       0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Fo ld E nr ic hm en t TNFSF10 promoter DMSO BYL719 GSK2334470 COMBO IgG FOXO3A p=0.0009 ChIP: 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 E nr ic hm en t S co re (E S ) FOXO3 signature BYL+GSK DMSO Rank in ordered dataset NES: 2.1 p value ≤ 10-4 FDR ≤ 10-4 0.6 A B C D M S O G S K 7 0 B Y L 7 1 9 C O M B O KLHL24 GABARAPL1 GRB7 CITED2 HBP1 GBP2 BTG1 SESTD1 ULK1 YPEL3 KIAA1370 CCDC28A BRD8 TMEM140 ARID5B TXNIP PIM1 TNFSF10 FAM46A BCL6 IRS2 CABC1 C5orf41 TM4SF1 WSB1 ZSWIM4 FAM100B SPSB3 MMD MNT ZFAND5 ABTB1 CBLB FBXL20 CTDSP2 CHMP1B C16orf75 CCDC77 FBXO32 KLHDC2 YPEL5 KIAA0430 PCMTD1 ATG14 CYBRD1 MXD3 FOXO4 PIK3IP1 FRAT2 RNF103 KLF6 HMGB2 HSD17B11 CCNG2 C2orf55 PAIP2 FBXO33 CYTH2 HECA C20orf108 ZNF217 JMJD1C PILRB P2RY2 VAV3 LFNG RARA TGIF1 ERBB3 F3 FJX1 LYAR AMD1 RRS1 MAT2A ADORA2B CCND1 BRIX1 CHSY1 GRPEL1 C7orf40 NIP7 F12 AEN MARS2 CDC25A PLD6 NOP56 POLR1E GFOD1 PNP NOP16 PNO1 ID3 DYNLL1 MRPS17 C12orf5 LOC653506 GPATCH4 PAK1IP1 Average gene expression JIMT1 −1 −0.5 0 0.5 1 Value 0 2 0 5 0 Color Key and Histogram C o u n t DM SO GS K7 0 BY L7 19 CO MB O LHL24 ABARAPL1 RB7 ITED2 BP1 BP2 TG1 ESTD1 LK1 PEL3 IAA1370 CDC28A RD8 MEM140 RID5B XNIP IM1 NFSF10 AM46A CL6 IRS2 ABC1 5orf41 M4SF1 SB1 SWIM4 AM100B PSB3 MD NT FAND5 BTB1 BLB BXL20 TDSP2 HMP1B 16orf75 CDC77 BXO32 LHDC2 PEL5 IAA0430 CMTD1 TG14 YBRD1 XD3 OXO4 IK3IP1 RAT2 NF103 LF6 MGB2 SD17B11 CNG2 2orf55 AIP2 BXO33 YTH2 ECA 20orf108 NF217 MJD1C ILRB 2RY2 AV3 FNG ARA GIF1 RBB3 3 JX1 YAR MD1 RS1 AT2A DORA2B CND1 RIX1 HSY1 RPEL1 7orf40 IP7 12 EN ARS2 DC25A LD6 OP56 OLR1E FOD1 NP OP16 NO1 ID3 YNLL1 RPS17 12orf5 OC653506 PATCH4 AK1IP1 Average gene expression JIMT1 −1 −0.5 0 0.5 1 Value 0 20 50 Color Key and Histogram Co un t D M S O G S K 70 B Y L7 19 B Y L+ G S K JIMT1 E D DM SO G SK 70 BY L7 19 CO M BO KLHL24 GABARAPL1 GRB7 CITED2 HBP1 GBP2 BTG1 SESTD1 ULK1 YPEL3 KIAA1370 CCDC28A BRD8 TMEM140 ARID5B TXNIP PIM1 TNFSF10 FAM46A BCL6 IRS2 CABC1 C5orf41 TM4SF1 WSB1 ZSWIM4 FAM100B SPSB3 MMD MNT ZFAND5 ABTB1 CBLB FBXL20 CTDSP2 CHMP1B C16orf75 CCDC77 FBXO32 KLHDC2 YPEL5 KIAA0430 PCMTD1 ATG14 CYBRD1 MXD3 FOXO4 PIK3IP1 FRAT2 RNF103 KLF6 HMGB2 HSD17B11 CCNG2 C2orf55 PAIP2 FBXO33 CYTH2 HECA C20orf108 ZNF217 JMJD1C PILRB P2RY2 VAV3 LFNG RARA TGIF1 ERBB3 F3 FJX1 LYAR AMD1 RRS1 MAT2A ADORA2B CCND1 BRIX1 CHSY1 GRPEL1 C7orf40 NIP7 F12 AEN MARS2 CDC25A PLD6 NOP56 POLR1E GFOD1 PNP NOP16 PNO1 ID3 DYNLL1 MRPS17 C12orf5 LOC653506 GPATCH4 PAK1IP1 Average gene expression JIMT1 −1 −0.5 0 0.5 1 Value 0 20 50 Color Key and Histogram Co un t DM SO GS K7 0 BY L7 19 CO MB O GFOD1 AMD1 MAT2A FAM46A PNP C7orf40 LYAR F3 MRPS17 PLD6 RRS1 BRIX1 NIP7 ID3 C12orf5 ADORA2B FJX1 CCND1 LOC653506 MARS2 GRPEL1 DYNLL1 CHSY1 CDC25A NOP56 NOP16 F12 POLR1E PNO1 GPATCH4 PAK1IP1 AEN CHMP1B CABC1 ZSWIM4 ERBB3 CYBRD1 CTDSP2 C2orf55 YPEL5 C20orf108 PAIP2 MMD HSD17B11 JMJD1C VAV3 RARA TGIF1 PILRB LFNG P2RY2 TXNIP TNFSF10 GRB7 KLHL24 HBP1 ZNF217 FBXO32 BTG1 KLF6 HMGB2 FBXO33 IRS2 CITED2 CCDC28A FRAT2 CYTH2 KIAA0430 GABARAPL1 GBP2 ATG14 MNT PCMTD1 FBXL20 YPEL3 SESTD1 CBLB ABTB1 KLHDC2 FOXO4 TM4SF1 HECA KIAA1370 ULK1 TMEM140 ZFAND5 C16orf75 RNF103 WSB1 SPSB3 MXD3 CCDC77 BCL6 BRD8 FAM100B C5orf41 CCNG2 PIM1 PIK3IP1 ARID5B Average gene expression HCC1954 −1 −0.5 0 0.5 1 Value 0 20 50 Color Key and Histogram Co un t D M S O G S K 7 0 B Y L 7 1 9 C O M B O GFOD1 AMD1 MAT2A FAM46A PN C7orf40 LYAR F3 MRPS17 PLD6 RRS1 BRIX1 NIP7 ID3 C12orf5 ADORA2B FJX1 CCND1 LOC653506 MARS2 GRPEL1 DYNLL1 CHSY1 CD 25A NOP56 NOP16 F12 POLR1E PNO1 GPATCH4 PAK1IP1 AEN CHMP1B CABC1 ZSWIM4 ERBB3 CYBRD1 CTDSP2 C2orf55 YPEL5 C20orf108 PAI 2 MMD HSD17B11 JM D1C VA 3 RA A TGIF1 PILRB LFNG P2RY2 TXNIP TNFSF10 GRB7 KLHL24 HBP1 ZNF217 FBXO32 BTG1 KLF6 HMGB2 FBXO33 IRS2 CITED2 CCDC28A FRAT2 CYTH2 KIAA0430 GABARAPL1 GBP2 ATG14 MNT PCMTD1 FBXL20 YPEL3 SE TD1 CBLB ABTB1 KLHDC2 FOXO4 TM4SF1 HECA KIAA1370 ULK1 TMEM140 ZFAND5 C16orf75 RNF103 WSB1 SP B3 MXD3 CCDC77 BCL6 BRD8 FAM100B C5orf41 CCNG2 PIM1 PIK3IP1 ARID5B Average gene expression HCC1954 −1 −0.5 0 0.5 1 Value 0 2 0 5 0 Color Key and Histogram C o u n t D M S O G S K 70 B Y L7 19 HCC1954 B Y L+ G S K DAPI FOXO3 DMSO BYL719 GSK2334470 BYL+GSK Castel et al. Figure S3 D M S O B Y L7 19 G S K 23 34 47 0 C O M BO 0 1 2 3 4 R el at iv e m R N A le ve ls ERBB3 D M S O B Y L7 19 G S K 23 34 47 0 C O M BO 0 10 20 30 40 50 R el at iv e m R N A le ve ls TNFSF10 D M S O B Y L7 19 G S K 23 34 47 0 C O M BO 0 1 2 3 4 R el at iv e m R N A le ve ls BCL6 D M S O B Y L7 19 G S K 23 34 47 0 C O M BO 0 1 2 3 4 5 R el at iv e m R N A le ve ls IRS2 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Fo ld E nr ic hm en t IRS2 promoter DMSO BYL719 GSK2334470 COMBO IgG FOXO3A p=0.0059 ChIP:   139             JIM T1 HC C1 95 4 MD A- MB -45 3 T4 7D 0 5 10 15 20 25 (% In pu t) IgG Pol II CTD Pol II CTD (pS5) Methylated Unmethylated SGK1 promoter Sen sitiv e Res ista nt 0 20 40 60 80 100 M et hy la tio n (% ) p=0.014 CpG 1 n=8 n=3 Sen sitiv e Res ista nt 0 20 40 60 80 100 M et hy la tio n (% ) p=0.009 CpG 2 n=8 n=3 Sen sitiv e Res ista nt 0 20 40 60 80 100 M et hy la tio n (% ) CpG 3 p=0.014 n=8 n=3 GTCCGTTCCGCAT ACACGTGA! TSS MDA-MB-361 MDA-MB-453 HCC1954 BT-474 CAMA-1 T47D JIMT1 CAL51 Sensitive Resistant TSS -56 bp 391bp A Low SGK1 High SGK1 0.0 0.2 0.4 0.6 0.8 1.0 1.2 BY L7 19 V ia bi lit y (v s D M SO ) p< 0.0001 B Sensitive Resistant 0 2 4 6 8 10 SG K2 re la tiv e m R N A ex pr es si on C Sensitive Resistant 0 2 4 6 8 10 12 14 SG K3 re la tiv e m R N A ex pr es si on D T4 7D M D A -M B -3 61 M D A -M B -4 53 B T4 74 C A L- 14 8 H C C 19 54 JI M T1 B T2 0 C A L5 10 2 4 6 8 10 12 14 N D R G 1 (p ho sp ho /to ta l) Fo ld C ha ng e Sensitive Resistant E F 0 20 40 60 80 100 0 2 4 6 8 10 12 14 Methylation (%) m R N A le ve ls R=0.4304 CpG 1 0 20 40 60 80 100 0 2 4 6 8 10 12 14 Methylation (%) m R N A le ve ls R=0.4352 CpG 2 0 20 40 60 0 2 4 6 8 10 12 14 Methylation (%) m R N A le ve ls R=0.4026 CpG 3 G H I MCF7 MDA361 MDA453 T47D 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 6.0 6.5 7.0 SG K1 re la tiv e m R N A ex pr es si on DMSO AZA+LBH p<0.0005 p<0.005 p<0.001 p<0.0001 J pNDRG1 (T346) Actin p4EBP1 (S65) pAKT (S473) AZD8055: DMSO HCC1954 GSK2334470 0 0.2 1 2 0 0.2 1 2 pS6 (S235/6) pFoxO (T32) AZD80 5: K HCC1954 -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 20 40 60 80 100 120 140 GSK2334470 (Log2 µM) Vi ab ilit y (% ) DMSO AZD8055 JIMT1 -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 20 40 60 80 100 120 140 160 GSK2334470 (Log2 µM) Vi ab ilit y (% ) DMSO AZD8055 L Rictor pNDRG1 (T346) Actin shRictor shGFP GSK2334470 (µM): 0 1 1.5 2 0 1 1.5 2 JIMT1 Castel et al. Figure S4 Low SGK1 High SGK1    140   Figure S4. Relative to Figure 3 A) Cell viability of breast cancer cell lines treated with BYL719 (2 µM) and classified according to the SGK1 mRNA expression in high (>median expression) and low ( 5000 nM F -1 0 1 2 3 4 0 20 40 60 80 100 120 log [SGK1-inh], nM Ac tiv ity (% ) SGK1 kinase activity 10 µM ATP 100 µM ATP 250 µM ATP 350 µM ATP 500 µM ATP G H I G V112 K127 V149 L150 V160 L159 L176 Y178 I171 Y220 L238 T239 F241 Residues ΔΔ G (k ca l/m ol ) J SGK1-inh ATP K pAKT (S473) pNDRG1 (T346) p4EBP1 (S65) JIMT1 pS6 (S240/4) pS6 (S235/6) Actin D M S O B Y L7 19 S G K 1- in h B Y L+ S G K 1- in h L M D M S O B Y L7 19 S G K 1- in h C O M B O A ZD 80 55 0 1 2 3 4 5 Fo ld c ha ng e 4EBP1 D M S O B Y L7 19 S G K 1- in h C O M B O A ZD 80 55 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Fo ld c ha ng e eIF4G D M S O B Y L7 19 S G K 1- in h C O M B O A ZD 80 55 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Fo ld c ha ng e eIF4A D M S O B Y L7 19 S G K 1- in h C O M B O A ZD 80 55 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Fo ld c ha ng e eIF4E p4 E B P 1 (T 37 /4 6) VEHICLE BYL719 SGK1-inh Combination N JIMT1 -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 20 40 60 80 100 120 BYL719 (Log2 µM) V ia bi lit y (% ) DMSO SGK1-inh BT20 -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 20 40 60 80 100 120 BYL719 (Log2 µM) V ia bi lit y (% ) DMSO SGK1-inh O Castel et al. Figure S5    142 of Dundee (http://www.kinase-screen.mrc.ac.uk/kinase-inhibitors). B) In vitro S6K1 kinase assay using recombinant KKRNRTLTK peptide as a substrate in the presence of increasing concentrations of SGK1-inh. IC50 value is indicated. C) In vitro S6K1 kinase assay using constitutively active S6K kinase immunoprecipitated from 293T cells expressing HA-S6K (ΔCT T389E) and treated with increasing concentrations of SGK1- inh. Recombinant GST-S6 was used as a substrate and phosphorylated S6 (S235/6) antibody was used for the detection of phosphorylated substrate by Western blot. IC50 value is indicated. D) Western blot analysis of S6K targets in TSC2 knockout mouse embryonic fibroblasts (MEF) and fibroblasts derived from a TSC2null Lymphangioleiomyomatosis (LAM) patient treated with increasing concentrations of SGK1-inh for 4 hours. Everolimus was used as a positive control at 200 nM. E) In vitro mTOR kinase assay using recombinant 4EBP1 as a substrate in the presence of increasing concentrations of SGK1-inh. IC50 value is indicated. F) SGK1-inh IC50 determination in an ATP competition assay using increasing concentration of ATP. G) Residues involved in the interaction between SGK1 (active conformation) and SGK1-inh. Hydrogen bonds are shown as purple dotted lines. H) Residues involved in the interaction between ATP and the active conformation of SGK1. Hydrogen bonds are shown as purple dotted lines and Mg+2 as a grey sphere. I) Distribution of free energies (ΔG) of the conformations sampled during MD simulations of SGK1 bound to SGK1-inh (black) or ATP (red). Distribution of van der Waals’ interactions and electrostatic solvation contribution for the total binding energy are shown. J) Upper panel. Decomposition of binding free energy on per-residue basis for SGK1 (DFG-out conformation) and SGK1-inh complex. Lower panel. Alanine scanning results for the selected residues are shown. Results are expressed in the change of free energy when the indicated residues are mutated to alanine ΔΔG. K) Western blot analysis of JIMT1 cells treated with BYL719 (1µM), SGK1-inh (10µM), or the combination of both agents for 4h. L) m7GTP pull down assay for HCC1954 cells treated with BYL719 (1µM), SGK1- inh (10 µM), or the combination of both agents for 4h. Quantification of the m7GTP- precipitated proteins is indicated in fold change. AZD8055 is used as a control at 1µM. M) Dose-response curves from JIMT1 cells treated with increasing concentrations of BYL719 in the presence or absence of SGK1-inh (2µM) for 6 days. N) Dose-response curves from JIMT1 cells treated with increasing concentrations of BYL719 in the presence or absence of SGK1-inh (2µM) for 6 days. O) p4EBP1 (T37/46) IHC from the tumors harvested from Figure 4L.   143                 0 1 2 3 4 5 6 7 8 9 10 11 0 10 20 30 40 Fractions P er ce nt ag e of to ta l p ro te in TSC2 TSC1 TBC1D7 SGK1 TSC complex TSC2 pAKT (S473) pS6K (T389) pS6 (S235) pS6 (S240) p4EBP1 (S65) Actin D M S O D M S O B Y L7 19 B Y L7 19 T47D shGFP shTSC2 A H. sapiens DSFRARSTSLNE! M. musculus DSFRARSTSLNE! R. norvegicus DSFRARSTSLNE! B. taurus DSFRARSTSLNE! G. gallus DSFRARSTSLNE! X. tropicalis DSFRARSTSLNE! D. rerio DSFRARSTSLNE! AGC consensus RXRXXS! ! ! ! ! EAFRCRSISVSE! EAFRCRSISVSE! ------------! DAFRCRSISVSE! DAFRSRSISVSE! ------------! DAFRSRSISVSE! RXRXXS! HVVRSRIQTSLT! HVVRSRIQTSLT! ----SRIQTSLT! HVVRSRIQTSLT! HVVRSRIQTSIT! ----SRMQTSVT! HAV-RRMQTSST! RXRXXT! RGARDRVRSMSGGH! RGARDRVRSMSGGH! RGARDRVRSMSGGH! HGARDRVRSMSGGH! RSSRNRVRSMSGGH! QTPRHRVRSMSGGT! TNTRTRVRSISGGH! RXRXXS! RXRXXS! ! ! SGLRPRGYTISD! SGLRPRGYTISD! SGLRPRGYTISD! SGLRPRGYTISD! SGHRPRGYTISD! TGHRPRGHTISD! TGHRPRGHTISV! RXRXXT! VGQRKRLISSVE! TGQRKRLISSVD! TGQRKRLISSVD! TGQRKRLVSSVD! TGQRKRLISSVD! TGQRTRLISAVD! AGQRKRLVSTVD! RXRXXS! ! - S9 39 ! - S9 81 ! - T9 93 ! - S1 13 0 ! - S1 13 2 ! - T1 46 2! - S1 79 8 ! B HA-SGK1 Flag-mTOR: - + HA Flag HA Flag IP: Flag WCL C D Flag Flag HA HA IP: Flag WCL Flag-TSC2: HA-SGK1 - + TSC2 TBC1D7 SGK1 TSC1 JIMT1 Fractions: 1 2 3 4 5 6 7 8 9 10 11 1M 0.4M Sucrose: E F GAP CC CC LZ ERK RSK AMPK GSK3β S540 S664 S1345 T1329 S1333 S1337 S1341 S1798 T1462 S939 G pTSC2 (S939) TSC2 pAKT (S473) pRSK (T359) pS6 (S235/6) p4EBP1 (S65) Actin HCC1954 D M S O P D 03 25 90 1 G S K 11 20 21 2 D M S O P D 03 25 90 1 G S K 11 20 21 2 DMSO BYL719 pS6 (S240/4) pERK1/2 (T202/Y204) -6 -5 -4 -3 -2 -1 0 1 2 3 4 0 20 40 60 80 100 120 140 160 180 BYL719 (log2 µM) Vi ab ilit y (% ) T47D shTSC2 shGFP shTSC2 H pACC (S79) Actin pAKT (S473) pS6 (S235/6) pS6 (S240/4) D M S O 2D G A 76 96 62 D M S O 2D G A 76 96 62 DMSO BYL719 HCC1954 DMSO BYL719 0 0.02 0.2 0 0.02 0.2 DKK-1 (µg/mL): pS6 (S235/6) pS6 (S240/4) pAKT (S473) pβ-Catenin (S33/7) Actin HCC1954 pACC (S79) Actin pAKT (S473) pS6 (S235/6) pS6 (S240/4) D M S O 2D G A 76 96 62 D M S O 2D G A 76 96 62 DMSO BYL719 HCC1954 DMSO BYL719 0 0.02 0.2 0 0.02 0.2 DKK-1 (µg/mL): pS6 (S235/6) pS6 (S240/4) pAKT (S473) pβ-Catenin (S33/7) Actin I Castel et al. Figure S6    144 Figure S6. Relative to Figure 5 A) Co-immunoprecipitation assay using Flag-mTOR (left) or Flag-TSC2 (right) and HA- SGK1 in 293T cells. B) Western blot analysis of sucrose gradient fractions collected upon ultracentrifugation. Columns were packed in densities ranging from 0.4 to 1M of sucrose, and a small aliquot of each fraction was analyzed. Densitometry quantification of the sucrose gradient results are represented and dotted lines represent the fractions in which the TSC complex is highly enriched, as assessed by the immunodetection of the three components TSC1, TSC2, and TBC1D7. C) Alignment of the sequence of TSC2 comprising the AGC phosphorylation motifs RXRXX(S/T). R-1 and R-3 are highlighted in blue and phosphorylatable S or T in red. Alignment was performed with ClustalW2 using the protein sequence from mouse (Mus musculus), rat (Rattus norvegicus), cattle (Bos taurus), chicken (Gallus gallus domesticus), frog (Xenopus tropicalis), and zebrafish (Danio rerio). D) Dose-response curves from T47D cells transduced with lentivirus expressing shGFP and shTSC2 and treated with increasing concentrations of BYL719 for 6 days. E) Western blot from T47D shGFP and shTSC2 cells treated with BYL719 (1 µM) for 4h. F) Representative signaling integration of other kinases involved in the phosphorylation of TSC2. Residues previously identified to be phosphorylated by the indicated kinases are shown. Domains are indicated: LZ (leucine zipper); CC (coiled coil); GAP (GTP-ase activation protein). G) Western blot analysis of HCC1954 cells treated with the MEK inhibitors PD0325901 (1 µM) and GSK1120212 (50 nM) in the presence or absence of BYL719 (1 µM) for 4 hours. H) Western blot analysis of HCC1954 cells treated with the AMPK inductors 2-deoxyglucose (50 mM) and A769662 (300 µM) in the presence or absence of BYL719 (1 µM) for 4 hours. Phosphorylation of the previously described substrate Acetyl-CoA Carboxylase (ACC) S79 is shown as control for AMPK activation. I) Western blot analysis of HCC1954 cells treated with increasing concentrations of the WNT antagonist Dickkopf WNT signaling pathway inhibitor 1 (DKK-1) for 30 minutes in the presence or absence of BYL719 (1 µM) for 4 hours. Phosphorylation of the previously described substrate β-catenin S33/7 is shown as control for GSK3β activation. J) Immunofluorescence analysis of β-catenin (green) in HCC1954 cells at basal conditions or treated with 0.2 µg/mL of DKK-1 for 30 minutes. Nuclei are shown in blue (DAPI). Scale bar represents 50 µm.       145 ARTICLE 3     Title: Somatic PIK3CA mutations as a driver of sporadic venous malformations Journal: Science Translational Medicine Impact Factor: 15.843                                                    146                                                                 147         VASCULAR MALFORMAT IONS Somatic PIK3CA mutations as a driver of sporadic venous malformations Pau Castel,1 F. Javier Carmona,1 Joaquim Grego-Bessa,2 Michael F. Berger,1,3 Agnès Viale,4 Kathryn V. Anderson,2 Silvia Bague,5 Maurizio Scaltriti,1,3 Cristina R. Antonescu,3 Eulàlia Baselga,6 José Baselga1,7* Venous malformations (VM) are vascular malformations characterized by enlarged and distorted blood vessel channels. VM grow over time and cause substantial morbidity because of disfigurement, bleeding, and pain, representing a clinical challenge in theabsenceofeffective treatments (Nguyenetal., 2014;Uebelhoeretal., 2012). Somaticmutationsmayactas drivers of these lesions, as suggestedby the identificationof TEKmutations in a proportionof VM (Limaye et al., 2009).We report that activating PIK3CA mutations gives rise to sporadic VM in mice, which closely resemble the histology of the humandisease. Furthermore, we identifiedmutations in PIK3CA and related genes of the PI3K (phosphatidylinositol 3-kinase)/AKT pathway in about 30% of human VM that lack TEK alterations. PIK3CAmutations promote downstream signaling and proliferation in endothelial cells and impair normal vasculogenesis in embryonic development. We suc- cessfully treated VM in mouse models using pharmacological inhibitors of PI3Ka administered either systemically or topically. This study elucidates the etiology of a proportion of VM and proposes a therapeutic approach for this disease. INTRODUCTION Venous malformations (VM) [Online Mendelian Inheritance in Man (OMIM) #600195] are the most frequent form of vascular malformation and are characterized by the presence of a single endothelial layer forming distended blood vessels of variable diameter that are surrounded by a disorganized mural cell layer containing both smooth muscle cells and pericytes (1, 2). Sporadic VM are particularly evident when they involve skin or mucosae. These lesions grow over time, causing substantial mor- bidity such as disfigurement, bleeding, and pain. They may also affect other tissues including muscles, joints, or the intestine (3). Furthermore, they represent a clinical challenge because of the lack of effective treat- ments, although some patients derive limited benefits from surgery or sclerotherapy (4, 5). Currently, there are no approved pharmacological treat- ments for these lesions that often tend to recur after conventional therapy. Previous studies in sporadic VM have identified activating somatic mutations in the gene coding for the endothelial-specific tyrosine kinase receptor TEK (TIE2) in almost half of the cases analyzed (6, 7). Activating mutations in TEK result in enhanced activation of the downstream PI3K (phosphatidylinositol 3-kinase)/AKT andMAPK (mitogen-activated protein kinase) pathways (8–11) and have been shown to promote the growth of human umbilical vein endothelial cells (HUVECs) in xenograft assays (11), deregulate the expression of genes involved in vascular develop- ment (12), compromise the endothelial cell (EC) monolayer as a result of the loss of fibronectin (13), and decrease the expression of the mural cell attractant PDGFB (platelet-derived growth factor, b polypeptide) (12). However, in a large proportion of VM that do not harbor mutations in TEK, the underlying pathogenesis remains unknown. The PI3K pathway is a central regulator of cell survival, growth, and metabolism (14), and its deregulation as a result of genetic or epi- genetic perturbations is observed in a variety of human tumors and overgrowth syndromes (15, 16). PI3K is regulated by virtually all receptor tyrosine kinases, including TIE2, and mediates the phosphorylation of phosphatidylinositol 4,5-biphosphate, giving rise to the second mes- senger phosphatidylinositol 3,4,5-trisphosphate that triggers down- stream signaling. AKT and mTOR (mammalian target of rapamycin) are two well-studied effectors of the PI3K pathway and are responsible for the cellular phenotypes such as cell cycle progression, proliferation, anabolism, and others (14). PI3K also plays a critical role in vascular homeostasis, and this pathway is essential for angiogenesis and main- tenance of the mature endothelium (17, 18). Hotspot mutations in PIK3CA, the gene encoding for the catalytic p110a subunit of PI3K, cause pathway hyperactivation resulting in cellular transformation and uncontrolled growth in epithelial and mesenchymal cells (15). This aberrant PI3K/AKT signaling is asso- ciated with overgrowth syndromes that are often accompanied by dif- ferent types of vascular malformations including lymphatic, venous, and arteriovenous malformations (19), as well as isolated lymphatic malformations (20). We report the existence of gain-of-function mutations in the PI3K/ AKT pathway in clinical specimens of isolated sporadic VM and provide a genetically engineered mouse model (GEMM) of this disease. The presence of PIK3CAmutations in sporadic VM, together with the obser- vation that these mutations are mutually exclusive with TEK mutations, suggests that VM may be defined as a disease state characterized by the presence of somatic activating mutations in the TIE2-PI3K-AKT axis. Together, these findings could result in the development of efficacious therapies for this challenging disease and could contribute to a genomi- cally based classification of vascular lesions. RESULTS PIK3CASprr2f-Cre mice develop spinal and cutaneous VM An unexpected finding pointed us toward the critical role of PI3K in the pathogenesis of VM. We were originally interested in studying the role of PIK3CA, the gene encoding the catalytic p110a subunit of PI3K 1Human Oncology and Pathogenesis Program (HOPP), Memorial Sloan Kettering Cancer Center, 1275 York Avenue, New York, NY 10065, USA. 2Developmental Biology Program, Sloan Kettering Institute, New York, NY 10065, USA. 3Department of Pathology, Memorial Sloan Ket- tering Cancer Center, New York, NY 10065, USA. 4Genomics Core Laboratory, Memorial Sloan Kettering Cancer Center, New York, NY 10065, USA. 5Department of Pathology, Hospital de la Santa Creu i Sant Pau, 167 Sant Antoni M. Claret, Barcelona 08025, Spain. 6Department of Dermatology, Hospital de la Santa Creu i Sant Pau, Barcelona 08025, Spain. 7Department of Medicine, Memorial Sloan Kettering Cancer Center, New York, NY 10065, USA. *Corresponding author. E-mail: baselgaj@mskcc.org R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 30 March 2016 Vol 8 Issue 332 332ra42 1 on M arch 30, 2016 http://stm.sciencemag.org/ Downloaded from    148         (PI3Ka), in uterine cancer, which is char- acterized by the presence of these muta- tions in approximately half of the cases (21). To investigate the role of PIK3CA oncogenicity in this disease, we took ad- vantage of the previously reported trans- genic mouse strain LoxP-STOP-LoxP (LSL)–PIK3CAH1047R, which allows the expression of the activating PIK3CA mu- tation H1047R in a tissue-specific manner using the Cre-loxP technology upon re- moval of the floxed synthetic transcription- al/translational STOP cassette (22). These animals were crossed with the Sprr2f-Cre strain, shown to drive Cre recombinase expression in both luminal and glandular uterine epithelial cells (fig. S1A) (23). Un- expectedly, although PIK3CASprr2f-WT mice were viable and normal, PIK3CASprr2f-Cre littermates exhibited hindlimb paralysis at an early age (4 to 10 weeks) (Fig. 1A). Because this phenotype was observed in both males and females, we decided to further explore the pathologic events under- lying this phenotype. Histologic exami- nation revealed lesions in the spinal cord resembling human vascular malforma- tions that were not present in wild-type animals (Fig. 1B). Specifically, these ab- normalities showed dilated “cavernous” vascular spaces with extensive blood pools and hemorrhage involving both white and gray matter. We further examined the spinal lesions in PIK3CASprr2f-Cre mice injected intra- venously with gold nanoparticles, using the in vivo x-ray computed tomography imaging to confirm the presence of hy- perdense lesions in the spine. These vas- cular lesions were present in animals with both advanced and milder phenotypes but not in wild-type littermates (Fig. 1C) and showed slow blood flow and extravasa- tion, radiological features of vascular malformations of the spine (24). The ab- normal vascular channels, represented by cavernous spaces and capillary prolifera- tion, were consistent with a diagnosis of VM according to the current classifica- tion of the International Society for the Study of Vascular Anomalies (25). Among the observed alterations, cutaneous VM were the most frequent, exhibiting high penetrance in the PIK3CASprr2f-Cre mice (about 90%) (Fig. 1D). Microscopically, the skin lesions resembled hu- man VM with positivity for CD31 (Fig. 1E and fig. S1B) and Perl’s Prussian blue staining (Fig. 1F), indicative of EC lining and hemosid- erin deposition, respectively. To further characterize the observed lesions, we stained our mouse VM for both glucose transporter 1 (GLUT-1) andWilms tumor 1 (WT-1), which are immunophenotyping markers of infantile hemangioma (IH), a different vascular disease with a distinct natural history that responds Fig. 1. PIK3CASprr2fCRE mice develop spinal vascular malformations. (A) Hindlimb paresis phenotype observed in PIK3CASprr2fCRE mice. WT, wild type. (B) Gross and detailed histology of the spinal cord of PIK3CASprr2fCRE mice compared to a normal WT spine. Arrows indicate the multiple focal hemorrhages found in the spinal cord. (C) microCT (micro-computed tomography) scan of a WTmouse compared to PIK3CASprr2fCRE mice littermates showing an early and an advanced phenotype. Arrows indicate the slow flow and extrav- asation lesions observed in the spinal cord. (D) Hematoxylin and eosin (H&E) histology from normal skin and cutaneous VM. Dashed line delimits the dermis (lower) from the epidermis (upper). Arrows indicate normal blood vessels. (E) CD31 immunohistochemistry (IHC) of the skin VM lesions. Arrows indicate normal blood vessels. (F) Prussian blue staining. Dashed line delimits the dermis (lower) from the epidermis (upper). Arrow indicates normal blood vessel. R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 30 March 2016 Vol 8 Issue 332 332ra42 2 on M arch 30, 2016 http://stm.sciencemag.org/ Downloaded from   149         to the b-blocker propranolol (26–28). Both types of staining were negative in mouse VM samples as compared to positive controls from human IH specimens and mouse tissue (fig. S1, C and D). Lymphatic malforma- tions, which can histologically resemble VM, can also harbor PIK3R1 and PIK3CA mutations (29). In fact, mice with a knockout of PIK3R1, encoding the PI3K regulatory subunits p85a, p55a, and p50a, have defects in normal lymphangiogenesis and develop lymphatic malfor- mations in the intestines and skin (30). Thus, to assess whether our VM model might exhibit a substantial lymphatic component, we stained for the lymphatic-specific markers lymphatic vessel endothelial hyaluronan receptor 1 (LYVE-1) (31) and prospero homeobox 1 (PROX-1) (32). Our IHC staining with these lymphatic markers did not detect any relevant reactivity in the areas comprising the malformation, indicating that these lesions are entirely VM (fig. S1, E and F). We hypothesized that the Sprr2f-Cre strain drives the expression of the Cre recombinase in mature or precursor ECs, in addition to the reported endometrial epithelial cells. Thus, we crossed the LSL-LacZ re- porter strain with the Sprr2f-Cre mouse and performed b-galactosidase staining in spinal sections. In LacZSprr2f-Cre sections, we detected discrete positive cells resembling ECs that were sparsely distributed within the white and gray matter of the spinal cord (fig. S1G). Because of technical constraints, we were not able to obtain double staining for LacZ and CD31. However, double immunofluorescence staining against Cre and CD31 on VM from the Sprr2f-Cre mice confirmed the presence of Cre recombinase in the CD31-positive lesions that explain the vascular phe- notype observed in these mice (fig. S1H). PIK3CA activating mutation affects normal ECs Analogous to recent studies (11), we transduced primary human skin ECs with retrovirus encoding the PIK3CAwild type or H1047R variants to study the cellular mechanisms by which PIK3CA mutations might alter EC function. PIK3CAH1047R mutant cells exhibited amplified down- stream PI3K/AKT/mTOR signaling with increased phosphorylation of AKT at S473 and T308, and the mTOR downstream targets S6 kinase at T389 and ribosomal S6 protein at S235/6 and S240/4 (Fig. 2A).We undertook tube formation assays to assess the ability of these cells to form a normal capillary network in a three-dimensional matrix, an approach widely used to assess the normal function of EC (33). PIK3CAH1047R mutant cells formed aberrant EC clusters, as opposed to their wild-type counter- parts, which generated normal vascular tubes in vitro (Fig. 2B). We further confirmed these results using HUVECs infected with PIK3CA wild type and the H1047R mutant, which recapitulated the increased PI3K/AKT signaling (fig. S2A) and aberrant tube formation in vitro (fig. S2B). Because PI3K regulates cell proliferation (18), we tested the proliferation ratio of our primary cells in vitro using 5-ethynyl-2′- deoxyuridine (EdU) incorporation assays and found that the mutant cells exhibited a slightly higher, but reproducible, proliferation rate as compared with wild-type cells and empty vector controls, which was reversed upon treatment with a PI3Ka inhibitor (Fig. 2C) in a dose- dependent manner (fig. S2, C and D). ECs, which are key players in the development of vascular malfor- mations, create a pathological niche that involves the mural cell com- partment, probably in part as a result of aberrant cytokine secretion (12, 34). To test the impact of the PIK3CA activating mutation H1047R on the secretion of angiogenic factors, we performed antibody arrays in our primary ECs carrying wild-type PIK3CA or the H1047R mu- tation. We found that angiopoietin-2 (ANG2) protein expression was decreased in PIK3CAH1047R but not in the PIK3CAWT or control cells (fig. S2E). Because ANG2 is a cytokine that is regulated by forkhead O (FOXO) transcription factors downstream of the PI3K/AKT pathway, inhibits blood vessel leakage (35), and plays a role in the pathogenesis of lym- phatic malformations and VM (12, 20, 36), we sought to confirm whether our primary ECs displayed decreased expression of ANG2. Consistently, PIK3CAH1047R mutant ECs had lower expression of ANGPT2 mRNA and secreted ANG2 protein compared to PIK3CAWT cells and empty vector controls (fig. S2F). Next, we treated PIK3CAH1047R ECs with the different inhibitors of the PI3K/AKT/mTOR pathway, namely, BYL719 (PI3Ka), MK2206 (AKT), and everolimus (mTOR). We observed that both PI3Ka and AKT inhibitors were able to rescue the mRNA and protein expression of ANG2, but the mTOR inhibitor everolimus was not (fig. S2F). These results are in agreement with the previous evidence describing this secretory phenotype in VM, where PDGFB and ANG2 are down-regulated in TEK mutant ECs (12). Human VM harbor PIK3CA mutations Next, to ascertain whether the same genetic alterations triggering the phenotype in our mouse and cellular models were also present in the human condition, we examined clinical specimens from 32 patients, mainly adults (median age, 36 years), diagnosed with VM (table S1). Patients diagnosed with VM at our institution, a cancer center, mainly presented with deep-seated and infiltrative masses in the skeletal muscle (53% of the cases were intramuscular, 34% involved skin, and 13% were in other locations) (table S1). Histologically, these lesions displayed a mixed pattern of vascular proliferation, including thick-walled mal- formed vessels, cavernous spaces filled with erythrocytes, and capillary areas (Fig. 2D), and these were radiologically detected by routine magnetic resonance imaging (MRI) scans (Fig. 2E). We analyzed these VM by targeted exome sequencing of 341 cancer-related genes using the MSK-IMPACT (Memorial Sloan Kettering–Integrated Mutation Profiling of Actionable Cancer Targets) assay (37) developed at our in- stitution, yielding a median coverage of 588X. This assay was complemented with the next-generation deep sequencing targeted to the TEK locus (table S2). Deep sequencing detected PIK3CAmutations in 25% (8 of 32) of cases in previously described (38) hotspots encod- ing for the gain-of-function mutations H1047R (3 of 32) and E542K (3 of 32) with an allele frequency ranging from 3 to 15% (Fig. 2F and fig. S2G). We also identified two other mutations in PIK3CA, coding for C420R and I143V. In addition, we found gain-of-function mutations in other genes related to the PI3K pathway, such as AKT2, AKT3, and IRS2, resulting in an overall frequency of about 30% of mutations in the PI3K/AKT pathway (Fig. 2G and table S1). Because it was im- possible to asses copy number variations in our cohort as a result of low cellularity and considering that amplification of PIK3CA and loss of PTEN are two common alterations of the PI3K pathway in hu- man cancer, we performed fluorescence in situ hybridization anal- ysis in all the VM patient samples sequenced. We did not find any amplification of PIK3CA or deletion of PTEN in any of the samples analyzed (fig. S2H). Furthermore, we found isolated mutations in genes involved in the MAPK pathway (GNAQ, NF1,MAP2K1, andMAP3K1) in 13% of the cases (Fig. 2G). Previously described mutations in the tyrosine kinase receptor TEK (6, 7) were found in 35% of the patients of our cohort, with allele fre- quencies ranging from 4 to 15%. These mutations were mutually ex- clusive with the mutations in the PI3K pathway, with the exception of R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 30 March 2016 Vol 8 Issue 332 332ra42 3 on M arch 30, 2016 http://stm.sciencemag.org/ Downloaded from    150         one case (Fig. 2H). Considering that in ECs, the TIE2 receptor, en- coded by TEK, is immediately upstream from PI3K and signals via PI3K itself (9, 13, 39), VM may be defined as a disease state character- ized by the presence of somatic activating mutations in the TIE2- PI3K-AKT axis. We were not able to detect any mutation in 5 of the 32 samples ana- lyzed (table S1). This could be a result of low mutation allele frequency that is below the detection limit of our assay, the presence of a muta- tion that is not represented in our MSK-IMPACT assay, or a non- genetic etiology. Ubiquitous expression of PIK3CAH1047R spontaneously induces VM in mice We hypothesized that the cell of origin giving rise to VM might be particularly sensitive to oncogenic PIK3CA transformation. Thus, we generated PIK3CACAG-CreER mice, in which the PIK3CAH1047R allele is ubiquitously expressed upon tamoxifen administration (fig. S3A) (40). Six- to eight-week-old mice fed with tamoxifen rapidly developed cuta- neous VM with 100% penetrance compared to their PIK3CAWT litter- mates (Fig. 3A). Histologic assessment confirmed a combined capillary and cavernous phenotype exhibiting dilated blood channels filled with Fig. 2. PIK3CAmutations cause sporadic VM inhumans. (A) Western blot of human skin ECs infected with empty vector (EV), PIK3CA WT, and H1047R mutation probed with the indicated antibodies. Cells were serum-starved overnight before lysis. S6K, S6 kinase; pS6K, phosphorylated S6K; pS6, phosphorylated S6; pAKT, phospshorylated AKT; HFSEC, human female skin endothe- lial cell. (B) Representative images from the tube formation assays of primary ECs in- fected with EV, PIK3CAWT, and H1047R mu- tation and serum-starved overnight before seeding. Pictures were taken 6 hours after seeding. Note the reticular network formed in the EV and PIK3CA WT cells that fails to form in the PIK3CA H1047R mutant cells. Scale bars, 200 mm. (C) EdU incorporation assay quantification of ECs infected with EV, PIK3CA WT, and H1047R mutation, serum- starved overnight, and labeled with EdU for 4 hours. Graph shows mean fold change ± SEM. n = 3 biological replicates. P values were calculated using Student’s t test. (D) H&E staining highlighting the representative morphology of one of the human VM pa- tients sequenced in this study. Blood pools and thick mural cell layer are evident in the histological sections of these patients. (E) Characteristic MRI scan from an intra- muscular sporadic VM patient sequenced in this study. T1 axial and coronal sections are shown. Dashed line delimits the radio- logical extension of the malformation. [R] and [L] indicate right and left, respectively. (F) PI3Ka domains and specific sites found to be mutated in this study. The p85-binding domain is represented in green, the Ras- binding domain in red, the C2 domain in blue, the helical domain in yellow, and the kinase domain in purple. aa, amino acid. (G) Schematic pathway depicting TIE2, PI3K, and MAPK pathway gene components found to be mutated in sporadic VM by MSK-IMPACT in this study. Activating muta- tions are indicated in red and inactivating mutations in blue. Dark red designates mu- tations found in more than 20% of the pa- tients. Unknownmutations are shown in gray. (H) Mutual exclusivity of the gene mutations present in the TEK and PIK3CA pathways. The activating mutations in TEK are indicated in blue, and the activating mutations in PIK3CA are in red. The alterations affecting genes involved in the PI3K or MAPK pathway are represented in green. R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 30 March 2016 Vol 8 Issue 332 332ra42 4 on M arch 30, 2016 http://stm.sciencemag.org/ Downloaded from   151         erythrocytes (Fig. 3B) and immunoreactivity for CD31 (Fig. 3C) and phosphorylated AKT (S473), a surrogate marker of PI3K activation (Fig. 3D). Similar to human VM, murine vascular lesions from the PIK3CACAG-CreER mice were negative for GLUT-1, WT-1, LYVE-1, and PROX-1 (fig. S3, B to E) and contained high amounts of hemosiderin deposition (Fig. 3E). Although the skin phenotype was readily evident, additional lesions were observed at necropsy at multiple sites, including mesentery, genitourinary tract, kidney, and retina (Fig. 3F), with no apparent difference in the incidence by anatomic site. Histological analy- ses of these lesions revealed large spaces filled with blood and lined by flattened ECs, with a similar immunophenotype, positive for CD31 and Prussian blue staining (Fig. 3, G and H). We confirmed our findings using the UBC-CreER strain, another transgenic strain in which the ubiquitin C promoter drives the expression of a tamoxifen-inducible Cre recombinase in all the cells of the organism Fig. 3. Ubiquitous expression of PIK3CAH1047R induces VM. (A) Disease- free survival plot of PIK3CACAG-CreER (n= 20) and PIK3CAWT (n= 25) littermates on tamoxifen diet assessed by the appearance of visible cutaneous VM. Dotted line represents the time when tamoxifen diet was administered (day 21). P value was calculated using log-rank test. (B) H&E staining of a representative normal blood vessel and a VM lesion developed in the skin of PIK3CACAG-CreER mice. Arrow indicates normal blood vessel. (C) CD31 IHC staining showingpositivity for the ECs of a normal blood vessel andVM.Note that erythrocytes exhibit nonspecific staining. Arrow indicates normal blood vessel. (D) Phosphorylated AKT (S473) IHC. The arrow in the bottom panel indicates lining ECs that show positivity for the staining. Note the negativity for phosphorylatedAKT in thenormal bloodvessel (top, indicatedbyanarrow). (E) Prussian blue staining of normal blood vessels and the PIK3CACAG-CreER mouse skin VM lesions. Arrow indicates a normal blood vessel. (F) Histological representation of mesenteric vasculature and VM harvested during necropsy and detailed view to highlight the blood pools observed in the preparations. Arrows indicate normal blood vessels. (G andH) CD31 (G) andPrussianblue (H) positivity for theVMdescribed in (F). (I) BrdU incorporation (red) inPIK3CACAG-CreER VMcompared to normal blood vessels. CD31 (green) andDAPI (4′,6-diamidino- 2-phenylindole) (blue) show ECs and nuclei, respectively. Note the encased BrdU-positive nuclei in the CD31-positive lining EC layer. (J) Quantification of BrdU-positive nuclei in normal blood vessels and VM. P value was calculated usingStudent’s t test. Graph showsmeans±SD.n=45 fields from fivebiological replicates. (K) Morphological quantification of the maximal blood vessel diam- eter of normal vessels and VM. P value was calculated using Student’s t test. Graph shows means ± SD. n = 45 fields from five biological replicates. R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 30 March 2016 Vol 8 Issue 332 332ra42 5 on M arch 30, 2016 http://stm.sciencemag.org/ Downloaded from    152         (fig. S4) (41). Consistently, our results indicated that upon ubiquitous expres- sion of the oncogenic PIK3CA trans- gene, the cell of origin for VM might be more sensitive to transformation than other cell types, resulting in the genesis of VM. To test whether the formation of VM in our mouse model results, at least in part, from increased proliferation caused by hyperactive PI3K signaling (42, 43), we measured 5-bromo-2′-deoxyuridine (BrdU) incorporation and Ki-67 staining in both PIK3CAWT and PIK3CACAG-CreER littermates. Whereas normal blood ves- sels were negative for BrdU incorpora- tion as a consequence of EC quiescence (44), VM displayed a marked increase in proliferative cells (Fig. 3, I and J). In agreement with the BrdU data, Ki-67 positivity was found in our VM (fig. S5, A and B). We cannot rule out wheth- er the proliferative enhancement ob- served is caused by the direct effect of the PIK3CA mutant allele or a result of autocrine and/or paracrine signaling from proliferating ECs, which plays an important role in the complex develop- ment of human VM (12, 45). At the mor- phological level, quantification of the lumen diameter from normal blood ves- sels and VM revealed a 10-fold increase in the size of these structures in VM samples (Fig. 3K). PI3K inhibitors are effective in the treatment of PIK3CA- induced VM The presence of oncogenic PIK3CAmutations in human specimens of VM, together with the observed phenotypes in mice, prompted us to evaluate the full growth potential of these lesions, despite the fact that they are considered to be vascular malformations. To this end, we in- jected PIK3CACAG-CreER VM cells into recipient immunocompromised nude mice. These cells formed highly vascularized and proliferative masses a few weeks after injection, with a histology and appearance highly resembling that of the original lesions (Fig. 4A). Although VM do not have metastatic potential in patients, our finding that they may be successfully transplanted and grown in animals suggests that these lesions display some tumorigenic behaviors and highlight the fine line between malignant and benign tumors in some cases. Allo- transplanted VM formed new cystic structures that contained blood and exhibited intravascular coagulopathy, as measured by the increased concentration of D-dimers in plasma (Fig. 4B), a useful tool for the dif- ferential diagnosis of VM in human patients (46). Of clinical relevance, the presence of activating PIK3CA mutations in VM opens the door for the treatment of this condition with PI3Ka inhibitors, currently under clinical development for several cancer indica- tions (47). Treatment of VM with the PI3Ka selective inhibitor BYL719 resulted in a marked response as measured by a decrease in VM volume (Fig. 4C), reduced proliferation, and increased apoptosis (Fig. 4D and fig. S5C). On the contrary, treatment with the b-adrenergic antagonist pro- pranolol, an active agent against IH (28), did not yield any effect (Fig. 4, E and F, and fig. S5D). In support of the role of the aberrant activa- tion of the PI3K/mTOR pathway in VM, treatment with the mTOR inhibitor everolimus (48, 49) partially decreased VM size and prolifera- tion in a similar fashion as PI3K inhibition, although it did not increase apoptosis (Fig. 4, E and F, and fig. S5D). As mentioned above, VMmay be defined as a disease characterized by the presence of somatic activating mutations in the TIE2-PI3K- AKT axis (Fig. 2, G and H). Because the TIE2 receptor, encoded by TEK, is immediately upstream from PI3K and signals via PI3K itself, we postulated that treatment with PI3K inhibitors might also be effica- cious in VM harboring TEKmutations. Indeed, treatment of HUVECs stably expressing the TEK mutation L914F with the PI3Ka inhibitor BYL719 decreased the amount of phosphorylated AKT (T308 and S473) (fig. S6A), suggesting that PI3K inhibitors may be efficacious for VM with either PIK3CA or TEK mutations. Given that a large number of VM are detected in the skin or su- perficial tissues, together with the substantial toxicity of systemic ad- ministration of PI3K inhibitors in patients (50), we wondered whether the topical application of PI3K inhibitors might be of therapeutic in- terest in this context. To this end, we formulated two different cream Fig. 4. PI3K inhibitors are effective for the treatment of VM. (A) Schematic representation and images from the allotransplantation assays. (B) Quantification of plasma D-dimers measured in animals with or without VM. P value was calculated using Student’s t test. (C) VM volume measured in PIK3CACAG-CreER VM-derived allografts treated with vehicle or PI3Ka inhibitor for 1 week [BYL719, 50 mg kg−1, daily, per os (po)]. P value was calculated using Student’s t test. (D) Quantification of BrdU incorporation and cleaved caspase-3 (clC3) in CD31-positive cells from (B). P value was calculated using Student’s t test. n = 10. (E) VM volume measured in PIK3CACAG-CreER VM-derived allografts treated with vehicle, everolimus (10 mg kg−1, daily, po), or propranolol (40 mg kg−1, daily, po) for 1 week. P value was calculated using Student’s t test. (F) Quantification of BrdU incorporation and cleaved caspase-3 in CD31-positive cells from (E). P value was calculated using Student’s t test. n = 8. (G) VM volume in PIK3CACAG-CreER VM-derived allografts treated topically with BYL719 at 1% (w/w) using two different formulations (free and soluble BYL719) for 3 weeks. The pretreatment time point indicates when the treatment was started. All treatments in weeks 1, 2, and 3 have a P < 0.001 as compared to the vehicle control VM. P values were calculated using Student’s t test. R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 30 March 2016 Vol 8 Issue 332 332ra42 6 on M arch 30, 2016 http://stm.sciencemag.org/ Downloaded from   153         preparations containing the PI3Ka inhibitor BYL719 at 1% (w/w): one preparation with the inhibitor dispersed directly into the cream base and another one with the inhibitor presolubilized in dimethyl sulfoxide. Topical administration of the PI3Ka inhibitor using these two different cream formulas achieved a rapid and sustained regres- sion of skin lesions (Fig. 4G and fig. S6B). Together, our results indi- cate that VM have tumorigenic growth potential as evidenced by their ability to engraft in nude mice and that the treatment with PI3K in- hibitors either systemically or locally is a suitable pharmacological ap- proach to control the disease. Expression of mutant PIK3CA impairs normal vasculogenesis VM may occur as a result of defects during angiogenesis, a process in which PI3Ka is actively involved (17, 18, 51). To explore the biological relevance of PI3K hyperactivation specifically in blood vessels, we crossed PIK3CAH1047R mice with the Tie2-Cre strain (52), which drives the expression of the transgene in ECs (fig. S7A). PIK3CATie2-Cre mice were not viable because of early embryonic lethality [embryonic day 10 (E10)] resulting from vascular defects (Fig. 5A). CD31 staining of coronal sections revealed dilated blood vessels and vascular anomalies (Fig. 5B, upper panel) present in meningeal vessels, cardinal vein, and dorsal aorta. Moreover, small intersomitic vessels failed to form (Fig. 5B, lower panel), suggesting that deregulated PI3K activity results in lethal impairment of small vessel formation (17). These malformations were also evident in whole-mount embryo CD31 staining, with aberrant for- mation of the cephalic and intersomitic vessels (Fig. 5, C and D, upper panels). Morphology, proliferation, and apoptosis were not altered in PIK3CATie2-Cre embryos’ hearts (fig. S7, B to D), suggesting that the observed phenotype is caused by a defect specifically affecting blood vessel formation. Aiming to validate the implication of excessive PI3K signaling in aberrant vasculogenesis, as well as to evaluate whether phar- macological inhibition could overcome this effect, we attempted to re- vert the phenotype by treating pregnant mice with the PI3Ka inhibitor BYL719. PIK3CATie2-Cre E9.5 embryos treated with the PI3Ka inhibitor showed an overall body size comparable to PIK3CAWT littermates, sug- gesting improved vascular function (fig. S7E). CD31 whole-mount stain- ing revealed restored cephalic and intersomitic small blood vessel formation (Fig. 5, C and D, lower panels). Phosphorylated AKT staining showed a strong reduction in both PIK3CATie2-Cre and PIK3CAWT embryos after PI3Ka inhibitor treatment, in contrast with untreated control embryos (fig. S7F). At the histologic level, treatment reestab- lished meningeal, cardinal vein, and dorsal aorta blood vessel morphol- ogy (Fig. 5E), indicating that aberrant PI3K pathway hyperactivation impairs normal embryonic angiogenesis in mice. DISCUSSION VM are the most common vascular malformation in humans (5) and are a cause of pain, functional limitations of the affected areas, aesthet- ic disfigurements, and coagulopathies. In severe cases, sclerotherapy or surgical resection may be considered; however, these procedures often involve complications such as cutaneous necrosis or extended inflam- matory reactions (53), and depending on the anatomic location and ex- tension may have limited applicability. Moreover, VM are prone to recur and recanalize (54), raising the need for developing more effec- tive therapies. GEMMs represent reliable tools for investigating the etiology, bio- logy, and progression of human diseases, as well as for exploring new therapeutic approaches (55, 56). The first somatic molecular altera- tions linked to the development of sporadic VM were the acquisition of gain-of-function mutations in the gene encoding the EC-specific tyrosine kinase receptor TIE2 (TEK) (6, 8, 57, 58). Ligand-independent receptor activation drives constitutive activation of the PI3K/AKT and MAPK pathways, resulting in increased proliferation and survival of EC that could account for increased EC accumulation in VM and ab- normal recruitment of smooth muscle cells. However, only a subgroup of VM harbor defects in TEK, suggesting that other genomic or mo- lecular alterations may be at play in this disease. In this line, it would Fig. 5. PIK3CAH1047R impairs embryonic angiogenesis. (A) Embryonic phenotype observed in PIK3CAWT (left) and PIK3CATie2-Cre (right) littermates at E10. For morphological studies, a minimum of four dissections for each geno- type were performed yielding n ≥ 15 embryos. (B) CD31 staining of coronal sections from PIK3CAWT and PIK3CATie2-Cre embryos at E9.5. Arrowheads indi- cate blood vessel enlargement defects in the meningeal vessels (upper panel), and arrows indicate the defects in the intersomitic vessels (lower panel). For CD31 histologic studies, a minimum of four embryos for each phenotype obtained from two different dissections were used. Scale bars, 100 mm. (C) Whole-embryo CD31 staining of PIK3CAWT and PIK3CATie2-Cre em- bryos from mice treated with vehicle or PI3Ka inhibitor (BYL719, 50 mg kg−1, daily, po; 48, 24, and 2 hours before embryos are harvested). For CD31 his- tologic studies, a minimum of four embryos for each condition were used. (D) Detailed view of the cephalic and intersomitic blood vessels from (C). Arrow indicates defects in the meningeal (upper panel) and the intersomitic vessels (lower panel). (E) CD31 coronal sections from embryos in (D). CV, cardinal vein; DA, dorsal aorta. Scale bars, 100 mm. R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 30 March 2016 Vol 8 Issue 332 332ra42 7 on M arch 30, 2016 http://stm.sciencemag.org/ Downloaded from    154         be interesting to develop a GEMM expressing activating TEK muta- tions in the endothelial compartment to characterize in detail their role in the histopathology and the mechanism of pathogenesis of this disease. Recent studies performing xenograft experiments with HUVECs transduced with the most frequent TEK mutation, L914F, have dem- onstrated its functional relevance in inducing VM (11). Treatment of murine xenografts with rapamycin proved the efficacy of inhibiting mTOR activity, which also showed clinical activity in VM patients in a pilot trial. Intriguingly, three of five patients that responded to mTOR inhibition in this study did not harbor any genetic defect in TEK (11). It is thus tempting to speculate that additional molecular al- terations enhancing the activity of the PI3K/AKT/mTOR pathway could be driving the formation of VM in these patients. We report the generation of a GEMM for VM by inducing the expression of the gain-of-function PIK3CAH1047R mutant allele in mice. The histopathologic resemblance of the lesions arising in mice to those affecting humans prompted us to evaluate the existence of similar al- terations in clinical specimens. Through targeted exome sequencing, we found that 25% of the evaluated samples bear activating mutations in PIK3CA or additional genetic defects predicted to stimulate con- stitutive downstream signaling. To reconcile our findings with those previously reported, we confirmed that 35% of the patients harbored mutations in TEK, yet these were mutually exclusive with the presence of activating PI3K mutations, consistent with a functional redundancy. These results are in agreement with the high prevalence of TEKmuta- tions reported by others, mainly in pediatric patients (6, 7). TIE2 activated as a result of TEKmutations mainly signals through PI3K (9, 39, 59), consistent with the unifying hypothesis that aberrant activation of PI3K signaling by mutations in the upstream receptor or in the pathway itself causes the development of VM. Somatic mutation of PIK3CA is frequently detected in several cancer types, and genetic alterations driving hyperactivation of the PI3K/AKT pathway have also been reported in nonhereditary postzygotic tissue overgrowth syn- dromes that often exhibit mixed capillary, lymphatic, and venous anomalies. Because of the clinical overlap of these overgrowth syn- dromes with PIK3CAmutations, the term PIK3CA-related overgrowth syndrome has been proposed (19). Patients suffering from congenital lipomatous overgrowth, vascular malformations, epidermal nevi, and skeletal/spinal abnormalities (CLOVES) syndrome harbor somatic mo- saicism for activating PIK3CA mutations resulting in hyperactive PI3K/ AKT signaling (60). The presence of somatic mutations in PIK3CA was also detected in patients affected by Klippel-Trenaunay-Weber syndrome, an overgrowth condition with features overlapping those of CLOVES syndrome: isolated lymphatic malformations, fibroadipose hyperpla- sia, and fibroadipose vascular anomalies (61, 62). Additional genetic alterations in PTEN, GNAQ, AKT isoforms, or the regulatory subunit of PI3K PIK3R1, which enhance PI3K/AKT/mTOR andMAPK pathway activation, have also been reported in other malformation syndromes including Proteus (63), megalencephaly-capillary malformation (MCAP) (64), Sturge-Weber (65), and Bannayan-Riley-Ruvalcaba (66) syn- dromes, underscoring the involvement of aberrant PI3K/AKT/mTOR signaling in developmental disorders. The MCAP syndrome exhibits a predominant brain overgrowth phenotype in which PIK3CA mutations are also involved. A recent report has described the first mouse model for brain overgrowth using GEMM of PIK3CA mutants E545K and H1047R, validating the importance of these mouse models in the study of PIK3CA-driven syndromes (67). Nevertheless, sporadic and solitary VM lesions are a different and much more prevalent entity that is not necessarily associated with overgrowth. Despite the fact that most lymphatic malformations carry PIK3CA mutations, our mouse model does not present detectable lymphatic anomalies. This is a limitation of our model that could be explained by a number of factors including the moment and time of recombina- tion, or possibly different sets of precursor cells giving rise to each mal- formation. Our observations are further supported by those made by Castillo and colleagues, who found that mosaic somatic mutations in- duced in a PIK3CAH1047R mouse model cause VM that are associated with neither tissue overgrowth nor lymphatic malformations (45), sug- gesting that the cell of origin giving rise to VM may be more suscep- tible to hyperactive PI3K signaling than other cell lineages and that additional genetic or environmental cues are required to reproduce the complex phenotypes observed in overgrowth syndromes. In the fu- ture, it would also be interesting to identify the cell of origin of VM by means of lineage-tracing experiments. Given the ability of our mouse model to recapitulate the patho- genesis of human VM, we determined whether it could be used as a platform for testing pharmacological inhibition using PI3K inhibitors currently under clinical development. To put it into context, we also evaluated the efficacy of other agents that have been proposed to in- hibit the growth of VM, including rapamycin analogs and propranolol (11, 68). The greatest growth inhibition was achieved when treating allograft transplants with a PI3Ka inhibitor or the rapamycin analog everolimus, but no effect was observed with propranolol. Rapamycin im- proves quality of life in patients with VM (11), lymphatic malformations (69), and other vascular syndromes (48), probably because long-term treatment with rapamycin has the ability to inhibit AKT in ECs; this observation is also seen in adipocytes, but not in all epithelial cells (70). In contrast to the antiproliferative effect of rapamycin analogs, we pro- pose that the proapoptotic effect achieved upon PI3K inhibition is likely to yield improved therapeutic efficacy by diminishing the recurrence of VM. Topical administration of PI3Ka inhibitor further demonstrated the ef- ficacy of this treatment with an approach that would be devoid of the substantial side effects associated with systemic drug administration (hyperglycemia, nausea, gastrointestinal effects, and fatigue) (50). The impaired vasculogenesis observed in mouse embryos as a consequence of endothelial-restricted expression of the PIK3CAH1047R allele was also rescued when pregnant mice were treated with the PI3Ka inhibitor, further supporting a functional requirement for controlled PI3K signaling in normal embryonic vasculogenesis as has been demonstrated by others (51). In summary, our study provides a GEMM recapitulating human VM caused by hyperactivation of the PI3K/AKT pathway, reveals the impact of PIK3CA somatic mutations in the pathogenesis of VM, and provides a potential therapeutic approach to treat advanced or recurrent lesions in these patients. MATERIALS AND METHODS Materials and methods can be found in the Supplementary Materials. SUPPLEMENTARY MATERIALS www.sciencetranslationalmedicine.org/cgi/content/full/8/332/332ra42/DC1 Materials and Methods R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 30 March 2016 Vol 8 Issue 332 332ra42 8 on M arch 30, 2016 http://stm.sciencemag.org/ Downloaded from   155         Fig. S1. Histologic characterization of PIK3CASprr2f-Cre mice. Fig. S2. PIK3CA mutation in ECs. Fig. S3. Histologic characterization of PIK3CACAG-CreER mice. Fig. S4. Histologic characterization of PIK3CAUBC-CreER mice. Fig. S5. Cell proliferation in mouse VM with or without in vivo treatments. Fig. S6. Treatment of VM with PI3K inhibitors. Fig. S7. Histological assessment of PIK3CATie2-Cre embryos. Table S1. Clinical features and genomic findings in VM patients. Table S2. Bait sequences used for TEK targeted sequencing. REFERENCES AND NOTES 1. P. Brouillard, M. Vikkula, Genetic causes of vascular malformations. Hum. Mol. Gen. 16, R140–R149 (2007). 2. P. Brouillard, M. Vikkula, Vascular malformations: Localized defects in vascular morphogenesis. Clin. Genet. 63, 340–351 (2003). 3. A. Dompmartin, M. Vikkula, L. M. Boon, Venous malformation: Update on aetiopathogenesis, diagnosis and management. Phlebology 25, 224–235 (2010). 4. H.-L. Nguyen, L. M. Boon, M. Vikkula, Genetics of vascular malformations. Semin. Pediatr. Surg. 23, 221–226 (2014). 5. M. Uebelhoer, L. M. Boon, M. Vikkula, Vascular anomalies: From genetics toward models for therapeutic trials. Cold Spring Harb. Perspect. Med. 2, a009688 (2012). 6. N. Limaye, V. Wouters, M. Uebelhoer, M. Tuominen, R. Wirkkala, J. B. Mulliken, L. Eklund, L. M. Boon, M. Vikkula, Somatic mutations in angiopoietin receptor gene TEK cause solitary and multiple sporadic venous malformations. Nat. Genet. 41, 118–124 (2009). 7. J. Soblet, N. Limaye, M. Uebelhoer, L. M. Boon, M. Vikkula, Variable somatic TIE2 mutations in half of sporadic venous malformations. Mol. Syndromol. 4, 179–183 (2013). 8. M. Vikkula, L. M. Boon, K. L. Carraway III, J. T. Calvert, A. J. Diamonti, B. Goumnerov, K. A. Pasyk, D. A. Marchuk, M. L. Warman, L. C. Cantley, J. B. Mulliken, B. R. Olsen, Vascular dysmorphogenesis caused by an activating mutation in the receptor tyrosine kinase TIE2. Cell 87, 1181–1190 (1996). 9. C. D. Kontos, T. P. Stauffer, W.-P. Yang, J. D. York, L. Huang, M. A. Blanar, T. Meyer, K. G. Peters, Tyrosine 1101 of Tie2 is the major site of association of p85 and is required for activation of phosphatidylinositol 3-kinase and Akt. Mol. Cell. Biol. 18, 4131–4140 (1998). 10. P. N. Morris, B. J. Dunmore, A. Tadros, D. A. Marchuk, D. C. Darland, P. A. D’Amore, N. P. J. Brindle, Functional analysis of a mutant form of the receptor tyrosine kinase Tie2 causing venous mal- formations. J. Mol. Med. 83, 58–63 (2005). 11. E. Boscolo, N. Limaye, L. Huang, K.-T. Kang, J. Soblet, M. Uebelhoer, A. Mendola, M. Natynki, E. Seront, S. Dupont, J. Hammer, C. Legrand, C. Brugnara, L. Eklund, M. Vikkula, J. Bischoff, L. M. Boon, Rapamycin improves TIE2-mutated venous malformation in murine model and human subjects. J. Clin. Invest. 125, 3491–3504 (2015). 12. M. Uebelhoer, M. Nätynki, J. Kangas, A. Mendola, H.-L. Nguyen, J. Soblet, C. Godfraind, L. M. Boon, L. Eklund, N. Limaye, M. Vikkula, Venous malformation-causative TIE2 mutations mediate an AKT-dependent decrease in PDGFB. Hum. Mol. Genet. 22, 3438–3448 (2013). 13. M. Nätynki, J. Kangas, I. Miinalainen, R. Sormunen, R. Pietilä, J. Soblet, L. M. Boon, M. Vikkula, N. Limaye, L. Eklund, Common and specific effects of TIE2 mutations causing venous mal- formations. Hum. Mol. Genet. 24, 6374–6389 (2015). 14. D. A. Fruman, C. Rommel, PI3K and cancer: Lessons, challenges and opportunities. Nat. Rev. Drug Discov. 13, 140–156 (2014). 15. J. A. Engelman, Targeting PI3K signalling in cancer: Opportunities, challenges and limita- tions. Nat. Rev. Cancer 9, 550–562 (2009). 16. K. M. Keppler-Noreuil, J. J. Rios, V. E. R. Parker, R. K. Semple, M. J. Lindhurst, J. C. Sapp, A. Alomari, M. Ezaki, W. Dobyns, L. G. Biesecker, PIK3CA-related overgrowth spectrum (PROS): Diagnostic and testing eligibility criteria, differential diagnosis, and evaluation. Am. J. Med. Genet. A 167A, 287–295 (2015). 17. M. Graupera, J. Guillermet-Guibert, L. C. Foukas, L.-K. Phng, R. J. Cain, A. Salpekar, W. Pearce, S. Meek, J. Millan, P. R. Cutillas, A. J. H. Smith, A. J. Ridley, C. Ruhrberg, H. Gerhardt, B. Vanhaesebroeck, Angiogenesis selectively requires the p110a isoform of PI3K to control endothelial cell migra- tion. Nature 453, 662–666 (2008). 18. M. Graupera, M. Potente, Regulation of angiogenesis by PI3K signaling networks. Exp. Cell Res. 319, 1348–1355 (2013). 19. H.-C. Kang, S. T. Baek, S. Song, J. G. Gleeson, Clinical and genetic aspects of the segmental overgrowth spectrum due to somatic mutations in PIK3CA. J. Pediatr. 167, 957–962 (2015). 20. A. J. Osborn, P. Dickie, D. E. Neilson, K. Glaser, K. A. Lynch, A. Gupta, B. H. Dickie, Activating PIK3CA alleles and lymphangiogenic phenotype of lymphatic endothelial cells isolated from lymphatic malformations. Hum. Mol. Genet. 24, 926–938 (2015). 21. Cancer Genome Atlas Research Network, C. Kandoth, N. Schultz, A. D. Cherniack, R. Akbani, Y. Liu, H. Shen, A. G. Robertson, I. Pashtan, R. Shen, C. C. Benz, C. Yau, P. W. Laird, L. Ding, W. Zhang, G. B. Mills, R. Kucherlapati, E. R. Mardis, D. A. Levine, Integrated genomic char- acterization of endometrial carcinoma. Nature 497, 67–73 (2013). 22. J. R. Adams, K. Xu, J. C. Liu, N. M. R. Agamez, A. J. Loch, R. G. Wong, W. Wang, K. L. Wright, T. F. Lane, E. Zacksenhaus, S. E. Egan, Cooperation between PIK3CA and p53 mutations in mouse mammary tumor formation. Cancer Res. 71, 2706–2717 (2011). 23. C. M. Contreras, E. A. Akbay, T. D. Gallardo, J. M. Haynie, S. Sharma, O. Tagao, N. Bardeesy, M. Takahashi, J. Settleman, K. K. Wong, D. H. Castrillon, Lkb1 inactivation is sufficient to drive endometrial cancers that are aggressive yet highly responsive to mTOR inhibitor monotherapy. Dis. Model. Mech. 3, 181–193 (2010). 24. T. Krings, Vascular malformations of the spine and spinal cord*: Anatomy, classification, treatment. Clin. Neuroradiol. 20, 5–24 (2010). 25. M. Wassef, F. Blei, D. Adams, A. Alomari, E. Baselga, A. Berenstein, P. Burrows, I. J. Frieden, M. C. Garzon, J. C. Lopez-Gutierrez, D. J. Lord, S. Mitchel, J. Powell, J. Prendiville, M. Vikkula; ISSVA Board and Scientific Committee, Vascular anomalies classification: Recommendations from the International Society for the Study of Vascular Anomalies. Pediatrics 136, e203–e214 (2015). 26. P. E. North, M. Waner, A. Mizeracki, M. C. Mihm Jr., GLUT1: A newly discovered immuno- histochemical marker for juvenile hemangiomas. Hum. Pathol. 31, 11–22 (2000). 27. L. P. Lawley, F. Cerimele, S. W. Weiss, P. North, C. Cohen, H. P. W. Kozakewich, J. B. Mulliken, J. L. Arbiser, Expression of Wilms tumor 1 gene distinguishes vascular malformations from proliferative endothelial lesions. Arch. Dermatol. 141, 1297–1300 (2005). 28. C. Léauté-Labrèze, P. Hoeger, J. Mazereeuw-Hautier, L. Guibaud, E. Baselga, G. Posiunas, R. J. Phillips, H. Caceres, J. C. Lopez Gutierrez, R. Ballona, S. F. Friedlander, J. Powell, D. Perek, B. Metz, S. Barbarot, A. Maruani, Z. Z. Szalai, A. Krol, O. Boccara, R. Foelster-Holst, M. I. Febrer Bosch, J. Su, H. Buckova, A. Torrelo, F. Cambazard, R. Grantzow, O. Wargon, D. Wyrzykowski, J. Roessler, J. Bernabeu-Wittel, A. M. Valencia, P. Przewratil, S. Glick, E. Pope, N. Birchall, L. Benjamin, A. J. Mancini, P. Vabres, P. Souteyrand, I. J. Frieden, C. I. Berul, C. R. Mehta, S. Prey, F. Boralevi, C. C. Morgan, S. Heritier, A. Delarue, J.-J. Voisard, A randomized, controlled trial of oral propranolol in infantile hemangioma. N. Engl. J. Med. 372, 735–746 (2015). 29. P. Brouillard, L. Boon, M. Vikkula, Genetics of lymphatic anomalies. J. Clin. Invest. 124, 898–904 (2014). 30. C. Mouta-Bellum, A. Kirov, L. Miceli-Libby, M. L. Mancini, T. V. Petrova, L. Liaw, I. Prudovsky, P. E. Thorpe, N. Miura, L. C. Cantley, K. Alitalo, D. A. Fruman, C. P. H. Vary, Organ-specific lymphangiectasia, arrested lymphatic sprouting, and maturation defects resulting from gene-targeting of the PI3K regulatory isoforms p85a, p55a, and p50a. Dev. Dyn. 238, 2670–2679 (2009). 31. S. Banerji, J. Ni, S. X. Wang, S. Clasper, J. Su, R. Tammi, M. Jones, D. G. Jackson, LYVE-1, a new homologue of the CD44 glycoprotein, is a lymph-specific receptor for hyaluronan. J. Cell Biol. 144, 789–801 (1999). 32. E. C. Castro, C. Galambos, Prox-1 and VEGFR3 antibodies are superior to D2–40 in identifying endothelial cells of lymphatic malformations—A proposal of a new immunohistochemical panel to differentiate lymphatic from other vascular malformations. Pediatr. Dev. Pathol. 12, 187–194 (2009). 33. I. Arnaoutova, J. George, H. K. Kleinman, G. Benton, The endothelial cell tube formation assay on basement membrane turns 20: State of the science and the art. Angiogenesis 12, 267–274 (2009). 34. X. L. Tian, R. Kadaba, S. A. You, M. Liu, A. A. Timur, L. Yang, Q. Chen, P. Szafranski, S. Rao, L. Wu, D. E. Housman, P. E. DiCorleto, D. J. Driscoll, J. Borrow, Q. Wang, Identification of an angiogenic factor that when mutated causes susceptibility to Klippel–Trenaunay syndrome. Nature 427, 640–645 (2004). 35. C. Daly, E. Pasnikowski, E. Burova, V. Wong, T. H. Aldrich, J. Griffiths, E. Ioffe, T. J. Daly, J. P. Fandl, N. Papadopoulos, D. M. McDonald, G. Thurston, G. D. Yancopoulos, J. S. Rudge, Angiopoietin-2 functions as an autocrine protective factor in stressed endothelial cells. Proc. Natl. Acad. Sci. U.S.A. 103, 15491–15496 (2006). 36. G. Thurston, C. Daly, The complex role of angiopoietin-2 in the angiopoietin–tie signaling pathway. Cold Spring Harb. Perspect. Med. 2, a006550 (2012). 37. D. T. Cheng, T. N. Mitchell, A. Zehir, R. H. Shah, R. Benayed, A. Syed, R. Chandramohan, Z. Y. Liu, H. H. Won, S. N. Scott, A. R. Brannon, C. O’Reilly, J. Sadowska, J. Casanova, A. Yannes, J. F. Hechtman, J. Yao, W. Song, D. S. Ross, A. Oultache, S. Dogan, L. Borsu, M. Hameed, K. Nafa, M. E. Arcila, M. Ladanyi, M. F. Berger, Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT): A hybridization capture-based next-generation sequencing clinical assay for solid tumor molecular oncology. J. Mol. Diagn. 17, 251–264 (2015). 38. Y. Samuels, L. A. Diaz Jr., O. Schmidt-Kittler, J. M. Cummins, L. Delong, I. Cheong, C. Rago, D. L. Huso, C. Lengauer, K. W. Kinzler, B. Vogelstein, V. E. Velculescu, Mutant PIK3CA promotes cell growth and invasion of human cancer cells. Cancer Cell 7, 561–573 (2005). 39. N. Jones, Z. Master, J. Jones, D. Bouchard, Y. Gunji, H. Sasaki, R. Daly, K. Alitalo, D. J. Dumont, Identification of Tek/Tie2 binding partners. Binding to a multifunctional docking site mediates cell survival and migration. J. Biol. Chem. 274, 30896–30905 (1999). 40. S. Hayashi, A. P. McMahon, Efficient recombination in diverse tissues by a tamoxifen- inducible form of Cre: A tool for temporally regulated gene activation/inactivation in the mouse. Dev. Biol. 244, 305–318 (2002). R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 30 March 2016 Vol 8 Issue 332 332ra42 9 on M arch 30, 2016 http://stm.sciencemag.org/ Downloaded from    156         41. Y. Ruzankina, C. Pinzon-Guzman, A. Asare, T. Ong, L. Pontano, G. Cotsarelis, V. P. Zediak, M. Velez, A. Bhandoola, E. J. Brown, Deletion of the developmentally essential gene ATR in adult mice leads to age-related phenotypes and stem cell loss. Cell Stem Cell 1, 113–126 (2007). 42. R. H. Adams, K. Alitalo, Molecular regulation of angiogenesis and lymphangiogenesis. Nat. Rev. Mol. Cell Biol. 8, 464–478 (2007). 43. M. C. Smith, D. Y. Li, K. J. Whitehead, Mechanisms of vascular stability and the relationship to human disease. Curr. Opin. Hematol. 17, 237–244 (2010). 44. S. P. Herbert, D. Y. R. Stainier, Molecular control of endothelial cell behaviour during blood vessel morphogenesis. Nat. Rev. Mol. Cell Biol. 12, 551–564 (2011). 45. S. D. Castillo, E. Tzouanacou, M. Zaw-Thin, I. M. Berenjeno, V. Parker, I. Chivite, M. Milà-Guasch, W. Pearce, I. Solomon, A. Angulo-Urarte, A. M. Figueiredo, R. E. Dewhurst, R. G. Knox, G. R. Clark, C. L. Scudamore, A. Badar, T. L. Kalber, J. Foster, D. J. Stuckey, A. L. David, W. A. Phillips, M. F. Lythgoe, V. Wilson, R. K. Semple, N. J. Sebire, V. A. Kinsler, M. Graupera, B. Vanhaesebroeck, Somatic activating mutations in Pik3ca cause sporadic venous malformations in mice and humans. Sci. Transl. Med. 8, 332ra43 (2016). 46. A. Dompmartin, F. Ballieux, P. Thibon, A. Lequerrec, C. Hermans, P. Clapuyt, M.-T. Barrellier, F. Hammer, D. Labbé, M. Vikkula, L. M. Boon, Elevated D-dimer level in the differential diagnosis of venous malformations. Arch. Dermatol. 145, 1239–1244 (2009). 47. C. Fritsch, A. Huang, C. Chatenay-Rivauday, C. Schnell, A. Reddy, M. Liu, A. Kauffmann, D. Guthy, D. Erdmann, A. De Pover, P. Furet, H. Gao, S. Ferretti, Y. Wang, J. Trappe, S. M. Brachmann, S.-M. Maira, C. Wilson, M. Boehm, C. Garcia-Echeverria, P. Chene, M. Wiesmann, R. Cozens, J. Lehar, R. Schlegel, G. Caravatti, F. Hofmann, W. R. Sellers, Characterization of the novel and specific PI3Ka inhibitor NVP-BYL719 and development of the patient stratification strategy for clinical trials. Mol. Cancer Ther. 13, 1117–1129 (2014). 48. H. Lackner, A. Karastaneva, W. Schwinger, M. Benesch, P. Sovinz, M. Seidel, D. Sperl, S. Lanz, E. Haxhija, F. Reiterer, E. Sorantin, C. E. Urban, Sirolimus for the treatment of children with various complicated vascular anomalies. Eur. J. Pediatr. 174, 1579–1584 (2015). 49. A. M. Hammill, M. Wentzel, A. Gupta, S. Nelson, A. Lucky, R. Elluru, R. Dasgupta, R. G. Azizkhan, D. M. Adams, Sirolimus for the treatment of complicated vascular anomalies in children. Pediatr. Blood Cancer 57, 1018–1024 (2011). 50. J. Rodon, R. Dienstmann, V. Serra, J. Tabernero, Development of PI3K inhibitors: Lessons learned from early clinical trials. Nat. Rev. Clin. Oncol. 10, 143–153 (2013). 51. L. M. Hare, Q. Schwarz, S. Wiszniak, R. Gurung, K. G. Montgomery, C. A. Mitchell, W. A. Phillips, Heterozygous expression of the oncogenic PIK3CA(H1047R) mutation during murine develop- ment results in fatal embryonic and extraembryonic defects. Dev. Biol. 404, 14–26 (2015). 52. Y. Y. Kisanuki, R. E. Hammer, J.-i. Miyazaki, S. C. Williams, J. A. Richardson, M. Yanagisawa, Tie2-Cre transgenic mice: A new model for endothelial cell-lineage analysis in vivo. Dev. Biol. 230, 230–242 (2001). 53. J. A. Cox, E. Bartlett, E. I. Lee, Vascular Malformations: A review. Semin. Plast. Surg. 28, 58–63 (2014). 54. A. A. de Lorimier, Sclerotherapy for venous malformations. J. Pediatr. Surg. 30, 188–194 (1995). 55. J. Heyer, L. N. Kwong, S. W. Lowe, L. Chin, Non-germline genetically engineered mouse models for translational cancer research. Nat. Rev. Cancer 10, 470–480 (2010). 56. M. H. van Miltenburg, J. Jonkers, Using genetically engineered mouse models to validate candidate cancer genes and test new therapeutic approaches. Curr. Opin. Genet. Dev. 22, 21–27 (2012). 57. J. T. Calvert, T. J. Riney, C. D. Kontos, E. H. Cha, V. G. Prieto, C. R. Shea, J. N. Berg, N. C. Nevin, S. A. Simpson, K. A. Pasyk, M. C. Speer, K. G. Peters, D. A. Marchuk, Allelic and locus hetero- geneity in inherited venous malformations. Hum. Mol. Genet. 8, 1279–1289 (1999). 58. V. Wouters, N. Limaye,M. Uebelhoer, A. Irrthum, L.M. Boon, J. B.Mulliken, O. Enjolras, E. Baselga, J. Berg, A. Dompmartin, S. A. Ivarsson, L. Kangesu, Y. Lacassie, J.Murphy, A. S. Teebi, A. Penington, P. Rieu, M. Vikkula, Hereditary cutaneomucosal venous malformations are caused by TIE2muta- tions withwidely variable hyper-phosphorylating effects. Eur. J. Hum. Genet.18, 414–420 (2010). 59. I. Kim, H. G. Kim, J.-N. So, J. H. Kim, H. J. Kwak, G. Y. Koh, Angiopoietin-1 regulates endothelial cell survival through the phosphatidylinositol 3′-kinase/Akt signal transduction pathway. Circ. Res. 86, 24–29 (2000). 60. K. C. Kurek, V. L. Luks, U. M. Ayturk, A. I. Alomari, S. J. Fishman, S. A. Spencer, J. B. Mulliken, M. E. Bowen, G. L. Yamamoto, H. P. Kozakewich, M. L. Warman, Somatic mosaic activating mutations in PIK3CA cause CLOVES syndrome. Am. J. Hum. Genet. 90, 1108–1115 (2012). 61. V. L. Luks, N. Kamitaki, M. P. Vivero, W. Uller, R. Rab, J. V. M. G. Bovée, K. L. Rialon, C. J. Guevara, A. I. Alomari, A. K. Greene, S. J. Fishman, H. P. W. Kozakewich, R. A. Maclellan, J. B. Mulliken, R. Rahbar, S. A. Spencer, C. C. Trenor III, J. Upton, D. Zurakowski, J. A. Perkins, A. Kirsh, J. T. Bennett, W. B. Dobyns, K. C. Kurek, M. L. Warman, S. A. McCarroll, R. Murillo, Lymphatic and other vascular malformative/overgrowth disorders are caused by somatic mutations in PIK3CA. J. Pediatr. 166, 1048–1054 (2015). 62. M. J. Lindhurst, V. E. R. Parker, F. Payne, J. C. Sapp, S. Rudge, J. Harris, A. M. Witkowski, Q. Zhang, M. P. Groeneveld, C. E. Scott, A. Daly, S. M. Huson, L. L. Tosi, M. L. Cunningham, T. N. Darling, J. Geer, Z. Gucev, V. R. Sutton, C. Tziotzios, A. K. Dixon, T. Helliwell, S. O’Rahilly, D. B. Savage, M. J. O. Wakelam, I. Barroso, L. G. Biesecker, R. K. Semple, Mosaic overgrowth with fibroadipose hyperplasia is caused by somatic activating mutations in PIK3CA. Nat. Genet. 44, 928–933 (2012). 63. M. J. Lindhurst, J. C. Sapp, J. K. Teer, J. J. Johnston, E. M. Finn, K. Peters, J. Turner, J. L. Cannons, D. Bick, L. Blakemore, C. Blumhorst, K. Brockmann, P. Calder, N. Cherman, M. A. Deardorff, D. B. Everman, G. Golas, R. M. Greenstein, B. M. Kato, K. M. Keppler-Noreuil, S. A. Kuznetsov, R. T. Miyamoto, K. Newman, D. Ng, K. O’Brien, S. Rothenberg, D. J. Schwartzentruber, V. Singhal, R. Tirabosco, J. Upton, S. Wientroub, E. H. Zackai, K. Hoag, T. Whitewood-Neal, P. G. Robey, P. L. Schwartzberg, T. N. Darling, L. L. Tosi, J. C. Mullikin, L. G. Biesecker, A mosaic activating mutation in AKT1 associated with the Proteus syndrome. N. Engl. J. Med. 365, 611–619 (2011). 64. J.-B. Rivière, G. M. Mirzaa, B. J. O’Roak, M. Beddaoui, D. Alcantara, R. L. Conway, J. St-Onge, J. A. Schwartzentruber, K. W. Gripp, S. M. Nikkel, T. Worthylake, C. T. Sullivan, T. R. Ward, H. E. Butler, N. A. Kramer, B. Albrecht, C. M. Armour, L. Armstrong, O. Caluseriu, C. Cytrynbaum, B. A. Drolet, A. M. Innes, J. L. Lauzon, A. E. Lin, G. M. S. Mancini, W. S. Meschino, J. D. Reggin, A. K. Saggar, T. Lerman-Sagie, G. Uyanik, R. Weksberg, B. Zirn, C. L. Beaulieu; Finding of Rare Disease Genes (FORGE) Canada Consortium, J. Majewski, D. E. Bulman, M. O’Driscoll, J. Shendure, J. M. Graham Jr., K. M. Boycott, W. B. Dobyns, De novo germline and postzygotic mutations in AKT3, PIK3R2 and PIK3CA cause a spectrum of related megalencephaly syndromes. Nat. Genet. 44, 934–940 (2012). 65. M. D. Shirley, H. Tang, C. J. Gallione, J. D. Baugher, L. P. Frelin, B. Cohen, P. E. North, D. A. Marchuk, A. M. Comi, J. Pevsner, Sturge–Weber syndrome and port-wine stains caused by somatic mu- tation in GNAQ. N. Engl. J. Med. 368, 1971–1979 (2013). 66. D. J. Marsh, P. L. M. Dahia, Z. Zheng, D. Liaw, R. Parsons, R. J. Gorlin, C. Eng, Germline mutations in PTEN are present in Bannayan-Zonana syndrome. Nat. Genet. 16, 333–334 (1997). 67. A. Roy, J. Ni, J. Skibo, S. Rankin, W. B. Dobyns, F. Kalume, S. J. Baker, J. Zhao, K. J. Millen, Modeling human PIK3CA-related congenital brain overgrowth and epilepsy in mice. Int. J. Dev. Neurosci. 47, 46 (2015). 68. C. Pföhler, E. Janssen, A. Buecker, T. Vogt, C. S. L. Müller, Successful treatment of a congenital extra-truncal vascular malformation by orally administered propranolol. J. Dermatol. Treat. 26, 59–62 (2015). 69. J. Blatt, T. W. McLean, S. M. Castellino, C. N. Burkhart, A review of contemporary options for medical management of hemangiomas, other vascular tumors, and vascular malformations. Pharmacol. Ther. 139, 327–333 (2013). 70. D. D. Sarbassov, S. M. Ali, S. Sengupta, J.-H. Sheen, P. P. Hsu, A. F. Bagley, A. L. Markhard, D. M. Sabatini, Prolonged rapamycin treatment inhibits mTORC2 assembly and Akt/PKB. Mol. Cell 22, 159–168 (2006). Acknowledgments: We thank P. Zanzonico and V. Longo from the Animal Imaging Core; M. Turkekul, V. Boyko, and D. Yarilin from the Molecular Cytology Core; and G. Nanjangud from the Molecular Cytogenetics Core. We also thank M. Lepherd from the Laboratory of Comparative Pathol- ogy. We thank the MSKCC (Memorial Sloan Kettering Cancer Center) Center for Molecular Oncology, the Integrated Genomics Operation, K. Huberman, and N. Socci for the assistance with DNA sequen- cing and data analysis. We thank J. Bischoff and M. Valiente for providing cell lines and G. Minuesa for help with flow cytometry. Funding: Supported by Geoffrey Beene Cancer Center. The Molecular Cy- tology Core is supported by grant P30 CA008748J. The Integrated Genomics Operation is supported by grant P30 CA008748. J.G.-B. is supported by the Secretary for Universities and Research of the Government of Catalonia and the COFUND program of the Marie Curie Actions of the 7th R&D Framework Programme of the European Union. F.J.C. holds a fellowship from the Terri Brodeur Breast Cancer Foundation. K.V.A. was funded by NIH R01 NS 044385. Author contributions: P.C. and J.B. conceived the project. P.C., M.S., K.V.A., E.B., and J.B. supervised the research. P.C., F.J.C., and J.G.-B. performed the in vitro and in vivo experiments. C.R.A. reviewed the human and mice cases. C.R.A., S.B., and E.B. provided patients’ samples. M.F.B. and A.V. supervised and analyzed the se- quencing. P.C. prepared the figures. P.C., F.J.C., E.B., and J.B. wrote the manuscript. Competing interests: J.B. has consulted for Novartis Pharmaceuticals and serves on the board of Infinity Phar- maceuticals. The authors (P.C., E.B., and J.B.) have applied for a patent for the use of PI3K inhibitors for the treatment of vascular malformations. Data and materials availability: MSK-IMPACT raw sequencing data are available upon request. For the reagents, please contact the corresponding author. Submitted 29 October 2015 Accepted 2 March 2016 Published 30 March 2016 10.1126/scitranslmed.aaf1164 Citation: P. Castel, F. J. Carmona, J. Grego-Bessa, M. F. Berger, A. Viale, K. V. Anderson, S. Bague, M. Scaltriti, C. R. Antonescu, E. Baselga, J. Baselga, Somatic PIK3CA mutations as a driver of sporadic venous malformations. Sci. Transl. Med. 8, 332ra42 (2016). R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 30 March 2016 Vol 8 Issue 332 332ra42 10 on M arch 30, 2016 http://stm.sciencemag.org/ Downloaded from   157 ARTICLE 4     Title: Antagonism of EGFR and HER3 Enhances the Response to Inhibitors of the PI3K- Akt Pathway in Triple-Negative Breast Cancer Journal: Science Signaling Impact Factor: 6.279    158   159         C A N C E R Antagonism of EGFR and HER3 Enhances the Response to Inhibitors of the PI3K-Akt Pathway in Triple-Negative Breast Cancer Jessica J. Tao,1* Pau Castel,2* Nina Radosevic-Robin,3,4* Moshe Elkabets,2 Neil Auricchio,1 Nicola Aceto,1 Gregory Weitsman,5 Paul Barber,6,7 Borivoj Vojnovic,6,8 Haley Ellis,2 Natasha Morse,2 Nerissa Therese Viola-Villegas,9 Ana Bosch,2 Dejan Juric,1 Saswati Hazra,10 Sharat Singh,10 Phillip Kim,10 Anna Bergamaschi,11 Shyamala Maheswaran,1 Tony Ng,5,12 Frédérique Penault-Llorca,3,4 Jason S. Lewis,9 Lisa A. Carey,13 Charles M. Perou,14 José Baselga,2† Maurizio Scaltriti2† Both abundant epidermal growth factor receptor (EGFR or ErbB1) and high activity of the phosphatidyl- inositol 3-kinase (PI3K)–Akt pathway are common and therapeutically targeted in triple-negative breast cancer (TNBC). However, activation of another EGFR family member [human epidermal growth factor recep- tor 3 (HER3) (or ErbB3)] may limit the antitumor effects of these drugs. We found that TNBC cell lines cultured with the EGFR or HER3 ligand EGF or heregulin, respectively, and treated with either an Akt inhibitor (GDC- 0068) or a PI3K inhibitor (GDC-0941) had increased abundance and phosphorylation of HER3. The phospho- rylation of HER3 and EGFR in response to these treatments was reduced by the addition of a dual EGFR and HER3 inhibitor (MEHD7945A). MEHD7945A also decreased the phosphorylation (and activation) of EGFRand HER3 and the phosphorylation of downstream targets that occurred in response to the combination of EGFR ligands and PI3K-Akt pathway inhibitors. In culture, inhibition of the PI3K-Akt pathway combined with either MEHD7945A or knockdown of HER3 decreased cell proliferation compared with inhibition of the PI3K-Akt pathway alone. Combining eitherGDC-0068 orGDC-0941withMEHD7945A inhibited the growthof xenografts derived from TNBC cell lines or from TNBC patient tumors, and this combination treatment was also more effective thancombiningeitherGDC-0068orGDC-0941withcetuximab,anEGFR-targetedantibody.After ther- apy with EGFR-targeted antibodies, some patients had residual tumors with increased HER3 abundance and EGFR/HER3 dimerization (an activating interaction). Thus, we propose that concomitant blockade of EGFR, HER3, and the PI3K-Akt pathway in TNBC should be investigated in the clinical setting. INTRODUCTION Triple-negative breast cancer (TNBC) is clinically defined by the absence of estrogen receptor (ER), progesterone receptor, and human epidermal growth factor receptor (EGFR) 2 (HER2) overexpression or amplification. It represents 15 to 20% of newly diagnosed breast cancer, affects women in the reproductive age, and often follows an aggressive clinical course, with early recurrences in the form of distant visceral metastases, including to the brain (1–3). On the other hand, this tumor type has been demon- strated to be more responsive to cytotoxic therapy than ER-positive breast cancers (4–6). The current neoadjuvant strategies for TNBC use taxane/ anthracycline-based regimens, which reportedly achieve “pathological complete response” (pCR; defined as no invasive and no in situ residual tumors in breast and nodes) in about 20% of patients in unselected cohorts (7). TNBC has been described as having a high frequency of inactivation or decreased expression of the gene encoding phosphatase and tensin homolog deleted on chromosome 10 (PTEN) (1, 8), as well as overexpres- sion of the gene encoding human EGFR in up to about 50% of cases (9, 10). These biochemical features offer the opportunity to explore novel potential therapeutic strategies in this breast cancer subtype. Clinical benefits from the EGFR inhibitor cetuximab (11, 12) and the pan–phosphatidylinositol 3-kinase (PI3K) inhibitor NVP-BKM120 (13) have been reported in TNBC patients. However, none of these studies showed durable responses. Preclinical evidence suggests that inhibition of the PI3K-Akt-mTOR (mammalian target of rapamycin) axis induces compensatory genetic expres- sion and activation of upstream receptor tyrosine kinases (RTKs), includ- ing EGFR and, most prominently, HER3 (also known as ErbB3) (14–17). This may reduce the antitumor effects of single-agent PI3K pathway blockade. Furthermore, studies using cellular models of cetuximab resistance 1Massachusetts General Hospital Cancer Center and Harvard Medical School, 149 13th Street, Charlestown, MA 02129, USA. 2Human Oncology & Patho- genesis Program, Memorial Sloan Kettering Cancer Center, 1275 York Avenue, Box 20, New York, NY 10065, USA. 3Department of Biopathology, Centre Jean Perrin, 58 rue Montalembert, 63011 Clermont-Ferrand, France. 4ERTICA EA4677, University of Auvergne, 63000 Clermont-Ferrand, France. 5Richard Dimbleby Department of Cancer Research, Randall Division of Cell & Molecular Biophysics and Division of Cancer Studies, King’s College London, London SE1 1UL, UK. 6Gray Institute for Radiation Oncology and Biology, Department of Oncology, University of Oxford, Old Road Campus Research Building, Roose- velt Drive, Oxford OX3 7DQ, UK. 7Institute for Mathematical and Molecular Bio- medicine,King’sCollegeLondon, LondonSE11UL,UK. 8RandallDivisionofCell& Molecular Biophysics, King’s College London, London SE1 1UL, UK. 9Depart- ment of Radiology and Program inMolecular Pharmacology andChemistry,Me- morial Sloan Kettering Cancer Center, New York, NY 10065, USA. 10Prometheus Therapeutics & Diagnostics, 9410 Carroll Park Drive, San Diego, CA 92121, USA. 11Department of Molecular and Integrative Physiology, University of Illi- nois at Urbana-Champaign, 524 Burrill Hall, 407 South Goodwin Avenue, Urba- na, IL 61801, USA. 12UCL Cancer Institute, Paul O’Gorman Building, University College London, London WC1E 6DD, UK. 13Department of Medicine, University of North Carolina at Chapel Hill, 170 Manning Drive, Chapel Hill, NC 27599, USA. 14Department of Genetics, Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599, USA. *These authors contributed equally to this work. †Corresponding author. E-mail: scaltrim@mskcc.org (M.S.); baselgaj@mskcc.org (J.B.) R E S E A R C H A R T I C L E www.SCIENCESIGNALING.org 25 March 2014 Vol 7 Issue 318 ra29 1 on February 3, 2016 http://stke.sciencemag.org/ Downloaded from    160         suggest that HER3 itself can limit the sensitivity to cetuximab by increasing EGFR-HER3 heterodimerization and activation of downstream pathways (18). Although HER3 targeting is being explored in other breast cancer subtypes (19, 20), no rationale has yet been provided for the inhibition of this RTK in TNBC. Here, we hypothesized that targeting both EGFR and HER3 in combination with inhibition of the PI3K-Akt pathway would enhance the therapeutic response in EGFR-positive TNBC. RESULTS Blockade of EGFR and HER3 combined with inhibition of the PI3K-Akt pathway results in superior antitumor activity HCC70 andMDA-MB-468 TNBC cell lines, characterized by increased abundance of EGFR and loss of PTEN expression (fig. S1), were treated with GDC-0068 [a selective inhibitor of the Akt1, 2, and 3 isoforms (21)], GDC-0941 [a class I selective pan-PI3K inhib- itor (22)], MEHD7945A [an antibody targeting both EGFR and HER3 (23)], or a combina- tion of these inhibitors in the presence of either EGF or heregulin (NRG1), ligands for EGFR and HER3, respectively. Consistent with other reports (14–16), treatment with either GDC-0068 or GDC-0941 increased the abundance of HER3 and, in HCC70 cells, induced the phosphorylation (activation) of both EGFR and HER3 (Fig. 1A). The ad- dition of MEHD7945A prevented the EGF- or NRG1-induced activation of EGFR and HER3 and reduced the phosphorylation of the downstream mTOR effector ribosomal protein S6 and extracellular signal–regulated kinase (ERK) pathways in both cell lines (Fig. 1A). The effects of MEHD7945A on the phosphorylation of ERK in cells trig- gered by EGF are mild, likely because of the high abundance of EGFR in these cells. Notably, GDC-0068 competes for the aden- osine 5′-triphosphate (ATP)–binding site of Akt and is known to cause increased phos- phorylation of the enzyme at its two reg- ulatory sites [Thr308 and Ser473 (21)], as is evident in the blots. Given its effects on Akt and ERK ac- tivation, we tested whether combining MEHD7945A with either GDC-0068 or GDC-0941would enhance the antiprolifer- ative response in HCC70 and MDA-MB- 468 cells. In cells treated with single or double agents for 5 days, we observed vary- ing sensitivity to the single agents GDC- 0068 and GDC-0941, but in every case, the combination of the PI3K- or Akt-targeted agents and MEHD7945A considerably inhibited cell proliferation more effectively than did either single agent (Fig. 1B). To expand our findings in vivo, we first tested the efficacy of MEHD7945A in combination with either GDC-0068 or GDC-0941 in both HCC70- and MDA-MB-468–derived xenografts (fig. S2). Whereas both types of tumors responded only modestly to any of the single agents (GDC-0068, GDC- 0941, or MEHD7945A), the combination of GDC-0068 or GDC-0941 and MEHD7945A yielded significantly superior tumor growth inhibi- tion compared to monotherapy. Moreover, one-third of the animals in the cohorts of both combination regimens achieved complete tumor shrink- age, with no relapses observed 90 days after treatment cessation. We next GDC-0941 GDC-0941 + MEHD7945A 0 2×10–7 4×10–7 6×10–7 8× 10–7 0.0 0.2 0.4 0.6 0.8 1.0 1.2 0 2×10–7 4×10–7 6×10–7 8× 10–7 0.0 0.2 0.4 0.6 0.8 1.0 1.2 0 2×10–7 4×10–7 6×10–7 8×10–7 0.0 0.2 0.4 0.6 0.8 1.0 1.2 0 2×10–7 4×10–7 6×10–7 8×10–7 0.0 0.2 0.4 0.6 0.8 1.0 1.2 GDC-0068 GDC-0068 + MEHD7945A GDC-0068 GDC-0068 + MEHD7945A GDC-0941 GDC-0941 + MEHD7945A Time (days) 0 20 40 60 Tu m or v ol um e (m m 3 ) 2500 2000 1500 1000 500 0 Control MEHD7945A GDC-0941 GDC-0068 GDC-0941 + MEHD GDC-0068 + MEHD R el at iv e lu m is ce nc e un its C U s C U s C U s C U s C on tr ol G D C - 0 06 8 G D C -0 94 1 M E H D 79 45 A 00 68 + M E H D 09 41 + M E H D C on tr ol G D C -0 06 8 G D C -0 94 1 M E H D 79 45 A 00 68 + M E H D 09 41 + M E H D pEGFR pHER3 tEGFR tHER3 25 20 15 10 5 0 .10 .08 .06 .04 .02 0 4 3 2 1 40 30 20 10 0 0 Log(M) GDC-0068 Log(M) GDC-0941 Log(M) GDC-0068 Log(M) GDC-0941 pEGFR tEGFR pAkt (308) pAkt (473) pERK pS6 (240/4) Actin pHER3 tHER3 pAkt (308) pAkt (473) pERK pS6 (240/4) Actin pEGFR tEGFR pAkt (308) pAkt (473) pERK pS6 (240/4) Actin pHER3 tHER3 pAkt (308) pAkt (473) pERK pS6 (240/4) Actin MEHD7945A GDC-0941 GDC-0068 EGF MEHD7945A GDC-0941 GDC-0068 NRG1 MEHD7945A GDC-0941 GDC-0068 EGF MEHD7945A GDC-0941 GDC-0068 NRG1 HCC70 MDA468A B C D ** * Fig. 1. Therapeutic activity of combined inhibition of EGFR, HER3, and the PI3K-Akt pathway in TNBC preclinical models. (A) Western blot for total and phosphorylated EGFR, HER3, and downstream proteins HCC70 and MDA-MB-468 (MDA468) cells after the indicated treatments for 24 hours: EGF or NRG1, 4 ng/ml; MEHD7945A, 10 nM; GDC-0068 and GDC-0941, 1 mM. (B) Analysis of the proliferation of HCC70 (left) and MDA-MB-468 cells (right) treated for 5 days as indicated; concentrations as in (A). (C) Tumor growth curves of TNBC PDX treated as indicated: MEHD7945A (MEHD), 10 mg/kg twice weekly; GDC-0941, 75 mg/kg daily; GDC-0068, 40 mg/kg daily. (D) CEER analysis of total and phosphorylated EGFR and HER3 in PDXs treated as indicated. Blots in (A) are representative of and data in (B) are means ± SEM from two experiments. n ≥ 8 and n ≥ 3 for each treatment arms in (B) and (D), respectively; *P = 0.048 in (C), GDC-0941 + MEHD versus GDC-0941, and *P = 0.007, GDC-0068 + MEHD versus GDC-0068. (D) **P = 0.054, GDC-0941 + MEHD versus GDC-0941, two-sided Student’s t test. R E S E A R C H A R T I C L E www.SCIENCESIGNALING.org 25 March 2014 Vol 7 Issue 318 ra29 2 on February 3, 2016 http://stke.sciencemag.org/ Downloaded from   161         investigated the abundance and activation of EGFR and HER3 in HCC70- derived xenografts collected at the end of each experiment. The technical challenge of obtaining reliable detection of phosphorylated HER3 by im- munohistochemistry (IHC) and the small amount of tissue available from the tumors treated with the combination regimens prompted us to assess these using an alternative methodology. Frozen tissue was analyzed by collaborative enzyme enhanced reactive-immunoassay (CEER), a platform that uses reversed-phase detection of nanogram quantities of protein (22). Treatment with GDC-0068 or GDC-0941 increased the abundance and phosphorylation of both EGFR and HER3 (fig. S3 and table S1A). The increased phosphorylation of EGFR after GDC-0068 treat- ment was most likely a result of increased EGFR/HER3 het- erodimerization because we observed no changes in the total abundance of EGFR. As expected, the cotreatment of MEH- D7945A prevented receptor phosphorylation induced by either GDC-0068 or GDC-0941. To test the response to these treatments in patient-derived xenografts (PDXs) of TNBC, we used available patient tu- mors in our laboratory. These tumors were characterized by IHC to have undetectable abundance of PTEN, high abun- dance of EGFR, and generally high (more than 50%) staining for the proliferation marker Ki67 (fig. S4). These features predicted a particularly aggressive phenotype, confirmed by the rapid growth of the tumor xenografts in untreated mice (Fig. 1C). Single-agent treatment with GDC-0068, GDC-0941, or MEHD7945A delayed tumor growth, whereas the combi- nation of either GDC drug with MEHD7945A caused durable tumor stasis (Fig. 1C). Consistent with the cell line–derived xenografts, the abundance of HER3 and EGFR increased after inhibition of the PI3K-Akt pathway. The addition of MEH- D7945A effectively prevented HER3 phosphorylation and kept that of EGFR at the same abundance as in control tumors (Fig. 1D and table S1B). The capability of Akt inhibitors to increase the abundance of EGFR and HER3 in PDXs was also confirmed by positron emission tomography (PET). From the PET scans, we found a visibly higher accumulation of 89Zr- MEHD7945A in the tumor of the GDC-0068–treated cohort of mice compared to the control group (fig. S5). Quantifica- tion of the uptake of 89Zr-MEHD7945A in GDC-0068–treated or untreated tumors revealed a nearly twofold higher tracer (89Zr) accumulation in the treated group compared to control mice. Liver accumulation of 89Zr-MEHD7945A is considered as the main route of excretion of the probe. Tumor cell proliferation was also measured using the Ki67 index in specimens from xenografts collected at the experimental endpoints. We found that the percentage of Ki67-positive cells was significantly lower only in the combination therapy co- horts (fig. S6A). These results were further confirmed mea- suring the number of Ki67-positive circulating tumor cells (CTCs) in mice bearing PDXs greater than 1 cm3 in volume and treated for 6 days with GDC-0068, GDC-0941, MEH- D7945A, or the combination of these agents (fig. S6B). Col- lectively, these data show that targeting both EGFR and HER3 enhanced the antitumor effects of PI3K-Akt inhibitors. HER3 suppression improves the antitumor activity of PI3K-Akt inhibition To dissect the role of HER3 inhibition in these models, we compared the activity of cetuximab (an antibody targeting exclusively EGFR) with MEHD7945A, each in combina- tion with either GDC-0068 or GDC-0941 in HCC70 cells. Both cetuxi- mab and MEHD7945A enhanced the antiproliferative activity mediated by PI3K-Akt pathway inhibition in cells stimulated with EGF; however, MEHD7945A was more effective than cetuximab in cooperating with GDC-0068 and GDC-0941 in cells stimulated with NRG1 (Fig. 2A). The importance of specifically blocking HER3 in this setting was con- firmed by testing the activity of PI3K-Akt inhibitors after HER3 knock- down by small interfering RNA (siRNA). HER3 depletion sensitized MDA-MB-468 cells to the antiproliferative activity of either GDC-0068 or GDC-0941 (fig. S7). 0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 0 2 × 10–7 4 × 10–7 6 × 10–78 × 10–7 0.0 0.2 0.4 0.6 0.8 1.0 1.2 R el at iv e lu m is ce nc e un its R el at iv e lu m is ce nc e un its Log(M) GDC-0068 Log(M) GDC-0068 Log(M) GDC-0941 Log(M) GDC-0941 pEGFR tEGFR pHER3 tHER3 GDC-0068 GDC-0068 + cetuximab GDC-0068 + MEHD7945A GDC-0941 GDC-0941 + cetuximab GDC-0941 + MEHD7945A C on tr ol C on tr ol C et ux im ab C et ux im ab +G D C - 00 68 C et ux im ab +G D C - 09 41 C on tr ol C on tr ol C et ux im ab C et ux im ab + G D C -0 06 8 C et ux im ab + G D C -0 94 1 Time (days) 0 10 20 30 40 50 T um or v ol um e (m m 3 ) 1500 1000 500 0 Control Cetuximab GDC-0941 + Cetux GDC-0941 GDC-0941 + MEHD +NRG1 +EGF +E G F +N R G 1 C U s C U s pEGFR pHER3 1.5 1 0.5 0 .008 .006 .004 .002 0 ** EGFR HER3 C U s .8 .6 .4 .2 0 C U s 8 6 4 2 0 10 ** *** * C on tr ol C et ux im ab G D C -0 94 1 + C et ux G D C -0 94 1 + M E H D G D C -0 94 1 C on tr ol C et ux im ab G D C -0 94 1 + C et ux G D C -0 94 1 + M E H D G D C -0 94 1 A B C 2 × 10–7 4 × 10–7 6 × 10–78 × 10–7 00 2 × 10–74 × 10–7 6 × 10–78 × 10–7 2 × 10–7 4 × 10–7 6 × 10–78 × 10–7 Fig. 2. Efficacy of MEHD7945A or cetuximab in combination with PI3K inhibition. (A) Left: Proliferation analysis of HCC70 cells treated for 5 days as indicated in the presence of EGF (top) or NRG1 (bottom). MEHD7945A (MEHD) and cetuximab (Cetux), 10 nM; GDC-0068 andGDC-0941, 1 µM. Right: Western blot for phosphorylated and total EGFR and HER3 in HCC70 cells treated as indicated. Blots are representative of two experiments. (B) Tumor growth curves of HCC70 xenografts treated as indicated, doses as in Fig. 1A. (C) CEER analysis of active and total EGFR and HER3 in HCC70 xenografts treatedas indicated.Dataaremeans±SEM.n≥8andn≥3 for each treatment arms in (B) and (D), respectively, at least three tumors per condition; (B) *P = 0.020, GDC-0941 + MEHDversusGDC-0941, andP=0.0054,GDC-0941+MEHDversusGDC-0941+Cetux. (C) **P = 0.0004, GDC-0941 + MEHD versus control, and P < 0.0001, GDC-0941 + Cetux versus control. (C) ***P = 0.054, GDC-0941 + MEHD versus GDC-0941, Student’s t test. R E S E A R C H A R T I C L E www.SCIENCESIGNALING.org 25 March 2014 Vol 7 Issue 318 ra29 3 on February 3, 2016 http://stke.sciencemag.org/ Downloaded from    162         We next compared the antitumor effect of combining cetuximab versus MEHD7945A with GDC-0941 in HCC70-derived xenografts. Whereas the combination of cetuximab and GDC-0941 did not further inhibit tu- mor growth compared to either single-agent treatment, the combination of MEHD7945A and GDC-0941 (concomitantly targeting EGFR, HER3, and PI3K) was superior to either cetuximab alone or cetuximab in com- bination with GDC-0941 (Fig. 2B), with no palpable tumor present in four of nine cases. The combination of GDC-0068 with cetuximab (tested in both HCC70 and PDX models) appeared to inhibit tumor growth in some cases compared to either single agent, but the effect was not statistically significant (fig. S8), suggesting that adding cetuximab had no benefit over either PI3K-Akt pathway inhibitor alone. Biochemically, both MEHD7945A and cetuximab combination treatments with GDC-0941 decreased the phos- phorylation of EGFR in HCC70 xenografts, but only the MEHD7945A combination decreased HER3 activation, although this was not statistically significant (Fig. 2C and table S1C). These results suggest that HER3 plays an important role in limiting the efficacy of PI3K-Akt pathway inhibitors in this model. Decreased EGFR and increased HER3 abundance are associated with lower response to EGFR antagonists in TNBC patients To investigate whether changes in EGFR and HER3 abundance can affect the response to anti-EGFR therapy in TNBC patients, we evaluated the abundance of these receptors in samples from patients enrolled in two pilot neoadjuvant clinical trials testing the antitumor activity of the EGFR antibodies panitumumab (47 patients) and cetuximab (29 patients) in com- bination with various standard chemotherapies. Of the 47 TNBC patients enrolled in the study that combined panitumumab with four standard cy- totoxic agents (DNA-damaging agents 5-fluorouracil, epirubicin, or cyclo- phosphamide, or the mitotic inhibitor docetaxel), 22 patients (46.8%) achieved pCR at the time of surgery (24 weeks after treatment commenced), whereas 25 patients (53.2%) showed residual disease (NCT00933517). This was a twofold increase in pCR compared to TNBC patients treated only with cytotoxic-based neoadjuvant chemotherapy (6), which underscores the benefit of adding EGFR-targeted agents in this setting. Of 29 patients enrolled in the study testing the antitumor activity of cetuximab combined with docetaxel, 8 experienced pCR (27.5%) (NCT00600249). IHC assessment of EGFR and HER3 abundance was possible on pre- treatment samples from 16 patients who achieved pCR with panitumumab combination therapy at time of surgery and 24 patients who did not (table S2). We observed a trend toward a higher probability to achieve pCR after panitumumab treatment in patients with a pretreatment EGFR score higher than 70 (Fig. 3A). No statistical correlation was found between the pre- treatment HER3 score and the likelihood to achieve pCR in this cohort. For the cetuximab-treated patients, the analysis of EGFR and HER3 abun- dance before and after treatment was possible only for six of the eight patients who reached pCR; thus, because of the low sample size, we did not perform the pCR correlation analysis for the patients enrolled in the cetuximab trial. The results from the panitumumab trial at least suggest that high EGFR abundance before treatment predicts an optimal response to EGFR-targeted antibodies. However, a substantial portion (42%) of pa- tients who had high EGFR abundance in the pretreated tumor did not show complete response; therefore, we investigated whether changes in the abundance or activity of EGFR and HER3 occurred after treatment. For patients who did not achieve pCR at the time of surgery, IHC assess- ment was possible on 24 and 22 paired samples for EGFR and HER3, respectively, from the panitumumab trial, and on 19 and 20 paired sam- ples, respectively, from the cetuximab trial (table S2). When we consid- ered only the tumors with a pretreatment EGFR score higher than 150 (12 from the panitumumab trial and 7 from the cetuximab trial), we found decreased EGFR abundance in residual tumors after treatment in 12 of 19 cases compared with that in the paired pretreatment specimens (Fig. 3, B and C). On a side note, we observed a similar trend in a different cohort of samples obtained from metastatic TNBC patients enrolled in the TBCRC 001 clinical trial (12), whowere treated with cetuximab in combination with carboplatin. pCR could not be used as a parameter of response given the metastatic nature and the low response rate observed in the study, so we examined overall survival. On the basis of gene expression data from 16 paired biopsies (before and after 1 to 2 weeks of treatment), we observed 10 H 02 83 11 F0 40 7 09 H 24 20 10 H 17 76 09 H 25 87 10 F0 62 1 10 H 05 40 11 F0 65 1 11 H 02 83 10 H 15 65 10 F0 38 3 10 H 10 53 12 F0 15 7 11 H 11 78 11 H 15 12 12 F0 15 5A 09 H 21 10 12 H 01 43 12 H 01 45 Panitumumab Cetuximab Pre Post Pre Post E G F R h is to sc or e 300 200 100 0 CR Non- CR 10 4 11 15 28% 58% 30 25 20 15 10 5 0 N um be r of p at ie nt s EGFR < 70 EGFR > 70 Pretreatment Posttreatment 10 H 02 83 11 F 04 07 A B C Fig. 3. EGFR expression and response to anti-EGFR therapy in TNBC patients. (A) Correlation between the abundance of EGFR (table S2) and pCR in 47 TNBC patients treated with panitumumab combination therapy (receiver operating characteristic curve, P = 0.08). (B) EGFR abundance in tumors before and after treatment with either panitumumab or cetuximab combination therapy in patients who did not achieve pCR (P = 0.0048). (C) Representative IHC images (×25 magnification) of EGFR abundance in residual tumors (posttreatment) versus baseline specimens (pre- treatment). Scale bars, 80 µm. R E S E A R C H A R T I C L E www.SCIENCESIGNALING.org 25 March 2014 Vol 7 Issue 318 ra29 4 on February 3, 2016 http://stke.sciencemag.org/ Downloaded from   163         decreased overall survival in patients with decreased EGFR mRNA in the posttreatment specimens (fig. S9). The abundance of HER3 in the 22 residual (posttreatment) tumors from panitumumab-treated patients compared with their pretreatment counterparts was increased in 12 patients (Fig. 4, A and B). A similar trend was observed in cetuximab-treated patients: HER3 abundance was increased in the residual tumors of 13 of 20 available non-pCR patient samples compared with that in paired pretreatment specimens (fig. S10). When the data from both the panitumumab- and the cetuximab-treated pa- tient samples were pooled, the increase in HER3 abundance was statisti- cally significant (table S2). Using fluorescence resonance energy transfer (FRET) to analyze EGFR-HER3 dimerization (a mechanism of activa- tion), we investigated whether there was increased activation in addition to increased abundance of HER3 in the residual tumors of patients not achieving pCR from EGFR-targeted therapy. Enough tissue was available in a small cohort of specimens, three patients treated with cetuximab and four patients treated with panitumumab. Five of seven of the pooled cases displayed increased dimerization of HER3 and EGFR (Fig. 4, C and D). Together, these results suggest that increased HER3 expres- sion and HER3 activation may mediate residual tumor growth after EGFR-targeted therapy. DISCUSSION Here, we found that HER3 may play a critical role in limiting the antitumor effect of inhibitors targeting either the PI3K- Akt or EGFR pathway. We demonstrated that simultaneously targeting EGFR and HER3 by MEHD7945A enhanced the efficacy of PI3K-Akt pathway inhibitors in preclinical models of EGFR-positive TNBC. Furthermore, our clinical analysis suggests that HER3 abundance and activation are induced in TNBC patients by EGFR-targeted therapies and that this change may prevent therapy-induced tumor regression. In a study testing the clinical activity of cetuximab in TNBC, only a minority of patients whose tumors showed EGFR pathway inhibition derived clinical benefit from the therapy, suggest- ing that different mechanisms of receptor activation may oc- cur in this subtype of breast cancer (12). Preclinically, low abundance of EGFR and high abundance of HER3, among others, may determine cetuximab resistance in PDX models (24). Moreover, HER3 abundance and phosphorylation (a marker of activation) are induced after Akt suppression. The first evidence of this feedback activation was reported by pio- neering work from Sergina et al., who also postulated that HER3 plays a pivotal role in limiting the efficacy of HER ki- nase inhibitors (17). These results were extensively validated using other HER inhibitors (25) or specific molecules directly targeting PI3K or Akt (14, 16). We have previously shown that this phenomenon also occurs in patients treated with the Akt inhibitor GDC-0068 (26). Therefore, our rationale behind simul- taneously targeting EGFR, HER3, and PI3K in TNBC is de- rived from the observations that nearly half of TNBC have a high abundance of EGFR, that TNBC often has a low abun- dance of the endogenous Akt inhibitor PTEN, and that in- creased abundance and activation of HER3 appear to limit the sensitivity of TNBC to targeted therapy. It is plausible that lower doses of PI3K-Akt pathway inhibitors are required to achieve pathway suppression (and, consequently, tumor growth inhibition) when both EGFR and HER3 are inhibited. In addi- tion, tumor cells that have a relatively high abundance of these receptors can function as molecular “flags” for immune-mediated antibody-dependent cytotoxicity, intrinsic characteristics of im- munoglobulin G type I antibodies such as MEHD7945A in vivo. Thus, the antibody-mediated immune response may also ex- plain, at least in part, the higher response to the drug combi- nations in nude mice compared to the in vitro setting. In conclusion, we believe that simultaneous inhibition of EGFR, HER3, and the PI3K-Akt pathway has the potential to greatly expand the percentage of TNBC patients who can Pre Post Pre Post H E R 3 hi st os co re 300 200 100 0 09 H 24 20 09 H 26 79 10 F 02 26 10 F0 38 3 10 F 06 21 11 F 04 07 10 F 03 52 10 F 05 39 10 F 07 12 09 H 25 87 10 F0 54 2 11 F0 65 1 10 H 02 83 10 H 05 40 11 F 04 08 11 F 04 10 10 H 15 65 10 H 10 53 10 F0 62 6 12 F 00 13 10 F 03 20 12 F0 01 2 Pre Post 12 R 00 19 11 F 04 10 10 F 05 42 10 F0 35 2 09 H 24 20 12 H 00 80 10 H 02 83 F R E T e ffi ci en cy (% ) .06 .04 .02 0 10F0352 10F0542 P re tre at m en t P os tt re at m en t 10F0352 P os ttr ea tm en t P re tr ea tm en t P os tt re at m en t P re tre at m en t 10F0542 Anti-EGFR FRET map Anti-EGFR FRET map A B D C 0 0 00 20 2020 20 % % % % Fig. 4. HER3 expression and response to anti-EGFR therapy in TNBC patients. (A) HER3 abundance by IHC in tumors before and after panitumumab-based treatment in pa- tients who did not achieve pCR (P = 0.0028). (B) Representative IHC (×25 magnifi- cation) from two patient tumors analyzed in (A). Scale bars, 80 µm. (C) FRET analysis of HER3-EGFR dimerization in residual tumors from a subset of patients who did not achieve pCR after treatment with panitumumab/cetuximab-based therapy. (D) Rep- resentative time-resolved immunofluorescence images (×20 magnification) from two tumors analyzed in (A) to (C). Grayscale image shows intensity of the donor fluorophore, and pseudocolor image shows the pixel-by-pixel FRET efficiency values. Scale bar, 50 µm. R E S E A R C H A R T I C L E www.SCIENCESIGNALING.org 25 March 2014 Vol 7 Issue 318 ra29 5 on February 3, 2016 http://stke.sciencemag.org/ Downloaded from    164         benefit from targeted therapy. Given that both pharmacokinetic and pharmacodynamic data for MEHD7945A, GDC-0068, and GDC-0941 are already available, the design of phase 2 clinical trials testing the activ- ity of these possible combinations in TNBC would be straightforward. Patients may be enrolled on the basis of EGFR abundance in the tumors, and HER3 abundance and activation after PI3K pathway inhibition may be measured with either biopsies collected during treatment or live imag- ing techniques. MATERIALS AND METHODS Study design The objective of our study was to test the activity of concomitant blockade of EGFR, HER3, and the PI3K-Akt pathway in preclinical models of TNBC. Moreover, we aimed to evaluate whether the expression of both EGFR and HER3 was influencing the clinical response to anti-EGFR ther- apy in TNBC patients. We planned to treat with GDC-0068 (Akt inhibitor), GDC-0941 (PI3K inhibitor), and MEHD7945A (antibody binding to both EGFR and HER34) TNBC cell lines and tumors to test the antitumor activ- ity of these compounds separately and in combination. Moreover, we planned to test by IHC the expression of both EGFR and HER3 in TNBC patients who underwent cetuximab (antibody anti-EGFR)–based therapy. In vitro experiments were performed at least two times and at least in triplicate for each replica. Cell lines and chemical compounds MDA-MB-468 and HCC70 were purchased from the American Type Culture Collection and maintained at 37°C in Dulbecco’s modified Eagle’s medium/Ham’s F-12 1:1 and RPMI 1640, respectively, with 10% fetal calf serum, L-glutamine (2 mM), penicillin (20 U/ml), and streptomycin (20 µg/ml) in a humidified atmosphere and 5% CO2. The pan-PI3K inhibitor GDC-0941 was obtained from the Stand Up To Cancer/PI3K Dream Team Mouse Pharmacy. The Akt inhibitor GDC-0068 and the dual EGFR-HER3 in- hibitor MEHD7945A were provided by Genentech. All compounds were dissolved in dimethyl sulfoxide for in vitro experiments. Cell viability and proliferation For proliferation, 5 × 103 to 8 × 103 cells were seeded in 96-well plates and treated with the indicated concentrations of GDC-0068, GDC-0941, and/or MEHD7945A. After 5 days, cells were fixed and stained with crystal violet. Cell proliferation was also analyzed with CellTiter-Glo Luminescent Cell Viability Assay (Promega) as described by the manufacturer. For prolifera- tion in response to heregulin (NRG1; PeproTech) and EGF (PeproTech), 5 × 104 cells were treated with GDC-0068, GDC-0941, and/or MEHD7945A in the presence of NRG1 (4 ng/ml) or EGF for 5 days and then stained with crystal violet. For siRNA experiments, 5 × 103 cells were seeded in 96-well plates and transfected with siRNA (Silencer, Ambion) control or a pool of two hairpins targeting human HER3 mRNA using DharmaFECT transfec- tion reagent (Thermo Scientific). Western blotting Cells were washed with ice-cold phosphate-buffered saline (PBS) and scraped into ice-cold radioimmunoprecipitation assay lysis buffer (Cell Signaling Technology) supplemented with phosphatase inhibitor cocktails (Complete Mini and PhosphoStop, Roche). Lysates were cleared by centrifugation at 13,000 rpm for 10 min at 4°C, and supernatants were removed and assayed for protein concentration using the Pierce BCA Protein Assay Kit (Thermo Scientific). Thirty-five micrograms of total lysate was resolved on NuPAGE 4 to 12% bis-tris gels (Life Technologies) and electrophoretically transferred to Immobilon transfer membranes (Millipore). Membranes were blocked for 1 hour in 5% nonfat dry milk in tris-buffered saline (TBS)–Tween and then hybridized using the following primary antibodies in 5% bovine serum albumin (BSA) TBS-Tween: phospho-Akt (Ser473), phospho-Akt (Thr308), Akt, phospho-S6 (Ser240/4), phospho-S6 (Ser235/6), S6, phospho- ERK (Thr202/Tyr204), ERK, phospho-EGFR (Tyr1068), EGFR, phospho- HER3 (Tyr1289), and HER3 (1:500 to 1:1000; Cell Signaling Technology). b-Actin was used as a loading control (1:5000; Sigma), also in 5% BSA TBS-Tween. Mouse and rabbit horseradish peroxidase (HRP)–conjugated secondary antibodies (1:50,000; Amersham Biosciences) were diluted in 2% nonfat dry milk in TBS-Tween. Protein-antibody complexes were detected by chemiluminescence with SuperSignal West Femto Chemiluminescent Substrate (Thermo Scientific), and images were captured with a G:BOX camera system. Establishment of tumor xenografts and in vivo treatments All mouse studies were conducted through Institutional Animal Care and Use Committee–approved animal protocols in accordance with institution- al guidelines. Six-week-old female athymic nude mice were purchased from Charles River Laboratories and housed in air-filtered laminar flow cabinets with a 12-hour light cycle and food and water ad libitum. The size of the animal groups was calculated to measure means difference be- tween placebo and treatment groups of 25% with a power of 80% and a P value of 0.01. Host mice carrying xenografts were randomly and equally assigned to either control or treatment groups. Animal experiments were conducted in a controlled and nonblinded manner. For cell line–derived xenograft studies, mice were injected subcutaneously with 1 × 107 HCC70 or MDA-MB-468 suspended in 150 ml of culture medium/Matrigel (BD Biosciences) in a 4:1 ratio. 17b-Estradiol (1 mM) was supplemented in the mouse drinking water as described (27). For PDX studies, tumors were subcutaneously implanted in 6-week- old female athymic nude mice. Upon xenograft growth, tumor tissue was reimplanted into recipient mice, which were randomized upon implant growth. For the collection of CTCs, tumors were implanted into the mam- mary pad of athymic nude mice. Once tumors reached an average volume of ~150 to 250 mm3, mice were randomized into treatment arms, with 7 to 11 tumors per group. GDC-0068 (40 mg/kg) or GDC-0941 (75 mg/kg) was dissolved in 0.5% methylcellulose and 0.2% Tween-80 (MCT) solution and administered once daily via oral gavage. MEHD7945A (10 mg/kg) and cetuximab (10 mg/kg) were diluted in PBS and injected intraperitoneally twice weekly. Tumors were measured by digital caliper over the entire treatment period and har- vested 2 hours after the last administration of the drug. Tumor volume was determined using the following formula: (length × width2) × (p/6). Tumor volumes are plotted as means ± SEM. Small-animal immuno-PET Preparation of 89Zr-MEHD7945A: The MEHD7945A monoclonal anti- body (mAb) was functionalized with p-isothiocyanatobenzyl-desferrioxamine (DFO-Bz-NCS, Macrocyclics Inc.) with a 1:7 mAb/DFO-Bz-NCS ratio. The reaction was incubated at 37°C for 1 hour. The modified antibodies were purified using a 10-kD centrifugal filter (GE Vivaspin 500). 89Zr was produced through proton beam bombardment of yttrium foil and isolated in high purity as 89Zr-oxalate at Memorial Sloan Kettering Cancer Center ac- cording to previously established procedure (28). Labeling of theMEHD7945A- DFO conjugate proceeded as described (29) with an obtained specific activity of 2.5 to 3 mCi/mg and >95% purities. PET imaging: Scans were recorded with a microPET Focus 120 (Concorde Microsystems). Mice (n = 3 for each group, bearing two tumors on each flank) were administered with 89Zr-MEH- D7945A (150 to 200 mCi, 50 to 68 mg) in 100 ml of 0.9% saline formulations R E S E A R C H A R T I C L E www.SCIENCESIGNALING.org 25 March 2014 Vol 7 Issue 318 ra29 6 on February 3, 2016 http://stke.sciencemag.org/ Downloaded from   165         via lateral tail vein injections. Whole-body acquisitions were acquired on mice while anesthetized with 1.5 to 2.0% isoflurane (Baxter Healthcare) in oxygen at 24 to 120 hours after injection. Images were reconstructed via filtered back projection. The images were analyzed using ASIPro VM software (Concorde Microsystems). Volumes of interest (VOIs) were measured on various planar sections of the acquired image by manu- ally drawing on the tumor site. The average VOI was calculated and ex- pressed as percent injected dose per gram of tumor tissue (%ID/g). Data values were expressed as means ± SD unless otherwise stated. Statis- tical analysis was performed with GraphPad Prism version 6.02 soft- ware using Student’s t test. A P value of <0.05 is considered statistically significant. Collaborative enzyme enhanced reactive-immunoassay The abundance and phosphorylation of EGFR and HER3 in xenografts were determined by CEER (30, 31). CEER uses the formation of unique immunocomplexes between capture antibodies printed on a nitrocellulose microarray surface, with which the target molecule in cell lysates reacts, and two independent detector antibodies. One detector antibody is con- jugated to glucose oxidase, and the other is conjugated to HRP. Target de- tection [expressed as computational unit (CU)] requires the presence of both detector antibodies, and the enzyme channeling event between glu- cose oxidase and HRP will not occur unless both antibodies are in close proximity. For each assay, a standard curve was generated from eight con- centrations of serially diluted reference lysates from cell lines, well char- acterized for the abundance and phosphorylation of RTKs. Each assay included controls along with sample lysates. When control lysates pro- vided acceptable values, signals generated from samples were quantified against the standard curve. One computational unit for EGFR represented about 106 molecules, whereas 1 CU for HER3 represented about 5 × 104 molecules. Raw data were normalized by the total amount of cyto- keratins (CKs) to include only protein expressed in the epithelial com- partment (table S1). Circulating tumor cells CTCs were captured on the herringbone chip, fixed, and permeabilized as previously described (32). For capture, the herringbone chip was coated with antibodies against EpCAM (epithelial cell adhesion molecule) (R&D Systems) and EGFR (cetuximab; Eli Lilly) (33). The CTC-containing chip was incubated with primary antibodies against wide-spectrum CKs (Abcam), CD45 (Santa Cruz Biotechnology), and Ki67 (Life Technologies), and sec- ondary antibodies were conjugated with Alexa Fluor 647, Alexa Fluor 555, and Alexa Fluor 488 (all from Life Technologies). Nuclei were stained with 4′,6′-diamidino-2-phenylindole (DAPI). We used an automated fluo- rescence microscopy scanning system (BioView) to identify Ki67-positive CTCs (CK+/CD45–/Ki67+), Ki67-negative CTCs (CK+/CD45–/Ki67–), and contaminating white blood cells (CD45+). Immunohistochemistry For IHC on xenografts, dissected tissues were fixed immediately after re- moval in a 10% buffered formalin solution for a maximum of 24 hours at room temperature before being dehydrated and paraffin-embedded under vacuum conditions. Samples were blocked with normal goat serum and incubated with Ki67 (Life Technologies), EGFR (Cell Signaling Technol- ogy), and PTEN (Cell Signaling Technology) antibodies. The antigen-antibody reaction was revealed by SignalStain Boost IHC Detection Reagent (8114, Cell Signaling Technology) with 3,3′-diaminobenzidine (DAB) as a sub- strate (Dako). For IHC on patient samples, tumor tissue was fixed in 10% buffered formalin for 48 hours and embedded in paraffin. Four-micrometer sections were deparaffinized in xylene and hydrated in graded alcohols. For EGFR detection, the antigen was retrieved by protease treatment (8 min at 37°C), and the sections were further incubated at 37°C for 1 hour with a prediluted, ready-to-use mouse mAb to EGFR (clone 3C6, Ventana). The antigen-antibody reaction was visualized by ultraView DAB reveal system in a BenchMark XT automated IHC stainer (all from Ventana). For HER3, the antigen retrieval was performed by heating the sections at 97°C for 20 min in EnVision Target Retrieval Solution, High pH (Dako) in PT Link apparatus (Dako). The tissues were then incubated at 37°C for 2 hours with a mouse mAb to HER3 (clone DAK-H3-IC, Dako) diluted 1:50. The antigen-antibody reaction was revealed using EnVision FLEX DAB system in a Dako Autostainer Plus automate. For each patient, the pre- and posttreatment tumor samples were run together. IHC staining was interpreted by an expert pathologist who was blind to patient information. Both EGFR and HER3 abundances were quantified using an arbitrary scale having 0, 0.5, 1, 1.5, 2, 2.5, and 3 as measures of increasing staining intensity. EGFR and HER3 histoscores were defined as a sum of products obtained by multiplying the staining intensity with the percentage of stained cells. Patient samples For PDX establishment, fresh tissue was obtained from the Massachusetts General Hospital under Institutional Review Board approval and patient’s informed consent. Triple-negative status was determined by the Massachu- setts General Hospital Clinical Laboratory and Department of Pathology. Formalin-fixed paraffin-embedded (FFPE) specimens for IHC analyses of EGFR and HER3 abundance were obtained from the institutions par- ticipating in two French multicenter pilot phase 2 neoadjuvant trials that tested the efficacy of an anti-EGFR antibody combined to chemotherapy in TNBC stage II to IIIA patients (NCT00933517 and NCT00600249). pCR was the primary endpoint (with clinical response and toxicity as secondary endpoints), for which 47 and 29 patients were evaluated in the panitumumab and cetuximab trials, respectively (table S2). Tumor tissue samples were sys- tematically collected before and at the end of the neoadjuvant treatment at the Jean Perrin Comprehensive Cancer Center, where molecular and pathologi- cal analyses were performed. pCR was evaluated using Chevallier et al.’s (34) and Sataloff et al.’s (35) classifications. Fluorescence resonance energy transfer To monitor FRET between EGFR and HER3, we used fluorescence life- time imaging microscopy (FLIM), which is a gold standard technique for measuring protein proximity within the typically less than 10-nm range (36–38), and which we pioneered in its application to FFPE cancer sam- ples (39, 40). Two consecutive slices were placed on the same glass slide, and antigen retrieval was performed with the Ventana BenchMark ULTRA system according to the manufacturer’s instructions. One slice was stained for EGFR alone [detected by an Alexa Fluor 546–conjugated mAb to EGFR (F4), from the Cancer Research UK (CRUK) repository], and the second slice for EGFR and HER3 [detected by a Cy5-conjugated mAb to HER3 (2F12), Thermo Scientific]. Antibodies were directly labeled according to the manufacturer’s protocol with Alexa Fluor 546 and Cy5, respec- tively. Both antibodies were shown to be specific in either cells overex- pressing untagged EGFR (plasmid provided by A. Reynolds, Tumor Angiogenesis Group, The Breakthrough Breast Cancer Research Centre, London) or enhanced green fluorescent protein–tagged HER3 (a gift from S. Roberts, Gray Cancer Institute, Mount Vernon Hospital), or in FFPE sections (fig. S11). Samples were imaged on an “open” micro- scope automated FLIM system (41). Image analysis was done using newly developed algorithm to create the lifetime filter to eliminate auto- fluorescence, so any lifetime reduction on the masked tumor image will indicate true FRET (42). R E S E A R C H A R T I C L E www.SCIENCESIGNALING.org 25 March 2014 Vol 7 Issue 318 ra29 7 on February 3, 2016 http://stke.sciencemag.org/ Downloaded from    166         Statistical analysis Two-way t tests were performed with GraphPad Prism (GraphPad Software). Error bars represent the SEM, and P values are indicated by *P < 0.05 and **P < 0.01. All cellular experiments were repeated at least three times. All the in vivo experiments were run with at least seven mice for each treatment arm. Statistical analysis on data related to the French clin- ical trial samples was performed with Microsoft Excel and Statistics Epi- demiology Medicine, a biomedical statistical analysis software created by Kwiatkowski et al. (43). SUPPLEMENTARY MATERIALS www.sciencesignaling.org/cgi/content/full/7/318/ra29/DC1 Fig. S1. EGFR and PTEN abundance in HCC70 and MDA-MB-468 xenografts. Fig. S2. Growth of HCC70 and MDA-MB-468 xenografts treated with MEHD7945A and PI3K-Akt pathway inhibitors. Fig. S3. CEER analysis of HCC70 xenografts. Fig. S4. Immunostaining for EGFR, PTEN, and Ki67 in patient-derived TNBC xenografts. Fig. S5. Imaging of EGFR and HER3 in patient-derived TNBC treated with GDC-0068. Fig. S6. Ki67 staining of tumors and CTCs from PDXs treated with MEHD7945A and PI3K-Akt pathway inhibitors. Fig. S7. HER3 knockdown in MDA-MB-468 cells treated with GDC-0068 or GDC-0941. Fig. S8. Tumor growth of HCC70 and PDXs treated with cetuximab and GDC-0068. Fig. S9. Correlation between EGFR mRNA expression after cetuximab treatment and overall survival in TNBC metastatic patients. Fig. S10. Quantification of HER3 abundance in paired samples from patients treated with cetuximab. Fig. S11. Antibody specificity for FRET analysis. Table S1. CEER values (single replicas). Table S2. Patient clinicopathological information. REFERENCES AND NOTES 1. S. P. Shah, A. Roth, R. Goya, A. Oloumi, G. Ha, Y. Zhao, G. Turashvili, J. Ding, K. Tse, G. Haffari, A. Bashashati, L. M. Prentice, J. Khattra, A. Burleigh, D. Yap, V. Bernard, A. McPherson, K. Shumansky, A. Crisan, R. Giuliany, A. Heravi-Moussavi, J. Rosner, D. Lai, I. Birol, R. Varhol, A. Tam, N. Dhalla, T. Zeng, K. Ma, S. K. Chan, M. Griffith, A. Moradian, S.W. Cheng, G. B. Morin, P.Watson, K. Gelmon, S. Chia, S. F. Chin, C. Curtis, O. M. Rueda, P. D. Pharoah, S. Damaraju, J. Mackey, K. Hoon, T. Harkins, V. Tadigotla, M.Sigaroudinia,P.Gascard,T.Tlsty, J.F.Costello, I.M.Meyer,C.J.Eaves,W.W.Wasserman, S. Jones, D. Huntsman, M. Hirst, C. Caldas, M. A. Marra, S. Aparicio, The clonal and mu- tational evolution spectrum of primary triple-negative breast cancers.Nature 486, 395–399 (2012). 2. W. D. Foulkes, I. E. Smith, J. S. Reis-Filho, Triple-negative breast cancer. N. Engl. J. Med. 363, 1938–1948 (2010). 3. R. Dent, W. M. Hanna, M. Trudeau, E. Rawlinson, P. Sun, S. A. Narod, Pattern of meta- static spread in triple-negative breast cancer. Breast Cancer Res. Treat. 115, 423–428 (2009). 4. F. André, C. C. Zielinski, Optimal strategies for the treatment of metastatic triple-negative breast cancer with currently approved agents. Ann. Oncol. 23 (Suppl. 6), vi46–vi51 (2012). 5. L. A. Carey, Directed therapy of subtypes of triple-negative breast cancer. Oncologist 16 (Suppl. 1), 71–78 (2011). 6. C. Liedtke, C. Mazouni, K. R. Hess, F. André, A. Tordai, J. A. Mejia, W. F. Symmans, A. M. Gonzalez-Angulo, B. Hennessy, M. Green, M. Cristofanilli, G. N. Hortobagyi, L. Pusztai, Response to neoadjuvant therapy and long-term survival in patients with triple-negative breast cancer. J. Clin. Oncol. 26, 1275–1281 (2008). 7. G. von Minckwitz, M. Martin, Neoadjuvant treatments for triple-negative breast cancer (TNBC). Ann. Oncol. 23 (Suppl. 6), vi35–vi39 (2012). 8. B. Marty, V. Maire, E. Gravier, G. Rigaill, A. Vincent-Salomon, M. Kappler, I. Lebigot, F. Djelti, A. Tourdès, P. Gestraud, P. Hupé, E. Barillot, F. Cruzalegui, G. C. Tucker, M. H. Stern, J. P. Thiery, J. A. Hickman, T. Dubois, Frequent PTEN genomic alterations and activated phosphatidylinositol 3-kinase pathway in basal-like breast cancer cells. Breast Cancer Res. 10, R101 (2008). 9. T. O. Nielsen, F. D. Hsu, K. Jensen, M. Cheang, G. Karaca, Z. Hu, T. Hernandez-Boussard, C. Livasy, D. Cowan, L. Dressler, L. A. Akslen, J. Ragaz, A. M. Gown, C. B. Gilks, M. van de Rijn, C. M. Perou, Immunohistochemical and clinical characterization of the basal-like subtype of invasive breast carcinoma. Clin. Cancer Res. 10, 5367–5374 (2004). 10. L. Carey, E. Winer, G. Viale, D. Cameron, L. Gianni, Triple-negative breast cancer: Disease entity or title of convenience? Nat. Rev. Clin. Oncol. 7, 683–692 (2010). 11. J. Baselga, P. Gomez, A. Awada, R. Greil, S. Braga, M. Climent, A. Wardley, A. Zubel, J. Groos, B. Kaufmann, The addition of cetuximab to cisplatin increases overall re- sponse rate (ORR) and progression-free survival (PFS) in metastatic triple-negative breast cancer (TNBC): Results of a randomized phase II study (BALI-1), paper presented at the 35th Congress of the European Society for Medical Oncology, Milan, Italy, 2010. 12. L. A. Carey, H. S. Rugo, P. K. Marcom, E. L. Mayer, F. J. Esteva, C. X. Ma, M. C. Liu, A. M. Storniolo, M. F. Rimawi, A. Forero-Torres, A. C. Wolff, T. J. Hobday, A. Ivanova, W. K. Chiu, M. Ferraro, E. Burrows, P. S. Bernard, K. A. Hoadley, C. M. Perou, E. P. Winer, TBCRC 001: Randomized phase II study of cetuximab in combination with carboplatin in stage IV triple-negative breast cancer. J. Clin. Oncol. 30, 2615–2623 (2012). 13. J. C. Bendell, J. Rodon, H. A. Burris, M. de Jonge, J. Verweij, D. Birle, D. Demanse, S. S. De Buck, Q. C. Ru, M. Peters, M. Goldbrunner, J. Baselga, Phase I, dose-escalation study of BKM120, an oral pan-class I PI3K inhibitor, in patients with advanced solid tumors. J. Clin. Oncol. 30, 282–290 (2012). 14. S. Chandarlapaty, A. Sawai, M. Scaltriti, V. Rodrik-Outmezguine, O. Grbovic-Huezo, V. Serra, P. K. Majumder, J. Baselga, N. Rosen, AKT inhibition relieves feedback sup- pression of receptor tyrosine kinase expression and activity. Cancer Cell 19, 58–71 (2011). 15. V. S. Rodrik-Outmezguine, S. Chandarlapaty, N. C. Pagano, P. I. Poulikakos, M. Scaltriti, E. Moskatel, J. Baselga, S. Guichard, N. Rosen, mTOR kinase inhibition causes feedback- dependent biphasic regulation of AKT signaling. Cancer Discov. 1, 248–259 (2011). 16. V. Serra, M. Scaltriti, L. Prudkin, P. J. Eichhorn, Y. H. Ibrahim, S. Chandarlapaty, B. Markman, O. Rodriguez, M. Guzman, S. Rodriguez, M. Gili, M. Russillo, J. L. Parra, S. Singh, J. Arribas, N. Rosen, J. Baselga, PI3K inhibition results in enhanced HER signaling and acquired ERK dependency in HER2-overexpressing breast cancer. Oncogene 30, 2547–2557 (2011). 17. N. V. Sergina, M. Rausch, D. Wang, J. Blair, B. Hann, K. M. Shokat, M. M. Moasser, Escape from HER-family tyrosine kinase inhibitor therapy by the kinase-inactive HER3. Nature 445, 437–441 (2007). 18. D. L. Wheeler, S. Huang, T. J. Kruser, M. M. Nechrebecki, E. A. Armstrong, S. Benavente, V. Gondi, K. T. Hsu, P. M. Harari, Mechanisms of acquired resistance to cetuximab: Role of HER (ErbB) family members. Oncogene 27, 3944–3956 (2008). 19. M. M. Morrison, K. Hutchinson, M. M. Williams, J. C. Stanford, J. M. Balko, C. Young, M. G. Kuba, V. Sánchez, A. J. Williams, D. J. Hicks, C. L. Arteaga, A. Prat, C. M. Perou, H. S. Earp, S. Massarweh, R. S. Cook, ErbB3 downregulation enhances luminal breast tumor response to antiestrogens. J. Clin. Invest. 123, 4329–4343 (2013). 20. D. C. Kirouac, J. Y. Du, J. Lahdenranta, R. Overland, D. Yarar, V. Paragas, E. Pace, C. F. McDonagh, U. B. Nielsen, M. D. Onsum, Computational modeling of ERBB2- amplified breast cancer identifies combined ErbB2/3 blockade as superior to the com- bination of MEK and AKT inhibitors. Sci. Signal. 6, ra68 (2013). 21. T. Okuzumi, D. Fiedler, C. Zhang, D. C. Gray, B. Aizenstein, R. Hoffman, K. M. Shokat, Inhibitor hijacking of Akt activation. Nat. Chem. Biol. 5, 484–493 (2009). 22. A. J. Folkes, K. Ahmadi, W. K. Alderton, S. Alix, S. J. Baker, G. Box, I. S. Chuckowree, P. A. Clarke, P. Depledge, S. A. Eccles, L. S. Friedman, A. Hayes, T. C. Hancox, A. Kugendradas, L. Lensun, P. Moore, A. G. Olivero, J. Pang, S. Patel, G. H. Pergl-Wilson, F. I. Raynaud, A. Robson, N. Saghir, L. Salphati, S. Sohal, M. H. Ultsch, M. Valenti, H. J. Wallweber, N. C. Wan, C. Wiesmann, P. Workman, A. Zhyvoloup, M. J. Zvelebil, S. J. Shuttleworth, The identification of 2-(1H-indazol-4-yl)-6-(4-methanesulfonyl-piperazin- 1-ylmethyl)-4-morpholin-4-yl-thieno[3,2-d]pyrimidine (GDC-0941) as a potent, selective, orally bioavailable inhibitor of class I PI3 kinase for the treatment of cancer. J. Med. Chem. 51, 5522–5532 (2008). 23. G. Schaefer, L. Haber, L. M. Crocker, S. Shia, L. Shao, D. Dowbenko, K. Totpal, A. Wong, C. V. Lee, S. Stawicki, R. Clark, C. Fields, G. D. Lewis Phillips, R. A. Prell, D. M. Danilenko, Y. Franke, J.-P. Stephan, J. Hwang, Y. Wu, J. Bostrom, M. X. Sliwkowski, G. Fuh, C. Eigenbrot, A two-in-one antibody against HER3 and EGFR has superior inhibitory activity compared with monospecific antibodies. Cancer Cell 20, 472–486 (2011). 24. R. Krumbach, J. Schüler, M. Hofmann, T. Giesemann, H. H. Fiebig, T. Beckers, Primary resistance to cetuximab in a panel of patient-derived tumour xenograft models: Acti- vation of MET as one mechanism for drug resistance. Eur. J. Cancer 47, 1231–1243 (2011). 25. J. T. Garrett, M. G. Olivares, C. Rinehart, N. D. Granja-Ingram, V. Sánchez, A. Chakrabarty, B. Dave, R. S. Cook, W. Pao, E. McKinely, H. C. Manning, J. Chang, C. L. Arteaga, Tran- scriptional and posttranslational up-regulation of HER3 (ErbB3) compensates for in- hibition of the HER2 tyrosine kinase. Proc. Natl. Acad. Sci. U.S.A. 108, 5021–5026 (2011). 26. Y. Yan, V. Serra, L. Prudkin, M. Scaltriti, S. Murli, O. Rodríguez, M. Guzman, D. Sampath, M. Nannini, Y. Xiao, M. C.Wagle, J.Wu, G. M. Hampton, M.Wongchenko, V. Ramakrishnan, M. R. Lackner, C. Saura, D. Roda, A. Cervantes, J. Tabernero, P. Patel, J. Baselga, Eval- uation and clinical analyses of downstream targets of the Akt inhibitor GDC-0068. Clin. Cancer Res. 19, 6976–6986 (2013). 27. C. García-García, Y. H. Ibrahim, V. Serra, M. T. Calvo, M. Guzmán, J. Grueso, C. Aura, J. Pérez, K. Jessen, Y. Liu, C. Rommel, J. Tabernero, J. Baselga, M. Scaltriti, Dual mTORC1/2 and HER2 blockade results in antitumor activity in preclinical models of breast cancer resistant to anti-HER2 therapy. Clin. Cancer Res. 18, 2603–2612 (2012). 28. J. P. Holland, Y. Sheh, J. S. Lewis, Standardized methods for the production of high specific-activity zirconium-89. Nucl. Med. Biol. 36, 729–739 (2009). R E S E A R C H A R T I C L E www.SCIENCESIGNALING.org 25 March 2014 Vol 7 Issue 318 ra29 8 on February 3, 2016 http://stke.sciencemag.org/ Downloaded from   167         29. N. T. Viola-Villegas, S. L. Rice, S. Carlin, X. Wu, M. J. Evans, K. K. Sevak, M. Drobjnak, G. Ragupathi, R. Sawada, W. W. Scholz, P. O. Livingston, J. S. Lewis, Applying PET to broaden the diagnostic utility of the clinically validated CA19.9 serum biomarker for oncology. J. Nucl. Med. 54, 1876–1882 (2013). 30. M. Elkabets, S. Vora, D. Juric, N. Morse, M. Mino-Kenudson, T. Muranen, J. Tao, A. B. Campos, J. Rodon, Y. H. Ibrahim, V. Serra, V. Rodrik-Outmezguine, S. Hazra, S. Singh, P. Kim, C. Quadt, M. Liu, A. Huang, N. Rosen, J. A. Engelman, M. Scaltriti, J. Baselga, mTORC1 inhibition is required for sensitivity to PI3K p110a inhibitors in PIK3CA-mutant breast cancer. Sci. Transl. Med. 5, 196ra199 (2013). 31. P. Kim, X. Liu, T. Lee, L. Liu, R. Barham, R. Kirkland, G. Leesman, A. Kuller, B. Ybarrondo, S. C. Ng, S. Singh, Highly sensitive proximity mediated immunoassay reveals HER2 status conversion in the circulating tumor cells of metastatic breast cancer patients. Proteome Sci. 9, 75 (2011). 32. S. L. Stott, C. H. Hsu, D. I. Tsukrov, M. Yu, D. T. Miyamoto, B. A.Waltman, S. M. Rothenberg, A. M. Shah, M. E. Smas, G. K. Korir, F. P. Floyd, A. J. Gilman, J. B. Lord, D. Winokur, S. Springer, D. Irimia, S. Nagrath, L. V. Sequist, R. J. Lee, K. J. Isselbacher, S. Maheswaran, D. A. Haber, M. Toner, Isolation of circulating tumor cells using a microvortex-generating herringbone-chip. Proc. Natl. Acad. Sci. U.S.A. 107, 18392–18397 (2010). 33. M. Yu, A. Bardia, B. S. Wittner, S. L. Stott, M. E. Smas, D. T. Ting, S. J. Isakoff, J. C. Ciciliano, M. N. Wells, A. M. Shah, K. F. Concannon, M. C. Donaldson, L. V. Sequist, E. Brachtel, D. Sgroi, J. Baselga, S. Ramaswamy, M. Toner, D. A. Haber, S. Maheswaran, Circulating breast tumor cells exhibit dynamic changes in epithelial and mesenchymal composition. Science 339, 580–584 (2013). 34. B. Chevallier, H. Roche, J. P. Olivier, P. Chollet, P. Hurteloup, Inflammatory breast cancer. Pilot study of intensive induction chemotherapy (FEC-HD) results in a high his- tologic response rate. Am. J. Clin. Oncol. 16, 223–228 (1993). 35. D. M. Sataloff, B. A. Mason, A. J. Prestipino, U. L. Seinige, C. P. Lieber, Z. Baloch, Path- ologic response to induction chemotherapy in locally advanced carcinoma of the breast: A determinant of outcome. J. Am. Coll. Surg. 180, 297–306 (1995). 36. E. M. Bublil, G. Pines, G. Patel, G. Fruhwirth, T. Ng, Y. Yarden, Kinase-mediated quasi-dimers of EGFR. FASEB J. 24, 4744–4755 (2010). 37. J. R. Morris, C. Boutell, M. Keppler, R. Densham, D. Weekes, A. Alamshah, L. Butler, Y. Galanty, L. Pangon, T. Kiuchi, T. Ng, E. Solomon, The SUMO modification pathway is involved in the BRCA1 response to genotoxic stress. Nature 462, 886–890 (2009). 38. J. W. Legg, C. A. Lewis, M. Parsons, T. Ng, C. M. Isacke, A novel PKC-regulated mech- anism controls CD44–ezrin association and directional cell motility. Nat. Cell Biol. 4, 399–407 (2002). 39. T. Ng, A. Squire, G. Hansra, F. Bornancin, C. Prevostel, A. Hanby, W. Harris, D. Barnes, S. Schmidt, H. Mellor, P. I. H. Bastiaens, P. J. Parker, Imaging protein kinase Ca activa- tion in cells. Science 283, 2085–2089 (1999). 40. M. Parsons, T. Ng, Intracellular coupling of adhesion receptors: Molecular proximity measurements. Methods Cell Biol. 69, 261–278 (2002). 41. P. R. Barber, I. D. Tullis, G. P. Pierce, R. G. Newman, J. Prentice, M. I. Rowley, D. R.Matthews, S. M. Ameer-Beg, B. Vojnovic, The Gray Institute ‘open’ high-content, fluorescence lifetime microscopes. J. Microsc. 251, 154–167 (2013). 42. P. R. Barber, I. D. Tullis, M. I. Rowley, C. D. Martins, G. Weitsman, K. Lawler, M. Coffey, N. Woodman, C. E. Gillett, T. Ng, B. Vojnovic, The Gray Institute Open Microscopes ap- plied to radiobiology and protein interaction studies. Proc. SPIE, in press. 43. F. Kwiatkowski, M. Girard, K. Hacene, J. Berlie, Sem: A suitable statistical software adaptated for research in oncology. Bull. Cancer 87, 715–721 (2000). Acknowledgments:We are grateful to C. Abrial and F. Kwiatkowski for statistical support and the Translational Breast Cancer Research Consortium for the TBCRC 001 study. Funding: The clinical panitumumab study conducted in France was sponsored by Amgen and the Jean Perrin Comprehensive Cancer Center. Preclinical work was funded by Susan G. Komen for the Cure (SAC110046) and the Breast Cancer Research Foundation. King’s College London–University College London Comprehensive Cancer Imaging Centre funding was provided by CRUK and the Engineering and Physical Sciences Research Coun- cil in association with the Medical Research Council and Department of Health (England; C1519/ A10331) and an FP7-HEALTH-2010 European Union grant entitled “IMAGINT” (grant number 259881). M.E. is an International Sephardic Education Foundation postdoctoral fellow; A.B. holds a Translational Research Fellowship from the Spanish Society of Medical Oncology (SEOM); C.M.P. is funded by the National Cancer Institute Breast SPORE (Specialized Program of Research Excellence) program (P50-CA58223-09A1) and the Breast Cancer Re- search Foundation; and N. Aceto is a fellow of the Human Frontiers Science Program, the Swiss National Science Foundation, and the Swiss Foundation for Grants in Biology and Medicine. Author contributions: J.J.T., P.C., D.J., J.B., and M.S. designed the research. J.J.T., P.C., N. Auricchio, M.E., A. Bosch, H.E., N.M., and N. Aceto performed the experiments. N.R.-R. and F.P.-L. collected and analyzed the human specimens. S.H., S.S., and P.K. performed the CEER arrays and their analyses. S.M. assisted with the design of the research. G.W., P.B., B.V., and T.N. performed and analyzed the FRET experiments on patient samples. L.A.C. and C.M.P. provided the gene expression data, which were analyzed by A. Bergamaschi, J.S.L., and N.T.V.-V. A. Bergamaschi, J.S.L., and N.T.V.-V. produced and analyzed the in vivo imaging data. N.R.-R., J.J.T., and M.E. prepared the figures and analyzed the cellular and in vivo data. J.J.T., J.B., and M.S. wrote the manuscript. Competing interests: S.H., S.S., and P.K. are employees of Prometheus Therapeutics & Diagnostics. M.S. is a consultant for Prometheus Therapeutics & Diagnostics. J.B. had consultant and advisory roles at Genentech. C.M.P. is an equity stock holder and a member of the Board of Directors of BioClassifier LLC and University Genomics. Submitted 27 January 2014 Accepted 18 February 2014 Final Publication 25 March 2014 10.1126/scisignal.2005125 Citation: J. J. Tao, P. Castel, N. Radosevic-Robin, M. Elkabets, N. Auricchio, N. Aceto, G. Weitsman, P. Barber, B. Vojnovic, H. Ellis, N. Morse, N. T. Viola-Villegas, A. Bosch, D. Juric, S. Hazra, S. Singh, P. Kim, A. Bergamaschi, S. Maheswaran, T. Ng, F. Penault-Llorca, J. S. Lewis, L. A. Carey, C. M. Perou, J. Baselga, M. Scaltriti, Antagonism of EGFR and HER3 enhances the response to inhibitors of the PI3K-Akt pathway in triple-negative breast cancer. Sci. Signal. 7, ra29 (2014). R E S E A R C H A R T I C L E www.SCIENCESIGNALING.org 25 March 2014 Vol 7 Issue 318 ra29 9 on February 3, 2016 http://stke.sciencemag.org/ Downloaded from    168                                                                 169 ARTICLE 5     Title: PI3K inhibition results in enhanced estrogen receptor function and dependence in hormone receptor-positive breast cancer Journal: Science Translational Medicine Impact Factor: 15.843    170   171 CANCER PI3K inhibition results in enhanced estrogen receptor function and dependence in hormone receptor–positive breast cancer Ana Bosch,1* Zhiqiang Li,1 Anna Bergamaschi,2 Haley Ellis,1 Eneda Toska,1 Aleix Prat,3,4 Jessica J. Tao,5 Daniel E. Spratt,6 Nerissa T. Viola-Villegas,6† Pau Castel,1 Gerard Minuesa,7 Natasha Morse,1 Jordi Rodón,8,9 Yasir Ibrahim,10 Javier Cortes,8 Jose Perez-Garcia,8 Patricia Galvan,3 Judit Grueso,10 Marta Guzman,10 John A. Katzenellenbogen,11 Michael Kharas,7 Jason S. Lewis,6,7 Maura Dickler,12 Violeta Serra,10 Neal Rosen,7 Sarat Chandarlapaty,1,12,13‡ Maurizio Scaltriti,1‡ José Baselga1,12,13‡ Activating mutations of PIK3CA are the most frequent genomic alterations in estrogen receptor (ER)–positive breast tumors, and selective phosphatidylinositol 3-kinase a (PI3Ka) inhibitors are in clinical development. The activity of these agents, however, is not homogeneous, and only a fraction of patients bearing PIK3CA-mutant ER-positive tumors benefit from single-agent administration. Searching for mechanisms of resistance, we observed that sup- pression of PI3K signaling results in induction of ER-dependent transcriptional activity, as demonstrated by changes in expression of genes containing ER-binding sites and increased occupancy by the ER of promoter regions of up- regulated genes. Furthermore, expression of ESR1 mRNA and ER protein were also increased upon PI3K inhibition. These changes in gene expression were confirmed in vivo in xenografts and patient-derived models and in tumors from patients undergoing treatment with the PI3Ka inhibitor BYL719. The observed effects on transcription were enhanced by the addition of estradiol and suppressed by the anti-ER therapies fulvestrant and tamoxifen. Fulvestrant markedly sensitized ER-positive tumors to PI3Ka inhibition, resulting in major tumor regressions in vivo. We propose that increased ER transcriptional activity may be a reactive mechanism that limits the activity of PI3K inhibitors and that combined PI3K and ER inhibition is a rational approach to target these tumors. INTRODUCTION The phosphatidylinositol 3-kinase (PI3K) pathway is essential for cell growth, proliferation, survival, and metabolism (1, 2). The PI3K family of enzymes is divided into three main classes (I to III), with class I being the most often implicated in human cancer (3). Members of the class IA PI3K are characterized by a heterodimer composed of a cat- alytic subunit (p110a, b, and d) and a regulatory subunit (p85) (4, 5). PIK3CA, the gene coding for p110a, is frequently mutated in human cancers (6, 7). In particular, hotspot mutations of this gene that reside in the helical (E542K and E545K) or catalytic (H1047R) domains are found in over a third of estrogen receptor (ER)–positive breast cancer, representing the most common genomic alteration in this group of tumors (7, 8). Selective PI3K p110a (PI3Ka) inhibitors are currently being tested in the clinic in patients with advanced malignancies, with promising results in patients with breast tumors harboring PIK3CA mutations (9, 10). However, not all the patients benefit equally from these agents, and even those that initially respond typically relapse after months of therapy. Although we have recently reported that the emergence of resistant clones with genomic alterations that activate PI3Kb may partially explain acquired resistance to PI3Ka inhibitors (11), alternative mecha- nisms may also be at play in primary or early resistance to these thera- pies. Among them, activation of alternative cellular compensatory pathways could explain primary resistance or the emergence of rapid resistance. For example, we have shown that pharmacological suppres- sion of mammalian target of rapamycin (mTOR), which is downstream from PI3K and a central node within the PI3K/AKT/mTOR axis, results in activation of both AKT (12) and extracellular signal–regulated ki- nases (ERK) (13) and can account for decreased efficacy of mTOR inhibitors. Similarly, inhibition of PI3K leads to compensatory activa- tion of upstream receptor tyrosine kinases that limit the effectiveness of these compounds (14, 15). Given that the vast majority of PIK3CA-mutant tumors are ER- positive, it is plausible to hypothesize that both pathways can drive proliferation and survival in these cells. A tangible evidence that the PI3K and ER pathways can cooperate in tumor progression came from a clinical study (16) that showed an impressive improvement 1Human Oncology and Pathogenesis Program and Memorial Sloan Kettering Cancer Center, 1275 York Avenue, Box 20, New York, NY 10065, USA. 2Department of Molecular and Integrative Physiology, University of Illinois at Urbana-Champaign, 524 Burrill Hall, Urbana, IL 61801, USA. 3Translational Genomics Group, Vall d’Hebron Institute of On- cology (VHIO), Passeig Vall d’Hebron 119-129, Barcelona 08035, Spain. 4Translational Ge- nomics and Targeted Therapeutics in Solid Tumors, August Pi i Sunyer Biomedical Research Institute, Hospital Clinic Barcelona, C/Rosselló 149-153, Barcelona 08035, Spain. 5Massachusetts General Hospital Cancer Center and Harvard Medical School, 425 13th Street, Charlestown, MA 02129, USA. 6Department of Radiology, Memorial Sloan Kettering Cancer Center, New York, NY 10065, USA. 7Molecular Pharmacology and Chemistry Program and Center for Cell Engineering, Memorial Sloan Kettering Cancer Center, New York, NY 10065, USA. 8Department of Medical Oncology, VHIO, Barcelona 08035, Spain. 9Universitat Autònoma de Barcelona, Plaza Cívica, Campus UAB, 08193 Bellaterra, Spain. 10Experimental Therapeutics Group, VHIO, Barcelona 08035, Spain. 11Department of Chemistry, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA. 12Breast Medicine Service, Department of Medicine, Memorial Sloan Kettering Cancer Center, New York, NY 10065, USA. 13Weill Cornell Medical College, New York, NY 10065, USA. *Present address: Division of Oncology and Pathology, Department of Clinical Sciences, Lund University, Medicon Village Building 404:C2, Scheelevägen 2, SE-223 81 Lund, Sweden. †Present address: Department of Oncology, Karmanos Cancer Institute, 4100 John R. Street, Detroit, MI 48201, USA. ‡Corresponding author. E-mail: chandars@mskcc.org (S.C.); scaltrim@mskcc.org (M.S.); baselgaj@mskcc.org (J.B.) R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 15 April 2015 Vol 7 Issue 283 283ra51 1 on February 2, 2016 http://stm.sciencemag.org/ Downloaded from    172 in progression-free survival in ER-positive breast cancer patients treated with the mTOR inhibitor everolimus in combination with the anti- estrogen aromatase inhibitor exemestane. These patients had failed previous endocrine therapy, and considering that activity of single- agent mTOR inhibitors is minimal, these results suggest a synergis- tic activity in targeting mTOR and ER signaling simultaneously. Nevertheless, little is known about the reciprocal regulation of these key pathways, although it has been described that chronic anti-estrogen therapy induces the activation of the PI3K pathway in vitro (17, 18). Perhaps more relevant to our interest in understanding the effects of PI3K inhibition on ER signaling, studies in prostate cancer have con- vincingly shown that the PI3K pathway regulates androgen receptor activity (19). On the basis of all these observations, our work was aimed at study- ing the effects of PI3Ka inhibition on ER signaling and at deciphering its potential effect on limiting the efficacy of PI3K inhibitors. We report that inhibition of the PI3K pathway triggers the activation of the ER- dependent transcription machinery. The importance of this adaptive response is underscored by the finding that suppression of ER activity can sensitize tumors to PI3K inhibition. RESULTS PI3K inhibition promotes ER activity To obtain a comprehensive view of the role of PI3K activation in modulating ER function, we examined the effects of pharmacological PI3K inhibition in ER-positive breast cancer cells that harbor activat- ing PIK3CA mutations (MCF7-E545K and T47D-H1047R). Previous studies have established that a large portion of the transcriptome in these cells is regulated by ER activity (20), so they were a good model to explore whether PI3K inhibition had any effect on ER-regulated genes. Cells were treated with the highly selective PI3Ka inhibitor BYL719 at 1 mM for 4, 8, 12, 24, or 48 hours, and cellular lysates were analyzed. As measured by intensity of phosphorylation of AKT (S473) and S6 (S235/6), BYL719 caused potent inhibition of PI3K signaling as early as 4 hours after drug exposure and lasting for 48 hours, with a slight rebound in activity observed at the 48-hour time point (fig. S1). We observed that at this concentration, BYL719 also induced glob- al changes in the transcription profile of both MCF7 and T47D cells. Perturbations in their gene expression profiles were evident 12 hours after the addition of BYL719 to the culture medium and were sustained until at least 48 hours (Fig. 1A and fig. S2, respectively). Strikingly, of the 383 and 706 genes significantly altered upon BYL719 treatment in MCF7 and T47D cells [false discovery rate (FDR) ≤ 1%], respectively, up to 60% contained an estrogen-responsive element (ERE) in their promoter (Fig. 1B). Gene Set Enrichment Analysis (GSEA) was con- ducted on the genes altered by PI3K inhibition and demonstrated that gene sets characterized by ER dependence were highly enriched in this group (FDR ≤ 25%) (Fig. 1C, fig. S3, and table S1). On the basis of these findings, we next decided to confirm the relationship between PI3K activity and ER transcriptional activity by analyzing the effects of drug inhibition on promoters known to be stimulated by estrogen. First, we examined the effect of BYL719 treatment on expression of a luciferase transgene linked to a 3X-ERE promoter. Sixteen hours after treatment, an about twofold higher luminescence was detected in the BYL719-treated cells than in the dimethyl sulfoxide (DMSO)–treated cells (Fig. 1D). We next used polymerase chain reaction (PCR) to confirm that drug treatment had an effect on the expression of endogenous mRNAs known to be regulated by ER, including progesterone receptor (PGR), growth regulation by estrogen in breast cancer 1 (GREB1), and insulin- like growth factor–binding protein 4 (IGFBP4) (Fig. 1E). The increased expression was due, at least in part, to ER transcriptional activity, be- cause ChIP-qPCR (quantitative PCR) experiments showed a two- to threefold enhancement in occupancy by ER of the promoter regions of these up-regulated genes (Fig. 1F). Together, our findings demon- strate that inhibition of activated PI3K signaling present in PIK3CA- mutant ER-positive breast cancer cells is associated with an increase in the transcriptional function of the ER. In addition to PIK3CA mutations, there are other genetic mecha- nisms that also result in aberrant PI3K pathway signaling. Specifically, phosphatase and tensin homolog (PTEN) function is lost in a subset of ER-positive tumors (21). Unlike PIK3CA mutation, PTEN loss appears to activate phosphatidylinositol 3,4,5-trisphosphate (PIP3)/ AKT signaling preferentially through the p110b subunit, and PI3Ka- selective inhibitors do not inhibit PI3K/AKT signaling in these cells (22). Because AKT is downstream from PI3K, we used MK2206, a se- lective AKT inhibitor, in CAMA1 cells (PTENmut D92H) (fig. S4) to test whether inhibition of the PI3K pathway promoted ER activation in an ER-positive PTEN-mutant model. Similar to MCF7 and T47D cells treated with BYL719, suppression of AKT in CAMA1 cells re- sulted in an overall increased expression of ER-dependent genes and in the induction of an ER-dependent signature (fig. S5 and table S2). The activation of ER as a consequence of PI3K pathway inhibition in the context of PTENnull/mut tumors was further corroborated by assess- ing the effect of the AKT inhibitor MK2206 on PGR and GREB1 ex- pression by qPCR in CAMA1, ZR-75-1, and MDA-MB-415 cells (fig. S6). Overall, the findings observed in both PTEN- and PIK3CA-mutant ER-positive breast cancer models are similar and demonstrate an in- creased ER-dependent transcription activity upon inhibition of the PI3K pathway by different strategies. PI3K inhibition increases ER expression As mentioned above, we found that inhibition of PI3K promotes ER activity, as manifested by increases in ER binding to target promoters and increases in ER target gene expression. We speculated that ER expression itself might also be increased in response to PI3K inhibi- tion and might partially explain the increases in ER activity. We ex- amined ESR1mRNA expression in a panel of ER-positive breast cancer cell lines (Fig. 2, A and B, and fig. S7) and found increases ranging from 1.5× to 3× upon PI3K inhibition with BYL719, with maximal accumu- lation observed at 24 hours in the MCF7 model. Coinciding with the initial rise in mRNA (8 hours), RNA polymerase II binding to the ESR1 promoter was observed to be increased about twofold by ChIP assay (fig. S8). A similar induction of the ESR1 transcript was observed with various PI3K inhibitors (GDC0032, GDC0941, BAY80-6946, and BKM120) in ER-positive/PIK3CA-mutant models (fig. S9). The increase in ESR1mRNA coincided with increases in the ER protein, which was also maximal at about 24 hours over a 48-hour time course (Fig. 2B). Comparing the effects of PI3K inhibition with mTORC1 (mTOR com- plex 1) inhibition on ESR1 regulation, we observed an appreciable increase in the expression of ESR1 and its target genes also during rapamycin treatment, albeit with a lower magnitude compared to BYL719 (fig. S10). Together, the data show an increase in ER expression as a result of PI3K inhibition in various cell culture systems. To determine whether R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 15 April 2015 Vol 7 Issue 283 283ra51 2 on February 2, 2016 http://stm.sciencemag.org/ Downloaded from   173 Fig. 1. PI3K inhibition promotes ER function. (A) MCF7 cells were treated with BYL719 (1 mM) over a period of 48 hours. RNA was iso- lated at specified time points, and expression microarray analysis was performed. Heat map represents genes whose expression differed sig- nificantly across different time points with an FDR ≤1%. Each of the columns under the experimental conditions represents one biological replicate. (B) PI3Ka inhibition leads to modu- lation of genes containing ER bind- ing sites (ERE). MCF7 cells were treated with BYL719 (1 mM), and gene ex- pression analysis was performed as described in (A). The diagram repre- sents the genes that were differen- tially regulated upon treatment across all the time points [Significance Anal- ysis of Microarrays (SAM) analysis FDR ≤ 1%] and the percentage of these genes that contained an ER- binding element [defined by ER ChIP (chromatin immunoprecipitation) sequencing (40)]. (C) GSEA was per- formed to determine which gene sets were enriched in our data set (FDR ≤ 25%). Graph represents enrichment for ER-associated signature as de- scribed in (41). ES, enrichment score; NES, normalized enrichment score. (D) MCF7 cells were transfected with 3X-ERE-TATA firefly-luciferase and pRL-TK Renilla luciferase plasmids and treated with vehicle (Ctrl) or BYL719 (BYL) (1 mM) for 16 hours. Results represent firefly luciferase ac- tivity measured by luminescence and normalized both to Renilla luciferase luminescence for transfection effi- ciency and to Ctrl. Two-tailed Stu- dent’s unpaired t test was performed to compare Ctrl versus BYL719-treated cells. (E) MCF7 cells were treated with BYL719 (1 mM) over a period of 48 hours, and RNA was isolated at the indicated times. qPCR was performed to detect bACTIN, PGR, GREB1, and IGFBP4 gene expression. The data are presented relative to bACTIN and to expression in vehicle-treated cells (Ctrl). One-way analysis of vari- ance (ANOVA) statistical test was used to compare gene expression be- tween each time point and vehicle- treated cells, applying the Bonferroni method to correct for multiple com- parisons. Error bars denote SEM of at least two biological replicates, each with three technical replicates. (F) MCF7 cells were treated with BYL719 (1 mM) or vehicle (Ctrl), and ChIP was performed with anti-ERa antibody or control im- munoglobulin G (IgG). Primers to amplify the ER-binding regions of the PGR, GREB1, and IGFBP4 promoters were used in qPCR to determine fold enrich- ment relative to a noncoding region. Two-tailed Student’s unpaired t test was performed to compare Ctrl versus BYL719-treated cells. Error bars represent SEM of three independent experiments. R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 15 April 2015 Vol 7 Issue 283 283ra51 3 on February 2, 2016 http://stm.sciencemag.org/ Downloaded from    174 Fig. 2. PI3K inhibition induces ER expression. (A) MCF7 cells were treated with BYL719 (1mM)over a 48-hour period, andRNAwas isolated at the indicated times. qPCR was performed to detect bACTIN and ESR1 expression. The data are presented relative to bACTIN and to expression of ESR1 in vehicle-treated control (Ctrl). One-way ANOVA statistical test was used to compare gene expression be- tween each time point and to vehicle-treated cells, applying the Bonferroni method to correct for multiple comparisons. Error bars denote SEM of at least twobiological replicates, eachwith three technical replicates. (B)MCF7 cellswere treated with BYL719 (1 mM) over a period of 48 hours, and total protein was isolatedat the indicated times. Immunoblottingwasperformed todetect expres- sion of ER, phosphorylation of AKT at Ser473 (pAKT S473), and bACTIN. Graph represents the fold change of total ERawith respect to bACTIN and to untreated samples (Ctrl) of two independent experiments. Statistical analysis was done using the one-way ANOVA statistical test with the Bonferroni method to correct for multiple comparisons. (C) 18F-FES uptake in T47D xenograft mouse models treated with vehicle or BYL719 daily. The uptake was measured after a 4-day treatment, 2 hours after the last dose, and is represented as % injected dose per gram of tumor tissue (% ID/g). Statistical analysis to compare 18F-FES uptake between theCtrl and theBYL719-treatedmicewasperformedbymeansof anon- parametric Kruskal-Wallis test. (D) Graphical representation of ESR1 transcript abundance in 20 paired breast cancer biopsies before (PRE) and on BYL719 treatment (ON) collected as part of two clinical trials with the p110a inhibitor BYL719. ESR1 was one of the 105 breast cancer–specific genes analyzed using the nCounter platform. (E) Graphical representation of the induction of a luminal A signature upon BYL719 treatment in the tumor samples used in (D). (F) MCF7 cells were treated with vehicle or BYL719 (1 mM) for 2 hours. ChIP was performedwith anti-FOXO3A antibody or control IgG. Primers to amplify the FOXO3A-binding regions of the ESR1 promoter were used in qPCR to determine fold enrichment relative to input. Two-tailed Student’s unpaired t test was per- formed to compare mean signal amplification between vehicle- and BYL719- treated samples. Error bars represent SEM of two independent experiments with three technical replicates each. (G)MCF7 cells were transfectedwith non- targeted siRNA (Ctrl) or FOXO3A siRNA. Forty-eight hours later, cellswere treated with vehicle or BYL719 (1 mM) for 24 hours. mRNA was isolated, and qPCR was performed to detect bACTIN, FOXO3A, and ESR1 expression. The data are presented relative to bACTIN and to expression in the samples treated with Ctrl siRNA and vehicle. One-way ANOVA statistical test was used to compare gene expression between each condition and Ctrl siRNA and vehicle-treated cells, ap- plying the Bonferronimethod to correct formultiple comparisons. Error bars de- note SEM of two independent experiments with three technical replicates each. R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 15 April 2015 Vol 7 Issue 283 283ra51 4 on February 2, 2016 http://stm.sciencemag.org/ Downloaded from   175 increases in ER expression and activity might be observed in vivo, we used a noninvasive probe of ER expression, 16a-18F-fluoro-17b- estradiol positron emission tomography (18F-FES-PET). This probe measures uptake of labeled estradiol (E2) as an indirect measure of ER expression. In T47D xenografts, we observed a selective increase in tumor uptake of 18F-FES in mice treated with BYL719 compared to those receiving vehicle (Fig. 2C). Although such differences could be due to other mechanisms, such as changes in tumor retention time or receptor affinity for E2, they are consistent with the findings of in- creased expression of the receptor observed in vitro (Fig. 2, A and B, and figs. S7 to S9). To assess the clinical relevance of our findings, we analyzed the gene expression profiles of tumor samples collected from patients treated with BYL719 as part of either the first-in-human clinical trial (9) or an ongoing clinical study testing the efficacy of BYL719 in combination with the aromatase inhibitors letrozole or exemestane (NCT01870505). Paired tumor biopsies were collected from patients before commencing BYL719 therapy and after a minimum of 14 days on treatment, be- tween 4 and 6 hours after the daily drug administration. Two patients were treated with BYL719 as a single agent, and eight patients were treated with the combination of BYL719 and an aromatase inhibitor. Table S3 shows the patients’ information, including breast tumor his- tology at diagnosis, PIK3CA mutation status, BYL719 dose, and the treatment in combination with BYL719 (if any). Notably, either as single agent or in combination, BYL719 was administered at clinically active doses in all cases (11, 23). The expression of 105 breast cancer–related genes, including the genes from the prediction analysis of microarray 50 (PAM50) intrinsic subtype predictor (24), was compared across the 20 paired biopsies. Thirty-nine genes were differentially expressed (FDR ≤ 25%; Table 1). Not surprisingly, proliferation-related genes such as MKI67, BIRC5, and CENPF were among the most highly down-regulated genes in the on-treatment samples, whereas the antiapoptotic genes BCL2 andMDM2 were up-regulated. Central to our work, ESR1 and its tar- get gene PGR were among the most highly induced genes upon PI3K inhibition (Table 1). The ESR1 transcript levels were shown to increase during PI3K inhibition in all but two patients evaluated (Fig. 2D). In accordance with this result, a switch from a non–luminal A phenotype [luminal B or human epidermal growth factor receptor 2 (HER2)–enriched] in the pretreatment sample to a luminal A subtype in the on-treatment sample was identified by the PAM50 subtype predictor in three (pa- tients 1, 4, and 6) of the six patients (patients 1, 3, 4, 6, 9, and 10) who showed the highest increase in ESR1 (table S3). There was a global shift in the transcriptional profile of the tumors toward a more lumi- nal A–like signature across all “on-treatment” samples (Fig. 2E). No- tably, the pretreatment sample from patient 2 was already identified as luminal A; however, we observed a higher expression of the luminal A signature in the on-treatment sample compared to the pretreatment sample. Furthermore, patient 9 presented a change in intrinsic subtype from basal-like in the pretreatment sample to luminal B in the on- treatment biopsy, which was accompanied by the highest increase in ESR1 expression (Fig. 2, D and E). Although these are small num- bers of patients, the data support the idea that ER mRNA increases upon PI3K inhibition, in conjunction with a more estrogen-dependent luminal A phenotype. In an attempt to identify the mechanisms responsible for the ob- served increase in ESR1 transcription upon PI3Ka inhibition, we hy- pothesized that the expression and/or activity of transcription factors Table 1. Differentially expressed genes upon BYL719 treatment in pa- tients with metastatic breast cancer. Paired biopsies of tumors from ER- positive breast cancer patients receiving BYL719 as part of a clinical trial were collected before and during treatment. RNA was extracted, and the expression of 105 breast cancer–specific genes was analyzed using the nCounter platform. Differentially expressed genes (FDR ≤ 25%) between on-treatment and pretreatment biopsies are shown. Gene ID Score (d) Fold change q (%) GRB7 2.177 1.273 0.000 BCL2 2.150 1.363 0.000 MDM2 2.046 1.277 0.000 CXXC5 1.770 1.277 11.478 PGR 1.560 1.359 14.667 ESR1 1.452 1.554 14.667 ACTR3B 1.390 1.176 14.667 SFRP1 1.289 1.266 17.742 ERBB2 1.209 1.144 17.742 SLC39A6 1.034 1.140 19.556 PHGDH 0.999 1.253 19.556 FOXA1 0.908 1.103 19.556 FGFR4 0.890 1.269 19.556 GPR160 0.874 1.134 19.556 FOXC1 0.653 1.157 22.564 MLPH 0.641 1.082 22.564 MAPT 0.622 1.124 22.564 BIRC5 −2.009 −1.219 11.478 MYBL2 −1.946 −1.226 11.478 EXO1 −1.544 −1.190 11.478 CENPF −1.424 −1.202 11.478 CEP55 −1.280 −1.172 11.478 TYMS −1.276 −1.130 11.478 RRM2 −1.265 −1.208 11.478 CDH3 −1.261 −1.317 11.478 MKI67 −1.227 −1.148 11.478 MELK −1.164 −1.149 11.478 CCNE1 −1.155 −1.121 11.478 UBE2T −1.137 −1.131 11.478 KIF2C −1.111 −1.164 11.478 KNTC2 −1.032 −1.104 11.478 CDCA1 −0.997 −1.130 11.478 ORC6L −0.988 −1.136 11.478 CDC6 −0.924 −1.143 11.478 CDC20 −0.886 −1.106 11.478 CCNB1 −0.871 −1.093 11.478 PTTG1 −0.759 −1.098 14.667 ANLN −0.673 −1.096 17.742 MMP11 −0.619 −1.114 17.742 R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 15 April 2015 Vol 7 Issue 283 283ra51 5 on February 2, 2016 http://stm.sciencemag.org/ Downloaded from    176 known to bind the promoter of ESR1 were augmented under these conditions. Hence, we investigated the possible role of forkhead box O3 (FOXO3A), a transcription factor known to be regulated by PI3K/AKT signaling, in regulating ER expression. We found that in- hibition of PI3K with BYL719 led to a >5× accumulation of FOXO3A at the ESR1 promoter by ChIP assay (Fig. 2F). To determine whether the induction of FOXO3A binding was necessary for the increase in ESR1 mRNA, we used small interfering RNA (siRNA) to knock down FOXO3A in MCF7 cells and found that the loss of FOXO3A prevented the induction of ESR1 caused by BYL719 (Fig. 2G). These data suggest that FOXO3A is necessary for the effect of PI3K inhibition upon ESR1 expression. PI3K inhibition–induced ER activity is enhanced by the presence of ligand The effects of ER on breast cancer progression are thought to be dependent on both ligand-dependent and ligand-independent mecha- nisms (25, 26). Because these different modes of activation have im- plications for the optimal means of pharmacologic inhibition of ER, we investigated whether the induction of ER activity required the pres- ence of E2. MCF7 cells were grown in medium depleted of steroidal hormones (charcoal-stripped serum) for 48 hours and then treated with DMSO (vehicle), BYL719, E2, or the combination of BYL719 and E2 for 4 hours, and ChIP assays were performed. As expected, the addition of E2 increased ER binding to the PGR and GREB1 pro- moters about threefold compared to control cells not exposed to E2. In the presence of E2, 1 mM BYL719 caused further enhancement of binding two- to threefold over E2 alone (Fig. 3, A and B). However, in the absence of E2, BYL719 caused a much smaller enhancement of binding (1.2× to 1.5×) over untreated cells. Similarly, when we examined the effect of PI3K inhibition on transcription of PGR and GREB1, we again found that addition of E2 or the combination of E2 and BYL719 promoted PGR and GREB1 accumulation. However, BYL719 treatment in the absence of E2 caused very little increase in PGR or GREB1 after 0 to 16 hours of drug exposure (Fig. 3C). We next analyzed the effects of the ER antagonists 4-hydroxytamoxifen (4-OHT) and fulvestrant by qPCR. Under normal serum conditions, both 4-OHT (1 mM) and fulvestrant (100 nM) attenuated the expres- sion of ER target genes with no impact on ESR1mRNA (fig. S11). The same inhibitors were sufficient to prevent BYL719-mediated increased expression of the four tested ER target genes (Fig. 3D and fig. S12). These data suggest that blockade of estrogen function mitigates the effects mediated by PI3K inhibition. PI3K inhibition in combination with ER inhibitor fulvestrant has profound antitumor activity in ER/PIK3CAmut models Because fulvestrant was the most potent antagonist of PI3K inhibitor– induced ER activity, we chose to use this agent to characterize the biologic consequences of the ER induction. We used two well-established ER-positive/PIK3CAmut xenograft models (MCF7 and T47D cells) to test the combination. Daily administration of BYL719 (25 mg/kg) re- sulted in modest reduction of tumor growth in both models (Fig. 4A and fig. S13). Fulvestrant monotherapy (200 mg/kg, twice weekly) was sufficient to prevent further tumor growth and, in some cases, to in- duce tumor shrinkage. However, the combination of both agents showed marked tumor regression and, in some cases, resulted in com- plete tumor remissions. Consistent with a tumor cell–autonomous ef- fect of the combination, fulvestrant and BYL719 in combination potently inhibited cell cycle progression in MCF7 cells in vitro (fig. S14). Not only were these effects seen in cell line–derived xenograft models but also when we further investigated the effects of these treat- ments in a patient-derived xenograft (PDX) model of ER-positive PIK3CAmut (H1047R) breast cancer (Fig. 4B). Notably, this PDX was established from a patient who had previously progressed on multiple lines of endocrine therapy, including fulvestrant. In this model, the combination of BYL719 and fulvestrant induced partial tu- mor regression, despite very limited single-agent activity for either BYL719 or fulvestrant. To confirm that the lack of tumor regressions from single-agent BYL719 was not due to failure of the drug to inhibit the target in vivo, we collected a set of MCF7 xenografts for analysis after 4 days of treatment. BYL719 effectively inhibited pAKT (S473), as well as phosphorylation of the ribosomal protein S6, a downstream effector of S6K, indicative of a therapeutic dose of BYL719 (Fig. 4C). Moreover, ER protein expression showed a marked increase with BYL719 treatment that was mitigated by the addition of fulvestrant. Finally, to confirm our observations that ER activity is induced by the PI3K inhibitor in vivo, gene expression analysis was conducted on a representative cohort of the PDX and MCF7 xenografts. Tumors were collected after short-term treatment (4 to 7 days), and gene ex- pression profiling revealed that BYL719 significantly varied the expres- sion of 190 genes (FDR ≤ 1%) in the PDX models, 61% of which have an ER-binding site in their promoter (Fig. 4D). This translated to en- richment for an ER-dependent signature, as confirmed with GSEA (Fig. 4E). A similar enrichment in the ER-dependent signature was seen in the treated MCF7 tumors (Fig. 4F). Together, these data indi- cate that the increase in ER expression and function mediated by PI3K suppression attenuates the benefit of the PI3K inhibitor, and strongly suggest that combinations of PI3K and ER inhibitors should be tested in the clinic. DISCUSSION Here, we show that inhibition of the PI3K pathway in ER-positive breast cancer results in induction of ER-dependent transcriptional activity. These effects on the transcriptomewere not restricted to a few selected ER target genes but rather expression of hundreds of genes controlled by ERE- containing promoters. The causative role of ER in rewiring gene expres- sion upon PI3K inhibition was underscored by its complete prevention when fulvestrant, a direct ER antagonist, was added to the system. These observations led us to consider whether one of the mecha- nisms for up-regulation of ER signaling was augmentation of ER itself. This indeed proved to be the case, and we found a consistent increase in ER transcript and induction of a luminal signature (typical of hormone- responsive breast cancers) in cell lines, murine models, and patient samples upon suppression of the PI3K pathway. However, it remains to be elucidated whether this increase in ER expression is the sole factor responsible for the induction of ER activity after PI3K inhibition. This increase in ER transcription is likely to be an adaptive re- sponse to the inhibition of the PI3K pathway. Indeed, we and others have shown that nongenetic activation of compensatory pathways is frequently observed in response to a variety of targeted therapies and that it may limit their efficacy (12–15). We surmise that the compen- satory activation of ER-dependent genes occurring early upon PI3K inhibition decreases the antitumor efficacy of PI3K inhibitors. This may explain the limited activity of PI3K inhibitors when used as R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 15 April 2015 Vol 7 Issue 283 283ra51 6 on February 2, 2016 http://stm.sciencemag.org/ Downloaded from   177 monotherapy in patients with ER-positive breast cancer and suggests that simulta- neous ER suppression would be a logical strategy to combine with PI3K inhibition. The clinical activity observed in ER-positive patients treated with the combination of the mTOR inhibitor everolimus and an aromatase inhibitor (16), as well as very early clinical data with PI3Ka inhibitors in combination with other anti-estrogen agents (27–29) seem to support this hy- pothesis. Consistent with our PDX data, these clinical studies indicate that dual PI3K and ER blockade is effective even in patients who had progressed on previous anti-estrogen therapies. We are aware that our findings have certain limitations. We have not tested the ligand dependency of ER in vivo, mainly because of a lack of preclinical models of aromatase inhibition, a standard of care in postmenopausal women. Moreover, we acknowledge that the clinical valida- tion of our findings would benefit from a larger cohort of patients treated with PI3K inhibitors than the one used in this work. The requirement of paired pre- treatment and on-treatment samples to study acute changes in gene expression, however, limits the number of patients suitable for these analyses. This is a cave- at that should be taken into consideration at the time of designing clinical trials, where mandatory on-treatment biopsies could provide important information on early adaptive response to therapy. Reflecting on our findings, one cannot escape from drawing a parallel with pros- tate cancer, where a reciprocal feedback regulation of PI3K and androgen receptor (AR) activity has been recently described (19). In the case of prostate cancer, inhibi- tion of the PI3K pathway results in activa- tion of AR and, conversely, blockade of AR activates PI3K signaling. This bidirectional crosstalk seems to also occur in the breast where, in addition to our findings showing ER activation upon PI3K inhibition, in- activation of ER appears to be associated with activation of PI3K signaling (18, 30). This speaks for a true interdependency be- tween these two pathways, where a state of equilibrium between PI3K and ER sig- naling is reached to ensure cell survival. Thus, both prostate and breast cancer cells may adapt to suppression of the PI3K pathway by increasing their dependence on the hormone receptor function. Fig. 3. PI3K inhibitor–mediated induction of hormone signaling is dependent on E2 and ER. MCF7 cells were treated with BYL719 (1 mM), E2 (10 nM), or the combination after 48 hours in estrogen-free medium. ChIP was performed with anti-ERa antibody or control IgG. (A and B) Primers to amplify the ER-binding regions of the PGR (A) and GREB1 (B) promoters were used in qPCR to determine fold enrich- ment relative to input. One-way ANOVA statistical test was used to compare mean signal amplification between each treatment and vehicle-treated samples, applying the Bonferroni method to correct for multiple comparisons. Error bars represent SEM of two independent experiments with three technical repli- cates each. (C) MCF7 cells were preincubated for 48 hours in steroid hormone–depleted medium and sub- sequently treated with BYL719 (1 mM), E2 (10 nM), or the combination over a period of 16 hours, and RNA was isolated at the indicated times. qPCR was performed to detect bACTIN, PGR, and GREB1 expression. The data are presented relative to bACTIN and to expression at time 0 in the BYL719-treated samples. One-way ANOVA statistical test was used to compare gene expression between each treatment and vehicle-treated cells, applying the Bonferroni method to correct for multiple comparisons. The results presented are for the comparisons at the 16-hour time point. Error bars denote SEM of two independent experiments with three technical replicates each. (D) MCF7 cells, grown under normal serum conditions, were treated with BYL719 (1 mM), alone or in combination with 4-OHT (1 mM) or fulvestrant (FULV) (100 nM) over a period of 16 hours. mRNA was isolated at the indicated times. qPCR was performed to detect bACTIN, PGR, and GREB1 ex- pression. The data are presented relative to bACTIN and to expression at time 0 in the vehicle-treated samples. One-way ANOVA statistical test was used to compare gene expression between each treatment and vehicle-treated cells, applying the Bonferroni method to correct for multiple comparisons. The anal- ysis results presented are for the comparisons at the 16-hour time point. Error bars denote SEM of two independent experiments with three technical replicates each. R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 15 April 2015 Vol 7 Issue 283 283ra51 7 on February 2, 2016 http://stm.sciencemag.org/ Downloaded from    178 Fig. 4. Combination of BYL719 and fulves- trant in vivo induces prolonged responses. (A) MCF7 in vivo xenograft was treated with vehicle, BYL719, fulvestrant, or the combina- tion at the indicated doses and schedule. Graph shows the fold change in tumor volume with respect to day 0 of treatment. One-way ANOVA statistical test was performed to com- pare tumor volume fold change on the last day of treatment between each treatment arm and vehicle, applying the Bonferroni method to correct for multiple comparisons. Er- ror bars represent SEM. (B) ER-positive/PIK3CAmut PDX-bearingmice were randomized to receive treatment with indicated doses and schedules of vehicle, BYL719, and/or fulvestrant. Graph shows the fold change in tumor volume with respect today 0of treatment.One-wayANOVA statistical test was performed to compare tu- mor volume fold change on the last day of treatment between each treatment arm and vehicle, applying theBonferronimethod tocor- rect for multiple comparisons. Error bars rep- resent SEM. (C) Pharmacodynamic study of MCF7 mouse xenograft. Mice were treated with vehicle, BYL719, fulvestrant, or the com- bination with the same dosing and schedule as in (A) for 4 days. Animalswere sacrificed 2 hours after the last dose, and tumors were processed for immunohistochemistry and stained with the indicated antibodies. The figure shows rep- resentative images for each of the treatment arms. Scale bars, 50 mm. (D) A parallel pharma- codynamic studywas performedwith the ER- positive/PIK3CAmut PDX mice, which were treated with either vehicle or BYL719 with the same dosing and schedule as described in (B).Micewere sacrificed, and tumors were obtained on day 4, 2 hours after the last dose, and processed to obtain RNA for micro- array gene expression analysis. Graph repre- sents genes whose expression differed significantly across different treatments with an FDR ≤1%. Each of the columns under the experimental conditions represents one biological replicate. Venn diagram represents differentially regulated genes upon treatment with BYL719 (SAM analysis FDR ≤ 1%) and the percentage of these that contained an ER- binding site, defined by ER ChIP-sequencing data available from (40). (E) GSEA analysis was per- formed to determine which gene sets were enriched in the PDX microarray expression data set obtained in (D). Graph represents enrichment for ER-associated signature (FDR ≤ 25%) as described in (42). (F) Pharmacodynamic studies on theMCF7 xenografts from (A) were performed on day 7 of treatment by means of a punch biopsy in both vehicle- and BYL719-treated mice. A representative number of biopsies (at least two biological replicates per condition)was processed to obtain RNA and submitted for gene expression analysis. GSEA was performed to determine which gene sets were enriched in our data set (FDR≤ 25%). Graph represents enrich- ment for ER-associated signature as described in (42). R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 15 April 2015 Vol 7 Issue 283 283ra51 8 on February 2, 2016 http://stm.sciencemag.org/ Downloaded from   179 In our hands, ER degradation by fulvestrant treatment was more effective than the ER modulator 4-OHT for resensitizing tumors to PI3K inhibitors. A plausible explanation for these results is that ER degradation may potentially prevent both the estrogen-dependent and estrogen-independent activities (31, 32) mediated by PI3K inhibi- tion. However, further studies will be required to confirm which of the currently available anti-estrogen therapies, if any, is superior when given in combination with PI3K inhibitors. In summary, our results suggest that PI3K blockade in ER-positive breast cancer triggers an ER-dependent transcriptional program that ultimately may be reversed with ER-targeting therapies. Thus, simulta- neous blockade of PI3K and ER may be needed for optimal treatment of ER-positive breast tumors with aberrant activation of the PI3K pathway. MATERIALS AND METHODS Study design The aim of our study was to explore the mechanism by which the com- bination of PI3K pathway inhibitors and ER function blockade results in superior antitumor activity.We aimed to evaluatewhether changes in ER function were influencing the clinical response to anti-PI3K therapy in ER-positive breast tumors that harbor PI3K pathway activation. For this purpose, we planned to use various specific PI3K inhibitors, namely, BYL719 (p110a-specific catalytic inhibitor),GDC0941 andBKM120 (pan- PI3K inhibitors), and GDC0032 and BAY80-6946 (p110b-sparing PI3K inhibitors) in a panel of ER-positive breast cancer cell lines and xeno- grafts that harbor PIK3CA activatingmutations.We also usedMK2206 (pan-AKT allosteric inhibitor) to inhibit the PI3K pathway in ER- positive cell lines that activate this pathway through PTEN loss. Finally, to evaluate the role of ER up-regulation as a prosurvival signal in our in vitro and in vivo models, we planned to use the selective ERmodulator 4-OHT and the degrader fulvestrant. For the in vivo experiments, the number of animals in each group was calculated to measure a 25% difference between the means of placebo and treatment groups with a power of 80% and a P value of 0.01. Host mice carrying xenografts were randomly and equally assigned to either control or treatment groups. Animal experiments were conducted in a controlled and nonblinded manner. We also used RNAseq to evaluate gene expression changes in breast cancer patients who underwent BYL719-based therapy to val- idate our in vitro findings on ER expression. In vitro experiments were performed at least two times and at least in triplicate. Plasmids pRL-TK Renilla luciferase plasmid was obtained from Promega, and 3X-ERE-TATA firefly-luciferase reporter was obtained from Addgene [plasmid 28230 deposited by D. McDonnell (33)]. Establishment of tumor xenografts and in vivo treatments All mouse studies were conducted through Memorial Sloan Kettering Cancer Center, Massachusetts General Hospital Cancer Center, and Vall d’Hebron Institute of Oncology Institutional Animal Care and Use Com- mittees approved animal protocols in accordance with institutional guide- lines. Six-week-old female athymic nude mice were purchased from Charles River Laboratories and housed in air-filtered laminar flow cabi- nets with a 12-hour light cycle and food and water ad libitum. The size of the animal groups was calculated to measure a difference of means of 25% between placebo and treatment groups with a power of 80% and a P value of 0.01. Host mice carrying xenografts were randomly as- signed to either control or treatment groups. Animal experiments were conducted in a nonblinded manner. For cell line–derived xenograft stu- dies, mice were injected subcutaneously with 1 × 107 MCF7 or T47D sus- pended in 150 ml of culture medium/Matrigel (BD Biosciences) in a 1:1 ratio. For PDXs, patient consent for tumor use in animals was obtained under a protocol approved by the Vall d’Hebron Hospital Clinical In- vestigation Ethical Committee and Animal Use Committee. PDXs were derived from an ER-positive PR-positive HER2-negative breast cancer patient previously treated with both chemotherapy (taxanes and vinor- elbine) and anti-endocrine therapy (exemestane and fulvestrant). Tumors (~2 × 2–mm size) were subcutaneously implanted in 6-week-old female HsdCpb:NMRI-Foxn1nu mice (Harlan Laboratories). All animals were supplemented with 1 mM 17b-E2 (Sigma) in their drinking water as described (34). Once tumors reached an average volume of ~75 to 250 mm3, mice were randomized into treatment arms, with n = 7 to 10 tumors per group. BYL719 was dissolved in 0.5% carboxymethylcellulose solution and administered once daily via oral gavage at 25 mg/kg. Fulvestrant was diluted in castor oil and administered subcutaneously twice week- ly at 200 mg/kg. Tumors were measured by digital caliper twice per week over the entire treatment period and, where indicated, harvested 2 hours after the last drug administration. Tumor volume was determined using the following formula: (length × width2) × (p/6). Tumor volumes are plotted as means ± SEM. mRNA nCounter gene expression procedure A section of the formalin-fixed paraffin-embedded (FFPE) breast tis- sue was first examined with hematoxylin and eosin staining to deter- mine the tumor surface area and cellularity. For RNA purification (Roche High Pure FFPET RNA Isolation Kit), one to three 10-mm FFPE slides were cut for each tumor, and macrodissection was per- formed, when needed, to avoid normal breast contamination. A minimum of ~100 ng of total RNA was used to measure the expres- sion of 105 breast cancer–related genes and 5 housekeeping genes using the nCounter platform (NanoString Technologies) (35). Data were log2-transformed and normalized using five housekeeping genes (ACTB, MRPL19, PSMC4, RPLP0, and SF3A1). Raw gene expression data for patient samples were deposited in Gene Expression Omnibus (GSE63579). The list of 105 genes includes genes from the following three sig- natures: PAM50 intrinsic subtype predictor (n = 50) (24), Claudin-low subtype predictor (n = 43) (36), and 13-VEGF/Hypoxia signature (n = 13) (37). In addition, we included eight individual genes that have been found to play an important role in breast cancer (CD24, CRYAB, ERBB4, PIK3CA, PTEN, RAD17, RAD50, and RB1). All tumors were assigned to an intrinsic molecular subtype of breast cancer (luminal A, luminal B, HER2-enriched, basal-like, or the normal-like group) using the previously reported PAM50 subtype predictor (24). Analyses of microarray and mRNA nCounter gene expression data Illumina IDAT files were preprocessed using the IlluminaExpres- sionFileCreator module on GenePattern (www.broad.mit.edu/cancer/ software/genepattern). Supervised analysis to find genes associated with PI3K pathway inhibition treatments was performed using SAM (38). The multiclass unpaired or the two-class paired method of SAM was used to identify genes whose expression differed significantly R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 15 April 2015 Vol 7 Issue 283 283ra51 9 on February 2, 2016 http://stm.sciencemag.org/ Downloaded from    180 among treatments and/or time points. FDR was set as less than or equal to 1% for microarray-based analysis or to 25% for nCounter-based anal- ysis. GSEA was used to determine the extent to which expression pro- files were enriched for a priori defined sets of genes from biologically coherent pathways (39). GSEA was performed using version 2.0 of GSEA run on all the gene sets in version 2.5 of the Molecular Signatures Database and to correct for multiple hypotheses testing; the FDR threshold was set at ≤0.25. A list of the specific signatures used for graphical representation and their specific description has been added to the Excel file with raw data (table S4). Statistical analysis Statistical analysis for in vitro and in vivo experiments was performed using GraphPad Prism (GraphPad Software). When comparing two groups (control versus treated), two-tailed Student’s unpaired t test was performed (significance level set at P < 0.05). When comparing various groups, one-wayANOVA statistical test was used, applying the Bonferroni method to correct for multiple comparisons. Independent experiments were conductedwith aminimumof two biological replicates per condition to allow for statistical comparison. Error bars represent the SEM, and P values are indicated. All cellular experiments were repeated at least two times. ForChIPanalysis (Fig. 1F), thedata arepresented as fold enrichment relative to bACTIN as a control gene region. Error bars represent SEM of three independent experiments. For theChIP analyses in Figs. 2G and3 (A and B) and fig. S8, data were normalized to their respective input signals and thus represented as % input. Two-tailed Student’s unpaired t test was performed (P < 0.05) comparing control versus treated samples. Error bars represent SEM of at least two independent experiments. Raw data for the figures are provided in table S4. SUPPLEMENTARY MATERIALS www.sciencetranslationalmedicine.org/cgi/content/full/7/283/283ra51/DC1 Materials and Methods Fig. S1. Western blot of MCF7 and T47D cells treated in vitro with BYL719 for a series of time points. Fig. S2. T47D transcriptional profile upon p110a inhibition. Fig. S3. GSEA for T47D microarray expression data set. Fig. S4. Western blot of CAMA1 cells treated with BYL719 or MK2206 for 48 hours. Fig. S5. CAMA1 transcriptional profile after AKT inhibition. Fig. S6. ER target genes induced by AKT inhibition in ER-positive/PTENmut/null breast cancer cells. Fig. S7. ESR1 expression induced by PI3Ka inhibition in ER-positive/PIK3CAmut breast cancer cells. Fig. S8. ESR1 transcription increased by PI3Ka inhibition. Fig. S9. Induction of ESR1 and its target genes by different PI3K inhibitors. Fig. S10. Comparison of induction of ESR1 and its target genes between BYL719 and the mTORC1 allosteric inhibitor rapamycin. Fig. S11. Decreased expression of ER target genes after anti-ER therapy, with no effect on ESR1mRNA. Fig. S12. Up-regulation of ER target genes reversed by combining BYL719 with anti-ER treatment. Fig. S13. Better tumor control in vivo after combining BYL719 with fulvestrant. Fig. S14. Analysis of the effect of PI3Ka inhibition alone or with anti-ER therapy on the cell cycle. Table S1. GSEA to assess ER-dependent signatures enriched in MCF7 cells treated with BYL719. Table S2. GSEA to assess ER-dependent signatures enriched in CAMA1 cells treated with MK2206. Table S3. Clinical and pathologic features corresponding to paired pretreatment and BYL719- treated tumor samples. Table S4. Raw data (provided as an Excel file). References (43–45) REFERENCES AND NOTES 1. J. A. Engelman, J. Luo, L. C. Cantley. The evolution of phosphatidylinositol 3-kinases as regulators of growth and metabolism. Nat. Rev. Genet. 7, 606–619 (2006). 2. L. C. Cantley. The phosphoinositide 3-kinase pathway. Science 296, 1655–1657 (2002). 3. J. A. Engelman. Targeting PI3K signalling in cancer: Opportunities, challenges and limitations. Nat. Rev. Cancer 9, 550–562 (2009). 4. S. M. Brachmann, K. Ueki, J. A. Engelman, R. C. Kahn, L. C. Cantley. Phosphoinositide 3- kinase catalytic subunit deletion and regulatory subunit deletion have opposite effects on insulin sensitivity in mice. Mol. Cell. Biol. 25, 1596–1607 (2005). 5. L. Zhao, P. K. Vogt. Helical domain and kinase domain mutations in p110a of phosphati- dylinositol 3-kinase induce gain of function by different mechanisms. Proc. Natl. Acad. Sci. U.S.A. 105, 2652–2657 (2008). 6. S. Kang, A. G. Bader, P. K. Vogt. Phosphatidylinositol 3-kinase mutations identified in human cancer are oncogenic. Proc. Natl. Acad. Sci. U.S.A. 102, 802–807 (2005). 7. Y. Samuels, Z. Wang, A. Bardelli, N. Silliman, J. Ptak, S. Szabo, H. Yan, A. Gazdar, S. M. Powell, G. J. Riggins, J. K. Willson, S. Markowitz, K. W. Kinzler, B. Vogelstein, V. E. Velculescu. High frequency of mutations of the PIK3CA gene in human cancers. Science 304, 554 (2004). 8. S. J. Isakoff, J. A. Engelman, H. Y. Irie, J. Luo, S. M. Brachmann, R. V. Pearline, L. C. Cantley, J. S. Brugge. Breast cancer–associated PIK3CA mutations are oncogenic in mammary epithelial cells. Cancer Res. 65, 10992–11000 (2005). 9. D. Juric, J. Rodon, A. M. Gonzalez-Angulo, H. A. Burris, J. Bendell, J. D. Berlin, M. R. Middleton, D. Bootle, M. Boehm, A. Schmitt, N. Rouyrre, C. Quadt, J. Baselga. Abstract CT-01: BYL719, a next generation PI3K alpha specific inhibitor: Preliminary safety, PK, and efficacy results from the first-in-human study. Cancer Res. 72, CT-01 (2012). 10. D. Juric, I. Krop, R. K. Ramanathan, J. Xiao, S. Sanabria, T. R. Wilson, Y. Choi, H. Parmar, J. Hsu, J. Baselga, D. D. Von Hoff. Abstract LB-64: GDC-0032, a beta isoform-sparing PI3K inhibitor: Results of a first-in-human phase Ia dose escalation study. Cancer Res. 73, LB-64 (2013). 11. D. Juric, P. Castel, M. Griffith, O. L. Griffith, H. H. Won, H. Ellis, S. H. Ebbesen, B. J. Ainscough, A. Ramu, G. Iyer, R. H. Shah, T. Huynh, M. Mino-Kenudson, D. Sgroi, S. Isakoff, A. Thabet, L. Elamine, D. B. Solit, S. W. Lowe, C. Quadt, M. Peters, A. Derti, R. Schegel, A. Huang, E. R. Mardis, M. F. Berger, J. Baselga, M. Scaltriti. Convergent loss of PTEN leads to clinical resistance to a PI(3)Ka inhibitor. Nature 518, 240–244 (2015). 12. K. E. O’Reilly, F. Rojo, Q. B. She, D. Solit, G. B. Mills, D. Smith, H. Lane, F. Hofmann, D. J. Hicklin, D. L. Ludwig, J. Baselga, N. Rosen. mTOR inhibition induces upstream receptor tyrosine ki- nase signaling and activates Akt. Cancer Res. 66, 1500–1508 (2006). 13. A. Carracedo, L. Ma, J. Teruya-Feldstein, F. Rojo, L. Salmena, A. Alimonti, A. Egia, A. T. Sasaki, G. Thomas, S. C. Kozma, A. Papa, C. Nardella, L. C. Cantley, J. Baselga, P. P. Pandolfi. Inhi- bition of mTORC1 leads to MAPK pathway activation through a PI3K-dependent feedback loop in human cancer. J. Clin. Invest. 118, 3065–3074 (2008). 14. V. Serra, M. Scaltriti, L. Prudkin, P. J. Eichhorn, Y. H. Ibrahim, S. Chandarlapaty, B. Markman, O. Rodriguez, M. Guzman, S. Rodriguez, M. Gili, M. Russillo, J. L. Parra, S. Singh, J. Arribas, N. Rosen, J. Baselga. PI3K inhibition results in enhanced HER signaling and acquired ERK dependency in HER2-overexpressing breast cancer. Oncogene 30, 2547–2557 (2011). 15. S. Chandarlapaty, A. Sawai, M. Scaltriti, V. Rodrik-Outmezguine, O. Grbovic-Huezo, V. Serra, P. K. Majumder, J. Baselga, N. Rosen. AKT inhibition relieves feedback suppression of receptor tyrosine kinase expression and activity. Cancer Cell 19, 58–71 (2011). 16. J. Baselga, M. Campone, M. Piccart, H. A. Burris III, H. S. Rugo, T. Sahmoud, S. Noguchi, M. Gnant, K. I. Pritchard, F. Lebrun, J. T. Beck, Y. Ito, D. Yardley, I. Deleu, A. Perez, T. Bachelot, L. Vittori, Z. Xu, P. Mukhopadhyay, D. Lebwohl, G. N. Hortobagyi. Everolimus in postmenopausal hormone-receptor–positive advanced breast cancer. N. Engl. J. Med. 366, 520–529 (2012). 17. T. W. Miller, J. M. Balko, E. M. Fox, Z. Ghazoui, A. Dunbier, H. Anderson, M. Dowsett, A. Jiang, R. A. Smith, S. M. Maira, H. C. Manning, A. M. González-Angulo, G. B. Mills, C. Higham, S. Chanthaphaychith, M. G. Kuba, W. R. Miller, Y. Shyr, C. L. Arteaga. ERa-dependent E2F transcription can mediate resistance to estrogen deprivation in human breast cancer. Cancer Discov. 1, 338–351 (2011). 18. M. M. Morrison, K. Hutchinson, M. M. Williams, J. C. Stanford, J. M. Balko, C. Young, M. G. Kuba, V. Sánchez, A. J. Williams, D. J. Hicks, C. L. Arteaga, A. Prat, C. M. Perou, H. S. Earp, S. Massarweh, R. S. Cook. ErbB3 downregulation enhances luminal breast tumor response to antiestrogens. J. Clin. Invest. 123, 4329–4343 (2013). 19. B. S. Carver, C. Chapinski, J. Wongvipat, H. Hieronymus, Y. Chen, S. Chandarlapaty, V. K. Arora, C. Le, J. Koutcher, H. Scher, P. T. Scardino, N. Rosen, C. L. Sawyers. Reciprocal feedback regula- tion of PI3K and androgen receptor signaling in PTEN-deficient prostate cancer. Cancer Cell 19, 575–586 (2011). 20. J. S. Carroll, C. A. Meyer, J. Song, W. Li, T. R. Geistlinger, J. Eeckhoute, A. S. Brodsky, E. K. Keeton, K. C. Fertuck, G. F. Hall, Q. Wang, S. Bekiranov, V. Sementchenko, E. A. Fox, P. A. Silver, T. R. Gingeras, X. S. Liu, M. Brown. Genome-wide analysis of estrogen receptor binding sites. Nat. Genet. 38, 1289–1297 (2006). 21. Cancer Genome Atlas Network, Comprehensive molecular portraits of human breast tumours. Nature 490, 61–70 (2012). 22. S. Wee, D. Wiederschain, S. M. Maira, A. Loo, C. Miller, R. deBeaumont, F. Stegmeier, Y. M. Yao, C. Lengauer, PTEN-deficient cancers depend on PIK3CB. Proc. Natl. Acad. Sci. U.S.A. 105, 13057–13062 (2008). 23. M. Elkabets, S. Vora, D. Juric, N. Morse, M. Mino-Kenudson, T. Muranen, J. Tao, A. B. Campos, J. Rodon, Y. H. Ibrahim, V. Serra, V. Rodrik-Outmezguine, S. Hazra, S. Singh, P. Kim, C. Quadt, M. Liu, R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 15 April 2015 Vol 7 Issue 283 283ra51 10 on February 2, 2016 http://stm.sciencemag.org/ Downloaded from   181 A. Huang, N. Rosen, J. A. Engelman, M. Scaltriti, J. Baselga. mTORC1 inhibition is required for sen- sitivity to PI3Kp110a inhibitors inPIK3CA-mutant breast cancer. Sci. Transl.Med.5, 196ra199 (2013). 24. J. S. Parker, M. Mullins, M. C. Cheang, S. Leung, D. Voduc, T. Vickery, S. Davies, C. Fauron, X. He, Z. Hu, J. F. Quackenbush, I. J. Stijleman, J. Palazzo, J. S. Marron, A. B. Nobel, E. Mardis, T. O. Nielsen, M. J. Ellis, C. M. Perou, P. S. Bernard. Supervised risk predictor of breast cancer based on intrinsic subtypes. J. Clin. Oncol. 27, 1160–1167 (2009). 25. S. Massarweh, C. K. Osborne, C. J. Creighton, L. Qin, A. Tsimelzon, S. Huang, H. Weiss, M. Rimawi, R. Schiff. Tamoxifen resistance in breast tumors is driven by growth factor receptor signaling with repression of classic estrogen receptor genomic function. Cancer Res. 68, 826–833 (2008). 26. R. L. Yamnik, A. Digilova, D. C. Davis, Z. N. Brodt, C. J. Murphy, M. K. Holz. S6 kinase 1 reg- ulates estrogen receptor a in control of breast cancer cell proliferation. J. Biol. Chem. 284, 6361–6369 (2009). 27. D. Juric, C. Saura, A. Cervantes, C. Kurkjian, M. Patel, J. Sachdev, I. Mayer, I. E. Krop, M. Oliveira, S. Sanabria, S. Cheeti, R. S. Lin, R. A. Graham, T. R. Wilson, H. Parmar, J. Y. Hsu, D. D. Von Hoff, J. Baselga. Ph1b study of the PI3K inhibitor GDC-0032 in combination with fulvestrant in patients with hormone receptor-positive advanced breast cancer, San Antonio Breast Cancer Symposium, San Antonio, TX, 10 to 14 December 2013, abstract PD1-3. 28. F. Janku, D. Juric, J. Cortes, H. Rugo, H. A. Burris, M. Schuler, B. Deschler-Baier, M. R. Middleton, M. Gil-Martin, J. Berlin, E. Winer, D. Bootle, L. Blumenstein, D. Demanse, C. Coughlin, C. Quadt, J. Baselga. Phase I study of the PI3Ka inhibitor BYL719 plus fulvestrant in patients with PIK3CA-altered and wild type ER+/HER2-locally advanced or metastatic breast cancer. J. Clin. Oncol. abstract PD5-5 (2014). 29. I. A. Mayer, V. G. Abramson, S. J. Isakoff, A. Forero, J. M. Balko, M. G. Kuba, M. E. Sanders, J. T. Yap, A. D. Van den Abbeele, Y. Li, L. C. Cantley, E. Winer, C. L. Arteaga. Stand up to cancer phase Ib study of pan-phosphoinositide-3-kinase inhibitor buparlisib with letrozole in estrogen receptor- positive/human epidermal growth factor receptor 2-negative metastatic breast cancer. J. Clin. Oncol. 32, 1202–1209 (2014). 30. X. Fu, C. J. Creighton, N. C. Biswal, V. Kumar, M. Shea, S. Herrera, A. Contreras, C. Gutierrez, T. Wang, S. Nanda, M. Giuliano, G. Morrison, A. Nardone, K. L. Karlin, T. F. Westbrook, L. M. Heiser, P. Anur, P. Spellman, S. M. Guichard, P. D. Smith, B. R. Davies, T. Klinowska, A. V. Lee, G. B. Mills, M. F. Rimawi, S. G. Hilsenbeck, J. W. Gray, A. Joshi, C. Osborne, R. Schiff. Overcoming endocrine resistance due to reduced PTEN levels in estrogen receptor-positive breast cancer by co- targeting mammalian target of rapamycin, protein kinase B, or mitogen-activated protein kinase kinase. Breast Cancer Res. 16, 430 (2014). 31. C. K. Osborne, A. Wakeling, R. I. Nicholson. Fulvestrant: An oestrogen receptor antagonist with a novel mechanism of action. Br. J. Cancer 90 (Suppl. 1), S2–S6 (2004). 32. N. Patani, A. K. Dunbier, H. Anderson, Z. Ghazoui, R. Ribas, E. Anderson, Q. Gao, R. A’Hern, A. Mackay, J. Lindemann, R. Wellings, J. Walker, I. Kuter, L. A. Martin, M. Dowsett. Differ- ences in the transcriptional response to fulvestrant and estrogen deprivation in ER-positive breast cancer. Clin. Cancer Res. 20, 3962–3973 (2014). 33. J. M. Hall, D. P. McDonnell. The estrogen receptor b-isoform (ERb) of the human estrogen receptor modulates ERa transcriptional activity and is a key regulator of the cellular re- sponse to estrogens and antiestrogens. Endocrinology 140, 5566–5578 (1999). 34. C. García-García, Y. H. Ibrahim, V. Serra, M. T. Calvo, M. Guzmán, J. Grueso, C. Aura, J. Pérez, K. Jessen, Y. Liu, C. Rommel, J. Tabernero, J. Baselga, M. Scaltriti. Dual mTORC1/2 and HER2 blockade results in antitumor activity in preclinical models of breast cancer resistant to anti-HER2 therapy. Clin. Cancer Res. 18, 2603–2612 (2012). 35. G. K. Geiss, R. E. Bumgarner, B. Birditt, T. Dahl, N. Dowidar, D. L. Dunaway, H. P. Fell, S. Ferree, R. D. George, T. Grogan, J. J. James, M. Maysuria, J. D. Mitton, P. Oliveri, J. L. Osborn, T. Peng, A. L. Ratcliffe, P. J. Webster, E. H. Davidson, L. Hood, K. Dimitrov. Direct multiplexed measure- ment of gene expression with color-coded probe pairs. Nat. Biotechnol. 26, 317–325 (2008). 36. A. Prat, J. S. Parker, O. Karginova, C. Fan, C. Livasy, J. I. Herschkowitz, X. He, C. M. Perou. Phenotypic and molecular characterization of the claudin-low intrinsic subtype of breast cancer. Breast Cancer Res. 12, R68 (2010). 37. Z. Hu, C. Fan, C. Livasy, X. He, D. S. Oh, M. G. Ewend, L. A. Carey, S. Subramanian, R. West, F. Ikpatt, O. I. Olopade, M. van de Rijn, C. M. Perou. A compact VEGF signature associated with distant metastases and poor outcomes. BMC Med. 7, 9 (2009). 38. V. G. Tusher, R. Tibshirani, G. Chu. Significance analysis of microarrays applied to the ionizing radiation response. Proc. Natl. Acad. Sci. U.S.A. 98, 5116–5121 (2001). 39. A. Subramanian, P. Tamayo, V. K. Mootha, S. Mukherjee, B. L. Ebert, M. A. Gillette, A. Paulovich, S. L. Pomeroy, T. R. Golub, E. S. Lander, J. P. Mesirov. Gene set enrichment analysis: A knowledge- based approach for interpreting genome-wide expression profiles. Proc. Natl. Acad. Sci. U.S.A. 102, 15545–15550 (2005). 40. A. Hurtado, K. A. Holmes, C. S. Ross-Innes, D. Schmidt, J. S. Carroll. FOXA1 is a key deter- minant of estrogen receptor function and endocrine response. Nat. Genet. 43, 27–33 (2011). 41. J. M. Gozgit, B. T. Pentecost, S. A. Marconi, R. S. Ricketts-Loriaux, C. N. Otis, K. F. Arcaro. PLD1 is overexpressed in an ER-negative MCF-7 cell line variant and a subset of phospho- Akt-negative breast carcinomas. Br. J. Cancer 97, 809–817 (2007). 42. L. J. van ‘t Veer, H. Dai, M. J. van de Vijver, Y. D. He, A. A. Hart, M. Mao, H. L. Peterse, K. van der Kooy, M. J. Marton, A. T. Witteveen, G. J. Schreiber, R. M. Kerkhoven, C. Roberts, P. S. Linsley, R. Bernards, S. H. Friend. Gene expression profiling predicts clinical outcome of breast cancer. Nature 415, 530–536 (2002). 43. J. D. Nelson, O. Denisenko, K. Bomsztyk. Protocol for the fast chromatin immuno- precipitation (ChIP) method. Nat. Protoc. 1, 179–185 (2006). 44. E. Toska, H. A. Campbell, J. Shandilya, S. J. Goodfellow, P. Shore, K. F. Medler, S. G. Roberts. Repression of transcription by WT1-BASP1 requires the myristoylation of BASP1 and the PIP2-dependent recruitment of histone deacetylase. Cell Rep. 2, 462–469 (2012). 45. A. S. Doane, M. Danso, P. Lal, M. Donaton, L. Zhang, C. Hudis, W. L. Gerald. An estrogen receptor-negative breast cancer subset characterized by a hormonally regulated transcription- al program and response to androgen. Oncogene 25, 3994–4008 (2006). Funding: This work was funded by a Stand Up To Cancer Dream Team Translational Cancer Research Grant, a program of the Entertainment Industry Foundation (SU2C-AACR-DT0209 to J.B.), the Breast Cancer Research Foundation (to J.B.), the European Research Council (AdG09250244 to J.B.), the Instituto de Salud Carlos III (Intrasalud PSO9/00623 to J.B.), and Banco Bilbao Vizcaya Argentaria Foundation (Tumor Biomarker Research Program). A. Bosch holds a Translational Research Fellowship from the Spanish Society of Medical Oncology. V.S. is a recipient of an Instituto de Salut Carlos III grant (FIS PI13/01714) and a GHD/FERO grant. S.C. receives research funding from a Louis Gerstner YIA and the Damon Runyon Foundation. N.R. receives research funding from Bayer. N.T.V.-V. held a K99/R00 Pathway to Independence award (1K99CA181492-01A1). J.A.K. was funded by PHS 5R01 CA 025836. Author contributions: A. Bosch, S.C., M.S., and J.B. designed the research. A. Bosch, Z.L., H.E., E.T., J.J.T., P.C., G.M., and N.M. performed the experiments. A. Bergamaschi analyzed microarray gene expression data and per- formed GSEA. D.E.S., N.T.V.-V., and J.A.K. performed and analyzed 18F-FES-PET imaging. J.R., V.S., J.C., J.P.-G., and M.D. collected the human specimens. P.G. and A.P. performed and analyzed mRNA nCounter gene expression procedure. V.S., Y.I., J.G., and M.G. established and performed experiments with PDXs. J.A.K., M.K., J.S.L., and N.R. assisted with research design. A. Bosch prepared the figures. A. Bosch, S.C., M.S., and J.B. wrote the manuscript. Competing interests: J.B. and N.R. have consulted for Novartis Pharmaceuticals. J.B. is a past member of the scientific advisory board of Seragon. N.R. serves on advisory boards for AstraZeneca and Millennium. J.C. serves as an ad- viser for Roche/Genentech and has received speaking fees from Novartis and Roche. A.P. has served in an advisory role for NanoString Technologies Inc. J.R. serves on advisory boards for Novartis, Lilly, Leti, Servier, and KPS. Data and materials availability: BAY80-6946 was provided by Bayer [MTA (material transfer agreement) SK2014-1248]. 18F-FES was provided by the Radiochemistry and Molecular Imaging Probes Core Facility at Memorial Sloan Kettering Cancer Center. Requests for materials will be accommodated with MTAs. Raw gene expression data for cell lines and xenografts were deposited in Gene Expression Omnibus (GSE64033). Raw gene expression data for patient samples were also deposited in Gene Expression Omnibus (GSE63579). Submitted 7 December 2014 Accepted 18 March 2015 Published 15 April 2015 10.1126/scitranslmed.aaa4442 Citation: A. Bosch, Z. Li, A. Bergamaschi, H. Ellis, E. Toska, A. Prat, J. J. Tao, D. E. Spratt, N. T. Viola-Villegas, P. Castel, G. Minuesa, N. Morse, J. Rodón, Y. Ibrahim, J. Cortes, J. Perez-Garcia, P. Galvan, J. Grueso, M. Guzman, J. A. Katzenellenbogen, M. Kharas, J. S. Lewis, M. Dickler, V. Serra, N. Rosen, S. Chandarlapaty, M. Scaltriti, J. Baselga, PI3K inhibition results in enhanced estrogen receptor function and dependence in hormone receptor–positive breast cancer. Sci. Transl. Med. 7, 283ra51 (2015). R E S EARCH ART I C L E www.ScienceTranslationalMedicine.org 15 April 2015 Vol 7 Issue 283 283ra51 11 on February 2, 2016 http://stm.sciencemag.org/ Downloaded from    182   183 ARTICLE 6     Title: The tumor suppressor PTEN and the PDK1 kinase regulate formation of the columnar neural epithelium Journal: eLife Impact Factor: 9.322                                                  184                                                                 185         *For correspondence: k- anderson@ski.mskcc.org Present address: †Cancer Epigenetics and Biology Program, Bellvitge Biomedical Research Institute, Barcelona, Spain; ‡Division of Biological Sciences, University of California, San Diego, San Diego, United States Competing interests: The authors declare that no competing interests exist. Funding: See page 18 Received: 02 October 2015 Accepted: 02 December 2015 Published: 26 January 2016 Reviewing editor: Joseph G Gleeson, Howard Hughes Medical Institute, The Rockefeller University, United States Copyright Grego-Bessa et al. This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited. The tumor suppressor PTEN and the PDK1 kinase regulate formation of the columnar neural epithelium Joaquim Grego-Bessa1†, Joshua Bloomekatz1‡, Pau Castel2, Tatiana Omelchenko3, Jose´ Baselga2,4, Kathryn V Anderson1* 1Developmental Biology Program, Sloan Kettering Institute, Memorial Sloan Kettering Cancer Center, New York, United States; 2Human Oncology and Pathogenesis Program, Sloan Kettering Institute, Memorial Sloan Kettering Cancer Center, New York, United States; 3Cell Biology Program, Sloan Kettering Institute, Memorial Sloan Kettering Cancer Center, New York, United States; 4Department of Medicine, Memorial Sloan Kettering Cancer Center, New York, United States Abstract Epithelial morphogenesis and stability are essential for normal development and organ homeostasis. The mouse neural plate is a cuboidal epithelium that remodels into a columnar pseudostratified epithelium over the course of 24 hr. Here we show that the transition to a columnar epithelium fails in mutant embryos that lack the tumor suppressor PTEN, although proliferation, patterning and apical-basal polarity markers are normal in the mutants. The Pten phenotype is mimicked by constitutive activation of PI3 kinase and is rescued by the removal of PDK1 (PDPK1), but does not depend on the downstream kinases AKT and mTORC1. High resolution imaging shows that PTEN is required for stabilization of planar cell packing in the neural plate and for the formation of stable apical-basal microtubule arrays. The data suggest that appropriate levels of membrane-associated PDPK1 are required for stabilization of apical junctions, which promotes cell elongation, during epithelial morphogenesis. DOI: 10.7554/eLife.12034.001 Introduction Phosphoinositides are powerful second messengers in signaling pathways that also control epithelial organization and cell motility, placing them at a unique intersection of signaling and morphogenesis. The lipid phosphatase PTEN, which converts the membrane lipid phosphatidylinositol (3,4,5)-tri- sphosphate (PtdIns(3,4,5)P3) to phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P2), is the second most commonly mutated gene in human cancers. PtdIns(3,4,5)P3 and PtdIns(4,5)P2 act by recruiting specific sets of pleckstrin homology domain-containing proteins to the plasma membrane (e.g. Lietzke et al., 2000), where they become active. The best-studied functions of PTEN are as a negative regulator of proliferation and a positive reg- ulator of apoptosis through the PDPK1-AKT-mTOR pathway (Chalhoub and Baker, 2009; Song et al., 2012). In addition to its role in tumorigenesis, loss of one copy of the wild-type PTEN gene leads to complex human developmental disorders such as Cowden and Bannayan-Riley-Ruval- caba syndromes, which are characterized by macrocephaly, benign tumors, arteriovenous malforma- tions, and autism spectrum disorder (Blumenthal and Dennis, 2008; Zhou and Parada, 2012). Phosphoinositides play important roles in the architecture of epithelia (Shewan et al., 2011), consis- tent with the high frequency of PTEN mutations in carcinomas. Studies on lumen morphogenesis in a three-dimensional culture system showed that PtdIns(4,5)P2 is enriched in the apical membrane, whereas PtdIns(3,4,5)P3 is enriched in basolateral membranes (Martin-Belmonte et al., 2007), and Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 1 of 22 RESEARCH ARTICLE    186         this was proposed to be important in tumor development (Shewan et al., 2011). Mammalian PTEN regulates cellular processes as diverse as collective cell migration (Bloomekatz et al., 2012) and axon regeneration (Park et al., 2008), and some of the effects of PTEN are independent of the AKT pathway (e.g. Vasudevan et al., 2009). PTEN is essential for viability and Pten null mouse embryos arrest at midgestation with a complex set of morphological defects (Suzuki et al., 1998; Bloomekatz et al., 2012). We showed previously that PTEN is required for the directional collective migration of a population of extraembryonic cells, the anterior visceral endoderm (AVE), which must move from a distal to proximal position to define the anterior-posterior body axis of the embryo (Bloomekatz et al., 2012). PTEN is also required in the cells of the embryo proper: deletion of Pten in cells of the epiblast (the embryo proper) using the Sox2-Cre transgene (Hayashi et al., 2002) (Pten 4Epi) bypasses the requirement for AVE migra- tion but leads arrest at midgestation (~E9.0) with a syndrome of defects that included cardia bifida, abnormal mesoderm migration, and an abnormal open neural tube (Bloomekatz et al., 2012). Mammalian neural tube closure requires more than 100 genes that regulate a sequence of orchestrated morphogenetic processes that transform the neural epithelium into a closed tube (Copp and Greene, 2010; Harris and Juriloff, 2010; Colas and Schoenwolf, 2001). Failure of any one of these events can cause neural tube defects, the second most common type of human birth defect after cardiac malformations. Most genetic studies of neural tube closure have focused on the cell rearrangements in the ventral midline mediated by the planar cell polarity pathway (Murdoch et al., 2003; Ybot-Gonzalez et al., 2007; Nishimura et al., 2012; Williams et al., 2014) or on the actin-mediated apical constriction of neural epithelial cells required for neural tube closure (Suzuki et al., 2012; Grego-Bessa et al., 2015). Prior to apical constriction, the neural plate lateral to the midline is transformed from a cuboidal to a tightly packed pseudostratified columnar epithe- lium, so that by E9.5, up to 8 nuclei are stacked on top of each other, with each cell retaining con- nections to both the apical surface and the basement membrane of the epithelium. Here we define the cellular and biochemical basis of the neural tube closure defect seen in mouse embryos that lack PTEN. The Pten neural plate phenotype is not the result of changes in prolifera- tion, apoptosis, cell fate or loss of epithelial polarity. Instead, Pten mutants have a novel defect in neural morphogenesis: they fail to form a pseudostratified columnar epithelium. Cells do not elon- gate along their apical-basal axis; they fail to become compacted along the mediolateral axis of the embryo and they fail to pack into a stable hexagonal array. A combination of genetic and chemical genetic experiments demonstrate that these defects are due to the loss of the lipid phosphatase activity of PTEN and to the activation of 3-phosphoinositide-dependent protein kinase-1 (PDPK1 eLife digest In mammals, the brain and spinal cord develop from a flat sheet of cells called the neural plate, which bends around to create a structure known as the neural tube. This bending process occurs through a complex sequence of cell shape changes. The cells in the neural plate are initially short and wide, but transform into long, thin cells as the neural plate forms. Problems that prevent the neural tube from forming correctly are amongst the most common birth defects in humans. Many cancer cells contain a mutation that affects a gene that produces a protein called PTEN. This protein normally activates a tumor suppressor pathway, and so cancer cells that lack PTEN divide and grow uncontrollably. Grego-Bessa et al. have now examined mouse embryos that lack this gene, and found that the neural plate in such embryos forms irregular ruffles rather than a closed tube. Further investigation revealed that the neural tube defects are not due to the inactivation of the traditional tumor suppressor pathway. Instead, correct neural tube formation relies upon the ability of PTEN to remove phosphate groups from a target lipid, which is important for limiting the activity of an enzyme called PDK1. Unlimited PDK1 activity causes complex changes that prevent the neural plate cells from elongating and packing together correctly. Future work is now needed to investigate the exact molecules targeted by PDK1 and the roles they play in disorders and diseases caused by a lack of the PTEN protein. DOI: 10.7554/eLife.12034.002 Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 2 of 22 Research article Cell biology Developmental biology and stem cells   187         (PDK1)), but do not depend on the AKT-mTOR tumor suppressor pathway. The data suggest that PTEN activity is required for stabilization of cell packing in the neural plate, which is in turn required for formation of apical-basal microtubule arrays, apical-to-basal trafficking, and cell elongation in the neural plate. We suggest that the role of PTEN in epithelial morphogenesis contributes to the devel- opmental malformations in PTEN mutant syndromes and to the behavior of tumors that lack PTEN. Results PTEN is required for formation of the pseudostratified neural epithelium, but not for proliferation, patterning or apical-basal polarity The cephalic neural epithelium in Pten-/- or Pten 4Epi embryos does not close to make a neural tube (Bloomekatz et al., 2012). At E8.5, scanning electron micrographs showed that the wild-type cephalic neural plate was a smooth structure in which both sides have elevated to begin neural tube closure (Figure 1A,C). In contrast, irregular folds appeared in the Pten mutant neural plate as early as E8.0 and the neural plate was dramatically ruffled at E8.5 (Figure 1B,D); the position of the ectopic folds was highly variable between embryos. PTEN protein was strongly expressed in the E8.5 wild-type neural plate, where it was enriched both apically and basally (Figure 1—figure sup- plement 1A–F), consistent with a significant role for PTEN in morphogenesis of the neural tube. Phosphorylated AKT was not detectable in the wild-type neural plate, but was present in all mem- branes of Pten 4Epi neural plate cells (Figure 1—figure supplement 1G,H), consistent with strong activation of the PI3 kinase pathway in Pten mutants. The abnormal morphology of Pten-/- embryos was noted in previous experiments and was attrib- uted to increased proliferation (Stambolic et al., 1998); however we previously showed that prolifer- ation, cell number, and interkinetic nuclear migration are normal in the Pten-/-neural plate (Bloomekatz et al., 2012). Previous data suggested that there might be abnormalities in anterior- posterior patterning of cell types in the Pten-/- brain that could account for the abnormal morphol- ogy of the anterior neural tube (Suzuki et al., 1998). However, we found that anterior-posterior and dorsal-ventral neural patterning were normal in Pten 4Epi embryos (Figure 1—figure supplement 2A,B). It has also been reported that loss of Pten activates canonical Wnt signaling (Chen et al., 2015), but expression of the canonical Wnt reporter TOPGAL was normal in Pten 4Epi embryos (Figure 1—figure supplement 2C). Transverse sections of the cephalic neural plate showed striking differences in organization in the wild-type and Pten 4Epi cephalic neural epithelium. (For simplicity, we refer to Pten 4Epi in the text below as Pten.) The wild-type neural plate is a single-layered columnar epithelium; the cells of the neural epithelium are so tightly packed that the nuclei appear to stack on top of each other, cre- ating a pseudostratified epithelium. Nuclei in the cephalic neural plate, marked by expression of nuclear SOX2, were stacked in 3–5 rows at E8.5 (Figure 1E). In contrast, the SOX2+ nuclei of the E8.5 Pten cephalic neural plate were organized in only 1–3 rows (Figure 1F; Figure 1—figure sup- plement 3I). Apical recruitment of PTEN is required for apical-basal polarity during apical lumen formation by MDCK cells (Martin-Belmonte et al., 2007). In contrast, we found that global apical-basal organiza- tion in the mouse neural plate was normal in the absence of PTEN. Laminin was basal, and F-actin, N-cadherin, ZO1, aPKC and Par3 were correctly localized to the apical domain in the mutant neural plate (Figure 1G,H; Figure 1—figure supplement 3A–H). Thus the data indicate that the Pten neu- ral plate phenotype is not caused by abnormalities in proliferation, patterning or global apical-basal polarity; instead PTEN is required for normal morphogenesis of the neural plate. Pten-/- neuroepithelial cells are cuboidal rather than columnar and lack stable microtubule arrays Because cells are very tightly packed in the neural plate, we used the mosaic expression of a cyto- plasmic X-linked GFP transgene (Hadjantonakis et al., 2001) to visualize the shape of individual neural cells. In wild type, neural plate cells were highly elongated along the apical-basal axis, whereas Pten neuroepithelial cells were shorter and wider (Figure 2A,B). Accompanying the lack of pseudostratification, the Pten neural plate was 1.5 fold wider than the wild type: the mediolateral apical contour (from left to right) at the level of the mid-hindbrain junction in the E8.5 wild-type Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 3 of 22 Research article Cell biology Developmental biology and stem cells    188         neural plate was 668 ± 242 mm wide (n = 6) and 1023 ± 369 mm wide in Pten (n = 6). Despite this increase in width, the number of nuclei across the width of the cephalic neural plate was not changed in the mutant (275 ± 140 nuclei wide in wild type; 262 ± 88 nuclei in Pten mutants), indicat- ing that the same number of cells occupy more area in Pten. We measured the apical surface area of individual neural plate cells by en face imaging, with cell boundaries marked by expression of the tight junction marker ZO1 (Figure 2C). At the onset of neu- ral morphogenesis (head fold stage, E7.75), the apical surfaces of wild-type and Pten mutant cells were both variable in size and shape but had the same average area (approximately 30 mm2; Figure 1. Morphological defects in the Pten mutant cephalic neural plate. (A, B, C, D) Comparison of neural plate morphology of the dorsal head of wild-type (WT) and Pten 4Epi mutant embryos at E8.0 and E8.5 in scanning electron microscope images. Scale bar = 100 mm. (E, F) Transverse sections of E8.5 WT and Pten 4Epi embryos show the absence of pseudostratified columnar organization in the Pten mutant cephalic neural plate. Green is SOX2, red is phalloidin (F-actin), blue is DAPI. (G, H) Z-stack projection of three optical sections (total of 3 mm) from transverse sections of the cephalic neural plate of E8.5 WT and Pten 4Epi mutant embryos stained for phalloidin (red) and laminin (purple). Scale bar E–H = 10 mm. DOI: 10.7554/eLife.12034.003 The following figure supplements are available for figure 1: Figure supplement 1. PTEN expression in the cephalic neural plate. DOI: 10.7554/eLife.12034.004 Figure supplement 2. Normal neural patterning in Pten4Epi embryos. DOI: 10.7554/eLife.12034.005 Figure supplement 3. Apical markers in Pten 4Epi mutant embryos. DOI: 10.7554/eLife.12034.006 Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 4 of 22 Research article Cell biology Developmental biology and stem cells   189         Figure 2. Cellular defects of Pten 4Epi mutant neuroepithelial cells. (A) Comparison of WT and mutant cell shape in the E8.5 cephalic neural plate, using X-linked GFP-expression to mark individual cells. Schematic representations of individual cells for each genotype are shown (white box). Red is phalloidin. Scale bar is 10 mm. (B) Comparison of neural plate height in the cephalic region of WT and mutants. WT E7.75 = 23.9 ± 4.5 mm; Pten 4Epi E7.75 = 23.6 ± 4.1 mm: WT and mutant are not different, p = 0.86, by standard t-test. WT E8.0 = 32.5 ± 1.7 mm; Pten 4Epi E8.0 = 24.6 ± 3.7 mm: WT is significantly taller than the mutant, *p = 0.0164. WT E8.5 = 49.1 ± 9.6 mm; Pten 4Epi E8.5 = 32.6 ± 7.4 mm; WT is significantly taller than the mutant, ****p < 0.0001. For this and similar analyses below, >100 measurements were made from >3 embryos. (C) Comparison of apical cell shape in the Figure 2 continued on next page Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 5 of 22 Research article Cell biology Developmental biology and stem cells    190         Figure 2D). By ~6 hr later, at E8.0, the average apical surface area of wild-type neural cells had decreased to ~20 mm2, whereas the apical surface area of Pten mutant cells was unchanged (Figure 2D). At E8.5, the apical surface of wild-type neural plate cells had shrunk further, so that it was ~8 mm2, ~3.5 fold smaller than at E7.75. Between E8.0 and E8.5, the surface area of Pten neural plate cells also decreased, but the area of mutant cells was still ~40% greater than that of wild type (Figure 2D). At the same time as the apical surface of wild-type neural plate cells decreased, cell vol- ume remained constant, so the height of the cells increased ~2 fold in WT embryos from ~24 mm at E7.75 to ~50.0 mm at E8.5 (Figure 2B), while the height of Pten mutant cells increased only ~1.3 fold, to ~30 mm at E8.5 (Figure 2B). Formation of polarized columnar epithelia is accompanied by the formation of arrays of apicoba- sally polarized stable microtubules, with minus-ends apical (Bre´ et al., 1987; Jaulin and Kreitzer, 2010). For example, in the neural plate of the Xenopus embryo, multiple g-tubulin-positive apical centrioles nucleate stable arrays of parallel, acetylated microtubules that are thought to drive elon- gation of the cells along the apical-basal axis (Lee et al., 2007). In cells of the mouse embryo neural plate, there is only a single apical centrosome, but noncentrosomal microtubule arrays, marked by expression of a-tubulin, were present parallel to the apical-basal axis of cephalic neural plate cells in both the E8.5 wild-type neural plate, although a-tubulin arrays were not apparent in the Pten mutant (Figure 2—figure supplement 1A). Stable microtubules can become acetylated (Palazzo et al., 2003); wild-type microtubule arrays were not acetylated at E8.0 (0–2 somites) except in the floor plate but became acetylated by E8.5 (5–7 somites) and were strongly acetylated at E9.0 (11–13 somites) (Figure 2E; Figure 2—figure supplement 1B,C). In contrast, the Pten neural plate lacked acetylated microtubules at each of these stages; the only acetylated microtubule arrays in the mutant neural plate were located in the floor plate, the cells in the ventral midline (Figure 2E). Constitutive activation of PI3 kinase recapitulates the Pten neural plate phenotype Because PTEN has both lipid and protein phosphatase activities (Worby and Dixon, 2014), we tested whether the lipid phosphatase activity of PTEN mediated its role in epithelial morphogenesis. While PTEN dephosphorylates PtdIns(3,4,5)P3 to PtdIns(4,5)P2, phosphoinositide 3-kinase (PI3 kinase) carries out the reverse reaction and produces PtdIns (3,4,5)P3. We injected the pregnant mothers of Pten mutant embryos at E7.5 with LY294002, a small molecule inhibitor of PI3 kinase (Gharbi et al., 2007) and analyzed the embryonic phenotype 24 hr later. The development of wild- type embryos was not affected by this treatment, but the mutant neural plate appeared rescued: it was pseudostratified and showed acetylated microtubules arrays (Figure 3—figure supplement 1). Thus inhibition of PI3 kinase rescued Pten neural plate phenotype, suggesting that it is the lack of the lipid phosphatase activity that causes the Pten mutant phenotype. We used an independent genetic experiment to test whether increased levels of PtdIns(3,4,5)P3 were responsible for the defects in epithelial morphogenesis. Pik3ca encodes the p110 catalytic sub- unit of PI3 kinase that catalyzes the production of PtdIns(3,4,5)P3. Point mutations in PIK3CA are seen frequently in tumors and approximately 40% of breast cancer PIK3CA mutations are due to a single amino acid substitution allele, PIK3CAH1047R, which causes elevated kinase activity (Saal, 2005; Figure 2 continued cephalic neural epithelium of WT and Pten 4Epi embryos viewed en face at E7.75, E8.0 and E8.5. Cell borders are marked by expression of ZO1 (white). Scale bar = 20 mm. (D) Apical surface of cephalic neural epithelial cells, taken from images like those shown in (C). WT E7.75 = 29 ± 17 mm2; Pten 4Epi E7.75 = 30 ± 16 mm2: WT and mutant are not different, p = 0.79. WT E8.0 = 20 ± 10 mm2; Pten 4Epi E8.0 = 29 ± 15 mm2. The WT surface area is significantly smaller than in the mutant, ****p < 0.0001. WT E8.5 = 8 ± 4 mm2; Pten 4Epi E8.5 = 14 ± 9 mm2. The WT surface area is significantly smaller than in the mutant, ****p < 0.0001. (E) Acetylated microtubule arrays in the neural plate in stage-matched WT and mutant embryos. Transverse sections of cephalic regions of WT and Pten 4Epi embryos at E8.0 (0– 2 somites), E8.5 (5–7 somites) and E9.0 (11–13 somites). Green is acetylated tubulin; blue is DAPI. Arrows point to the apical surface of neural plate; arrowheads point to the floor plate. The first region of tubulin acetylation in WT is in the floor plate, which is only region of tubulin acetylation in the mutant. Scale bar = 25 mm. DOI: 10.7554/eLife.12034.007 The following figure supplement is available for figure 2: Figure supplement 1. Acetylated microtubules in the wild type cranial neural plate. DOI: 10.7554/eLife.12034.008 Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 6 of 22 Research article Cell biology Developmental biology and stem cells   191         Carson et al., 2008). We conditionally expressed a Pik3caH1047R allele in the epiblast under the con- trol of the Sox2 promoter (Pik3caH1047R-Epi). Western blot analysis confirmed that both pAKT Thr308 and pAKT Ser473, well-characterized targets of the PI3-kinase pathway (Sarbassov et al., 2005), were elevated in both Pten and Pik3caH1047R-Epi embryos (Figure 3A). Pik3caH1047R-Epi embryos had an open, ruffled cephalic neural plate, similar to that seen in Pten 4Epi (Figure 3B). Transverse sections of the cephalic neural plate showed that Pik3caH1047R-Epi neu- ral plate cells did not become columnar (height of E8.5 neural plate cells = 31.5 ± 7.2 mm), the nuclei failed to become pseudostratified, and there was reduced expression of acetylated tubulin (Figure 3C). The apical surface area of E8.5 Pik3caH1047R-Epi neural plate cells was ~15 mm2, ~40% larger than wild type (Figure 3C,D), and epithelial cell height was ~40% shorter than in wild type, as seen in Pten (Figure 3C,E). The common defects in Pik3caH1047R-Epi and Pten 4Epi embryos argue that elevated levels of PtdIns(3,4,5)P3 were responsible for the neural plate phenotypes of both mutants. Figure 3. Expression of an activated form of PI3 Kinase mimics the Pten mutant neural plate phenotype. (A) Loss of Pten (Pten 4Epi) or expression of the activating mutation Pik3caH1047R-Epi in the epiblast leads to phosphorylation of AKT in E8.5 embryos. Representative Western blots (n = 3) show the two phosphorylated forms of AKT in WT, Pten 4Epi and Pik3caH1047R–Epi embryos. Numbers indicate approximate MW. (B) Pik3caH1047R–Epi embryos phenocopy Pten 4Epi embryos. Whole embryos (inset) and expanded view of the cephalic region of E8.5 WT and Pik3caH1047R-Epi embryos; dorsal view. Scale bar = 120 mm. (C) The apical surface of the neural plate, viewed en face; cell borders marked by expression of ZO1 (white) (top row), and acetylated tubulin (green) in transverse sections of the cephalic neural epithelium of E8.5 WT and Pik3caH1047R-Epi embryos. Blue is DAPI. Scale bar = 20 mm. (D) Comparison of apical surface area of cephalic neural epithelial cells at E8.5. WT = 8 ± 4 mm2; Pten 4Epi = 14 ± 9 mm2; Pik3caH1047R-Epi = 15 ± 10 mm2. The surface areas of both mutants are significantly larger than wild type, ****p < 0.0001. (E) Cephalic neural plate height at E8.5. WT = 49.1 ± 9.6 mm; Pten 4Epi = 32.6 ± 7.4 mm; Pik3caH1047R-Epi = 31.5 ± 7.2 mm. Cells in both mutants are significantly shorter than in wild type, ****p < 0.0001. DOI: 10.7554/eLife.12034.009 The following figure supplement is available for figure 3: Figure supplement 1. Inhibition of PI3 kinase restores pseudostratification in the Pten 4Epi neural plate. DOI: 10.7554/eLife.12034.010 Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 7 of 22 Research article Cell biology Developmental biology and stem cells    192         Removal of 3-phosphoinositide dependent protein kinase 1 (PDPK1) rescues the Pten neural plate phenotype In the PTEN tumorigenesis pathway, elevated PtdIns(3,4,5)P3 recruits 3-phosphoinositide-dependent protein kinase-1 (PDPK1) to the plasma membrane through its PH domain, thereby allowing PDPK1 access to specific substrates, including AKT, an important target in tumorigenesis (Sommer et al., 2013). Pdpk1 null embryos die at midgestation with defects in morphogenesis of the brain and somites; proliferation and apoptosis are normal in null mutant MEFs, but Pdpk1 mutant cells are small (Lawlor et al., 2002). To assess the role of Pdpk1 in neural morphogenesis, we removed the gene in embryonic line- ages using a conditional Pdpk1 allele with Sox2-Cre (Pdpk1 4Epi). The general morphology of Pdpk1 4Epi embryos was similar to that previously described for the Pdpk1 null allele (Lawlor et al., 2002), although the conditionally deleted embryos appeared more healthy, formed recognizable somites and initiated embryonic turning, unlike the null mutants. The sides of the neu- ral plate in Pdpk1 4Epi failed to elevate at E8.5, but the neural tube closed by E9.5 (Figure 4—fig- ure supplement 1A,B). Transverse sections at E8.5 and E9.5 showed multiple layers of nuclei and strong acetylated tubulin staining in cephalic neural tube (Figure 4—figure supplement 1A,B), indi- cating that cell elongation and neural plate pseudostratification occurred in absence of PDPK1. To test whether the neural morphogenesis defects observed in Pten neural plate required the activity of PDPK1, we simultaneously removed both Pdpk1 and Pten in the epiblast using the Sox2- Cre transgene. While pAKT levels were increased in Pten embryos, the levels of both phosphory- lated forms of AKT were decreased in Pdpk1 4Epi single mutants (hereafter referred to as Pdpk1) and were present at approximately normal levels in Pten 4Epi Pdpk 14Epi double mutant embryos (referred to below as Pten Pdpk1 double mutants) (Figure 4A). Phosphorylation of the AKT target GSK3b (Ser9) was decreased (Figure 4A), confirming that activation of AKT by removal of PTEN depends on PDPK1, as in other cell types. We noted that phosphorylation of the downstream target ribosomal protein S6 was not affected in Pten embryos, while phosphorylation of S6 was abolished in Pdpk1 single and Pten Pdpk1 double mutant embryos (Figure 4A). The absence of increased phosphorylation of S6 in Pten embryos probably reflects the high rates of growth and cell division in the wild-type mouse embryo, which are not further increased by removal of PTEN. The global morphology of the Pten Pdpk1 double mutant embryos resembled that of the Pdpk1 single mutants (Figure 4B). The cells in the E8.5 double mutant cephalic neural plate were elongated similar to wild type (E8.5 Pten Pdpk1 neural plate height = 48.6 ± 8.8 mm), pseudostratified, and there were apical-basal arrays of acetylated microtubules in the double mutant neural plate (Figure 4C,D). The apical surface area of cells in E8.5 Pten Pdpk1 double mutant neural plate was 50% less than in Pten embryos (10 ± 7 mm2 compared to 15 ± 9 mm2), indicating a rescue of cell shape (Figure 4C,E). Thus these aspects of the Pten neural plate phenotype depend on PDPK1. PTEN acts in extraembryonic tissues to control polarized collective migration of the anterior vis- ceral endoderm that establishes the anterior-posterior body axis and in the epiblast to control move- ment of cardiac precursor cells to the midline (Bloomekatz et al., 2012). In double mutants that lack both Pten and Pdpk1 in all tissues (Pten-/-; Pdpk1-/-), the embryos showed the partial axis duplication seen in Pten single mutants (Figure 4—figure supplement 2A). Pten 4Epi Pdpk1 4Epi double mutants showed the cardia bifida phenotype seen in Pten 4Epi embryos (Figure 4—figure supple- ment 2B). Thus these cell migration phenotypes in Pten mutant embryos were not rescued by removal of PDPK1, in contrast to the PDPK1-dependent phenotype of the Pten neural plate. The neural plate defects in Pten mutants are independent of AKT and mTORC1 AKT is a direct substrate for phosphorylation by PDPK1 (Walker et al., 1998) and the biochemical assays showed that AKT phosphorylation was increased in Pten mutant embryos (e.g. Figure 3A), as expected. There are three Akt genes in the mouse with overlapping functions (Gonzalez and McGraw, 2009), prohibiting a classical genetic test of the role of Akt in neural morphogenesis. Therefore to test whether pAKT was required for the Pten 4Epi phenotype, we injected mothers of Pten mutant embryos with MK-2206, an allosteric inhibitor that blocks activation of the three AKT isoforms (Hirai et al., 2010), 24 and 48 hr before embryo dissection. Western blot analysis showed Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 8 of 22 Research article Cell biology Developmental biology and stem cells   193         that the treatment effectively blocked phosphorylation of AKT on both Thr308 and Ser473 (Figure 5A). Despite effective inhibition of AKT activation, treatment with MK-2206 had no detectable effect on the morphology of the neural plate of E8.5 Pten embryos (Figure 5B). En face imaging and trans- verse sections showed that blocking AKT activity did not rescue the neural plate height, pseudostra- tification or microtubule acetylation (Figure 5C,D). Quantification of apical surface area showed no significant difference between treated and untreated Pten embryos (Figure 5E). An important downstream target of AKT is mTORC1, which mediates its effects on growth and survival (Zoncu et al., 2011). To test whether mTORC1 activity plays a role in morphogenesis of the Figure 4. Removal of Pdpk1 rescues the pseudostratified columnar organization of the Pten neural plate. (A) Phosphorylation of downstream targets of the PI3 kinase pathway in E8.5 wild type, Pten 4Epi, Pdpk1 4Epi single mutant and Pten 4Epi Pdpk1 4Epi double mutant embryos. Representative western blot shown (n = 3). Numbers indicate approximate MW. (B) Dorsal views of E8.5 wild-type, Pten 4Epi, Pdpk14Epi and Pten 4Epi Pdpk1 4Epi embryos. The Pten Pdpk1 double mutants are similar in morphology to Pdpk1 single mutants, but are larger. Scale bar = 100 mm. (C) The apical surface of the neural plate, viewed en face. Cell borders marked by expression of ZO1 (white) (top row) and acetylated tubulin (green) in transverse sections of cephalic neural epithelium in E8.5 wild-type, Pten 4Epi, Pdpk1 4Epi and Pten 4Epi Pdpk1 4Epi embryos. Blue is DAPI. Scale bar = 20 mm. (D) Cephalic neural plate height at E8.5. WT = 49.1 ± 9.6 mm; Pten 4Epi = 32.6 ± 7.4 mm; Pdpk1 4Epi = 47.2 ± 8.9 mm; Pten 4Epi Pdpk1 4Epi = 48.6 ± 8.8 mm. Pten 4Epi cells are significantly shorter than in wild type, and Pten 4Epi Pdpk1 4Epi double mutant cells are significantly taller than in Pten 4Epi, ****p < 0.0001. (E) Apical surface area of E8.5 cephalic neuroepithelial cells. Wild type = 9 ± 6 mm2; Pten 4Epi = 15 ± 9 mm2. The surface area of Pten 4Epi is significantly greater than in wild type, ****p < 0.0001; Pdpk1 4Epi = 8 ± 5 mm2; Pten 4Epi Pdpk1 4Epi = 10 ± 7 mm2; the surface area of Pten 4Epi Pdpk1 4Epi double mutant cells is significantly less than in Pten 4Epi, ****p < 0.0001. DOI: 10.7554/eLife.12034.011 The following figure supplements are available for figure 4: Figure supplement 1. The Pdpk14Epi phenotype. DOI: 10.7554/eLife.12034.012 Figure supplement 2. Cell migration phenotypes in Pten Pdpk1 double mutants. DOI: 10.7554/eLife.12034.013 Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 9 of 22 Research article Cell biology Developmental biology and stem cells    194         neural plate, we injected pregnant females with the mTor inhibitor rapamycin. Western blot analysis of treated embryos showed that the rapamycin treatment blocked phosphorylation of ribosomal pro- tein S6, as expected (Figure 5—figure supplement 1A). Despite its clear biochemical activity, rapa- mycin did not rescue the cell shape, pseudostratification or tubulin acetylation in the Pten 4Epi neural plate (Figure 5—figure supplement 1B). Thus neither AKT nor mTORC1 mediated the effect of PDPK1 on neural morphogenesis. Many other direct substrates for phosphorylation by PDPK1 are known, including more than 20 protein kinases of the AGC family, in addition to AKT (Pearce et al., 2010). Atypical PKC (aPKC) and Figure 5. The Pten neural plate phenotype is independent of AKT. (A) Effect of the AKT inhibitor MK-2206 treatment on targets of the PI3 kinase pathway in E8.5 embryos. Western blot of the two phosphorylated forms of AKT and pS6 S240/4 in WT and Pten 4Epi at E8.5 in control embryos (vehicle) and after 24 or 48 hr of MK-2206 treatment in utero prior to embryo dissection. Numbers indicate approximate MW. (B) Dorsal view (inset) and enlarged image of the cephalic region of E8.5 wild-type and Pten 4Epi embryos. There is no change in the morphology of the mutant heads after 24 or 48 hr of MK-2206 treatment in utero. Scale bar = 120 mm. (C) The apical surface of the neural plate, viewed en face. Cell borders marked by expression of ZO1 (white) (top row); acetylated tubulin (green) in transverse sections of cephalic neural epithelium in wild type and Pten 4Epi at E8.5 after 24 or 48 hr of MK-2206 treatment in utero. Blue is DAPI. Scale bar = 10 mm. (D) Height of the E8.5 cephalic neural plate. Wild type, untreated (control) = 44.9 ± 5.7 mm; WT 48 hr treatment = 46.5 ± 9.9 mm; MK-2206 treatment had no significant effect. Pten 4Epi untreated (control) = 32.4 ± 7.3 mm; Pten 4Epi 24 hr = 33.9 ± 6.8 mm2; Pten 4Epi 48 hr = 30.3 ± 7.5 mm. Treated and untreated mutants were all significantly shorter than wild type, but MK-2206 treatment did not significantly rescue cell elongation in the mutant. (E) Apical surface area of E8.5 cephalic neuroepithelial cells. Control = 8 ± 5 mm2; WT 48 hr = 7 ± 4 mm2; Pten 4Epi Control = 14 ± 9 mm2; Pten 4Epi 24 hr = 13 ± 9 mm2; Pten 4Epi 48 hr = 14 ± 10 mm2. Treated and untreated mutant cells all had significantly larger surface area than wild type, but MK-2206 treatment did not significantly decrease cell surface area in the mutant. DOI: 10.7554/eLife.12034.014 The following figure supplements are available for figure 5: Figure supplement 1. Inhibition of mTORC1 by rapamycin does not rescue the Pten neural plate phenotype. DOI: 10.7554/eLife.12034.015 Figure supplement 2. Downstream targets of PDPK1. DOI: 10.7554/eLife.12034.016 Figure supplement 3. Myosin-II distribution and levels appear normal in the Pten neural plate. DOI: 10.7554/eLife.12034.017 Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 10 of 22 Research article Cell biology Developmental biology and stem cells   195         PKN family members are PDPK1 targets that are stimulated through PtdIns(3,4,5)P3 association (Balendran et al., 2000), and aPKC is an important regulator of epithelial polarity. However, we did not detect a change in localization or increased phosphorylation of aPKC in Pten mutants (Figure 1— figure supplement 3; Figure 5—figure supplement 2A). The Serum and Glucocorticoid-induced Kinase (SGK) protein family is also activated by phosphorylation by PDPK1. Phosphorylation of NDRG1 (T346) is mediated by SGK activity (Murray et al., 2004), and pNDRG1 was upregulated in Pten embryos and reduced in Pten Pdpk1 double mutants (Figure 5—figure supplement 2A). How- ever, in utero treatment of Pten embryos with the AKT inhibitor MK-2206 blocked phosphorylation of NDRG1 (Figure 5—figure supplement 2B), suggesting that activation of NDRG1 depends AKT and not on the pathway that regulates neural morphogenesis. Evidence suggests that PDPK1 can activate Rho kinase 1 (ROCK1) and phosphorylation of myosin light chain (Pinner and Sahai, 2008), which should increase the formation of myosin cables. However, myosin-II was anisotropically distrib- uted in neural plate cells of all genotypes (wild type, Pten-/-, Pdpk1-/- and Pten-/- Pdpk1-/-), there was no preferential enrichment of myosin-II at long or short cell edges in E8.0 embryos (Figure 5—figure supplement 3A–D) and phosphorylation of myosin light chain (MLC) was similar in wild type and Pten mutants (Figure 5—figure supplement 3E). PTEN and PDPK1 regulate cell packing in the neural plate To define the cellular processes regulated by PDPK1 in the neural plate, we examined the cellular basis of the Pten mutant phenotype at higher resolution. Pten has been implicated in planar topol- ogy of epithelial cells in Drosophila (Bardet et al., 2013) and cells in the amniote neural plate undergo dynamic cellular reorganization during neural tube closure as cells break and remake junc- tions with their neighbors (Schoenwolf and Alvarez, 1989; Nishimura et al., 2012). In stable epi- thelia, cells are hexagonally packed into a honeycomb-like array: each cell has six neighbors and three cells converge on each vertex (Zallen and Zallen, 2004). In dynamic epithelia, this pattern can be disrupted by cell division or by neighbor exchanges, so that each cell has fewer neighbors and a greater number of cells converge on each vertex (Zallen and Zallen, 2004). Visualizing cell borders with ZO1 (Figure 2C), b-Catenin (Figure 6A) or F-actin (Figure 6D; Fig- ure 6—figure supplement 1A,D), cells at the beginning of wild-type neural morphogenesis (E8.0) were not hexagonally packed: only ~45% had five or six edges (Figure 6B). Cell arrangements included rosette-like structures where as many as 8 cells converged at a single vertex (Figure 6A), similar to structures in epithelia undergoing active cell rearrangements (Blankenship et al., 2006) and previously described in the rearranging cells of the neural floor plate in chick and mouse embryos (Nishimura et al., 2012; Williams et al., 2014). The arrangement of cells in the Pten neural plate at E8.0 showed the same organization as seen in wild type, where !4 cells converging on ~60% of the vertices (Figure 6C). At E8.5, when pseudostratification was apparent, cells in the wild- type neural plate were packed in a more honeycomb-like arrangement: ~1.8 fold more cells with 5 and 6 edges, and the percentage of cases with !4 cells converging on a vertex was reduced by half (to ~30%), consistent with a more stable epithelium (Figure 6A–C). In contrast, these parameters did not change between E8.0 and E8.5 in Pten mutants. Thus PTEN appears to promote a more regular, hexagonal organization in the plane of the epithelium at the same stage when the epithelium becomes columnar. Cells in the E8.5 neural plate cells of the constitutively activate PI3 kinase mutant (Pik3caH1047R-Epi) showed the complex cells arrangements and rosettes seen in Pten mutants (Fig- ure 6—figure supplement 1A–C). The organization of E8.5 Pdpk1 single and the Pten Pdpk1 double mutant neural plates were sim- ilar to wild type, with similar distributions of neighbors per cell (~60% of cells with 5 or 6 edges) and the percentage of cases with !4 cells converging on a vertex was ~30% (Figure 6D–F). Blocking AKT activity with MK-2206 did not modify cell packing in the Pten neural plate (Figure 6—figure supplement 1D–F). Thus, as with cell elongation and pseudostratification, the failure of Pten mutant neural plate to assume a stable conformation was caused by elevated PtdIns(3,4,5)P3, and depended on PDPK1 but not AKT. PTEN and PDPK1 regulate apical-to-basal trafficking in the neural plate The bottle cells of the gastrulating Xenopus embryo share some characteristics with the early neural plate: they begin as cuboidal cells that elongate in an apical-basal direction while forming apical- Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 11 of 22 Research article Cell biology Developmental biology and stem cells    196         Figure 6. PTEN promotes stable cell packing in the neural plate. Panels (A) and (D) show high magnification views of the apical surface of the neural plate embryos, with magnification adjusted so that the cells appear to be approximately the same size, in order to highlight the difference in cell packing in the two genotypes. Scale bars in (A) and (D) = 15 mm. Orange arrowheads indicate examples of 3 cells/vertex, and yellow arrows indicate vertices formed by !4 cells. Cell borders marked by b-catenin (A) or F-actin (D) expression. (A) At E8.0, rosette-like structures are common in both WT and Pten. Fewer rosette-like arrangements are seen in WT at E8.5, but rosettes persist in the E8.5 Pten neural plate. (B) Quantification of percentage of Figure 6 continued on next page Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 12 of 22 Research article Cell biology Developmental biology and stem cells   197         basal arrays of microtubules and constricting their apical surfaces (Keller et al., 2003; Lee and Har- land, 2007). During the cuboidal-to-columnar transformation in Xenopus bottle cells, membrane from apical microvilli is endocytosed and trafficked to the basolateral membrane, creating a net movement of membrane from apical to basolateral domains (Lee and Harland, 2010). Because vesicle trafficking is highly active in dynamic epithelia and stable microtubules failed to form in the Pten mutant neural plate, we tested whether trafficking was affected by the loss of PTEN. Rab5, a marker of early endosomes, was distributed in an apical-to-basal gradient in the wild type neural plate. In contrast, Rab5+ vesicles were restricted to the most apical domain of the cells in Pten mutants (Figure 7A,B). Clathrin, a marker for coated endocytic vesicles, was also more api- cally restricted in Pten than in wild-type neural plate cells (Figure 7C,D). The normal distribution of Rab5+ and clathrin+ vesicles was restored in Pten Pdpk1 double mutant neural plates (Figure 7A– D). To test whether the change in vesicle distribution reflected changes in endocytosis or in apical-to- basal trafficking, we cultured E8.0 embryos in presence of transferrin coupled to Alexa-647 and ana- lyzed the localization of transferrin-647 after 8 hrours of culture (Christ et al., 2012). Total transfer- rin-647 uptake was similar in wild-type and Pten neural plate cells. However, while transferrin spread along the apical-basal extent of wild-type cells, transferrin accumulated in the apical region in Pten cells (Figure 7E,F), suggesting that defects in basal trafficking are coupled to the failure of Pten mutants to form a pseudostratified columnar neural epithelium. Similar to the other neural plate phe- notypes, basal transport of transferrin was rescued in Pten Pdpk1 double mutants, but was not res- cued by treatment of Pten with MK-2206 (Figure 7G,H). Discussion Mouse embryos that lack PTEN have an unprecedented defect in morphogenesis of the neural tube. In Pten mutant embryos, a SOX2+ neural epithelium forms, shows normal segregation of apical and basal markers, is patterned by developmental signals, and proliferates normally. However, the mutant cephalic neural epithelium fails to undergo the transition from a cuboidal to a tall, columnar pseudostratified epithelium; instead, the mutant neural plate is thin, wide and irregularly folded, and cephalic neural tube closure fails completely. Phosphoinositides have been described as key regulators of apical-basal polarity (Martin- Belmonte et al., 2007; Shewan et al., 2011), and indeed the Pten mutants have a profound defect in the organization of the third (apical-basal) dimension of the neural epithelium. However, the tradi- tional markers of apical-basal polarity are localized correctly in the Pten mutant neural plate: Par3, aPKC, ZO1, P-ERM, N-cadherin and F-actin are apically localized, and laminin is basally localized. Based on the enrichment of pAKT in both apical and basolateral membranes of the Pten mutant neu- ral plate, apical-basal polarity markers are localized correctly despite high levels of PtdIns(3,4,5)P3 throughout cell membranes. Despite the important roles of phosphoinositides in mTOR signaling, endocytic sorting, recycling and trafficking (Di Paolo and De Camilli, 2006; Shewan et al., 2011; Dibble and Cantley, 2015), the genetic and chemical genetic data demonstrate that all the phenotypes in the Pten neural plate are mediated by increased activity of PDPK1. Although phosphorylated AKT is enriched in all cellular membranes in the mutant neural plate, inhibition of the downstream kinases AKT or mTor does not Figure 6 continued cells with 3–8 edges. Between E8.0 and E8.5, the percentage of cells with 3–4 edges decreases ~45%, while the percentage with 5–6 edges increases ~1.6 fold in WT embryos, but these parameters are unchanged in E8.5 mutants. (C) The percentage of vertices plotted against the number of cells meeting at a vertex. In a honeycomb arrangement, 3 cells meet at a vertex; the number of cases where three cells meet at a vertex increases ~1.8 fold between E8.0 and E8.5, whereas the Pten neural plate does not changed in this interval. (D) At E8.5, Pdpk1 single and Pten Pdpk1 double mutants show packing similar to that in WT, compared to the more rosette-like packing in Pten. Quantification of % of cells with 3–8 edges (E) and % of vertices formed by 3–7 cells (F) showed similar values in E8.5 WT, Pdpk1 and Pten Pdpk1 embryos. Bars indicate %, lines indicate s.d. DOI: 10.7554/eLife.12034.018 The following figure supplement is available for figure 6: Figure supplement 1. Cell packing in the neural plate with constitutively active PI3 kinase and when AKT is inhibited with MK-2206. DOI: 10.7554/eLife.12034.019 Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 13 of 22 Research article Cell biology Developmental biology and stem cells    198         modify the Pten mutant phenotype, whereas removal of Pdpk1 rescues all aspects of the Pten phe- notype. We therefore conclude that it is the inappropriate PtdIns(3,4,5)P3-stimulated activity of PDPK1, and not changes in levels of other phosphoinositides or in the activity of AKT or mTorc1, that mediates all the morphogenetic defects seen in the Pten mutant neural epithelium. Perhaps the most striking cellular difference between the Pten and wild-type neural plate cells is the absence of stable apical-basal microtubule arrays in the mutant. The formation of noncentroso- mal apicobasal microtubule arrays, with apical minus-ends and basal plus-ends, is a hallmark of columnar epithelia (Bre´ et al., 1987; Jaulin and Kreitzer, 2010). Consistent with a requirement of Figure 7. Apical-basal trafficking in PI3 kinase pathway mutants. (A– D) Distribution of endosome markers along the apical-basal axis in transverse sections of the cephalic neural plate of E8.5 wild-type, Pten 4Epi, Pdpk1 4Epi and Pten 4Epi Pdpk1 4Epi embryos. (A) Localization of Rab5, an early endosome marker. (B) Distribution of Rab5 along the apical-basal axis, normalized to a maximum value of 100. (C) Localization of clathrin. (D) Distribution of clathrin along the apical-basal axis, normalized to a maximum value of 100. (E) Uptake of Transferrin-Alexa 647 after 8 hr of embryo culture. Transverse sections of cephalic neural plate of E8.5 wild-type, Pten 4Epi, Pdpk1 4Epi and Pten 4Epi Pdpk1 4Epi embryos. White signal is the native Alexa 647 fluorescence. (F) Distribution of Alexa-647 signal along the apical-basal axis. Transferrin-647 accumulates apically in the Pten 4Epi but not in Pten 4Epi Pdpk1 4Epi double mutants. (G) Transverse sections of cephalic neural plate of E8.5 wild-type and Pten 4Epiembryos treated in utero with MK-2206 for 48 hr and then cultured with 50 mg/ml of Transferrin-647 and MK-2206 for 8 hr. (H) Distribution of Alexa-647 along the apical- basal axis is not affected by MK-2206 treatment. Images are Z-projections of 3 optical sections of 1 mm each. Red is phalloidin. Blue is DAPI. Scale bars = 10 mm. DOI: 10.7554/eLife.12034.020 Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 14 of 22 Research article Cell biology Developmental biology and stem cells   199         microtubule arrays for apical-basal trafficking in columnar epithelia (Jaulin and Kreitzer, 2010; Rodriguez-Boulan and Macara, 2014), basal trafficking of apically endocytosed transferrin fails in the Pten neural plate. Recent work showed that PTEN can bind directly to microtubule-associated vesicles (Naguib et al., 2015), suggesting that PTEN could play a direct role in apical-to-basal traf- ficking in the neural plate. The data show that the PTEN is required for organization of stable arrays of apical-basally oriented microtubules, which may both stabilize the long axis of the cell and pro- mote the redistribution of membrane from the apical to the basolateral domains of neuroepithelial cells, leading to the transition from a cuboidal to a columnar epithelium. At the same stage (between E8.0 and E8.5) when wild-type neural cells begin to elongate and form arrays of apical-basal stable microtubules, cells of the neural plate are also reorganizing in the plane of the epithelium to become more hexagonally packed. At E8.0, cell packing in both the wild- type and Pten mutant anterior neural plate is irregular and includes the rosette-like arrangements that are a hallmark of dynamic epithelia (Blankenship et al., 2006). By E8.5, wild-type cells have resolved into a more regular packing pattern and fewer rosettes are observed, while the Pten neural plate continues to have many rosette-like cell arrangements. Pten-dependent, Akt-independent changes in cell packing have also been observed in the Dro- sophila wing disc, where the effect of Pten mutations was attributed to a defects in the remodeling of adherens junctions (Bardet et al., 2013). Similar to what we observed in the cephalic neural plate of the mouse Pten mutant, Drosophila Pten mutant wing disc epithelial cells have fewer neighbors than seen in a regular hexagonal array. In the Drosophila case, high levels of myosin-II are preferen- tially seen on short cell edges of Pten mutant cells. In contrast, myosin-II is anisotropically distributed in the both the wild-type and mutant E8.0 mouse neural plate, and it can be enriched at either long or short cell edges. The anisotropic distribution of myosin-II persists in the E8.5 Pten mutant, while myosin-II becomes enriched at all cell edges in the E8.5 wild-type neural plate, probably in prepara- tion for the next phase of neural tube closure, actomyosin-mediated apical constriction. Thus the loss of PTEN blocks the maturation of cell packing in the neural plate, but there is no simple relation- ship between the Pten phenotype and the distribution of myosin-II. The abnormal planar cell packing and the absence of apical-basal microtubule arrays in the Pten neural plate appear to be coupled: they occur simultaneously and both depend on regulated activity of PDPK1. The coupling of these two phenotypes is consistent with known links between apical junc- tions and microtubule arrays. Apical adherens junctions are sites for anchorage of noncentrosomal microtubule arrays (Meng et al., 2008; Gavilan et al., 2015). Microtubules dynamics, in turn, can regulate the stability of adherens junctions (Meng et al., 2008; Waterman-Storer et al., 2000), sup- porting the existence of a positive feedback loop that couples stable adherens junctions and micro- tubule arrays. We propose that a target of PDPK1 in the Pten mutant neural plate inhibits stabilization of apical junctions, which, in turn, blocks the formation of the noncentrosomal microtu- bule arrays required for elongation of cells in the neural plate (Figure 8). The direct target of PDPK1 in this process is not known; one possibility is that inappropriate activity of PDPK1 promotes dynamic fluctuations in the activity of aPKC and/or PKN. PtdIns(3,4,5)P3-tethered PDPK1 is sufficient to activate these two classes of kinases (Balendran et al., 2000) and aPKC can regulate both apical junctions and microtubule organization (Harris and Tepass, 2008; Harris and Peifer, 2007). PTEN has many roles in mammalian brain development, including control of cell size (Kwon et al., 2001), neuronal differentiation and migration (Yue et al., 2005), synapse structure and synaptic plasticity (Fraser et al., 2008; Sperow et al., 2012) and axon regeneration (Park et al., 2008). Human mutations in one copy of the PTEN gene are associated with a variety of abnormali- ties in brain development, including megalencephaly and focal cortical dysplasia, which can lead to autism and pediatric epilepsy (Hevner, 2015; Jansen, et al., 2015; Zhou and Parada, 2012). Our findings define a profound, very early role of PTEN in the organization of the brain that is likely to contribute to the human syndromes caused by PTEN haploinsufficiency. PDPK1-dependent changes in epithelial stability could also play an important role in tumors that lack PTEN. Mutations in PI3 kinase pathway are extremely common in tumors: for example, nearly 80% of cases of endometrial carcinoma (non-ultramutated samples) have inactivating mutations in PTEN (Cancer Genome Atlas Research Network et al., 2013) and 45% of human luminal A breast tumors harbor activating mutations in PIK3CA (Cancer Genome Atlas Network, 2012). Previous studies provided evidence that anchorage-independence and xenograft growth of breast cancer cells carrying the activated H1047R PI3KCA allele depended on PDPK1 but not AKT Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 15 of 22 Research article Cell biology Developmental biology and stem cells    200         (Gagliardi et al., 2012) and phosphoproteomic analysis of cell lines with activating PI3KCA muta- tions identified cases in which PDPK1 activity, but not AKT activity, was required for tumorigenicity (Vasudevan et al., 2009). The data presented here demonstrate that PtdIns(3,4,5)P3-dependent PDPK1 activity is an important consequence of the absence of PTEN in vivo, even in the absence of activation of AKT. Our findings highlight the importance of identifying the relevant PDPK1 targets during mouse development, in PTEN-associated developmental syndromes, and in tumors. Materials and methods Mouse strains The mutant alleles used here have been described previously: Ptenflox (Trotman et al., 2003), Pdk1flox (MGI designation: Pdpk1) (Lawlor et al., 2002), R26-Pik3caH1047R (Jackson Laboratories, Bar Harvor, ME. Stock #016977). The epiblast specific-expressing CRE line is Sox2-CRE (Hayashi et al., 2002). The Wnt-reporter line used was TOPGAL (DasGupta and Fuchs 1999). The genotype of the Pten DEpi (epiblast-deleted) embryos is Sox2-Cre/+; Ptenflox/Ptennull. The genotype of the Pdpk1 DEpi embryos is Sox2-Cre/+; Pdpk1flox/Pdpk1null. The genotype of the Pten Pdpk1 DEpi double mutants is Sox2-Cre/+; Ptenflox/Ptennull; Pdpk1flox/Pdpk1null. We generated the Pten and Pdpk1 deleted (null) alleles by crossing conditional mice with Sox2-Cre, taking advantage of Sox2 activity in the female germ line. The X-linked GFP transgene was a gift from Anna-Katerina Hadjantonakis (Hadjantonakis et al., 2001). Pten mutants were congenic in CD1, and all other lines, except R26-Pik3caH1047R (FVB), were backcrossed to CD1 for at least four generations before analy- sis. For timed pregnancies, noon on the day of the vaginal plug was E0.5. Figure 8. A model for the role of PTEN in the formation of the pseudostratified columnar epithelium. PDPK1 is anchored to the plasma membrane by PtdIns(3,4,5)P3 (PIP3), which is made by PI3 kinase (PI3K) and degraded by PTEN. In the Pten mutant, increased PIP3 recruits high levels of PDPK1 to the membrane, where it is activated. Activated membrane-associated PDPK1 has two targets: activated PDPK1 generates high levels of pAKT; in a separate pathway, high levels of membrane-associated PDPK1 inhibit the formation of stable apical junctions. Stable apical junctions are required for the formation of stable apical-basal microtubule arrays, which mediate apical-to-basal trafficking in the neural epithelium, allowing elongation and tight packing of cells in the neural epithelium. In WT, PDPK1 is not required for formation of the pseudostratified neural epithelium, although the delay in neural tube closure in Pdpk1 mutants may reflect a subtle role for the protein in epithelial organization. DOI: 10.7554/eLife.12034.021 Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 16 of 22 Research article Cell biology Developmental biology and stem cells   201         In utero embryo drug treatment Pregnant females were injected intraperitoneally (i.p.) following standard procedures. A final volume of 0.5 ml was injected. Treatments were as follows: 25 mg/kg/day of LY294002 (Selleckchem, Houston, TX) diluted in DMSO at E7.5; 120 mg/kg/day of MK-2206 (from the Baselga Laboratory; commercially available from Selleckchem) diluted in Captisol at E7.5 or E6.5 and E7.5; 3 mg/kg/day of Rapamycin (Sigma, St. Louis, MO) diluted DMSO at E6.5 and E7.5. Embryos were har- vested at E8.5. Scanning electron microscopy Embryos for SEM were fixed in 2.5% glutaraldehyde overnight at 4˚C, processed using standard pro- cedures and imaged with a Zeiss Supra 25 Field Emission Scanning Electron Microscope. LacZ staining and in situ hybridization b-Galactosidase activity was detected using standard described protocols (Hogan et al., 1994). Whole-mount in situ hybridization was performed on embryos following standard methods (Eggenschwiler and Anderson, 2000). The Brachyury (Wilkinson et al., 1990), En2 (Joyner and Martin, 1987), Krox20 (Wilkinson et al., 1989), EMX2 (Simeone et al., 1992), Fgf8 (Tanaka et al., 1992), and Axin2 (Jho et al., 2002) in situ probes were previously described. The embryos were photographed using an HRC Axiocam (Zeiss, Germany) fitted onto a stereomicroscope (Leica, Germany). Immunostaining Embryos were dissected in ice-cold or room temperature PBS/4% BSA and processed for imaging following established protocols (Lee et al., 2010). Immunofluorescence staining was performed with Alexa Fluor-conjugated secondary antibodies (Invitrogen, Waltham, MA) diluted 1:400. Sections were counterstained with DAPI (1:2000) to stain nuclei. All images shown are from the cephalic neu- ral plate. Rhodamine-phalloidin (Invitrogen) was used at 1:200. ARL13b antibody (Caspary et al., 2007) was used at 1:2000. Commercial antibodies were: Sigma: g-tubulin (T-6557), 1:1000 for immunofluo- rescence (IF); a-Tubulin (T5168) 1:1000 for IF, 1:3000 for western blots (WB); acetylated a-Tubulin (T7451) 1:1000 for IF and 1:3000 for WB. Santa Cruz, Dallas, TX: GAPDH (sc-32233), 1:5000 for WB. Invitrogen: ZO1 (33-9100), 1:200 for IF. Cascade Biosciences, Winchester, MA: Pten (ABM2052), 1:1000 for IF. Cell Signaling, Danvers, MA: Pten (9559) 1:500 for IF; S6 (2217) 1:2000 for WB; pS6 (2211) 1:1000 for WB; pAKT Ser473 (9271) 1:1000 for WB; pAKT Thr308 (2965) 1:1000 for WB; AKT (9272) 1:1000 for WB; Rab5 (3547) 1:100 for IF; Clathrin Heavy Chain (4796) 1:100 for IF; pMLC2 (3671), 1:1000 for WB; acetylated a-Tubulin (5335) 1:3000 for WB. Hybridoma Bank, Iowa City, IA: Nkx2.2 (74.5A5) 1:100 for IF; Nkx6.1 (F55A10) 1:50 for IF. Covance, Princeton, NJ: MHCIIB (CMII-23; PRB-445P), 1:50 for IF, and 1:1000 for WB. Abcam, Cambridge, MA: FOXA2 (AB40874) 1:800 for IF. Millipore, Billerica, MA: Olig2 (AB9610) 1:200 for IF; SOX2 (AB5603) 1:1000 for IF. For immunofluorescence, samples were mounted using Vectashield (Vector Labs, Burlingame, CA) or ProLong Gold (Life Technologies, Carlsbad, CA) mounting media, and slides were imaged with SP5 and SP8 confocal microscopes (Leica) with a 63 ! 0.5 NA lens, at a res- olution of 1024 ! 1024. In transverse sections, maximum intensity was set in the apical domain, and images with apical non-saturated signal on the neural plate were taken. En face images are Z-projec- tions of 3–5 single optical sections taken every 0.3 mm. Images were analyzed using Volocity soft- ware (PerkinElmer, Waltham, MA). The immunofluorescence data presented in the figures are representative images of at least three embryos. Fluorescence signal quantification Pixel intensity along the apicobasal axis of the neural plate was determined on Z-stack projections of 5 optical sections taken every 1 mm (grayscale). Pixel intensity values were taken from lines 20 pixels wide traced with ImageJ. Graphical distribution of pixel intensity average (n"3 embryos) was gener- ated using Prism6 with normalized values. Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 17 of 22 Research article Cell biology Developmental biology and stem cells    202         Transferrin uptake assay E8.0 embryos with intact yolk sac and ectoplacental cone were dissected in 37˚C DMEM/F12 con- taining 10% FBS. After dissection, 5 embryos were transferred to a glass bottle (Roller Bottle System) containing 5 ml of 50% rat serum/50% DMEM/F12 and incubated at 37˚C with 5% CO2 and 10% O2. Transferrin-Alexa 674 (Molecular Probes, Eugene, OR. #Ta3366) was diluted in the culture media to 50 mg/ml, as described (Christ et al., 2012). After 8 hr, the yolk sac was removed and the embryos were fixed in 4% PFA for 2 hr at 4˚C and mounted for cryosectioning following established protocols (Lee et al., 2010). Images were taken from transverse sections of the cephalic region using a SP5 Leica confocal microscope collecting the native signal from Transferrin-Alexa 674. Morphometric analysis Neural plate height of cephalic region was measured following a previously described method (Grego-Bessa et al., 2015). Apical surface area quantification of cephalic neuroepithelial cells was determined from en face images taken with a Leica SP5 inverted confocal microscope and 63 ! 0.5 NA lens, and analyzed by Volocity software (>100 measurements per embryo, n"3 embryos). For all analyses, n"3 embryos. Measurements are average ± s.d. Comparisons were made by standard t- test. Prism6 was used for statistical analysis. For analysis of cell packing, ZO-1, b-Catenin and Phalloidin-Rhodamine staining delineated the apical domain of cephalic neuroepithelial cells. En face images of the cephalic region were taken by confocal microscope at 63! of magnification. For two-dimensional cell patterns, the number of edges/cell and the number of vertices formed by 3–7 cells were quantitated manually from at least 3 embryos per genotype (>200 cell vertexes). Data analysis was performed with Excel and Prism6. Immunoblotting A pool of three E8.5 embryos, after removal of the heart, was lysed in Cell Lysis Buffer (Cytoskele- ton, Denver, CO. GL36) plus Complete Protease Inhibitor Cocktail (Roche, Germany). Western blots were performed according to standard protocols, and protein was detected with HRP-conjugated secondary antibodies and ECL detection reagents (Amersham, UK). Acknowledgements We thank Dr. Anna-Katerina Hadjantonakis for X-linked GFP transgenic mice and Dr. Dario Alessi for the Pdpk1 conditional mice. We thank Dr. Jennifer Zallen, Dr. Hadjantonakis and members of Ander- son laboratory for helpful conversations and thoughtful comments on the manuscript, Vitaly Boiko and the MSKCC Molecular Cytology Core Facility for valuable technical support. Monoclonal anti- bodies were obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242. The work was supported by R37 HD03455 and R01 NS044385 to KVA, the MSKCC Cancer Center Sup- port Grant (P30 CA008748), and a Beatriu de Pino´s postdoctoral Fellowship from Generalitat de Catalunya to JGB. Additional information Funding Funder Grant reference number Author National Institutes of Health HD03455 Kathryn V Anderson National Institutes of Health NS044385 Kathryn V Anderson National Institutes of Health P30 CA008748 Jose´ Baselga Kathryn V Anderson The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 18 of 22 Research article Cell biology Developmental biology and stem cells   203         Author contributions JGB, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article; JB, Conception and design, Acquisition of data, Analysis and interpretation of data; PC, JoB, Conception and design, Contributed unpublished essential data or reagents; TO, Conception and design, Analysis and interpretation of data, Contributed unpublished essential data or reagents; KVA, Conception and design, Analysis and interpretation of data, Drafting or revising the article Ethics Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All of the animals were handled according to approved Institutional Animal Care and Use Committee (IACUC) protocol (02-06-013) of Memorial Sloan Kettering Cancer Center References Balendran A, Biondi RM, Cheung PCF, Casamayor A, Deak M, Alessi DR. 2000. A 3-phosphoinositide-dependent protein kinase-1 (pDK1) docking site is required for the phosphorylation of protein kinase czeta (pKCzeta) and PKC-related kinase 2 by PDK1. Journal of Biological Chemistry 275:20806–20813. doi: 10.1074/jbc. M000421200 Bardet P-L, Guirao B, Paoletti C, Serman F, Le´opold V, Bosveld F, Goya Y, Mirouse V, Graner F, Bellaı¨che Y. 2013. PTEN controls junction lengthening and stability during cell rearrangement in epithelial tissue. Developmental Cell 25:534–546. doi: 10.1016/j.devcel.2013.04.020 Blankenship JT, Backovic ST, Sanny JSP, Weitz O, Zallen JA. 2006. Multicellular rosette formation links planar cell polarity to tissue morphogenesis. Developmental Cell 11:459–470. doi: 10.1016/j.devcel.2006.09. 007 Bloomekatz J, Grego-Bessa J, Migeotte I, Anderson KV. 2012. Pten regulates collective cell migration during specification of the anterior–posterior axis of the mouse embryo. Developmental Biology 364:192–201. doi: 10. 1016/j.ydbio.2012.02.005 Blumenthal GM, Dennis PA. 2008. PTEN hamartoma tumor syndromes. European Journal of Human Genetics 16: 1289–1300. doi: 10.1038/ejhg.2008.162 Bre´ MH, Kreis TE, Karsenti E. 1987. Control of microtubule nucleation and stability in madin-darby canine kidney cells: the occurrence of noncentrosomal, stable detyrosinated microtubules. The Journal of Cell Biology 105: 1283–1296. doi: 10.1083/jcb.105.3.1283 Cancer Genome Atlas Network. 2012. Comprehensive molecular portraits of human breast tumours. Nature 490:61–70. doi: 10.1038/nature11412 Cancer Genome Atlas Research Network, Kandoth C, Schultz N, Cherniack AD, Akbani R, Liu Y, Shen H, Robertson AG, Pashtan I, Shen R, Benz CC, Yau C, Laird PW, Ding L, Zhang W, Mills GB, Kucherlapati R, Mardis ER, Levine DA. 2013. Integrated genomic characterization of endometrial carcinoma. Nature 497:67–73. doi: 10.1038/nature12113 Carson JD, Van Aller G, Lehr R, Sinnamon RH, Kirkpatrick RB, Auger KR, Dhanak D, Copeland RA, Gontarek RR, Tummino PJ, Luo L. 2008. Effects of oncogenic p110alpha subunit mutations on the lipid kinase activity of phosphoinositide 3-kinase. Biochemical Journal 409:519–524. doi: 10.1042/BJ20070681 Caspary T, Larkins CE, Anderson KV. 2007. The graded response to sonic hedgehog depends on cilia architecture. Developmental Cell 12:767–778. doi: 10.1016/j.devcel.2007.03.004 Chalhoub N, Baker SJ. 2009. PTEN and the PI3-kinase pathway in cancer. Annual Review of Pathology 4:127– 150. doi: 10.1146/annurev.pathol.4.110807.092311 Chen Y, Huang W-C, Sejourne J, Clipperton-Allen AE, Page DT. 2015. Pten mutations alter brain growth trajectory and allocation of cell types through elevated beta-catenin signaling. Journal of Neuroscience 35: 10252–10267. doi: 10.1523/JNEUROSCI.5272-14.2015 Christ A, Christa A, Kur E, Lioubinski O, Bachmann S, Willnow TE, Hammes A. 2012. LRP2 is an auxiliary SHH receptor required to condition the forebrain ventral midline for inductive signals. Developmental Cell 22:268– 278. doi: 10.1016/j.devcel.2011.11.023 Colas Jean-Franc¸ois, Schoenwolf GC. 2001. Towards a cellular and molecular understanding of neurulation. Developmental Dynamics 221:117–145. doi: 10.1002/dvdy.1144 Copp AJ, Greene ND. 2010. Genetics and development of neural tube defects. The Journal of Pathology 220: 217–230. doi: 10.1002/path.2643 DasGupta R, Fuchs E. 1999. Multiple roles for activated LEF/TCF transcription complexes during hair follicle development and differentiation. Development (Cambridge, England) 126:4557–4568. Di Paolo G, De Camilli P. 2006. Phosphoinositides in cell regulation and membrane dynamics. Nature 443:651– 657. doi: 10.1038/nature05185 Dibble CC, Cantley LC. 2015. Regulation of mTORC1 by PI3K signaling. Trends in Cell Biology 25:545–555. doi: 10.1016/j.tcb.2015.06.002 Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 19 of 22 Research article Cell biology Developmental biology and stem cells    204         Eggenschwiler JT, Anderson KV. 2000. Dorsal and lateral fates in the mouse neural tube require the cell- autonomous activity of the open brain gene. Developmental Biology 227:648–660. doi: 10.1006/dbio.2000. 9918 Fraser MM, Bayazitov IT, Zakharenko SS, Baker SJ. 2008. Phosphatase and tensin homolog, deleted on chromosome 10 deficiency in brain causes defects in synaptic structure, transmission and plasticity, and myelination abnormalities. Neuroscience 151:476–488. doi: 10.1016/j.neuroscience.2007.10.048 Gagliardi PA, di Blasio L, Orso F, Seano G, Sessa R, Taverna D, Bussolino F, Primo L. 2012. 3-phosphoinositide- dependent kinase 1 controls breast tumor growth in a kinase-dependent but akt-independent manner. Neoplasia 14:719–IN19. doi: 10.1593/neo.12856 Gavilan MP, Arjona M, Zurbano A, Formstecher E, Martinez-Morales JR, Bornens M, Rios RM, Gumbiner BM. 2015. Alpha-catenin-dependent recruitment of the centrosomal protein CAP350 to adherens junctions allows epithelial cells to acquire a columnar shape. PLOS Biology 13:e1002087. doi: 10.1371/journal.pbio.1002087 Gharbi SI, Zvelebil MJ, Shuttleworth SJ, Hancox T, Saghir N, Timms JF, Waterfield MD. 2007. Exploring the specificity of the PI3K family inhibitor LY294002. Biochemical Journal 404:15–21. doi: 10.1042/BJ20061489 Gonzalez E, McGraw TE. 2009. The akt kinases: isoform specificity in metabolism and cancer. Cell Cycle 8:2502– 2508. doi: 10.4161/cc.8.16.9335 Greene NDE, Carmichael SL. 2010. An update to the list of mouse mutants with neural tube closure defects and advances toward a complete genetic perspective of neural tube closure. Birth Defects Research Part A: Clinical and Molecular Teratology 88:653–669. doi: 10.1002/bdra.20676 Grego-Bessa J, Hildebrand J, Anderson KV. 2015. Morphogenesis of the mouse neural plate depends on distinct roles of cofilin 1 in apical and basal epithelial domains. Development 142:1305–1314. doi: 10.1242/dev.115493 Hadjantonakis A-K, Cox LL, Tam PPL, Nagy A. 2001. An x-linked GFP transgene reveals unexpected paternal x- chromosome activity in trophoblastic giant cells of the mouse placenta. Genesis 29:133–140. doi: 10.1002/ gene.1016 Harris KP, Tepass U. 2008. Cdc42 and par proteins stabilize dynamic adherens junctions in the drosophila neuroectoderm through regulation of apical endocytosis. The Journal of Cell Biology 183:1129–1143. doi: 10. 1083/jcb.200807020 Harris TJC, Peifer M. 2007. APKC controls microtubule organization to balance adherens junction symmetry and planar polarity during development. Developmental Cell 12:727–738. doi: 10.1016/j.devcel.2007.02.011 Hayashi S, Lewis P, Pevny L, McMahon AP. 2002. Efficient gene modulation in mouse epiblast using a Sox2Cre transgenic mouse strain. Mechanisms of Development 119:S97–S101. doi: 10.1016/S0925-4773(03)00099-6 Hevner RF. 2015. Brain overgrowth in disorders of RTK–PI3K–AKT signaling: a mosaic of malformations. Seminars in Perinatology 39:36–43. doi: 10.1053/j.semperi.2014.10.006 Hirai H, Sootome H, Nakatsuru Y, Miyama K, Taguchi S, Tsujioka K, Ueno Y, Hatch H, Majumder PK, Pan B-S, Kotani H. 2010. MK-2206, an allosteric akt inhibitor, enhances antitumor efficacy by standard chemotherapeutic agents or molecular targeted drugs in vitro and in vivo. Molecular Cancer Therapeutics 9:1956–1967. doi: 10. 1158/1535-7163.MCT-09-1012 Hogan BL, Blessing M, Winnier GE, Suzuki N, Jones CM. 1994. Growth factors in development: the role of TGF- beta related polypeptide signalling molecules in embryogenesis. Development (Cambridge, England). Supplement:53–60. Jansen LA, Mirzaa GM, Ishak GE, O’Roak BJ, Hiatt JB, Roden WH, Gunter SA, Christian SL, Collins S, Adams C, Rivie`re J-B, St-Onge J, Ojemann JG, Shendure J, Hevner RF, Dobyns WB. 2015. PI3K/AKT pathway mutations cause a spectrum of brain malformations from megalencephaly to focal cortical dysplasia. Brain 138:1613– 1628. doi: 10.1093/brain/awv045 Jaulin F, Kreitzer G. 2010. KIF17 stabilizes microtubules and contributes to epithelial morphogenesis by acting at MT plus ends with EB1 and APC. The Journal of Cell Biology 190:443–460. doi: 10.1083/jcb.201006044 Jho E.-h., Zhang T, Domon C, Joo C-K, Freund J-N, Costantini F. 2002. Wnt/beta-Catenin/Tcf signaling induces the transcription of Axin2, a negative regulator of the signaling pathway. Molecular and Cellular Biology 22: 1172–1183. doi: 10.1128/MCB.22.4.1172-1183.2002 Joyner AL, Martin GR. 1987. En-1 and en-2, two mouse genes with sequence homology to the drosophila engrailed gene: expression during embryogenesis. Genes & Development 1:29–38. doi: 10.1101/gad.1.1.29 Keller R, Davidson LA, Shook DR. 2003. How we are shaped: the biomechanics of gastrulation. Differentiation 71:171–205. doi: 10.1046/j.1432-0436.2003.710301.x Kwon C-H, Zhu X, Zhang J, Knoop LL, Tharp R, Smeyne RJ, Eberhart CG, Burger PC, Baker SJ. 2001. Pten regulates neuronal soma size: a mouse model of lhermitte-duclos disease. Nature Genetics 29:404–411. doi: 10.1038/ng781 Lawlor MA, Mora A, Ashby PR, Williams MR, Murray-Tait V, Malone L, Prescott AR, Lucocq JM, Alessi DR. 2002. Essential role of PDK1 in regulating cell size and development in mice. The EMBO Journal 21:3728–3738. doi: 10.1093/emboj/cdf387 Lee C, Scherr HM, Wallingford JB. 2007. Shroom family proteins regulate gamma-tubulin distribution and microtubule architecture during epithelial cell shape change. Development 134:1431–1441. doi: 10.1242/dev. 02828 Lee J-Y, Harland RM. 2007. Actomyosin contractility and microtubules drive apical constriction in xenopus bottle cells. Developmental Biology 311:40–52. doi: 10.1016/j.ydbio.2007.08.010 Lee J-Y, Harland RM. 2010. Endocytosis is required for efficient apical constriction during xenopus gastrulation. Current Biology 20:253–258. doi: 10.1016/j.cub.2009.12.021 Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 20 of 22 Research article Cell biology Developmental biology and stem cells   205         Lee JD, Migeotte I, Anderson KV. 2010. Left–right patterning in the mouse requires Epb4.1l5-dependent morphogenesis of the node and midline. Developmental Biology 346:237–246. doi: 10.1016/j.ydbio.2010.07. 029 Lietzke SE, Bose S, Cronin T, Klarlund J, Chawla A, Czech MP, Lambright DG. 2000. Structural basis of 3- phosphoinositide recognition by pleckstrin homology domains. Molecular Cell 6:385–394. doi: 10.1016/S1097- 2765(00)00038-1 Martin-Belmonte F, Gassama A, Datta A, Yu W, Rescher U, Gerke V, Mostov K. 2007. PTEN-mediated apical segregation of phosphoinositides controls epithelial morphogenesis through Cdc42. Cell 128:383–397. doi: 10. 1016/j.cell.2006.11.051 Meng W, Mushika Y, Ichii T, Takeichi M. 2008. Anchorage of microtubule minus ends to adherens junctions regulates epithelial cell-cell contacts. Cell 135:948–959. doi: 10.1016/j.cell.2008.09.040 Murdoch JN, Henderson DJ, Doudney K, Gaston-Massuet C, Phillips HM, Paternotte C, Arkell R, Stanier P, Copp AJ. 2003. Disruption of scribble (scrb1) causes severe neural tube defects in the circletail mouse. Human Molecular Genetics 12:87–98. doi: 10.1093/hmg/ddg014 Murray JT, Campbell DG, Morrice N, Auld GC, Shpiro N, Marquez R, Peggie M, Bain J, Bloomberg GB, Grahammer F, Land F, Wulff P, Kuhl D, Cohen P. 2004. Exploitation of KESTREL to identify NDRG family members as physiological substrates for SGK1 and GSK3. Biochemical Journal 384:477–488. doi: 10.1042/ BJ20041057 Naguib A, Bencze G, Cho H, Zheng W, Tocilj A, Elkayam E, Faehnle CR, Jaber N, Pratt CP, Chen M, Zong W-X, Marks MS, Joshua-Tor L, Pappin DJ, Trotman LC. 2015. PTEN functions by recruitment to cytoplasmic vesicles. Molecular Cell 58:255–268. doi: 10.1016/j.molcel.2015.03.011 Nishimura T, Honda H, Takeichi M. 2012. Planar cell polarity links axes of spatial dynamics in neural-tube closure. Cell 149:1084–1097. doi: 10.1016/j.cell.2012.04.021 Palazzo A, Ackerman B, Gundersen GG. 2003. Cell biology (communication arising): tubulin acetylation and cell motility. Nature 421:230. doi: 10.1038/421230a Park KK, Liu K, Hu Y, Smith PD, Wang C, Cai B, Xu B, Connolly L, Kramvis I, Sahin M, He Z. 2008. Promoting axon regeneration in the adult CNS by modulation of the PTEN/mTOR pathway. Science 322:963–966. doi: 10. 1126/science.1161566 Pearce LR, Komander D, Alessi DR. 2010. The nuts and bolts of AGC protein kinases. Nature Reviews Molecular Cell Biology 11:9–22. doi: 10.1038/nrm2822 Pinner S, Sahai E. 2008. PDK1 regulates cancer cell motility by antagonising inhibition of ROCK1 by RhoE. Nature Cell Biology 10:127–137. Epub 2008 Jan 20. doi: 10.1038/ncb1675 Rodriguez-Boulan E, Macara IG. 2014. Organization and execution of the epithelial polarity programme. Nature Reviews Molecular Cell Biology 15:225–242. doi: 10.1038/nrm3775 Saal LH. 2005. PIK3CA mutations correlate with hormone receptors, node metastasis, and ERBB2, and are mutually exclusive with PTEN loss in human breast carcinoma. Cancer Research 65:2554–2559. doi: 10.1158/ 0008-5472-CAN-04-3913 Sarbassov DD, Guertin DA, Ali SM, Sabatini DM. 2005. Phosphorylation and regulation of Akt/PKB by the rictor- mTOR complex. Science 307:1098–1101. doi: 10.1126/science.1106148 Schoenwolf GC, Alvarez IS. 1989. Roles of neuroepithelial cell rearrangement and division in shaping of the avian neural plate. Development 106:427–439. Shewan A, Eastburn DJ, Mostov K. 2011. Phosphoinositides in cell architecture. Cold Spring Harbor Perspectives in Biology 3:a004796. doi: 10.1101/cshperspect.a004796 Simeone A, Gulisano M, Acampora D, Stornaiuolo A, Rambaldi M, Boncinelli E. 1992. Two vertebrate homeobox genes related to the drosophila empty spiracles gene are expressed in the embryonic cerebral cortex. The EMBO Journal 11:2541–2550. Sommer EM, Dry H, Cross D, Guichard S, Davies BR, Alessi DR. 2013. Elevated SGK1 predicts resistance of breast cancer cells to akt inhibitors. Biochemical Journal 452:499–508. doi: 10.1042/BJ20130342 Song MS, Salmena L, Pandolfi PP. 2012. The functions and regulation of the PTEN tumour suppressor. Nature Reviews Molecular Cell Biology 13:283–296. doi: 10.1038/nrm3330 Sperow M, Berry RB, Bayazitov IT, Zhu G, Baker SJ, Zakharenko SS. 2012. Phosphatase and tensin homologue (pTEN) regulates synaptic plasticity independently of its effect on neuronal morphology and migration. The Journal of Physiology 590:777–792. doi: 10.1113/jphysiol.2011.220236 Stambolic V, Suzuki A, de la Pompa JL, Brothers GM, Mirtsos C, Sasaki T, Ruland J, Penninger JM, Siderovski DP, Mak TW. 1998. Negative regulation of PKB/Akt-dependent cell survival by the tumor suppressor PTEN. Cell 95:29–39. doi: 10.1016/S0092-8674(00)81780-8 Suzuki A, de la Pompa JL, Stambolic V, Elia AJ, Sasaki T, Barrantes Ive´n del Barco, Ho A, Wakeham A, ltie A, Khoo W, Fukumoto M, Mak TW. 1998. High cancer susceptibility and embryonic lethality associated with mutation of the PTEN tumor suppressor gene in mice. Current Biology 8:1169–1178. doi: 10.1016/S0960-9822 (07)00488-5 Suzuki M, Morita H, Ueno N. 2012. Molecular mechanisms of cell shape changes that contribute to vertebrate neural tube closure. Development, Growth & Differentiation 54:266–276. doi: 10.1111/j.1440-169X.2012. 01346.x Tanaka A, Miyamoto K, Minamino N, Takeda M, Sato B, Matsuo H, Matsumoto K. 1992. Cloning and characterization of an androgen-induced growth factor essential for the androgen-dependent growth of mouse mammary carcinoma cells. Proceedings of the National Academy of Sciences of the United States of America 89:8928–8932. doi: 10.1073/pnas.89.19.8928 Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 21 of 22 Research article Cell biology Developmental biology and stem cells    206         Trotman LC, Niki M, Dotan ZA, Koutcher JA, Di Cristofano A, Xiao A, Khoo AS, Roy-Burman P, Greenberg NM, Dyke TV, Cordon-Cardo C, Pandolfi PP, Hastie N. 2003. Pten dose dictates cancer progression in the prostate. PLoS Biology 1:e59. doi: 10.1371/journal.pbio.0000059 Vasudevan KM, Barbie DA, Davies MA, Rabinovsky R, McNear CJ, Kim JJ, Hennessy BT, Tseng H, Pochanard P, Kim SY, Dunn IF, Schinzel AC, Sandy P, Hoersch S, Sheng Q, Gupta PB, Boehm JS, Reiling JH, Silver S, Lu Y, Stemke-Hale K, Dutta B, Joy C, Sahin AA, Gonzalez-Angulo AM, Lluch A, Rameh LE, Jacks T, Root DE, Lander ES, Mills GB, Hahn WC, Sellers WR, Garraway LA. 2009. AKT-independent signaling downstream of oncogenic PIK3CA mutations in human cancer. Cancer Cell 16:21–32. doi: 10.1016/j.ccr.2009.04.012 Walker KS, Deak M, Paterson A, Hudson K, Cohen P, Alessi DR. 1998. Activation of protein kinase B beta and gamma isoforms by insulin in vivo and by 3-phosphoinositide-dependent protein kinase-1 in vitro: comparison with protein kinase B alpha. Biochemical Journal 331:299–308. doi: 10.1042/bj3310299 Waterman-Storer CM, Salmon WC, Salmon ED. 2000. Feedback interactions between cell-cell adherens junctions and cytoskeletal dynamics in newt lung epithelial cells. Molecular Biology of the Cell 11:2471–2483. doi: 10.1091/mbc.11.7.2471 Wilkinson DG, Bhatt S, Chavrier P, Bravo R, Charnay P. 1989. Segment-specific expression of a zinc-finger gene in the developing nervous system of the mouse. Nature 337:461–464. doi: 10.1038/337461a0 Wilkinson DG, Bhatt S, Herrmann BG. 1990. Expression pattern of the mouse T gene and its role in mesoderm formation. Nature 343:657–659. doi: 10.1038/343657a0 Williams M, Yen W, Lu X, Sutherland A. 2014. Distinct apical and basolateral mechanisms drive planar cell polarity-dependent convergent extension of the mouse neural plate. Developmental Cell 29:34–46. doi: 10. 1016/j.devcel.2014.02.007 Worby CA, Dixon JE. 2014. PTEN. Annual Review of Biochemistry 83:641–669. doi: 10.1146/annurev-biochem- 082411-113907 Ybot-Gonzalez P, Savery D, Gerrelli D, Signore M, Mitchell CE, Faux CH, Greene NDE, Copp AJ. 2007. Convergent extension, planar-cell-polarity signalling and initiation of mouse neural tube closure. Development 134:789–799. doi: 10.1242/dev.000380 Yue Q, Groszer M, Gil JS, Berk AJ, Messing A, Wu H, Liu X. 2005. PTEN deletion in bergmann glia leads to premature differentiation and affects laminar organization. Development 132:3281–3291. doi: 10.1242/dev. 01891 Zallen JA, Zallen R. 2004. Cell-pattern disordering during convergent extension in Drosophila. Journal of Physics: Condensed Matter 16:S5073–S5080. doi: 10.1088/0953-8984/16/44/005 Zhou J, Parada LF. 2012. PTEN signaling in autism spectrum disorders. Current Opinion in Neurobiology 22:873– 879. doi: 10.1016/j.conb.2012.05.004 Zoncu R, Efeyan A, Sabatini DM. 2011. MTOR: from growth signal integration to cancer, diabetes and ageing. Nature Reviews Molecular Cell Biology 12:21–35. doi: 10.1038/nrm3025 Grego-Bessa et al. eLife 2015;5:e12034. DOI: 10.7554/eLife.12034 22 of 22 Research article Cell biology Developmental biology and stem cells 207 SUMMARY OF RESULTS 208 209 Research in our laboratory is aimed to identify mechanisms of resistance to targeted therapies. Among some of these targeted agents, we are currently testing the clinical antitumor activity of a novel PI3Kα inhibitor, BYL719, in patients with tumors harboring activating PI3Kα mutations. During the study, we studied the tumor genomic evolution in a patient with metastatic breast cancer bearing an activating PIK3CA mutation. The patient was treated with BYL719 for over 10 months, during which she achieved a partial response. However, the patient eventually progressed to treatment and developed new metastases in the lungs, dying shortly after termination of the study. A total of 14 metastatic sites were collected during a rapid “warm” autopsy. Importantly, some of these lesions were still responding to BYL719 treatment at the time of disease progression. According to the clinical protocol and RECIST criteria, when at least one tumor lesion increases in volume ≥20% compared to the baseline, the disease is classified as progressive and the experimental treatment has to be withdrawn. In order to systematically identify possible genetic determinants of acquired resistance to PI3Kα inhibition, we first examined both the archival primary tumor (prior BYL719 treatment) and the new lung metastasis by whole genome sequencing to analyze possible genetic alterations. Although both samples shared many somatic genetic aberrations, a new PTEN loss mutation was detected in the lung metastasis compared to the primary tumor. To confirm and expand our findings, we sequenced the primary tumor and all the metastatic lesions to >500-fold coverage using a custom targeted deep-sequencing assay available at our institution called IMPACT, which allows for the study of approximately 400 actionable genes related to cancer. We found that many mutations were shared between the primary and the metastatic tumor sites whereas others were observed only in some or all of the metastatic lesions. We confirmed that the PIK3CA E542K mutation of the primary tumor was conserved among the metastatic samples and detected another PIK3CA mutation (D725G) in all the samples. Moreover, we found increased copy number of FGFR1 and EI4EBP1 in all tumor samples, consistent with a relatively frequent amplification of the 8p11-12 loci described in breast cancers (Gelsi-Boyer et al., 2005). Central to our work, all metastatic lesions appeared to harbor a single copy loss of PTEN. In addition, we found that 10 out of 14 metastatic lesions harbored additional genomic alterations within PTEN. The spectrum of PTEN alterations was heterogeneous 210 across the 10 samples and included a splice site mutation at K342, a frameshift indel at P339, and 4 different exon-level deletions. We found that all lesions with genomic alteration of PTEN had lost the protein expression as measured by IHC, and these lesions were the same ones that did not respond during BYL719 administration. The four remaining samples with an intact PTEN were still sensitive to BYL719 at the time of study termination. Based on the sequencing data, we can predict the genomic progression of the disease and propose an approximate phylogeny of the tumor evolution, corroborating for the first time the now widely accepted hypothesis that tumors follow a phenomenon know as convergent evolution under selective therapeutic pressure. Tumors that are genetically different escape the pharmacologic inhibition using mechanisms that converge into the same phenotype; in this case, the escape mechanism is the loss of PTEN. Previous work has shown that in the absence of PTEN, cancer cells become predominantly dependent on the activity of the PI3K p110β subunit (PI3Kβ) to propagate signaling through the PI3K/AKT pathway (Jia et al., 2008). Furthermore, in PTEN- negative tumor xenografts, targeting of PI3Kβ is required to inhibit the PI3K/AKT pathway (Wee et al., 2008). Overall, these findings suggest that patients with PTEN negative tumors may preferentially benefit from inhibitors that are capable of optimally targeting PI3Kβ. To recapitulate our clinical findings that PTEN loss is a possible mechanism of acquired resistance to selective PI3Kα inhibition in the laboratory, we have generated doxycycline-inducible PTEN shRNA stable clones starting from three different BYL719- sensitive cell lines. In two PIK3CA-mutant breast cancer cell lines sensitive to BYL719 (T47D and MCF-7), when PTEN is knocked down we observed a consequent hyperactivation of the PI3K/AKT/mTOR pathway. In addition these cells become resistant to BYL719 when PTEN is knocked down. However, when these cells are treated with either a pan-PI3K inhibitor (which targets all PI3K isoforms) or the combination of a PI3Kα and a PI3Kβ inhibitor, PTEN knockdown does not affect cell viability. This suggests that PTEN loss indeed signals through PI3Kβ to activate downstream pathways such as AKT and mTORC1. Then, we tested whether these observations were also valid in the patient’s autopsy samples. To address this issue, we generated a patient-derived xenograft (PDX) from 211 the BYL719-resistant and PTEN loss lung metastasis. We found that these tumors were refractory to the antitumor activity of BYL719. Treatment with BKM120, which targets all isoforms of PI3K, resulted in strong inhibition of tumor growth. Next, we analyzed the tumors by IHC and found that only BKM120 was capable of inhibiting downstream signaling, as shown by pAKT (S473) and pS6 (S240/4) staining, two surrogates of AKT and mTORC1 activity, respectively. These results were also observed in vivo when we used the PI3Kβ inhibitor AZD6482. Moreover, to study if our findings are found in other patients, we analyzed paired samples (pre- and post-treatment) from nine additional patients treated with BYL719 by IHC and targeted sequencing and observed PTEN loss at therapy progression in two additional cases. These results can be found in Article 1. In patients treated with PI3Kα inhibitors, intrinsic resistance is also observed, indicating that although the tumors are driven by the oncogenic PIK3CA mutation, some molecular determinants might limit the efficacy of this agent (Elkabets et al., 2015; Elkabets et al., 2013). To study this process, we took advantage of two different breast cancer cell lines, HCC1954 and JIMT1, which are PIK3CA mutant and BYL719-resistant. As described previously, these cells maintain mTORC1 activity when treated with BYL719 although AKT signaling is completely inhibited (Elkabets et al., 2013). In order to study the mechanism underlying this process, we used these cells as a model and performed a synthetic lethal siRNA screening with a library against the human kinome and phosphatome. Cells were transfected using the kinome and phosphatome v4.0 libraries from Ambion, which contain an average of 3 different siRNA duplexes per each gene (approximately 3000 duplex). This screening was run using two experimental conditions: DMSO and 1 µM BYL719 and after 6 days of treatment cell viability was assessed. From the screen, we selected candidate genes whose knockdown in combination with BYL719 decreased viability. These candidates were chosen using an analysis strategy involving a stringent, five-step process called the Bhinder-Djaballah (BDA) analysis method (Bhinder and Djaballah, 2012). A total of five candidate genes that overlapped between the two cells lines were identified. Among them, MTOR (mTOR) was considered our positive control as it validated our previous finding that inhibition of both PI3Kα and mTOR is required to 212 achieve significant antiproliferative activity in these cells. We found 4 additional genes whose knockdown sensitized the resistant cells to BYL719: PDPK1, PIK3CA, PPP1R12A, and FLJ16165. Next, we validated our hits by running a second small-scale screening that contained active siRNA against the genes identified and assessed not only viability, but also pS6 (S240/4) as a read-out of mTORC1 activity. Amongst the genes tested, only MTOR and PDPK1 knockdown were able to decrease viability, cell count, and pS6 (S240/4) staining in the presence of BYL719. To validate our screening findings, we established a stable PDPK1 knockdown using shRNA in HCC1954 and JIMT1 cells. We found a decrease in cell viability and inhibition of mTORC1 upon BYL719 treatment. Moreover, in vivo treatment of HCC1954-derived xenografts with BYL719 translates into higher antitumor activity when PDK1 expression is knocked down. Although RNAi technology allows for genome-wide discovery and validation of cancer targets, translation into the clinical setting is still premature. Since the main goal of our laboratory is to translate lab discoveries into the clinically applicable findings, we have taken advantage of the commercially available PDK1 inhibitor GSK2334470 (GlaxoSmithKline). This is a highly specific PDK1 inhibitor with an IC₅₀ of ~10 nM that does not suppress the activity of other protein kinases (Najafov et al., 2011). We observed little inhibitory effects on AKT phosphorylation but a moderate suppression of p70S6K (pS6K) and S6 phosphorylation (pS6) using GSK2334470 in HCC1954 cells. However, only the combination of BYL719 and GSK2334470 achieved strong suppression of both pAKT and pS6, confirming that PDK1 is a key mediator for the sustained mTORC1 activity upon PI3Kα inhibition. These biochemical effects translated into synergistic inhibition of cell viability when JIMT1 and HCC1954 cells were treated with the combination of the two agents. More important, in vivo treatment of xenografts with GSK2334470 did not show major effects on tumor growth or mice health, but a profound antitumor activity in combination with BYL719. These results suggest that PDK1 inhibition might be a suitable target in combination with p110α inhibition in the clinical setting, especially for tumors that harbor PIK3CA mutations and remain insensitive to BYL719. We also performed gene expression analysis of both JIMT1 and HCC1954 cells treated with BYL719, GSK2334470, or the combination of both drugs. Gene Set Enrichment Analysis (GSEA) revealed a significant increase in the FOXO transcription factor 213 signature in the combination. In fact, most of the previously described FOXO gene targets are upregulated with the combination and were validated using RT-qPCR. The increase in gene transcription was found to be consistent with an increased occupancy of FOXO3 in the promoter of the target genes TNFSF10 and IRS2, as assessed by Chromatin Immunoprecipitation (ChIP) assays. Also, the combination of both drugs induces a decrease in the phosphorylation of FOXO1/3 at residues T24 and T32, respectively, which results in the nuclear translocation. AKT has been shown to phosphorylate FOXO1/3 at T24 and T32 residues, respectively, causing its dissociation from 14-3-3 proteins and consequent nuclear shuttling (Webb and Brunet, 2014). However, we observed that despite full inhibition of AKT by PI3Kα inhibition, FOXO3 was not efficiently primed to migrate to the nucleus and exert its transcriptional activity in cells resistant to BYL719, indicating that another kinase is responsible for the phosphorylation of FOXO3. Since PDK1 requires downstream AGC kinases as molecular effectors (Arencibia et al., 2013; Pearce et al., 2010), we reasoned that in BYL719-resistant cells, a downstream AGC kinase dependent on the PDK1 catalytic activity regulates both FOXO1/3 phosphorylation and mTORC1 activity, independently of AKT. Serum and glucocorticoid-induced kinase (SGK) is a family of AGC serine/threonine kinases that comprises three members (SGK1, SGK2, and SGK3) highly homologous to AKT, sharing 55% identity in the kinase domain (Kobayashi and Cohen, 1999; Kobayashi et al., 1999). SGK1 activation is mediated by mTORC2-dependent phosphorylation at the hydrophobic motif (S422) and subsequent PDK1 phosphorylation at the activation loop (T256) (Garcia-Martinez and Alessi, 2008). Earlier reports have demonstrated that SGK1 is able to directly phosphorylate FOXO1 at residues T32 and S315 (Brunet et al., 2001). Furthermore, SGK1 has been correlated with resistance to AKT inhibition (Sommer et al., 2013). Therefore, we hypothesized that SGK1 plays a critical role downstream of PDK1 in sustaining mTORC1 activity and inducing resistance to PI3Kα inhibition in our models. We analyzed the basal mRNA expression of 27 breast cancer cell lines, including cell lines previously characterized as sensitive or resistant to BYL719, and determined that resistant cell lines had higher levels of SGK1 mRNA compared to sensitive cells. This held true when only in breast cancer cells harboring PIK3CA-activating mutations, which are known to be sensitive to PI3Kα inhibition, were considered in the analysis. The 214 mRNA levels of SGK2 and SGK3 were similar between sensitive and resistant cell lines. The ratio of phosphorylated N-Myc Downstream Regulated 1 (NDRG1) (T346), a specific substrate of SGK1 (Murray et al., 2004), versus total NDRG1 was also higher in BYL719-resistant cells. We then aimed to investigate the variability of SGK1 expression in breast tumors. Analysis of SGK1 mRNA levels in the TCGA breast cancer patient cohort revealed that about 15% of breast tumors express high levels of SGK1, independently of the PIK3CA status (Fisher’s exact test, p=0.4) (Ciriello et al., 2015). To explore the correlation between the SGK1 activity surrogate pNDRG1 and clinical outcome to PI3Kα inhibition, we secured tumor samples from 17 breast cancer patients harboring PIK3CA mutations that are currently participating in a clinical trial at our institution evaluating the activity of BYL719 in combination with an aromatase inhibitor (NCT01870505). Three of these tumors were positive for pNDRG1 expression while the remaining fourteen tumors had no detectable staining. The patients with tumors positive for pNDRG1 did not respond to therapy and their disease progressed rapidly. On the contrary, in the pNDRG1 negative group, we observed three partial responses by RECIST criteria, in addition to a longer time to disease progression when compared to the pNDRG1 positive tumors. These results in this small sample size are suggestive of a role of SGK1 in mediating intrinsic resistance to PI3Kα inhibitors. In resistant cell lines, only the combination of BYL719 and GSK2334470 was able to decrease the phosphorylation of NDRG1, confirming that the combination of both drugs is required in order to effectively inhibit both endogenous SGK1 and AKT activity. When we immunoprecipitated endogenous SGK1, we found that BYL719 treatment was not sufficient to completely abolish the kinase activity of SGK1, in contrast to GSK2334470. On the other hand, immunoprecipitation of endogenous AKT revealed that while BYL719 treatment completely abrogated AKT kinase activity, this is not the case when cells are treated with GSK2334470, as previously observed. Due to its similarity with AKT, we reasoned that SGK1 could modulate mTORC1 activity by interacting with a component of the TSC/RHEB/mTORC1 axis. Immunoprecipitation of recombinant Flag-tagged TSC1, TSC2, RHEB, and mTOR in 293T cells revealed that SGK1 physically interacts with both mTOR and TSC2 proteins. While the interaction between SGK1 and mTOR has previously been described, to our knowledge, this is the first report showing an interaction between SGK1 and TSC2. We further confirmed the novel interaction between SGK1 and TSC2 by endogenous co-immunoprecipitation in 215 JIMT1 cells. Co-immunoprecipitation assays using five different fragments of TSC2 demonstrated that SGK1 binds its N-terminal region, found between amino acids 1-608. SGK1 has high similarity to AKT in the kinase domain and thus shares many substrates that contain the AGC-kinase consensus motif RXRXX(S/T), where R is Arginine, X is any amino acid, and (S/T) is a phosphorylatable Serine or Threonine. When we analyzed the TSC2 protein sequence searching for identifiable RXRXX(S/T) motifs, we found seven putative sites of phosphorylation: S939, S981, T993, S1130, S1132, T1462, and S1798. All these sites were conserved across lower species. To systematically test the ability of SGK1 to phosphorylate these residues, we established an in vitro kinase assay using recombinant active SGK1 and TSC2 as a substrate, immunoprecipitated from 293T cells expressing Flag-TSC2. The addition of recombinant SGK1 kinase increased the phosphorylation of the RXRXX(S/T) sites of TSC2. Using mass spectrometry to identify the phosphorylation status of the aforementioned residues in our in vitro kinase assay, we found increased phosphorylation in all these sites, except at T993. Mutation of these six phosphorylatable sites into the non-phosphorylatable amino acid alanine (TSC2 6A) completely abrogated the ability of SGK1 to phosphorylate TSC2 in vitro. SGK1 and AKT share the capability of phosphorylating five of these six sites in TSC2. It is well accepted that phosphorylation of these residues by AKT increases the downstream RHEB-GTP loading and mTORC1 signaling (Huang and Manning, 2008; Inoki et al., 2003). To confirm that our biochemical findings are consistent with the proposed mechanism of resistance to BYL719, we treated HCC1954 and JIMT1 cells with BYL719, GSK2334470, SGK1-inh, and the combination of these agents and found that the phosphorylation of endogenous TSC2 decreases only upon dual PI3Kα and PDK1 or SGK1 suppression. These results demonstrate that SGK1 can sustain mTORC1 activity in BYL719-resistant cells by phosphorylating and inhibiting the mTORC1 negative regulator TSC2. These results can be found in Article 2. Finally, to further pursue our objective of developing novel mouse models of PIK3CA to study the role of PI3K in oncogenicity, we used the previously described mouse model R26-LSL-PIK3CAH1047R, which expresses the PIK3CA H1047R mutation under the control of the constitutive Rosa26 promoter once the loxP-Stop-loxP sequence is removed by means of Cre recombinase (Adams et al., 2011). Breeding of the R26-LSL- 216 PIK3CAH1047R transgenic mouse with tissue specific Cre strains allows the expression of the activated PI3K allele in the tissue of interest. Aiming to develop a uterine mouse model, we crossed the R26-LSL-PIK3CAH1047R mouse with the Sprr2f-Cre strain, which has been reported to express Cre in the epithelial cells of the uterus (Contreras et al., 2010). Unexpectedly, while PIK3CASprr2f-WT mice were viable and normal, PIK3CASprr2f-Cre littermates showed hind limb paralysis at an early age (4-10 weeks). Histologic examination revealed lesions in the spinal cord resembling human venous malformations (VM). Specifically, these abnormalities showed dilated ‘cavernous’ vascular spaces with extensive blood pools and hemorrhage involving both white and grey matter in the spine. Intravenous injection of gold nanoparticles in PIK3CASprr2f-Cre mice confirmed the presence of hyperdense lesions in the spine by contrast and X-ray computed tomography imaging. These lesions were present in animals with both advanced and milder phenotype, showing slow blood flow and extravasation. Blood vessel alterations were also detected microscopically in other organs including lung, adrenal gland, epididymis, and skin. To initially study the cellular mechanisms by which PIK3CA mutations might alter endothelial cell (EC) function, primary human skin EC were transduced with PIK3CA WT or H1047R mutant expressing vectors. Mutant cells exhibited increased downstream PI3K signaling, failed to generate normal vascular tubes in vitro, and proliferated at a higher rate than WT cells. Next, in order to address if PIK3CA mutations are involved in human pathogenesis, we obtained VM samples from patients (n=32). VM lesions were analyzed by targeted- exome sequencing of 341 cancer genes (MSK-IMPACT assay) (Cheng et al., 2015), as well as targeted sequencing for the TEK locus. Deep sequencing detected PIK3CA mutations in 25% of cases in previously described hot spots (H1047R, E542K). In addition, we found gain-of-function mutations in other genes related to the PI3K pathway such as AKT2, AKT3, and IRS2, resulting in an overall 32% of mutations in the PI3K/AKT pathway (Fig. 10F). Furthermore, we found mutations in the MAPK pathway (GNAQ, NF1, MAP2K1, MAP3K1) in 13% of the cases. Previously described mutations in the tyrosine kinase receptor TEK were found in 35% of the patients and were mutually exclusive with the mutations in the PI3K and MAPK pathways (Limaye et al., 2009). 217 We then generated PIK3CACAG-CreER mice in which the PIK3CA H1047R mutation is ubiquitously expressed upon tamoxifen administration. Six to eight week-old mice fed with tamoxifen rapidly developed cutaneous VM compared to the PIK3CAWT littermates. Histology confirmed a combined capillary and cavernous phenotype, immunoreactive for CD31 and phosphorylated AKT (S473), a surrogate marker of PI3K activation. Similar to humans, mice VM contained high levels of hemosiderin. Although the skin phenotype was readily apparent, additional lesions were observed at necropsy at multiple sites including mesentery, genitourinary tract, kidney, and retina. The presence of PIK3CA mutations in VM, together with the observed phenotypes, prompted us to evaluate whether these lesions, despite being considered vascular malformations, are in fact tumorigenic. To this end, we subcutaneously injected PIK3CACAG-CreER VM cells into recipient immunocompromised nude mice. They induced highly vascular and proliferative tumors a few weeks after injection, with a histology highly resembling that of the original lesion. Of clinical relevance, the presence of activating PIK3CA mutations in VM opens the door to treatment of this condition with PI3Kα inhibitors, currently under clinical development for cancer indications. Treatment of these vascular tumors with the PI3Kα selective inhibitor BYL719 resulted in a marked response as measured by a decrease in tumor volume, decrease in proliferation, and increased apoptosis. These results can be found in Article 3. 218 219 DISCUSSION   220 221 Drug resistance is a way in which cancer cells develop tolerance to therapeutic amounts of an inhibitor, thereby hindering the development of novel targeted treatments. In the present dissertation we studied how cancer cells adapt to PI3K inhibition nullifying the antitumor activity of inhibitors of this pathway. The identification of the mechanisms involved in the resistance to these agents is required in order to propose combinatorial therapies that will improve their clinical benefit. PI3K inhibitors are currently being investigated in the clinical setting for the therapy of multiple malignancies that exhibit hyperactivation of the PI3K pathway (Rodon et al., 2013). Among these aberrations, activating mutations in PIK3CA, the gene encoding the PI3Kα isoform, are frequent in breast, lung, cervical, colorectal, and other cancers. These mutations are commonly found in the helical and catalytic domain of PI3Kα and have been shown to increase the downstream signaling of the pathway, leading to increased proliferation, cell growth, metabolism, and malignant properties in the cancer cell (Courtney et al., 2010). Therefore, the development of PI3K inhibitors that potently and selectively inhibit this oncogenic signal is therapeutically attractive (Engelman, 2009). Accordingly, initial first-in-human clinical trials have shown remarkable responses in metastatic patients harboring PIK3CA mutations (NCT01219699; NCT01296555). Phase II and III clinical trials are currently ongoing to assess the efficacy of these therapies in combination with other treatments, such as anti-estrogen therapy in hormone-positive breast cancers. Previous studies using the mTORC1 inhibitor everolimus in postmenopausal women with hormone-positive metastatic cancer has shown increased progression-free survival in Phase III clinical trials (Baselga et al., 2012). These results were corroborated by recent data elucidating the crosstalk between the ER and the mTORC1 pathway in preclinical models of breast cancer. The PI3K and ER pathways are the most commonly altered signaling networks in breast cancer (Bjornstrom and Sjoberg, 2005; Dhillon, 2013). Since both pathways play a role in promoting cell growth and survival in breast cancer cells, we hypothesized that the inhibition of one could lead to the activation of the other. This proved to be the case, since treatment with the PI3Kα isoform specific inhibitor BYL719 in ER-positive PIK3CA- mutant cell lines induces a significant ER-dependent transcriptional change (Article 5). The increase in the mRNA levels of the ER gene targets can be explained by both increase expression of ER itself and increased occupancy of ER transcription factor in the promoters of ER-target genes such as PGR and GREB1 (Hall et al., 2001). This transcriptional rewiring was very specific as the treatment of cells with the ER degrader 222 Fulvestrant abolished the induction of ER targets induced upon PI3K inhibition. The combination of PI3K inhibitor and Fulvestrant has the ability to overcome the acquired resistance of these cells and may become a new standard of care for this disease. A phase Ib clinical trial testing the combination of GDC0032 (taselisib), a potent PI3Kα/δ inhibitor, with fulvestrant has shown remarkable responses in the metastatic setting (Figure 17) and has been the basis for the design of a Phase III registration clinical trial. In this trial, also known as the SANDPIPER study (NCT02340221), two experimental arms (fulvestrant + placebo and fulvestrant + GDC0032) are tested in women with estrogen receptor-positive and HER2-negative locally advanced or metastatic breast cancer who present recurrent disease or progression during or after aromatase inhibitor therapy. This is a multicentric, double-blinded study in which the main outcome measured will be PFS and is expected to yield results in 3 to 4 years. Figure 17: Phase Ib clinical trial of fulvestrant with GDC0032 Waterfall plot indicates the best change in the tumor volume of patients treated with 6 or 9 mg QD of GDC0032 in combination with fulvestrant. The graph also indicates the presence of PIK3CA mutations (kinase or helical), the best tumor response (cPR: confirmed partial response, PR: Partial response, SD: Stable disease, and PD: Progressive disease), and the time in the study. Source: (Juric et al 2015, unpublished results). 223 A parallel with prostate cancer can be drawn, in which a reciprocal feedback regulation of PI3K and androgen receptor (AR) activity has been also described (Carver et al., 2011). In the case of prostate cancer, inhibition of the PI3K pathway results in activation of AR transcriptional activity and, conversely, blockade of AR activates PI3K signaling. This bidirectional crosstalk seems to occur also in the breast, where, in addition to our findings showing ER activation upon PI3K inhibition, it was reported that inactivation of ER appears to be associated with activation of PI3K signaling (Morrison et al., 2013). This highlights the interdependency between these two pathways, where a state of equilibrium between PI3K and ER signaling is reached to ensure cell survival. Thus, both prostate and breast cancer cells may adapt to suppression of the PI3K pathway by increasing their dependence on the hormone receptor function. Not all the breast cancers treated with PI3K inhibitors are driven by the presence of activating PIK3CA mutations and, certainly, not all the cases have the capability to upregulate the ER pathway to sustain cell survival. For instance, TNBC do not express hormone receptors and therefore the signaling through this survival pathway is suppressed (Mayer et al., 2014). TNBC have been described as having a high frequency of inactivation or decreased expression of the gene encoding PTEN, as well as overexpression of the gene encoding human EGFR in up to about 50% of cases (Hudis and Gianni, 2011). Therefore, the treatment with pan-PI3K or AKT inhibitors can be of interest in the cases in which the lack of PTEN drives the hyperactivation of the PI3K/AKT pathway (Courtney et al., 2010). Using TNBC-derived cell lines and patient- derived xenografts, we have shown that inhibition of the PI3K/AKT pathway induces the upregulation of EGFR/ERBB3-dependent signaling and this can be exploited as a molecular target in combination with a dual antibody targeting EGFR and HER3, namely MEHD7945A (Article 4). These results are in line with the observations made by Chandarlapaty et al, in which the inhibition of AKT kinase induces ERBB3 mRNA upregulation as a result of a FOXO-dependent transcription (Chandarlapaty et al., 2011). Treatment with cetuximab, a monoclonal antibody that targets EGFR, is currently approved for the treatment of several malignancies that are characterized by the overexpression of EGFR, such as head and neck and colorectal cancer (Bonner et al., 2006; Jonker et al., 2007). However, the treatment that would be more suitable for TNBC patients that exhibit EGFR overexpression and lack PTEN is not currently known. PTEN loss is a determinant of resistance to many anti-RTK therapies, since the hyperactivation 224 of the downstream PI3K/AKT pathway would compensate the inhibitory effects of targeting HER2 or EGFR (Dillon and Miller, 2014). On the other hand, inhibition of PI3K or AKT results in a hyperactivation of ERBB-dependent signaling, mainly by the upregulation of HER3, leading to the activation of downstream pathways such as MAPK and STAT that limit the sensitivity to PI3K/AKT inhibitors (Chandarlapaty et al., 2011). Hence, dual inhibition of EGFR and HER3 would be required to abolish the resistant phenotype. Therapy with antibodies is of particular interest because pharmacologically these agents are highly specific and do not show substantial toxicities. Moreover, the mechanism of action not only includes the inhibition of the target, but also is often mediated by Antibody-Dependent Cellular Cytotoxicity (ADCC) (Scott et al., 2012). Although most of the mechanisms of acquired resistance rely on the upregulation of parallel pathways that sustain tumor growth, such as the examples discussed above, sometimes these mechanisms are a result of a clonal selection of resistant populations (Schmitt et al., 2015). This concept is highly relevant in the case of advanced or metastatic tumors, in which the extent tumor burden is likely to contain heterogeneous populations that are genomically different (Arnedos et al., 2015). In order to study tumor evolution in response to PI3K inhibition in metastatic breast cancer, we analyzed the metastatic lesions from an exceptional responder patient enrolled in the BYL719 clinical trial. A rapid autopsy was performed and a total of 14 metastases with tumor cells present and differential response to the therapy were identified and collected for sequencing (Article 1). By means of NGS, we identified that the lesions that progressed to the therapy had loss of the tumor suppressor PTEN, while the lesions that still responded to the therapy had no alteration in this gene. Importantly, the mutations found in PTEN were different among the lesions (including SNP, CNV at different exons, and splice mutations) but shared the same phenotypic outcome — the loss of function of the phosphatase PTEN. This observation was not particularly surprising taking into account that PTEN is among the most commonly mutated tumor suppressors in human cancers along with p53 (Hollander et al., 2011). The novelty (and relevance) of our observations resides in the fact that, phylogenetically, all the lesions that were resistant to the PI3Kα inhibitor BYL719 exhibited a convergent phenotypic loss of PTEN. The findings presented are of significant interest in the field of tumor genetic evolution as this is the first report 225 providing evidences that upon therapeutic pressure, clones that are resistant to therapy are selected and accomplish this bypass via a similar mechanism. Parallel convergent evolution under selective pressure has been described in conditions where treatments are highly efficacious, such as in HIV (Calmy et al., 2004). Our case highlights that this tumor, despite its heterogeneity, was dependent on PI3K signaling, probably as a result of the presence of the same activating PIK3CA trunk mutation in all the tumor sites. Upon continued suppression of PI3Kα, diverse genomic alterations emerged, leading to PTEN loss as an alternative mechanism of PI3K activation. In the absence of PTEN, cancer cells become dependent mostly on the activity of the p110β isoform of PI3K (PI3Kβ) to propagate signaling through downstream pathway effectors (Thorpe et al., 2015). This was elegantly demonstrated in GEMM of prostate cancer that are driven by PTEN knock out. Prostate tissue-specific PTEN null mice develop hyperplasia and tumors in this gland, which are characterized by increased signaling of the downstream effector AKT (Jia et al., 2008). Crossing these mice with the tissue-specific PIK3CA knockout mice does not abolish the appearance of prostate tumors nor the pAKT signaling. On the other hand, when the mice are crossed with the PIK3CB knockout mice, prostate tumors no longer develop and downstream signaling of the pathway is attenuated. This is also the case in cultured cancer cells where shRNA- mediated knockdown of PIK3CB, but not PIK3CA, disrupts the growth and survival of the PTEN null cell lines U87MG (glioblastoma), PC3 (prostate cancer), and BT549 (TNBC breast cancer) (Wee et al., 2008). Therefore, we hypothesized that progressive decrease or loss of PTEN function in the presence of PI3Kα inhibition might restore PI3K/AKT signaling and survival through PI3Kβ activity. This highlights another important concept known as oncogenic addition, when cancer cells are dependent on the growth and survival advantage promoted by an oncoprotein. We validated our findings by generating isogenic PTEN shRNA breast cancer cell lines that are PIK3CA mutant and ER-positive (in order to mimic the phenotype observed in the patient studied). While all the tested cell lines were highly sensitive to the PI3Kα inhibitor BYL719, the induction of PTEN knockdown significantly increased the EC50 of this compound and re-activated the downstream pathway, as measured by pAKT and pS6 immunoblotting. This resistant phenotype was also observed in a patient-derived xenograft derived from a resistant metastatic lesion harvested at the time of autopsy. Dependency on the PI3Kβ isoform is the cause of the resistant phenotype, providing a therapeutic target that could be exploited with the use of isoform-specific PI3Kβ 226 inhibitors. In fact, our pre-clinical models responded to the combination of both agents or the administration of pan-PI3K inhibitors that target all the isoforms of the PI3K holoenzyme. This n=1 case provides proof-of-concept evidence that indicates PTEN loss as a biomarker of resistance to PI3Kα inhibitors and proposes a therapeutic strategy to overcome the resistant phenotype. Moreover, our study emphasizes the importance of tumor interrogation upon progression to therapy and the dynamic nature of tumor genomes under selective therapeutic pressure. As a follow-up study, the interrogation of samples from a large cohort of patients treated with PI3Kα inhibitors will be required to understand if this is a common mechanism of resistance. PTEN IHC is a feasible and economic histological test that could be used to predict resistance to these agents, providing a window to clinicians to amend the therapeutic approach. Combination of isoform-specific PI3K inhibitors seem to be an important combinatorial strategy, since other reports have suggested that resistant feedback loops aimed to reactivate the PI3K pathway occur upon monotherapy with either PI3Kα or PI3Kβ inhibitors (Cescon et al., 2015). For instance, in luminal breast cancer cells carrying normal PTEN levels, the p110β isoform is responsible for the reactivation of PIP3- dependent signaling after treatment with BYL719 (Costa et al., 2015). Another report in prostate cancer lead to the same conclusions, but in this case, treatment of PTEN loss- cell lines and GEMM with the PI3Kβ inhibitor AZD8186 induced a feedback loop that reactivated the pathway by a dependency on PI3Kα (Schwartz et al., 2015). In fact, PI3Kβ inhibitors are currently being tested in PTEN null malignancies and reactivation of the pathway through other isoforms is a mechanism of resistance that could certainly emerge. Despite the efforts made in trying to understand the acquired mechanisms of resistance to PI3K inhibitors, less is known about intrinsic refractoriness to these agents. This is a critical matter, since early clinical trials have revealed that a subset of patients that are PIK3CA mutant do not respond upfront to PI3Kα inhibition (Rodon et al., 2013). Our laboratory studied this effect in breast cancer cell lines and reported that mTORC1 inhibition is a determinant of sensitivity to these agents. When PI3Kα inhibitors are administered to sensitive cell lines, there is a reduction in the levels of phosphorylated AKT as well as downstream mTORC1 targets pS6 and p4EBP1, indicating that the pathway is fully inhibited. This is not the case in intrinsically resistant cell lines, where 227 the treatment with the PI3Kα inhibitor BYL719 completely blocked PI3Kα activity and downstream AKT signaling but not mTORC1 (Elkabets et al., 2013). In cells treated with BYL719, the levels of pS6, a robust read-out of mTORC1 activity, correlated with poor response to this agent. Similar to the case of PTEN, pS6 IHC can be used as a biomarker to assess the early response to PI3Kα inhibitors by anticipating the clinical benefit that these patients will achieve. These findings are similar to the observations made in other tumor types, in which the mechanism of resistance relies on the activation of downstream pathways. For instance, in the case of RAF inhibitors, a typical mechanism of resistance is the mutation in the immediate downstream effector kinase MEK (Lito et al., 2013), in the same way as BRAF mutations predict resistance to anti- EGFR therapies (Dienstmann et al., 2012). The activation of mTORC1 independently of the AKT axis offers a therapeutic opportunity since, as described in the Introduction, mTORC1 allosteric inhibitors are remarkably potent and specific (Courtney et al., 2010; Fruman and Rommel, 2014). Combination of PI3Kα inhibitors with a rapamycin analogue (everolimus) results in superior anti-tumor effects in xenografts, and a phase Ib clinical trial is currently recruiting patients to test the safety, tolerability, and early responses for this combination (NCT02077933). One might reason that simultaneously inhibiting PI3K and mTORC1, taking advantage of the dual PI3K/mTOR inhibitors, could result in a greater clinical benefit for patients. However, results from clinical trials testing these compounds have revealed that the toxicities associated with dual PI3K/mTOR inhibitors are substantial (Rodon et al., 2013). Moreover, the doses required for efficiently inhibiting mTOR and PI3K using a dual inhibitor are higher than combining a potent isoform-specific PI3Kα inhibitor with low doses of a rapamycin analogue. Despite the therapeutic implications of this combination, the mechanisms underlying the activation of mTORC1 independently of AKT remain elusive. Uncovering the effectors that contribute to this phenomenon would provide further insights into the biology of the PI3K pathway and offer potential pharmacological targets as well as novel biomarkers of resistance. We have undertaken a synthetic lethal screening using RNAi against the human kinome and phosphatome in two cell lines intrinsically resistant to BYL719 (HCC1954 and JIMT1) and identified two candidates (mTOR and PDK1) that, when inhibited with at least two different siRNA sequences, reduced the survival and decreased the levels of 228 pS6 in both resistant cells (Article 2). Further experiments validated that pharmacologic and genetic inactivation of PDK1 reduced the intrinsic resistance to the PI3K inhibitor BYL719. PDK1 can interact with multiple substrates but does not have the ability to phosphorylate effector proteins due to the lack of the consensus motif FCGT (where F is Phe, C is Cys, G is Gly, and T is phosphorylatable Thr) (Arencibia et al., 2013; Mora et al., 2004). These premises suggested that the mechanism of resistance to PI3Kα inhibitors is dependent on an AGC kinase that is inhibited upon PDK1 blockade and has the ability to activate mTORC1 in the absence of PIP3. PDK1 KO experiments in mouse embryonic stem cells revealed that PDK1 regulates multiple AGC kinases and, in the absence of PDK1, the activation loop is dephosphorylated and the activity is abrogated (Lawlor et al., 2002). Furthermore, pharmacological experiments showed that effective doses of PDK1 inhibitor are not sufficient to abolish AKT activity, due to a mechanism of resistance that involves mTORC2 and PIP3 levels, as previously discussed in the Introduction (Najafov et al., 2012). Our results indicate that AKT does not play a role in the resistant phenotype because treatment with the PI3Kα inhibitor reduces the levels of phosphorylated AKT (Ser473 and T308) and the activity of the enzyme, as measured by endogenous kinase assay. Also, the PDK1 inhibitor, which does not significantly inhibit AKT, is active in the resistant cells when combined with PI3Kα inhibition. Transcriptional analysis of cells treated with the combination of PI3Kα and PDK1 inhibitor revealed that FOXO transcription factors are the main transcriptional output. Brunet et al. reported that FOXO transcription factors are phosphorylated by SGK at different sites that are shared with AKT, as these two protein kinases share the consensus phosphorylation motif RXRXX(S/T) (where R is Arg, X is any amino acid, and S/T are phosphorylatable Ser or Thr) (Brunet et al., 1999; Brunet et al., 2001). Since AKT is inactive in our models treated with PI3Kα inhibitor, and SGK1 is downstream of PDK1, we speculated that SGK1 could be the effector kinase responsible for the sustained activation of mTORC1 and the consequent resistant phenotype in these cells. Moreover, another work has described a correlation between cell lines that are intrinsically resistant to AKT inhibitors and the mRNA and protein levels of SGK1 (Sommer et al., 2013). This is also the case in our cohort of BYL719/PI3K inhibitor resistant cell lines, where we observed increased levels of SGK1, but not SGK2 or SGK3. Some evidence also suggests that SGK3 may play an important role in the oncogenicity of PI3K-driven cells (Vasudevan et al., 2009). Our results indicate that this is not the case with intrinsic resistance to PI3K pathway 229 inhibitors since levels of SGK3 were similar between sensitive and resistant cell lines. Moreover, SGK3 is a direct downstream target of class III phosphoinositide 3-kinase and is indirectly inactivated upon class I PI3K inhibitor treatment, so it is an unlikely mediator of intrinsic resistance (Bago et al., 2014). Despite this, in the future it would be interesting to study whether the different members of the SGK family could induce resistance to PI3K inhibitors in models of acquired resistance. SGK1 is regulated by different mechanisms at both the transcriptional and post- transcriptional levels (Webster et al., 1993a). The half-life of the kinase is a result of the presence of a PEST domain, a peptide sequence that is rich in proline (P), glutamic acid (E), serine (S), and threonine (T), which has the ability to target proteins to the proteasome (Brickley et al., 2002). In the promoter of SGK1, there are several binding sites specific for the glucocorticoid receptor (GR), which promotes transcription when bound to glucocorticoids (i.e. dexamethasone) (Webster et al., 1993b). It is interesting to consider that treatment with these agents induces a rapid up-regulation of the mRNA and protein levels of SGK1, promoting the phosphorylation of its substrates, and is degraded rapidly through the proteasome. This creates an acute response to the corticoid stimuli. For example, cortisol secretion is mediated physiologically by the adrenal gland, producing a profound effect on the physiology on the kidney, among other organs. These effects are at least in part dependent on SGK1 since KO mice have defects in the regulation of renal Na+ excretion (Wulff et al., 2002). Therefore, SGK1 requires sustained transcription in order to maintain elevated levels of the protein. In our resistant cell lines, we found that this could be explained by the presence of a CpG site in the promoter of SGK1 that is methylated in the cell lines that exhibit low levels of mRNA and, conversely, low methylation levels are correlated with high SGK1 mRNA levels. Normal breast is characterized by the presence of low levels of SGK1 and high methylation status of this CpG site, indicating that hypo-methylation could be a consequence of the oncogenic transformation (Ciriello et al., 2015). Our results also provided further understanding on how SGK1 can activate the mTORC1 pathway. We analyzed the interaction of SGK1 with several components of the TSC complex and found that SGK1 has the ability to interact with TSC2. This interaction was also observed endogenously and it was determined to be mediated by the N-terminal domain of TSC2. It has been described that other protein-protein interactions take place in this region, such as the TSC1-TSC2 interaction (Li et al., 2004). Although the AKT and 230 TSC2 interaction has never been closely examined, it is also expected to take place at this molecular interphase as a result of the high sequence homology with SGK1. Importantly, SGK1 has the ability to phosphorylate TSC2 in the same RXRXX(S/T) motifs as AKT phosphorylation, compensating for the loss of activity of this enzyme. It is well accepted that phosphorylation of these sites increase downstream RHEB-GTP loading and mTORC1 signaling, as a result of a translocation from the lysosome to the cytoplasm, explaining how the high levels of SGK1 can sustain mTORC1-dependent signaling (Inoki et al., 2003; Inoki et al., 2002; Manning et al., 2002; Menon et al., 2014; Potter et al., 2002). The role of SGK1 in mediating mTORC1 activation upon PI3Kα inhibition can be explained by the differential regulation of AKT and SGK1 upon pharmacological stress. Although both kinases share the same upstream regulators, mTORC2 and PDK1, AKT contains a PH domain that is required for the PI3K-dependent plasma membrane translocation and subsequent activation (Pearce et al., 2010). In contrast, SGK1 does not require plasma membrane localization, which could at least partially explain why it remains active in the absence of PIP3. Moreover, in our resistant cell lines treated with PI3Kα inhibitor, we observe a substantial but incomplete decrease in SGK1 activity. This can be explained by the fact that PIP3 controls mTORC2 (Gan et al., 2011), although it would not be surprising that other PIP3-independent pools of mTORC2 are responsible for residual SGK1 activity (Alessi et al., 2009; Frias et al., 2006). While PDK1 is a constitutively active kinase and can be present in both the cytoplasm and the membrane (upon PIP3 synthesis), the subcellular localization of mTORC2 is ambiguous. Therefore, it is plausible that different pools of mTORC2 can be found in other cellular structures such as the membrane, the cytoplasm, and other organelles (Betz and Hall, 2013). The role of SGK1 in regulating signaling downstream of mTORC2 is intriguing but not entirely unexpected. From the evolutionary point of view, the SGK1 orthologue in Drosophila melanogaster is the dAkt gene, which shares the same similarity with human SGK1 (63%) and AKT1 (67%) genes. Thus, it is tempting to speculate that due to the high overlap of their substrates, Drosophila dAkt plays the role of both AKT and SGK1. In fact, Ypk2 and Gad8, the SGK1 orthologues in budding and fission yeast, respectively, are the main TORC2 downstream effectors. Similarly, in the worm Caenorhabditis elegans, sgk-1 appears to be essential for TORC2 signaling, lifespan, and growth (Casamayor et al., 1999; Jones et al., 2009; Kamada et al., 2005; Matsuo et 231 al., 2003; Soukas et al., 2009). The short half-life of SGK1 mRNA and the difficulties of studying this kinase in pre-clinical models (low stability, lack of effective compounds, resistance to genetic silencing tools such as RNAi) have hampered the suitability of targeting SGK1 in cancer therapy. We provided a step forward in this direction by characterizing novel small molecule inhibitors that were previously synthesized as sulfonamide-derivatives (Halland et al., 2015). Among the different inhibitors tested in our in vitro models, we found a compound named 21g, which we re-named as SGK1- inh, that exhibits high potency towards SGK1 (IC50 ≈ 4.8 nM) and selectivity when screened against a panel of 140 kinases. SGK1-inh was found to interact with the inactive conformation of SGK1 (DFG-out), as a type II kinase inhibitor, taking advantage of the allosteric pocket created next to the ATP pocket. In fact, this is a highly used strategy in the rational design of kinase inhibitors, since this unique pocket provides molecular features that increase the selectivity of the compounds (Muller et al., 2015). Treatment of resistant cell lines with the SGK1-inh in combination with BYL719 exhibited decreased cell viability, inhibition of mTORC1 signaling, and superior anti-tumor effects when tested in HCC1954 xenografts. Our report has pioneered the exploration of SGK1 inhibition as a target for cancer, and current efforts are aimed to develop novel chemical probes that exhibit better physicochemical and pharmaceutical properties. GEMMs provide a unique platform for the study of oncogene-induced transformation and experimental pharmacology. In terms of therapy, GEMM are a more reliable model since they take into account some key factors such as tumor heterogeneity, reliable PK and PD, microenvironment, a functional immune system, which is known to play a crucial role in both tumor initiation and progression, and compensatory or intrinsic mechanisms of resistance (Heyer et al., 2010). Recent studies using PIK3CA-driven GEMM have uncovered the cell-of-origin from breast cancers that have been developed using a knock-in model, explaining the heterogeneity of these tumors (Koren et al., 2015; Van Keymeulen et al., 2015). Experiments in our laboratory using a Cre-inducible PIK3CA H1047R mutation mouse model have shown that the expression of this oncogenic mutation in all the cells of the mouse is embryonically lethal (Article 6). When PIK3CAH1047R mice are crossed with the Cre deleter mouse (CAG-Cre), embryos carrying the transgene exhibit cardiovascular defects and fail to form a normal neural plate. While the vascular phenotype is a result of the important role of this lipid kinase in the normal angiogenesis program (Graupera et al., 2008; Graupera and Potente, 2013), 232 the cephalic phenotype was not very well understood. We conditionally expressed the PIK3CAH1047R allele under the control of the Sox2 promoter and found that embryos had an open, ruffled cephalic neural plate that was consistent with the results observed in the Pten KO mouse. The mutant cephalic neural epithelium exhibits a failure to undergo the transition from a cuboidal to a columnar pseudostratified epithelium, leading to a mutant neural plate (thin, wide, and irregularly folded). As a result, closure of the cephalic neural tube fails. These phenotypes were dependent on the presence of PIP3 levels and were not rescued by the treatment with AKT or mTORC1 inhibitor. Nevertheless, when Pten KO mice are crossed with the Pdpk1 (PDK1) KO mice, the phenotype is rescued. This is an interesting observation taking into account that PTEN loss and PIK3CA mutations in human cancers are mainly associated with the downstream activity of AKT (Courtney et al., 2010). However, this is in line with the observations made using the hypomorphic Pdpk1 mouse model, which exhibits decreased tumorigenesis when crossed with the heterozygous Pten+/-mice (Bayascas et al., 2005). While this certainly could be the result of impaired AKT activation, many other kinases are regulated by both PDK1 and, indirectly, increased PI3K activity. For instance, mTORC2 has been shown to be activated downstream of PI3K and several AGC kinases are phosphorylated by mTORC2 at the hydrophobic motif, including SGK1, SGK2, SGK3, and several isoforms of PKC (Pearce et al., 2010). Moreover, mTORC2 is shown to play an important role in cytoskeleton reorganization and actin polymerization (Alessi et al., 2009). To further understand the role of activated PI3K signaling in all the cells of the organism, we designed an inducible strategy using the conditional Cre deleter mouse CAG- CreERTM, in which the Cre recombinase is only active upon tamoxifen administration. Transgenic mice were born normally, but upon tamoxifen administration, PIK3CACAG- CreER but not the WT littermates developed vascular malformations that are consistent with the diagnosis of venous malformations according to the International Society for the Study of Vascular Anomalies (ISSVA) (Wassef et al., 2015). Venous malformations (VM) are the most common vascular malformation in humans and are cause of pain, functional limitations of the affected areas, aesthetic disfigurements, and coagulopathies (Killion et al., 2014). In these cases, sclerotherapy or surgical resection may be considered; however, these procedures often involve complications such as cutaneous necrosis or extended inflammatory reactions, and, depending on the anatomic location and extension, may be of limited application (Uebelhoer et al., 2012). 233 Moreover, VM are prone to recur and recanalize, demanding the need for development of more effective therapies (Dompmartin et al., 2010). The first somatic molecular alterations linked to the development of sporadic VM were the acquisition of gain-of-function mutations on the gene encoding for the EC-specific tyrosine-kinase receptor TIE2 (TEK) (Limaye et al., 2009). Ligand-independent receptor activation drives constitutive activation of the PI3K/AKT and MAPK pathways, leading to increased proliferation and survival of EC that could account for increased EC accumulation in VM and abnormal recruitment of smooth muscle cells (Nguyen et al., 2014b). However, only a subgroup of VM harbor defects in TEK, suggesting that other genomic or molecular alterations may be at play in this disease. In the future, it would be interesting to develop a GEMM expressing activating TEK mutations in the endothelial compartment to characterize their role in the histopathology and the mechanism of pathogenesis of the disease. Recent studies performing xenograft experiments of HUVEC transduced with the most frequent TEK mutation L914F have demonstrated their functional relevance in inducing VM (Boscolo et al., 2015). Treatment of murine xenografts with rapamycin proved the efficacy of inhibiting mTORC1 activity, and further showed clinical activity in VM patients in a pilot trial. Intriguingly, three out of five patients that responded to mTOR inhibition in this study did not harbor any genetic defect in TEK. It is thus tempting to speculate that additional molecular alterations enhancing the activity of the PI3K/AKT/mTOR pathway could be driving the formation of VM in these patients. In our mice, the histopathologic resemblance of the lesions arising in mice with those affecting humans prompted us to evaluate the existence of similar alterations in clinical specimens. Through targeted exome sequencing we found that 25% of the evaluated samples bear activating mutations in PIK3CA, or other genetic defects in the PI3K/AKT pathway. We also detected that 35% of the patients harbored mutations in TEK, yet these were mutually exclusive with the presence of activating PI3K mutations, consistent with a functional redundancy. These results are congruent with the high prevalence of TEK mutations (mainly in pediatric patients) reported by others (Limaye et al., 2009). Consistently, TIE2 activation as a result of TEK mutations mainly signals through PI3K, providing a unifying hypothesis that aberrant activation of PI3K signaling, either by upstream receptor mutations or in the pathway itself, leads to the development of VM. 234 Somatic mutation of PIK3CA is frequently detected in several cancer types, and genetic alterations driving hyperactivation of the PI3K/AKT pathway have also been reported in nonhereditary post-zygotic tissue overgrowth syndromes that often exhibit mixed capillary, lymphatic, and venous anomalies. Due to the clinical overlap of these overgrowth syndromes with PIK3CA mutations, the term PIK3CA-related overgrowth syndrome (PROS) has been proposed (Keppler-Noreuil et al., 2015). Patients suffering from Congenital Lipomatous Overgrowth, Vascular Malformations, Epidermal Nevi, And Skeletal/Spinal Abnormalities (CLOVES) syndrome harbor somatic mosaicism for activating PIK3CA mutations resulting in hyperactive PI3K/AKT signaling (Emrick et al., 2014; Kurek et al., 2012). The presence of somatic mutations in PIK3CA was also detected in patients affected by Klippel-Trenaunay-Weber syndrome (KTS) – an overgrowth condition with features overlapping those of CLOVES syndrome including isolated lymphatic malformations, fibro-adipose hyperplasia, and fibro-adipose vascular anomalies (Dompmartin et al., 2010). Additional genetic alterations in PTEN, GNAQ, AKT isoforms, or the regulatory subunit of PI3K, PIK3R1, that enhance PI3K/AKT/mTOR and MAPK pathway activation have also been reported in other malformative syndromes including Proteus, Megalencephaly capillary malformation (MCAP), Sturge-Weber, and Bannayan-Riley-Ruvalcaba syndromes, underscoring the involvement of aberrant PI3K/AKT/mTOR signaling in developmental disorders (Jansen et al., 2015; Lindhurst et al., 2012; Nakashima et al., 2014; Riviere et al., 2012; Shirley et al., 2013). Interestingly, MCAP syndrome exhibits a predominant brain overgrowth phenotype in which PIK3CA mutations are also involved. A recent report described the first mouse model for brain overgrowth using GEMM of PIK3CA mutants E545K and H1047R, validating the importance of these mouse models in the study of PIK3CA-driven syndromes (Roy et al., 2015). Nevertheless, sporadic and solitary VM lesions are a different and much more prevalent entity not associated with overgrowth. Despite the fact that most lymphatic malformations carry PIK3CA mutations, our mouse model does not present with any detectable lymphatic anomalies (Osborn et al., 2015). This is a limitation of our model that could be explained by a number of factors including the moment and time of genomic recombination, or different sets of precursor cells giving rise to each malformation. The use of lymphatic vessel-specific Cre strains may be used in the future to model lymphatic malformations. Moreover, it would be interesting to identify also the cell-of-origin of VM by lineage-tracing experiments, which could not be assessed in the current thesis. 235 Given the validity of our mouse model to recapitulate the pathogenesis of human VM, we asked whether it could be used as a platform for testing pharmacological inhibition using PI3K inhibitors currently under clinical development. To put this into clinical context, we also evaluated the efficacy of other agents that have been proposed to inhibit the growth of VM, including rapamycin analogues and propranolol. Greatest growth inhibition was achieved when treating allograft transplants with a PI3Kα inhibitor and the rapamycin analogue everolimus, as compared with no effect observed with propranolol. Rapamycin has been shown to improve quality of life in patients with VM, lymphatic malformations, and other vascular syndromes. A possible explanation for these therapeutic benefits is the ability to inhibit AKT upon long-term treatment with rapamycin in EC, an observation also seen in adipocytes, but not in all the epithelial cells (Boscolo et al., 2015; Sarbassov et al., 2006). In contrast to the anti-proliferative effect of rapamycin analogues, we propose that the pro-apoptotic effect achieved upon PI3K inhibition is likely to yield improved therapeutic efficacy by diminishing the recurrence of VM. One important caveat of systemic administration of PI3K inhibitors is the presence of secondary effects such as hyperglycemia, skin rash, fatigue, and gastrointestinal problems. Because most of the VM appear in the skin, we asked whether these lesions could be treated using topical formulations, an unprecedented administration for this class of compounds. Indeed, topical administration of PI3Kα inhibitor further demonstrated the efficacy of this treatment, an approach that would be devoid of systemic side effects. This paves the way for future clinical trials and opens a new avenue for the treatment of this disease. Importantly, the impaired vasculogenesis observed in mouse embryos as a result of endothelial-restricted expression of the PIK3CAH1047R allele was also rescued when pregnant mice were treated with the PI3Kα-inhibitor, further supporting a functional requirement for controlled PI3K signaling in normal embryonic vasculogenesis as has been demonstrated by others (Graupera et al., 2008; Hare et al., 2015). Our data indicates that the PI3K pathway is likely a key player in the development of VM, and pharmacological inhibition of this important node could be a suitable strategy for patients where the therapeutic options are limited. Further clinical trials will be required to investigate the safety and efficacy of PI3K inhibitors in this disease. 236                                                                                       237 CONCLUSIONS 238 239 The conclusions of this doctoral dissertation are that despite the therapeutic effectiveness of PI3K inhibitors, resistance can emerge and biomarkers of intrinsic or acquired resistance are required to determine the patient population that might benefit from these treatments. Combination with other agents will be required for the successful clinical development of PI3K inhibitors. Our results show that: a) Different mechanisms of acquired/intrinsic resistance take place under the selective pressure of PI3K oncoprotein inhibition. b) Studying each mechanism of resistance can lead to design better therapeutic regimes and drug combinations that can effectively reduce tumor growth and overcome drug resistance. c) The discovery of biomarkers, such as PTEN loss or SGK1 up-regulation, upon PI3K inhibition will likely be effective in predicting resistance to these therapies and will allow to propose novel combinatorial therapies. The conclusions for the specific goals of this thesis are: To characterize the mechanisms of sustained mTORC1 activation upon PI3K inhibition in BYL719 resistant cell lines - PDK1 inhibition reverts the resistant phenotype and decreases sustained mTORC1 activation - Pharmacological targeting of PDK1 in combination with PI3K inhibitors is feasible in preclinical models - Combined PI3Kα and PDK1 inhibition induces FOXO-dependent transcription - SGK1 is the main effector downstream of PDK1 mediating resistance to PI3Kα inhibitors - SGK1 is differentially regulated by promoter DNA methylation between BYL719 sensitive and resistant cell lines - SGK1 interacts with and phosphorylates TSC2 to activate mTORC1 240 - Pharmacological inhibition of SGK1 is possible and enhances anti-tumor effects in combination with PI3Kα inhibitors To define the role of secondary AKT-like kinases as mechanisms of resistance to PI3K inhibitors - PTEN loss is a mechanism of resistance to PI3Kα inhibitors - PTEN negative tumors also depend on PI3Kβ-mediated signaling - AKT reactivation is a mechanism of resistance to PI3Kα inhibitors - Under selective therapeutic pressure, tumors are selected for resistant clones undergoing a process that mimics features of population genetics - Combined inhibition of PI3Kα and PI3Kβ is effective in tumors that are PIK3CA mutant and present as loss of PTEN Development of a mouse model as a tool to study the role of these kinases in vivo - Expression of the PIK3CA H1047R allele is embryonically lethal - PIK3CA H1047R embryos exhibit defects in the development of the cephalic area and the vascular system - Conditional expression of PIK3CA H1047R in adult mice induces venous malformations - Human venous malformations harbor PIK3CA mutations - Treatment of mice with PIK3CA H1047R-induced venous malformations using PI3Kα inhibitors is effective in reducing morbidity and mortality - PIK3CA H1047R mutation induces changes in morphology, proliferation, secretion, and vascular physiology - Murine venous malformations are reverted with PI3Kα inhibitors administered either systemically or topically. 241 REFERENCES 242 243 Adams, J.R., Xu, K., Liu, J.C., Agamez, N.M., Loch, A.J., Wong, R.G., Wang, W., Wright, K.L., Lane, T.F., Zacksenhaus, E., et al. (2011). Cooperation between Pik3ca and p53 mutations in mouse mammary tumor formation. Cancer research 71, 2706- 2717. Agarwal, S., Sane, R., Gallardo, J.L., Ohlfest, J.R., and Elmquist, W.F. (2010). Distribution of gefitinib to the brain is limited by P-glycoprotein (ABCB1) and breast cancer resistance protein (ABCG2)-mediated active efflux. The Journal of pharmacology and experimental therapeutics 334, 147-155. Alessi, D.R., Andjelkovic, M., Caudwell, B., Cron, P., Morrice, N., Cohen, P., and Hemmings, B.A. (1996). Mechanism of activation of protein kinase B by insulin and IGF- 1. The EMBO journal 15, 6541-6551. Alessi, D.R., and Cohen, P. (1998). Mechanism of activation and function of protein kinase B. Current opinion in genetics & development 8, 55-62. Alessi, D.R., Kozlowski, M.T., Weng, Q.P., Morrice, N., and Avruch, J. (1998). 3- Phosphoinositide-dependent protein kinase 1 (PDK1) phosphorylates and activates the p70 S6 kinase in vivo and in vitro. Current biology : CB 8, 69-81. Alessi, D.R., Pearce, L.R., and Garcia-Martinez, J.M. (2009). New insights into mTOR signaling: mTORC2 and beyond. Science signaling 2, pe27. Aparicio, S., Hidalgo, M., and Kung, A.L. (2015). Examining the utility of patient-derived xenograft mouse models. Nature reviews Cancer 15, 311-316. Arencibia, J.M., Pastor-Flores, D., Bauer, A.F., Schulze, J.O., and Biondi, R.M. (2013). AGC protein kinases: from structural mechanism of regulation to allosteric drug development for the treatment of human diseases. Biochimica et biophysica acta 1834, 1302-1321. Arnedos, M., Vicier, C., Loi, S., Lefebvre, C., Michiels, S., Bonnefoi, H., and Andre, F. (2015). Precision medicine for metastatic breast cancer-limitations and solutions. Nature reviews Clinical oncology 12, 693-704. Bago, R., Malik, N., Munson, M.J., Prescott, A.R., Davies, P., Sommer, E., Shpiro, N., Ward, R., Cross, D., Ganley, I.G., et al. (2014). Characterization of VPS34-IN1, a selective inhibitor of Vps34, reveals that the phosphatidylinositol 3-phosphate-binding SGK3 protein kinase is a downstream target of class III phosphoinositide 3-kinase. The Biochemical journal 463, 413-427. Baguley, B.C. (2010). Multiple drug resistance mechanisms in cancer. Molecular biotechnology 46, 308-316. Balendran, A., Biondi, R.M., Cheung, P.C., Casamayor, A., Deak, M., and Alessi, D.R. (2000). A 3-phosphoinositide-dependent protein kinase-1 (PDK1) docking site is required for the phosphorylation of protein kinase Czeta (PKCzeta ) and PKC-related kinase 2 by PDK1. The Journal of biological chemistry 275, 20806-20813. 244 Baselga, J., Campone, M., Piccart, M., Burris, H.A., 3rd, Rugo, H.S., Sahmoud, T., Noguchi, S., Gnant, M., Pritchard, K.I., Lebrun, F., et al. (2012). Everolimus in postmenopausal hormone-receptor-positive advanced breast cancer. The New England journal of medicine 366, 520-529. Bayascas, J.R., Leslie, N.R., Parsons, R., Fleming, S., and Alessi, D.R. (2005). Hypomorphic mutation of PDK1 suppresses tumorigenesis in PTEN(+/-) mice. Current biology : CB 15, 1839-1846. Belanger, A.S., Tojcic, J., Harvey, M., and Guillemette, C. (2010). Regulation of UGT1A1 and HNF1 transcription factor gene expression by DNA methylation in colon cancer cells. BMC molecular biology 11, 9. Bendell, J.C., Rodon, J., Burris, H.A., de Jonge, M., Verweij, J., Birle, D., Demanse, D., De Buck, S.S., Ru, Q.C., Peters, M., et al. (2012). Phase I, dose-escalation study of BKM120, an oral pan-Class I PI3K inhibitor, in patients with advanced solid tumors. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 30, 282-290. Betz, C., and Hall, M.N. (2013). Where is mTOR and what is it doing there? The Journal of cell biology 203, 563-574. Bhinder, B., and Djaballah, H. (2012). A simple method for analyzing actives in random RNAi screens: introducing the "H Score" for hit nomination & gene prioritization. Comb Chem High Throughput Screen 15, 686-704. Biondi, R.M., Cheung, P.C., Casamayor, A., Deak, M., Currie, R.A., and Alessi, D.R. (2000). Identification of a pocket in the PDK1 kinase domain that interacts with PIF and the C-terminal residues of PKA. The EMBO journal 19, 979-988. Biondi, R.M., Kieloch, A., Currie, R.A., Deak, M., and Alessi, D.R. (2001). The PIF- binding pocket in PDK1 is essential for activation of S6K and SGK, but not PKB. The EMBO journal 20, 4380-4390. Bjornstrom, L., and Sjoberg, M. (2005). Mechanisms of estrogen receptor signaling: convergence of genomic and nongenomic actions on target genes. Molecular endocrinology 19, 833-842. Bonner, J.A., Harari, P.M., Giralt, J., Azarnia, N., Shin, D.M., Cohen, R.B., Jones, C.U., Sur, R., Raben, D., Jassem, J., et al. (2006). Radiotherapy plus cetuximab for squamous-cell carcinoma of the head and neck. The New England journal of medicine 354, 567-578. Borst, P., and Elferink, R.O. (2002). Mammalian ABC transporters in health and disease. Annual review of biochemistry 71, 537-592. Bosch, A., Li, Z., Bergamaschi, A., Ellis, H., Toska, E., Prat, A., Tao, J.J., Spratt, D.E., Viola-Villegas, N.T., Castel, P., et al. (2015). PI3K inhibition results in enhanced estrogen receptor function and dependence in hormone receptor-positive breast cancer. Science translational medicine 7, 283ra251. 245 Boscolo, E., Limaye, N., Huang, L., Kang, K.T., Soblet, J., Uebelhoer, M., Mendola, A., Natynki, M., Seront, E., Dupont, S., et al. (2015). Rapamycin improves TIE2-mutated venous malformation in murine model and human subjects. The Journal of clinical investigation 125, 3491-3504. Brannon, A.R., and Sawyers, C.L. (2013). "N of 1" case reports in the era of whole- genome sequencing. The Journal of clinical investigation 123, 4568-4570. Brickley, D.R., Mikosz, C.A., Hagan, C.R., and Conzen, S.D. (2002). Ubiquitin modification of serum and glucocorticoid-induced protein kinase-1 (SGK-1). The Journal of biological chemistry 277, 43064-43070. Brunet, A., Bonni, A., Zigmond, M.J., Lin, M.Z., Juo, P., Hu, L.S., Anderson, M.J., Arden, K.C., Blenis, J., and Greenberg, M.E. (1999). Akt promotes cell survival by phosphorylating and inhibiting a Forkhead transcription factor. Cell 96, 857-868. Brunet, A., Park, J., Tran, H., Hu, L.S., Hemmings, B.A., and Greenberg, M.E. (2001). Protein kinase SGK mediates survival signals by phosphorylating the forkhead transcription factor FKHRL1 (FOXO3a). Molecular and cellular biology 21, 952-965. Burke, J.E., and Williams, R.L. (2013). Dynamic steps in receptor tyrosine kinase mediated activation of class IA phosphoinositide 3-kinases (PI3K) captured by H/D exchange (HDX-MS). Advances in biological regulation 53, 97-110. Calcagno, A.M. (2011). Evolution of drug resistance in cancer: the emergence of unique mechanisms and novel techniques. Molecular pharmaceutics 8, 1993. Calmy, A., Pascual, F., and Ford, N. (2004). HIV drug resistance. The New England journal of medicine 350, 2720-2721. Carpten, J.D., Faber, A.L., Horn, C., Donoho, G.P., Briggs, S.L., Robbins, C.M., Hostetter, G., Boguslawski, S., Moses, T.Y., Savage, S., et al. (2007). A transforming mutation in the pleckstrin homology domain of AKT1 in cancer. Nature 448, 439-444. Carver, B.S., Chapinski, C., Wongvipat, J., Hieronymus, H., Chen, Y., Chandarlapaty, S., Arora, V.K., Le, C., Koutcher, J., Scher, H., et al. (2011). Reciprocal feedback regulation of PI3K and androgen receptor signaling in PTEN-deficient prostate cancer. Cancer cell 19, 575-586. Casamayor, A., Torrance, P.D., Kobayashi, T., Thorner, J., and Alessi, D.R. (1999). Functional counterparts of mammalian protein kinases PDK1 and SGK in budding yeast. Current biology : CB 9, 186-197. Cescon, D.W., Gorrini, C., and Mak, T.W. (2015). Breaking up is hard to do: PI3K isoforms on the rebound. Cancer cell 27, 5-7. Chabner, B.A., and Roberts, T.G., Jr. (2005). Timeline: Chemotherapy and the war on cancer. Nature reviews Cancer 5, 65-72. Chandarlapaty, S., Sawai, A., Scaltriti, M., Rodrik-Outmezguine, V., Grbovic-Huezo, O., Serra, V., Majumder, P.K., Baselga, J., and Rosen, N. (2011). AKT inhibition relieves 246 feedback suppression of receptor tyrosine kinase expression and activity. Cancer cell 19, 58-71. Chapman, P.B., Hauschild, A., Robert, C., Haanen, J.B., Ascierto, P., Larkin, J., Dummer, R., Garbe, C., Testori, A., Maio, M., et al. (2011). Improved survival with vemurafenib in melanoma with BRAF V600E mutation. The New England journal of medicine 364, 2507-2516. Cheng, D.T., Mitchell, T.N., Zehir, A., Shah, R.H., Benayed, R., Syed, A., Chandramohan, R., Liu, Z.Y., Won, H.H., Scott, S.N., et al. (2015). Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT): A Hybridization Capture-Based Next-Generation Sequencing Clinical Assay for Solid Tumor Molecular Oncology. J Mol Diagn 17, 251-264. Chresta, C.M., Davies, B.R., Hickson, I., Harding, T., Cosulich, S., Critchlow, S.E., Vincent, J.P., Ellston, R., Jones, D., Sini, P., et al. (2010). AZD8055 is a potent, selective, and orally bioavailable ATP-competitive mammalian target of rapamycin kinase inhibitor with in vitro and in vivo antitumor activity. Cancer research 70, 288-298. Ciriello, G., Gatza, M.L., Beck, A.H., Wilkerson, M.D., Rhie, S.K., Pastore, A., Zhang, H., McLellan, M., Yau, C., Kandoth, C., et al. (2015). Comprehensive Molecular Portraits of Invasive Lobular Breast Cancer. Cell 163, 506-519. Cohen, P., Alessi, D.R., and Cross, D.A. (1997). PDK1, one of the missing links in insulin signal transduction? FEBS letters 410, 3-10. Collins, B.J., Deak, M., Arthur, J.S., Armit, L.J., and Alessi, D.R. (2003). In vivo role of the PIF-binding docking site of PDK1 defined by knock-in mutation. The EMBO journal 22, 4202-4211. Collins, B.J., Deak, M., Murray-Tait, V., Storey, K.G., and Alessi, D.R. (2005). In vivo role of the phosphate groove of PDK1 defined by knockin mutation. Journal of cell science 118, 5023-5034. Contreras, C.M., Akbay, E.A., Gallardo, T.D., Haynie, J.M., Sharma, S., Tagao, O., Bardeesy, N., Takahashi, M., Settleman, J., Wong, K.K., et al. (2010). Lkb1 inactivation is sufficient to drive endometrial cancers that are aggressive yet highly responsive to mTOR inhibitor monotherapy. Dis Model Mech 3, 181-193. Costa, C., Ebi, H., Martini, M., Beausoleil, S.A., Faber, A.C., Jakubik, C.T., Huang, A., Wang, Y., Nishtala, M., Hall, B., et al. (2015). Measurement of PIP3 levels reveals an unexpected role for p110beta in early adaptive responses to p110alpha-specific inhibitors in luminal breast cancer. Cancer cell 27, 97-108. Costa, C., and Engelman, J.A. (2014). The double life of p85. Cancer cell 26, 445-447. Courtney, K.D., Corcoran, R.B., and Engelman, J.A. (2010). The PI3K pathway as drug target in human cancer. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 28, 1075-1083. Crowley, E., Di Nicolantonio, F., Loupakis, F., and Bardelli, A. (2013). Liquid biopsy: monitoring cancer-genetics in the blood. Nature reviews Clinical oncology 10, 472-484. 247 Currie, R.A., Walker, K.S., Gray, A., Deak, M., Casamayor, A., Downes, C.P., Cohen, P., Alessi, D.R., and Lucocq, J. (1999). Role of phosphatidylinositol 3,4,5-trisphosphate in regulating the activity and localization of 3-phosphoinositide-dependent protein kinase-1. The Biochemical journal 337 ( Pt 3), 575-583. Cybulski, N., and Hall, M.N. (2009). TOR complex 2: a signaling pathway of its own. Trends in biochemical sciences 34, 620-627. Dalton, W.S. (1990). Mechanisms of drug resistance in breast cancer. Seminars in oncology 17, 37-39. DeRose, Y.S., Wang, G., Lin, Y.C., Bernard, P.S., Buys, S.S., Ebbert, M.T., Factor, R., Matsen, C., Milash, B.A., Nelson, E., et al. (2011). Tumor grafts derived from women with breast cancer authentically reflect tumor pathology, growth, metastasis and disease outcomes. Nature medicine 17, 1514-1520. DeVita, V.T., Jr., and Chu, E. (2008). A history of cancer chemotherapy. Cancer research 68, 8643-8653. Dhillon, S. (2013). Everolimus in combination with exemestane: a review of its use in the treatment of patients with postmenopausal hormone receptor-positive, HER2-negative advanced breast cancer. Drugs 73, 475-485. Dickson, M., and Gagnon, J.P. (2004). Key factors in the rising cost of new drug discovery and development. Nature reviews Drug discovery 3, 417-429. Diehl, F., Schmidt, K., Choti, M.A., Romans, K., Goodman, S., Li, M., Thornton, K., Agrawal, N., Sokoll, L., Szabo, S.A., et al. (2008). Circulating mutant DNA to assess tumor dynamics. Nature medicine 14, 985-990. Dienstmann, R., De Dosso, S., Felip, E., and Tabernero, J. (2012). Drug development to overcome resistance to EGFR inhibitors in lung and colorectal cancer. Molecular oncology 6, 15-26. Dienstmann, R., Rodon, J., Markman, B., and Tabernero, J. (2011). Recent developments in anti-cancer agents targeting PI3K, Akt and mTORC1/2. Recent patents on anti-cancer drug discovery 6, 210-236. Dienstmann, R., Rodon, J., Serra, V., and Tabernero, J. (2014). Picking the point of inhibition: a comparative review of PI3K/AKT/mTOR pathway inhibitors. Molecular cancer therapeutics 13, 1021-1031. Dillon, L.M., and Miller, T.W. (2014). Therapeutic targeting of cancers with loss of PTEN function. Current drug targets 15, 65-79. Dittmann, A., Werner, T., Chung, C.W., Savitski, M.M., Falth Savitski, M., Grandi, P., Hopf, C., Lindon, M., Neubauer, G., Prinjha, R.K., et al. (2014). The commonly used PI3- kinase probe LY294002 is an inhibitor of BET bromodomains. ACS chemical biology 9, 495-502. 248 Dompmartin, A., Vikkula, M., and Boon, L.M. (2010). Venous malformation: update on aetiopathogenesis, diagnosis and management. Phlebology 25, 224-235. Eirew, P., Steif, A., Khattra, J., Ha, G., Yap, D., Farahani, H., Gelmon, K., Chia, S., Mar, C., Wan, A., et al. (2015). Dynamics of genomic clones in breast cancer patient xenografts at single-cell resolution. Nature 518, 422-426. Elkabets, M., Pazarentzos, E., Juric, D., Sheng, Q., Pelossof, R.A., Brook, S., Benzaken, A.O., Rodon, J., Morse, N., Yan, J.J., et al. (2015). AXL mediates resistance to PI3Kalpha inhibition by activating the EGFR/PKC/mTOR axis in head and neck and esophageal squamous cell carcinomas. Cancer cell 27, 533-546. Elkabets, M., Vora, S., Juric, D., Morse, N., Mino-Kenudson, M., Muranen, T., Tao, J., Campos, A.B., Rodon, J., Ibrahim, Y.H., et al. (2013). mTORC1 inhibition is required for sensitivity to PI3K p110alpha inhibitors in PIK3CA-mutant breast cancer. Science translational medicine 5, 196ra199. Emrick, L.T., Murphy, L., Shamshirsaz, A.A., Ruano, R., Cassady, C.I., Liu, L., Chang, F., Sutton, V.R., Li, M., and Van den Veyver, I.B. (2014). Prenatal diagnosis of CLOVES syndrome confirmed by detection of a mosaic PIK3CA mutation in cultured amniocytes. American journal of medical genetics Part A 164A, 2633-2637. Endicott, J.A., Noble, M.E., and Johnson, L.N. (2012). The structural basis for control of eukaryotic protein kinases. Annual review of biochemistry 81, 587-613. Engelman, J.A. (2009). Targeting PI3K signalling in cancer: opportunities, challenges and limitations. Nature reviews Cancer 9, 550-562. Engelman, J.A., Luo, J., and Cantley, L.C. (2006). The evolution of phosphatidylinositol 3-kinases as regulators of growth and metabolism. Nature reviews Genetics 7, 606-619. Engelman, J.A., Zejnullahu, K., Mitsudomi, T., Song, Y., Hyland, C., Park, J.O., Lindeman, N., Gale, C.M., Zhao, X., Christensen, J., et al. (2007). MET amplification leads to gefitinib resistance in lung cancer by activating ERBB3 signaling. Science 316, 1039-1043. Faber, A.C., Ebi, H., Costa, C., and Engelman, J.A. (2012). Apoptosis in targeted therapy responses: the role of BIM. Adv Pharmacol 65, 519-542. Falasca, M., and Maffucci, T. (2012). Regulation and cellular functions of class II phosphoinositide 3-kinases. The Biochemical journal 443, 587-601. Falkenburger, B.H., Jensen, J.B., Dickson, E.J., Suh, B.C., and Hille, B. (2010). Phosphoinositides: lipid regulators of membrane proteins. The Journal of physiology 588, 3179-3185. Folkes, A.J., Ahmadi, K., Alderton, W.K., Alix, S., Baker, S.J., Box, G., Chuckowree, I.S., Clarke, P.A., Depledge, P., Eccles, S.A., et al. (2008). The identification of 2-(1H- indazol-4-yl)-6-(4-methanesulfonyl-piperazin-1-ylmethyl)-4-morpholin-4-yl-t hieno[3,2- d]pyrimidine (GDC-0941) as a potent, selective, orally bioavailable inhibitor of class I PI3 kinase for the treatment of cancer. Journal of medicinal chemistry 51, 5522-5532. 249 Frias, M.A., Thoreen, C.C., Jaffe, J.D., Schroder, W., Sculley, T., Carr, S.A., and Sabatini, D.M. (2006). mSin1 is necessary for Akt/PKB phosphorylation, and its isoforms define three distinct mTORC2s. Current biology : CB 16, 1865-1870. Fritsch, C., Huang, A., Chatenay-Rivauday, C., Schnell, C., Reddy, A., Liu, M., Kauffmann, A., Guthy, D., Erdmann, D., De Pover, A., et al. (2014). Characterization of the novel and specific PI3Kalpha inhibitor NVP-BYL719 and development of the patient stratification strategy for clinical trials. Molecular cancer therapeutics 13, 1117-1129. Fruman, D.A., and Rommel, C. (2014). PI3K and cancer: lessons, challenges and opportunities. Nature reviews Drug discovery 13, 140-156. Furet, P., Guagnano, V., Fairhurst, R.A., Imbach-Weese, P., Bruce, I., Knapp, M., Fritsch, C., Blasco, F., Blanz, J., Aichholz, R., et al. (2013). Discovery of NVP-BYL719 a potent and selective phosphatidylinositol-3 kinase alpha inhibitor selected for clinical evaluation. Bioorganic & medicinal chemistry letters 23, 3741-3748. Gagan, J., and Van Allen, E.M. (2015). Next-generation sequencing to guide cancer therapy. Genome medicine 7, 80. Gan, X., Wang, J., Su, B., and Wu, D. (2011). Evidence for direct activation of mTORC2 kinase activity by phosphatidylinositol 3,4,5-trisphosphate. The Journal of biological chemistry 286, 10998-11002. Garcia-Martinez, J.M., and Alessi, D.R. (2008). mTOR complex 2 (mTORC2) controls hydrophobic motif phosphorylation and activation of serum- and glucocorticoid-induced protein kinase 1 (SGK1). The Biochemical journal 416, 375-385. Garcia-Olmo, D.C., Dominguez, C., Garcia-Arranz, M., Anker, P., Stroun, M., Garcia- Verdugo, J.M., and Garcia-Olmo, D. (2010). Cell-free nucleic acids circulating in the plasma of colorectal cancer patients induce the oncogenic transformation of susceptible cultured cells. Cancer research 70, 560-567. Gelsi-Boyer, V., Orsetti, B., Cervera, N., Finetti, P., Sircoulomb, F., Rouge, C., Lasorsa, L., Letessier, A., Ginestier, C., Monville, F., et al. (2005). Comprehensive profiling of 8p11-12 amplification in breast cancer. Molecular cancer research : MCR 3, 655-667. Geuna, E., Milani, A., Martinello, R., Aversa, C., Valabrega, G., Scaltriti, M., and Montemurro, F. (2015). Buparlisib , an oral pan-PI3K inhibitor for the treatment of breast cancer. Expert Opin Investig Drugs 24, 421-431. Gharwan, H., and Groninger, H. (2015). Kinase inhibitors and monoclonal antibodies in oncology: clinical implications. Nature reviews Clinical oncology. Gottesman, M.M. (2002). Mechanisms of cancer drug resistance. Annual review of medicine 53, 615-627. Gottesman, M.M., Fojo, T., and Bates, S.E. (2002). Multidrug resistance in cancer: role of ATP-dependent transporters. Nature reviews Cancer 2, 48-58. 250 Graupera, M., Guillermet-Guibert, J., Foukas, L.C., Phng, L.K., Cain, R.J., Salpekar, A., Pearce, W., Meek, S., Millan, J., Cutillas, P.R., et al. (2008). Angiogenesis selectively requires the p110alpha isoform of PI3K to control endothelial cell migration. Nature 453, 662-666. Graupera, M., and Potente, M. (2013). Regulation of angiogenesis by PI3K signaling networks. Exp Cell Res 319, 1348-1355. Guertin, D.A., and Sabatini, D.M. (2005). An expanding role for mTOR in cancer. Trends in molecular medicine 11, 353-361. Guertin, D.A., and Sabatini, D.M. (2007). Defining the role of mTOR in cancer. Cancer cell 12, 9-22. Guertin, D.A., and Sabatini, D.M. (2009). The pharmacology of mTOR inhibition. Science signaling 2, pe24. Guertin, D.A., Stevens, D.M., Thoreen, C.C., Burds, A.A., Kalaany, N.Y., Moffat, J., Brown, M., Fitzgerald, K.J., and Sabatini, D.M. (2006). Ablation in mice of the mTORC components raptor, rictor, or mLST8 reveals that mTORC2 is required for signaling to Akt-FOXO and PKCalpha, but not S6K1. Developmental cell 11, 859-871. Haber, D.A., Gray, N.S., and Baselga, J. (2011). The evolving war on cancer. Cell 145, 19-24. Hall, J.M., Couse, J.F., and Korach, K.S. (2001). The multifaceted mechanisms of estradiol and estrogen receptor signaling. The Journal of biological chemistry 276, 36869-36872. Halland, N., Schmidt, F., Weiss, T., Saas, J., Li, Z., Czech, J., Dreyer, M., Hofmeister, A., Mertsch, K., Dietz, U., et al. (2015). Discovery of N-[4-(1H-Pyrazolo[3,4-b]pyrazin-6- yl)-phenyl]-sulfonamides as Highly Active and Selective SGK1 Inhibitors. ACS medicinal chemistry letters 6, 73-78. Hare, L.M., Schwarz, Q., Wiszniak, S., Gurung, R., Montgomery, K.G., Mitchell, C.A., and Phillips, W.A. (2015). Heterozygous expression of the oncogenic Pik3ca(H1047R) mutation during murine development results in fatal embryonic and extraembryonic defects. Developmental biology 404, 14-26. Hawkins, P.T., and Stephens, L.R. (2014). PI3K signalling in inflammation. Biochimica et biophysica acta. Hayakawa, M., Kaizawa, H., Kawaguchi, K., Ishikawa, N., Koizumi, T., Ohishi, T., Yamano, M., Okada, M., Ohta, M., Tsukamoto, S., et al. (2007). Synthesis and biological evaluation of imidazo[1,2-a]pyridine derivatives as novel PI3 kinase p110alpha inhibitors. Bioorganic & medicinal chemistry 15, 403-412. Hayakawa, M., Kaizawa, H., Moritomo, H., Koizumi, T., Ohishi, T., Okada, M., Ohta, M., Tsukamoto, S., Parker, P., Workman, P., et al. (2006). Synthesis and biological evaluation of 4-morpholino-2-phenylquinazolines and related derivatives as novel PI3 kinase p110alpha inhibitors. Bioorganic & medicinal chemistry 14, 6847-6858. 251 Heyer, J., Kwong, L.N., Lowe, S.W., and Chin, L. (2010). Non-germline genetically engineered mouse models for translational cancer research. Nature reviews Cancer 10, 470-480. Hidalgo, M., Amant, F., Biankin, A.V., Budinska, E., Byrne, A.T., Caldas, C., Clarke, R.B., de Jong, S., Jonkers, J., Maelandsmo, G.M., et al. (2014). Patient-derived xenograft models: an emerging platform for translational cancer research. Cancer discovery 4, 998-1013. Hirai, H., Sootome, H., Nakatsuru, Y., Miyama, K., Taguchi, S., Tsujioka, K., Ueno, Y., Hatch, H., Majumder, P.K., Pan, B.S., et al. (2010). MK-2206, an allosteric Akt inhibitor, enhances antitumor efficacy by standard chemotherapeutic agents or molecular targeted drugs in vitro and in vivo. Molecular cancer therapeutics 9, 1956-1967. Hollander, M.C., Blumenthal, G.M., and Dennis, P.A. (2011). PTEN loss in the continuum of common cancers, rare syndromes and mouse models. Nature reviews Cancer 11, 289-301. Holohan, C., Van Schaeybroeck, S., Longley, D.B., and Johnston, P.G. (2013). Cancer drug resistance: an evolving paradigm. Nature reviews Cancer 13, 714-726. Hsieh, A.C., Liu, Y., Edlind, M.P., Ingolia, N.T., Janes, M.R., Sher, A., Shi, E.Y., Stumpf, C.R., Christensen, C., Bonham, M.J., et al. (2012). The translational landscape of mTOR signalling steers cancer initiation and metastasis. Nature 485, 55-61. Huang, C.H., Mandelker, D., Gabelli, S.B., and Amzel, L.M. (2008). Insights into the oncogenic effects of PIK3CA mutations from the structure of p110alpha/p85alpha. Cell cycle 7, 1151-1156. Huang, C.H., Mandelker, D., Schmidt-Kittler, O., Samuels, Y., Velculescu, V.E., Kinzler, K.W., Vogelstein, B., Gabelli, S.B., and Amzel, L.M. (2007). The structure of a human p110alpha/p85alpha complex elucidates the effects of oncogenic PI3Kalpha mutations. Science 318, 1744-1748. Huang, J., and Manning, B.D. (2008). The TSC1-TSC2 complex: a molecular switchboard controlling cell growth. The Biochemical journal 412, 179-190. Huang, J., and Manning, B.D. (2009). A complex interplay between Akt, TSC2 and the two mTOR complexes. Biochemical Society transactions 37, 217-222. Hudis, C.A., and Gianni, L. (2011). Triple-negative breast cancer: an unmet medical need. The oncologist 16 Suppl 1, 1-11. Hughes, J.P., Rees, S., Kalindjian, S.B., and Philpott, K.L. (2011). Principles of early drug discovery. British journal of pharmacology 162, 1239-1249. Hyman, D.M., Puzanov, I., Subbiah, V., Faris, J.E., Chau, I., Blay, J.Y., Wolf, J., Raje, N.S., Diamond, E.L., Hollebecque, A., et al. (2015). Vemurafenib in Multiple Nonmelanoma Cancers with BRAF V600 Mutations. The New England journal of medicine 373, 726-736. 252 Ihle, N.T., Williams, R., Chow, S., Chew, W., Berggren, M.I., Paine-Murrieta, G., Minion, D.J., Halter, R.J., Wipf, P., Abraham, R., et al. (2004). Molecular pharmacology and antitumor activity of PX-866, a novel inhibitor of phosphoinositide-3-kinase signaling. Molecular cancer therapeutics 3, 763-772. Inoki, K., Li, Y., Xu, T., and Guan, K.L. (2003). Rheb GTPase is a direct target of TSC2 GAP activity and regulates mTOR signaling. Genes Dev 17, 1829-1834. Inoki, K., Li, Y., Zhu, T., Wu, J., and Guan, K.L. (2002). TSC2 is phosphorylated and inhibited by Akt and suppresses mTOR signalling. Nature cell biology 4, 648-657. Ivy, S.P., Siu, L.L., Garrett-Mayer, E., and Rubinstein, L. (2010). Approaches to phase 1 clinical trial design focused on safety, efficiency, and selected patient populations: a report from the clinical trial design task force of the national cancer institute investigational drug steering committee. Clinical cancer research : an official journal of the American Association for Cancer Research 16, 1726-1736. Iyer, G., Hanrahan, A.J., Milowsky, M.I., Al-Ahmadie, H., Scott, S.N., Janakiraman, M., Pirun, M., Sander, C., Socci, N.D., Ostrovnaya, I., et al. (2012). Genome sequencing identifies a basis for everolimus sensitivity. Science 338, 221. Jacinto, E., Facchinetti, V., Liu, D., Soto, N., Wei, S., Jung, S.Y., Huang, Q., Qin, J., and Su, B. (2006). SIN1/MIP1 maintains rictor-mTOR complex integrity and regulates Akt phosphorylation and substrate specificity. Cell 127, 125-137. Jansen, L.A., Mirzaa, G.M., Ishak, G.E., O'Roak, B.J., Hiatt, J.B., Roden, W.H., Gunter, S.A., Christian, S.L., Collins, S., Adams, C., et al. (2015). PI3K/AKT pathway mutations cause a spectrum of brain malformations from megalencephaly to focal cortical dysplasia. Brain. Jia, S., Liu, Z., Zhang, S., Liu, P., Zhang, L., Lee, S.H., Zhang, J., Signoretti, S., Loda, M., Roberts, T.M., et al. (2008). Essential roles of PI(3)K-p110beta in cell growth, metabolism and tumorigenesis. Nature 454, 776-779. Johannessen, C.M., Boehm, J.S., Kim, S.Y., Thomas, S.R., Wardwell, L., Johnson, L.A., Emery, C.M., Stransky, N., Cogdill, A.P., Barretina, J., et al. (2010). COT drives resistance to RAF inhibition through MAP kinase pathway reactivation. Nature 468, 968- 972. Jones, K.T., Greer, E.R., Pearce, D., and Ashrafi, K. (2009). Rictor/TORC2 regulates Caenorhabditis elegans fat storage, body size, and development through sgk-1. PLoS Biol 7, e60. Jonker, D.J., O'Callaghan, C.J., Karapetis, C.S., Zalcberg, J.R., Tu, D., Au, H.J., Berry, S.R., Krahn, M., Price, T., Simes, R.J., et al. (2007). Cetuximab for the treatment of colorectal cancer. The New England journal of medicine 357, 2040-2048. Juric, D., and Baselga, J. (2012). Tumor genetic testing for patient selection in phase I clinical trials: the case of PI3K inhibitors. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 30, 765-766. 253 Juric, D., Castel, P., Griffith, M., Griffith, O.L., Won, H.H., Ellis, H., Ebbesen, S.H., Ainscough, B.J., Ramu, A., Iyer, G., et al. (2015). Convergent loss of PTEN leads to clinical resistance to a PI(3)Kalpha inhibitor. Nature 518, 240-244. Kamada, Y., Fujioka, Y., Suzuki, N.N., Inagaki, F., Wullschleger, S., Loewith, R., Hall, M.N., and Ohsumi, Y. (2005). Tor2 directly phosphorylates the AGC kinase Ypk2 to regulate actin polarization. Molecular and cellular biology 25, 7239-7248. Kartal-Yandim, M., Adan-Gokbulut, A., and Baran, Y. (2015). Molecular mechanisms of drug resistance and its reversal in cancer. Critical reviews in biotechnology, 1-11. Kaufmann, S.H. (2008). Paul Ehrlich: founder of chemotherapy. Nature reviews Drug discovery 7, 373. Keppler-Noreuil, K.M., Rios, J.J., Parker, V.E., Semple, R.K., Lindhurst, M.J., Sapp, J.C., Alomari, A., Ezaki, M., Dobyns, W., and Biesecker, L.G. (2015). PIK3CA-related overgrowth spectrum (PROS): diagnostic and testing eligibility criteria, differential diagnosis, and evaluation. American journal of medical genetics Part A 167A, 287-295. Killion, E., Mohan, K., and Lee, E.I. (2014). A review of vascular anomalies: genetics and common syndromes. Semin Plast Surg 28, 64-68. Knight, Z.A., Chiang, G.G., Alaimo, P.J., Kenski, D.M., Ho, C.B., Coan, K., Abraham, R.T., and Shokat, K.M. (2004). Isoform-specific phosphoinositide 3-kinase inhibitors from an arylmorpholine scaffold. Bioorganic & medicinal chemistry 12, 4749-4759. Kobayashi, S., Boggon, T.J., Dayaram, T., Janne, P.A., Kocher, O., Meyerson, M., Johnson, B.E., Eck, M.J., Tenen, D.G., and Halmos, B. (2005). EGFR mutation and resistance of non-small-cell lung cancer to gefitinib. The New England journal of medicine 352, 786-792. Kobayashi, T., and Cohen, P. (1999). Activation of serum- and glucocorticoid-regulated protein kinase by agonists that activate phosphatidylinositide 3-kinase is mediated by 3- phosphoinositide-dependent protein kinase-1 (PDK1) and PDK2. The Biochemical journal 339 ( Pt 2), 319-328. Kobayashi, T., Deak, M., Morrice, N., and Cohen, P. (1999). Characterization of the structure and regulation of two novel isoforms of serum- and glucocorticoid-induced protein kinase. The Biochemical journal 344 Pt 1, 189-197. Koren, S., Reavie, L., Couto, J.P., De Silva, D., Stadler, M.B., Roloff, T., Britschgi, A., Eichlisberger, T., Kohler, H., Aina, O., et al. (2015). PIK3CA(H1047R) induces multipotency and multi-lineage mammary tumours. Nature 525, 114-118. Kurek, K.C., Luks, V.L., Ayturk, U.M., Alomari, A.I., Fishman, S.J., Spencer, S.A., Mulliken, J.B., Bowen, M.E., Yamamoto, G.L., Kozakewich, H.P., et al. (2012). Somatic mosaic activating mutations in PIK3CA cause CLOVES syndrome. American journal of human genetics 90, 1108-1115. Lannutti, B.J., Meadows, S.A., Herman, S.E., Kashishian, A., Steiner, B., Johnson, A.J., Byrd, J.C., Tyner, J.W., Loriaux, M.M., Deininger, M., et al. (2011). CAL-101, a 254 p110delta selective phosphatidylinositol-3-kinase inhibitor for the treatment of B-cell malignancies, inhibits PI3K signaling and cellular viability. Blood 117, 591-594. Laplante, M., and Sabatini, D.M. (2009). mTOR signaling at a glance. Journal of cell science 122, 3589-3594. Laplante, M., and Sabatini, D.M. (2012). mTOR signaling in growth control and disease. Cell 149, 274-293. Lawlor, M.A., Mora, A., Ashby, P.R., Williams, M.R., Murray-Tait, V., Malone, L., Prescott, A.R., Lucocq, J.M., and Alessi, D.R. (2002). Essential role of PDK1 in regulating cell size and development in mice. The EMBO journal 21, 3728-3738. Lemmon, M.A. (2008). Membrane recognition by phospholipid-binding domains. Nature reviews Molecular cell biology 9, 99-111. Li, Y., Corradetti, M.N., Inoki, K., and Guan, K.L. (2004). TSC2: filling the GAP in the mTOR signaling pathway. Trends in biochemical sciences 29, 32-38. Limaye, N., Wouters, V., Uebelhoer, M., Tuominen, M., Wirkkala, R., Mulliken, J.B., Eklund, L., Boon, L.M., and Vikkula, M. (2009). Somatic mutations in angiopoietin receptor gene TEK cause solitary and multiple sporadic venous malformations. Nature genetics 41, 118-124. Lindhurst, M.J., Parker, V.E., Payne, F., Sapp, J.C., Rudge, S., Harris, J., Witkowski, A.M., Zhang, Q., Groeneveld, M.P., Scott, C.E., et al. (2012). Mosaic overgrowth with fibroadipose hyperplasia is caused by somatic activating mutations in PIK3CA. Nature genetics 44, 928-933. Lito, P., Rosen, N., and Solit, D.B. (2013). Tumor adaptation and resistance to RAF inhibitors. Nature medicine 19, 1401-1409. Luqmani, Y.A. (2005). Mechanisms of drug resistance in cancer chemotherapy. Medical principles and practice : international journal of the Kuwait University, Health Science Centre 14 Suppl 1, 35-48. Maira, S.M., Stauffer, F., Brueggen, J., Furet, P., Schnell, C., Fritsch, C., Brachmann, S., Chene, P., De Pover, A., Schoemaker, K., et al. (2008). Identification and characterization of NVP-BEZ235, a new orally available dual phosphatidylinositol 3- kinase/mammalian target of rapamycin inhibitor with potent in vivo antitumor activity. Molecular cancer therapeutics 7, 1851-1863. Majewski, I.J., Nuciforo, P., Mittempergher, L., Bosma, A.J., Eidtmann, H., Holmes, E., Sotiriou, C., Fumagalli, D., Jimenez, J., Aura, C., et al. (2015). PIK3CA mutations are associated with decreased benefit to neoadjuvant human epidermal growth factor receptor 2-targeted therapies in breast cancer. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 33, 1334-1339. Manning, B.D., Tee, A.R., Logsdon, M.N., Blenis, J., and Cantley, L.C. (2002). Identification of the tuberous sclerosis complex-2 tumor suppressor gene product tuberin as a target of the phosphoinositide 3-kinase/akt pathway. Molecular cell 10, 151-162. 255 Matsuo, T., Kubo, Y., Watanabe, Y., and Yamamoto, M. (2003). Schizosaccharomyces pombe AGC family kinase Gad8p forms a conserved signaling module with TOR and PDK1-like kinases. The EMBO journal 22, 3073-3083. Mayer, I.A., Abramson, V.G., Lehmann, B.D., and Pietenpol, J.A. (2014). New strategies for triple-negative breast cancer--deciphering the heterogeneity. Clinical cancer research : an official journal of the American Association for Cancer Research 20, 782-790. McManus, E.J., Collins, B.J., Ashby, P.R., Prescott, A.R., Murray-Tait, V., Armit, L.J., Arthur, J.S., and Alessi, D.R. (2004). The in vivo role of PtdIns(3,4,5)P3 binding to PDK1 PH domain defined by knockin mutation. The EMBO journal 23, 2071-2082. Menon, S., Dibble, C.C., Talbott, G., Hoxhaj, G., Valvezan, A.J., Takahashi, H., Cantley, L.C., and Manning, B.D. (2014). Spatial control of the TSC complex integrates insulin and nutrient regulation of mTORC1 at the lysosome. Cell 156, 771-785. Metzker, M.L. (2010). Sequencing technologies - the next generation. Nature reviews Genetics 11, 31-46. Montagut, C., Sharma, S.V., Shioda, T., McDermott, U., Ulman, M., Ulkus, L.E., Dias- Santagata, D., Stubbs, H., Lee, D.Y., Singh, A., et al. (2008). Elevated CRAF as a potential mechanism of acquired resistance to BRAF inhibition in melanoma. Cancer research 68, 4853-4861. Mora, A., Komander, D., van Aalten, D.M., and Alessi, D.R. (2004). PDK1, the master regulator of AGC kinase signal transduction. Semin Cell Dev Biol 15, 161-170. Morrison, M.M., Hutchinson, K., Williams, M.M., Stanford, J.C., Balko, J.M., Young, C., Kuba, M.G., Sanchez, V., Williams, A.J., Hicks, D.J., et al. (2013). ErbB3 downregulation enhances luminal breast tumor response to antiestrogens. The Journal of clinical investigation 123, 4329-4343. Motzer, R.J., Escudier, B., Oudard, S., Hutson, T.E., Porta, C., Bracarda, S., Grunwald, V., Thompson, J.A., Figlin, R.A., Hollaender, N., et al. (2010). Phase 3 trial of everolimus for metastatic renal cell carcinoma : final results and analysis of prognostic factors. Cancer 116, 4256-4265. Muller, B.A. (2009). Imatinib and its successors--how modern chemistry has changed drug development. Current pharmaceutical design 15, 120-133. Muller, S., Chaikuad, A., Gray, N.S., and Knapp, S. (2015). The ins and outs of selective kinase inhibitor development. Nature chemical biology 11, 818-821. Munos, B. (2009). Lessons from 60 years of pharmaceutical innovation. Nature reviews Drug discovery 8, 959-968. Murray, J.T., Campbell, D.G., Morrice, N., Auld, G.C., Shpiro, N., Marquez, R., Peggie, M., Bain, J., Bloomberg, G.B., Grahammer, F., et al. (2004). Exploitation of KESTREL to identify NDRG family members as physiological substrates for SGK1 and GSK3. The Biochemical journal 384, 477-488. 256 Najafov, A., Shpiro, N., and Alessi, D.R. (2012). Akt is efficiently activated by PIF- pocket- and PtdIns(3,4,5)P3-dependent mechanisms leading to resistance to PDK1 inhibitors. The Biochemical journal 448, 285-295. Najafov, A., Sommer, E.M., Axten, J.M., Deyoung, M.P., and Alessi, D.R. (2011). Characterization of GSK2334470, a novel and highly specific inhibitor of PDK1. The Biochemical journal 433, 357-369. Nakashima, M., Miyajima, M., Sugano, H., Iimura, Y., Kato, M., Tsurusaki, Y., Miyake, N., Saitsu, H., Arai, H., and Matsumoto, N. (2014). The somatic GNAQ mutation c.548G>A (p.R183Q) is consistently found in Sturge-Weber syndrome. Journal of human genetics 59, 691-693. Nazarian, R., Shi, H., Wang, Q., Kong, X., Koya, R.C., Lee, H., Chen, Z., Lee, M.K., Attar, N., Sazegar, H., et al. (2010). Melanomas acquire resistance to B-RAF(V600E) inhibition by RTK or N-RAS upregulation. Nature 468, 973-977. Nguyen, A., Ho, L., and Wan, Y. (2014a). Chemotherapy and Oncolytic Virotherapy: Advanced Tactics in the War against Cancer. Frontiers in oncology 4, 145. Nguyen, H.L., Boon, L.M., and Vikkula, M. (2014b). Genetics of vascular malformations. Semin Pediatr Surg 23, 221-226. Norman, B.H., Shih, C., Toth, J.E., Ray, J.E., Dodge, J.A., Johnson, D.W., Rutherford, P.G., Schultz, R.M., Worzalla, J.F., and Vlahos, C.J. (1996). Studies on the mechanism of phosphatidylinositol 3-kinase inhibition by wortmannin and related analogs. Journal of medicinal chemistry 39, 1106-1111. O'Reilly, K.E., Rojo, F., She, Q.B., Solit, D., Mills, G.B., Smith, D., Lane, H., Hofmann, F., Hicklin, D.J., Ludwig, D.L., et al. (2006). mTOR inhibition induces upstream receptor tyrosine kinase signaling and activates Akt. Cancer research 66, 1500-1508. Osborn, A.J., Dickie, P., Neilson, D.E., Glaser, K., Lynch, K.A., Gupta, A., and Dickie, B.H. (2015). Activating PIK3CA alleles and lymphangiogenic phenotype of lymphatic endothelial cells isolated from lymphatic malformations. Human molecular genetics 24, 926-938. Ostrem, J.M., Peters, U., Sos, M.L., Wells, J.A., and Shokat, K.M. (2013). K-Ras(G12C) inhibitors allosterically control GTP affinity and effector interactions. Nature 503, 548- 551. Owens, P.K., Raddad, E., Miller, J.W., Stille, J.R., Olovich, K.G., Smith, N.V., Jones, R.S., and Scherer, J.C. (2015). A decade of innovation in pharmaceutical R&D: the Chorus model. Nature reviews Drug discovery 14, 17-28. Pearce, L.R., Huang, X., Boudeau, J., Pawlowski, R., Wullschleger, S., Deak, M., Ibrahim, A.F., Gourlay, R., Magnuson, M.A., and Alessi, D.R. (2007). Identification of Protor as a novel Rictor-binding component of mTOR complex-2. The Biochemical journal 405, 513-522. Pearce, L.R., Komander, D., and Alessi, D.R. (2010). The nuts and bolts of AGC protein kinases. Nature reviews Molecular cell biology 11, 9-22. 257 Peifer, C., and Alessi, D.R. (2008). Small-molecule inhibitors of PDK1. ChemMedChem 3, 1810-1838. Potter, C.J., Pedraza, L.G., and Xu, T. (2002). Akt regulates growth by directly phosphorylating Tsc2. Nature cell biology 4, 658-665. Poulikakos, P.I., Persaud, Y., Janakiraman, M., Kong, X., Ng, C., Moriceau, G., Shi, H., Atefi, M., Titz, B., Gabay, M.T., et al. (2011). RAF inhibitor resistance is mediated by dimerization of aberrantly spliced BRAF(V600E). Nature 480, 387-390. Pylayeva-Gupta, Y., Grabocka, E., and Bar-Sagi, D. (2011). RAS oncogenes: weaving a tumorigenic web. Nature reviews Cancer 11, 761-774. Rexer, B.N., and Arteaga, C.L. (2012). Intrinsic and acquired resistance to HER2- targeted therapies in HER2 gene-amplified breast cancer: mechanisms and clinical implications. Critical reviews in oncogenesis 17, 1-16. Richardson, P.G., Eng, C., Kolesar, J., Hideshima, T., and Anderson, K.C. (2012). Perifosine , an oral, anti-cancer agent and inhibitor of the Akt pathway: mechanistic actions, pharmacodynamics, pharmacokinetics, and clinical activity. Expert opinion on drug metabolism & toxicology 8, 623-633. Riviere, J.B., Mirzaa, G.M., O'Roak, B.J., Beddaoui, M., Alcantara, D., Conway, R.L., St- Onge, J., Schwartzentruber, J.A., Gripp, K.W., Nikkel, S.M., et al. (2012). De novo germline and postzygotic mutations in AKT3, PIK3R2 and PIK3CA cause a spectrum of related megalencephaly syndromes. Nature genetics 44, 934-940. Rodon, J., Dienstmann, R., Serra, V., and Tabernero, J. (2013). Development of PI3K inhibitors: lessons learned from early clinical trials. Nature reviews Clinical oncology 10, 143-153. Rodriguez-Antona, C., and Ingelman-Sundberg, M. (2006). Cytochrome P450 pharmacogenetics and cancer. Oncogene 25, 1679-1691. Rodrik-Outmezguine, V.S., Chandarlapaty, S., Pagano, N.C., Poulikakos, P.I., Scaltriti, M., Moskatel, E., Baselga, J., Guichard, S., and Rosen, N. (2011). mTOR kinase inhibition causes feedback-dependent biphasic regulation of AKT signaling. Cancer discovery 1, 248-259. Roy, A., Skibo, J., Kalume, F., Ni, J., Rankin, S., Lu, Y., Dobyns, W.B., Mills, G.B., Zhao, J.J., Baker, S.J., et al. (2015). Mouse models of human PIK3CA-related brain overgrowth have acutely treatable epilepsy. eLife 4. Rubin, E.H., and Gilliland, D.G. (2012). Drug development and clinical trials--the path to an approved cancer drug. Nature reviews Clinical oncology 9, 215-222. Rubin, M.A., Putzi, M., Mucci, N., Smith, D.C., Wojno, K., Korenchuk, S., and Pienta, K.J. (2000). Rapid ("warm") autopsy study for procurement of metastatic prostate cancer. Clinical cancer research : an official journal of the American Association for Cancer Research 6, 1038-1045. 258 Sabatini, D.M. (2006). mTOR and cancer: insights into a complex relationship. Nature reviews Cancer 6, 729-734. Sancak, Y., Thoreen, C.C., Peterson, T.R., Lindquist, R.A., Kang, S.A., Spooner, E., Carr, S.A., and Sabatini, D.M. (2007). PRAS40 is an insulin-regulated inhibitor of the mTORC1 protein kinase. Molecular cell 25, 903-915. Sarbassov, D.D., Ali, S.M., Kim, D.H., Guertin, D.A., Latek, R.R., Erdjument-Bromage, H., Tempst, P., and Sabatini, D.M. (2004). Rictor, a novel binding partner of mTOR, defines a rapamycin-insensitive and raptor-independent pathway that regulates the cytoskeleton. Current biology : CB 14, 1296-1302. Sarbassov, D.D., Ali, S.M., Sengupta, S., Sheen, J.H., Hsu, P.P., Bagley, A.F., Markhard, A.L., and Sabatini, D.M. (2006). Prolonged rapamycin treatment inhibits mTORC2 assembly and Akt/PKB. Molecular cell 22, 159-168. Sarbassov, D.D., Guertin, D.A., Ali, S.M., and Sabatini, D.M. (2005). Phosphorylation and regulation of Akt/PKB by the rictor-mTOR complex. Science 307, 1098-1101. Schmitt, M.W., Loeb, L.A., and Salk, J.J. (2015). The influence of subclonal resistance mutations on targeted cancer therapy. Nat Rev Clin Oncol. Schwartz, S., Wongvipat, J., Trigwell, C.B., Hancox, U., Carver, B.S., Rodrik- Outmezguine, V., Will, M., Yellen, P., de Stanchina, E., Baselga, J., et al. (2015). Feedback suppression of PI3Kalpha signaling in PTEN-mutated tumors is relieved by selective inhibition of PI3Kbeta. Cancer cell 27, 109-122. Scott, A.M., Wolchok, J.D., and Old, L.J. (2012). Antibody therapy of cancer. Nature reviews Cancer 12, 278-287. Shaw, A.T., Friboulet, L., Leshchiner, I., Gainor, J.F., Bergqvist, S., Brooun, A., Burke, B.J., Deng, Y.L., Liu, W., Dardaei, L., et al. (2016). Resensitization to Crizotinib by the Lorlatinib ALK Resistance Mutation L1198F. The New England journal of medicine 374, 54-61. Shen, T., Pajaro-Van de Stadt, S.H., Yeat, N.C., and Lin, J.C. (2015). Clinical applications of next generation sequencing in cancer: from panels, to exomes, to genomes. Frontiers in genetics 6, 215. Shih, J.Y., Gow, C.H., and Yang, P.C. (2005). EGFR mutation conferring primary resistance to gefitinib in non-small-cell lung cancer. The New England journal of medicine 353, 207-208. Shirley, M.D., Tang, H., Gallione, C.J., Baugher, J.D., Frelin, L.P., Cohen, B., North, P.E., Marchuk, D.A., Comi, A.M., and Pevsner, J. (2013). Sturge-Weber syndrome and port-wine stains caused by somatic mutation in GNAQ. The New England journal of medicine 368, 1971-1979. Sommer, E.M., Dry, H., Cross, D., Guichard, S., Davies, B.R., and Alessi, D.R. (2013). Elevated SGK1 predicts resistance of breast cancer cells to Akt inhibitors. The Biochemical journal 452, 499-508. 259 Soukas, A.A., Kane, E.A., Carr, C.E., Melo, J.A., and Ruvkun, G. (2009). Rictor/TORC2 regulates fat metabolism, feeding, growth, and life span in Caenorhabditis elegans. Genes Dev 23, 496-511. Storz, P., and Toker, A. (2002). 3'-phosphoinositide-dependent kinase-1 (PDK-1) in PI 3- kinase signaling. Frontiers in bioscience : a journal and virtual library 7, d886-902. Thorpe, L.M., Yuzugullu, H., and Zhao, J.J. (2015). PI3K in cancer: divergent roles of isoforms, modes of activation and therapeutic targeting. Nature reviews Cancer 15, 7-24. Uebelhoer, M., Boon, L.M., and Vikkula, M. (2012). Vascular anomalies: from genetics toward models for therapeutic trials. Cold Spring Harb Perspect Med 2. Van Keymeulen, A., Lee, M.Y., Ousset, M., Brohee, S., Rorive, S., Giraddi, R.R., Wuidart, A., Bouvencourt, G., Dubois, C., Salmon, I., et al. (2015). Reactivation of multipotency by oncogenic PIK3CA induces breast tumour heterogeneity. Nature 525, 119-123. Vanhaesebroeck, B., and Alessi, D.R. (2000). The PI3K-PDK1 connection: more than just a road to PKB. The Biochemical journal 346 Pt 3, 561-576. Vanhaesebroeck, B., Guillermet-Guibert, J., Graupera, M., and Bilanges, B. (2010). The emerging mechanisms of isoform-specific PI3K signalling. Nature reviews Molecular cell biology 11, 329-341. Vanhaesebroeck, B., Stephens, L., and Hawkins, P. (2012). PI3K signalling: the path to discovery and understanding. Nature reviews Molecular cell biology 13, 195-203. Vasudevan, K.M., Barbie, D.A., Davies, M.A., Rabinovsky, R., McNear, C.J., Kim, J.J., Hennessy, B.T., Tseng, H., Pochanard, P., Kim, S.Y., et al. (2009). AKT-independent signaling downstream of oncogenic PIK3CA mutations in human cancer. Cancer cell 16, 21-32. Verrill, M. (2009). Chemotherapy for early-stage breast cancer: a brief history. British journal of cancer 101 Suppl 1, S2-5. Vivanco, I. (2014). Targeting molecular addictions in cancer. British journal of cancer 111, 2033-2038. Vlahos, C.J., Matter, W.F., Hui, K.Y., and Brown, R.F. (1994). A specific inhibitor of phosphatidylinositol 3-kinase, 2-(4-morpholinyl)-8-phenyl-4H-1-benzopyran-4-one (LY294002). The Journal of biological chemistry 269, 5241-5248. Wassef, M., Blei, F., Adams, D., Alomari, A., Baselga, E., Berenstein, A., Burrows, P., Frieden, I.J., Garzon, M.C., Lopez-Gutierrez, J.C., et al. (2015). Vascular Anomalies Classification: Recommendations From the International Society for the Study of Vascular Anomalies. Pediatrics. Webb, A.E., and Brunet, A. (2014). FOXO transcription factors: key regulators of cellular quality control. Trends in biochemical sciences 39, 159-169. 260 Webster, M.K., Goya, L., and Firestone, G.L. (1993a). Immediate-early transcriptional regulation and rapid mRNA turnover of a putative serine/threonine protein kinase. The Journal of biological chemistry 268, 11482-11485. Webster, M.K., Goya, L., Ge, Y., Maiyar, A.C., and Firestone, G.L. (1993b). Characterization of sgk, a novel member of the serine/threonine protein kinase gene family which is transcriptionally induced by glucocorticoids and serum. Molecular and cellular biology 13, 2031-2040. Wee, S., Wiederschain, D., Maira, S.M., Loo, A., Miller, C., deBeaumont, R., Stegmeier, F., Yao, Y.M., and Lengauer, C. (2008). PTEN-deficient cancers depend on PIK3CB. Proceedings of the National Academy of Sciences of the United States of America 105, 13057-13062. Wipf, P., Minion, D.J., Halter, R.J., Berggren, M.I., Ho, C.B., Chiang, G.G., Kirkpatrick, L., Abraham, R., and Powis, G. (2004). Synthesis and biological evaluation of synthetic viridins derived from C(20)-heteroalkylation of the steroidal PI-3-kinase inhibitor wortmannin. Organic & biomolecular chemistry 2, 1911-1920. Wojtkowiak, J.W., Verduzco, D., Schramm, K.J., and Gillies, R.J. (2011). Drug resistance and cellular adaptation to tumor acidic pH microenvironment. Molecular pharmaceutics 8, 2032-2038. Wulff, P., Vallon, V., Huang, D.Y., Volkl, H., Yu, F., Richter, K., Jansen, M., Schlunz, M., Klingel, K., Loffing, J., et al. (2002). Impaired renal Na(+) retention in the sgk1-knockout mouse. The Journal of clinical investigation 110, 1263-1268. Yao, J.C., Shah, M.H., Ito, T., Bohas, C.L., Wolin, E.M., Van Cutsem, E., Hobday, T.J., Okusaka, T., Capdevila, J., de Vries, E.G., et al. (2011). Everolimus for advanced pancreatic neuroendocrine tumors. The New England journal of medicine 364, 514-523. Yu, M., Stott, S., Toner, M., Maheswaran, S., and Haber, D.A. (2011). Circulating tumor cells: approaches to isolation and characterization. The Journal of cell biology 192, 373- 382. Zahreddine, H., and Borden, K.L. (2013). Mechanisms and insights into drug resistance in cancer. Frontiers in pharmacology 4, 28. Zhao, L., and Vogt, P.K. (2008a). Class I PI3K in oncogenic cellular transformation. Oncogene 27, 5486-5496. Zhao, L., and Vogt, P.K. (2008b). Helical domain and kinase domain mutations in p110alpha of phosphatidylinositol 3-kinase induce gain of function by different mechanisms. Proceedings of the National Academy of Sciences of the United States of America 105, 2652-2657. Zhao, Z., Wu, H., Wang, L., Liu, Y., Knapp, S., Liu, Q., and Gray, N.S. (2014). Exploration of type II binding mode: A privileged approach for kinase inhibitor focused drug discovery? ACS chemical biology 9, 1230-1241.