Molecular Characterization and Novel Therapeutic Approaches in Hepatocellular Carcinoma Laura Torrens Fontanals Aquesta tesi doctoral està subjecta a la llicència Reconeixement- NoComercial – CompartirIgual 4.0. Espanya de Creative Commons. Esta tesis doctoral está sujeta a la licencia Reconocimiento - NoComercial – CompartirIgual 4.0. España de Creative Commons. This doctoral thesis is licensed under the Creative Commons Attribution-NonCommercial- ShareAlike 4.0. Spain License. Faculty of Medicine Molecular Characterization and Novel Therapeutic Approaches in Hepatocellular Carcinoma Doctoral thesis report submitted by Laura Torrens Fontanals To obtain a doctoral degree by the University of Barcelona Supervised by: Prof. Josep M Llovet Bayer Full Professor of Medicine – University of Barcelona Professor of Research – Institut d'Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Hospital Clínic de Barcelona Doctoral Programme in Medicine and Translational Research Faculty of Medicine and Health Sciences. University of Barcelona December 2021 Professor of Research IDIBAPS Liver Unit, Hospital Clínic Barcelona Villaroel 170 08036 Barcelona Catalonia, Spain Office: +34-932279156 Lab: +34-932279155 Fax: +34-932275792 E-mail: jmllovet@clinic.cat Assistant: Ariadna Farré (afarrec@clinic.cat) Barcelona, 1 de Desembre de 2021 Jo, Dr. Josep M. Llovet Bayer, cap del “Laboratori de Recerca Translacional en Oncologia Hepàtica” de l’Institut d’investigacions Biomèdiques August Pi i Sunyer (IDIBAPS) – Universitat de Barcelona – Hospital Clínic, Professor ICREA i director de Máster Oficial de Medicina Translacional de la Universitat de Barcelona, CERTIFICO: Que la tesi doctoral titulada “Molecular Characterization and Novel Therapeutic Approaches in Hepatocellular Carcinoma” presentada per Laura Torrens Fontanals per optar al títol de Doctor en Medicina i Recerca Translacional per la Universitat de Barcelona s’ha realitzat sota la meva direcció i compleix tots el requisits necessaris per ser defensada davant el Tribunal d’Avaluació corresponent. Dr. Josep M Llovet i Bayer Conformitat de l’estudiant de doctorat: Laura Torrens Fontanals Vigila, esperit, vigila; no perdis mai el teu nord; no et deixis dur a la tranquil·la aigua mansa de cap port. Gira, gira els ulls enlaire, no miris les platges roïns, dóna el front en el gran aire; sempre, sempre mar endins. Sempre amb les veles suspeses del cel al mar transparent; sempre entorn aigües esteses que es moguin eternament. Fuig-ne de la terra immoble; fuig dels horitzons mesquins; sempre al mar, al gran mar noble: sempre, sempre mar endins. Fora terres, fora platja; oblida't de tot regrés; no s'acaba el teu viatge; no s'acabarà mai més. Joan Maragall ACKNOWLEDGEMENTS First and foremost, I would like to thank Dr. Josep M Llovet, my thesis director, for his support during my PhD, for the invaluable opportunity to participate in the research projects included in this thesis, and for always pushing me to be better. I also want to thank him for providing the ideal environment both in Barcelona and New York for me to grow as a researcher. I have been so fortunate to have the most supportive and amazing labmates, who have made the most enjoyable experience out of this PhD, even when experiments fail, grants get rejected and reviewer 3 doesn’t like your data. For this, I want to give my most heartfelt thank you to everyone at the Liver Cancer Translational Research laboratory at IDIBAPS. I would like to begin by thanking Agrin for guiding me in my first steps as a disoriented “little grasshopper”. To Roser, Victoria, and Catherine, for their continuous support and predisposition to answering all my questions. To Iris, Sara, and Robert for being my “senior PhD students” and reference during this journey. To Carmen, for being my partner in crime both inside and outside the lab and for making me smile even when there is an ocean in between us. To Florian, for all the priceless laughs, advice, and scientific discussions in the office. To all my “junior PhD students”, Jordi, Ugne, Agavni, Roger, Marta, and Júlia for making going to work a true pleasure, and particularly to the lenva/pembro and hepatoblastoma teams, for being a fundamental part of this thesis. To Judit and Laia, thanks for teaching me all the techniques and for being so patient with my absent- minded self – sorry for forgetting to turn off the centrifuge for the millionth time. Also to Ari, Marta, and Sara for always being so helpful and making the administrative side of research much easier. My stay in New York has been one of the most enriching experiences of this thesis, both professionally and personally. I would like to thank everyone at the Mount Sinai Liver Cancer Program and the ISMMS Department of Liver Diseases for their support and for making me feel at home since day one. To Daniela, for her mentoring and for taking the time to make sure that my projects were moving forward. To the relentless Mongolian team, Marc and Miguel, for cheering with me for all the victories and defeats – or just Fridays at Earl’s. Especially to Marc, per ser Nova York. To Philipp for all his advice and conversations on our way to the animal facility, and also for showing me how to sneak into Nowadays. To Carla and Miho for making all the hours spent at the lab more fun. Equally important for my cherished time in New York have been all those friends that reminded me there is a whole city outside of Mount Sinai. For this, I want to thank Vinit and Sara for including me in countless adventures and introducing me to the loveliest and most eclectic group of New Yorkers. Also, to Yago and Sami for completing the Earl’s gang and for helping me master my Jenga skills week after week. Com explicava el meu professor d’universitat Dr. David Bueno, la llengua materna està estretament lligada a l’emoció. Certament, no podria escriure els següents agraïments de cap altra manera que en català. A tots els amics i companys que m’heu acompanyat durant aquesta etapa, us estaré per sempre més agraïda. Gràcies, Paula, Irina, Irene, Jordi, Elena i Ania per ser el meu ATG. En especial a tu, Paula, per ser un pilar i la millor companya en els plans més descabellats. A la Blanca per ser-hi, des de fa gairebé 15 anys, cada estiu davant del Ceferino a les sis i cada cop que he necessitat una amiga. I també a la Júlia per les seves paraules d’ànims acompanyades sempre dels millors gin-tònics amb cardamom. Per últim, vull expressar el més profund agraïment a la meva família. En especial, als meus pares, Jordi i Núria, pel seu amor incondicional i per haver-me donat totes les eines i els valors que m’han permès arribar fins aquí. A la Mariona per ser la meva other half i per fer-me sentir sempre compresa. Aquest camí hauria sigut impossible si no l’hagués emprès amb tu. A l’Anna, pels seus bons consells de germana gran, i a la petita Berta per portar-nos joia i ajudar-me a retrobar la bellesa en les coses més essencials. Per últim, als avis Jordi i Carme, que ja no hi son, per fer-me sentir sempre estimada i encoratjada, i als avis Ernest i Carme per ser el meu referent i per construir, amb afecte i tenacitat, la nostra família que és el millor tresor que podria tenir. TABLE OF CONTENTS TABLE OF CONTENTS 5 INDEX ABBREVIATIONS ........................................................................................................................... 9 LIST OF ARTICLES COMPRISED IN THE THESIS ............................................................................ 13 THESIS SUMMARY (CATALAN) ................................................................................................... 17 INTRODUCTION .......................................................................................................................... 23 1. Principles of Cancer ............................................................................................................ 25 1.1. Worldwide Impact of Cancer ........................................................................................ 25 1.2. Biological Basis of Cancer .............................................................................................. 25 1.3. Genomic Basis of Cancer ............................................................................................... 28 1.4. Role of the Immune System in Cancer .......................................................................... 33 2. Translational Research in Cancer ....................................................................................... 34 2.1. Translational Research .................................................................................................. 34 2.2. Commonly Used Tools for Cancer Research ................................................................. 36 2.3. Current Challenges in Translational Oncology .............................................................. 41 3. Hepatocellular Carcinoma .................................................................................................. 41 3.1. Epidemiology and Risk Factors ...................................................................................... 41 3.2. Molecular Pathogenesis of HCC .................................................................................... 45 3.3. Tumor Microenvironment ............................................................................................ 52 3.4. Clinical Management of HCC Patients ........................................................................... 57 HYPOTHESES AND AIMS ............................................................................................................. 65 1. Hypotheses .......................................................................................................................... 67 2. Aims ..................................................................................................................................... 69 RESULTS ...................................................................................................................................... 71 Study 1 – Hepatocellular Carcinoma in Mongolia Delineates Unique Genomic Features .... 73 Summary .............................................................................................................................. 73 Publication ........................................................................................................................... 77 Study 2 – Liver Injury Increases the Incidence of HCC following AAV Gene Therapy in Mice ............................................................................................................................................... 105 Summary ............................................................................................................................ 105 Publication ......................................................................................................................... 105 Study 3 – Immunomodulatory Effects of Lenvatinib Plus Anti-Programmed Cell Death Protein 1 in Mice and Rationale for Patient Enrichment in Hepatocellular Carcinoma ..... 119 Summary ............................................................................................................................ 119 Publication ......................................................................................................................... 121 DISCUSSION .............................................................................................................................. 139 TABLE OF CONTENTS 6 1. Molecular Features of HCC in Mongolia ........................................................................... 141 2. Risk of AAV Integration in NAFLD Patients Undergoing Gene Therapy ........................ 147 3. Immunomodulatory Effects of Lenvatinib Plus Anti-PD1 Combination .......................... 152 4. Improving the Understanding of HCC Pathogenesis with Translational Approaches ..... 157 CONCLUSIONS ........................................................................................................................... 161 REFERENCES .............................................................................................................................. 165 FIGURES Figure 1. Hallmarks of cancer and enabling characteristics ........................................................ 27 Figure 2. Number of somatic mutations in representative human cancers ............................... 29 Figure 3. Active mutational processes over the course of cancer development ........................ 31 Figure 4. Example of mutational signature in cancer ................................................................. 32 Figure 5. The translational medicine pipeline. ............................................................................ 35 Figure 6. Age-standardized incidence rates in liver cancer ......................................................... 43 Figure 7. Mutational signatures identified in HCC. ..................................................................... 50 Figure 8. Molecular classification of HCC .................................................................................... 51 Figure 9. HCC immune-based classification ................................................................................ 53 Figure 10. Direct effects of VEGF pathway activation on tumor-infiltrated immune cells ......... 56 Figure 11. The Barcelona Clinic Liver Cancer (BCLC) staging system .......................................... 59 Figure 12. Cancer incidence in Mongolia .................................................................................. 142 Figure 13. Somatic mutation in HCC-driving signaling pathways .............................................. 144 Figure 14. Molecular classification of Mongolian HCC .............................................................. 147 Figure 15. Viral mechanisms of liver carcinogenesis ................................................................ 148 Figure 16. Overview of recombinant AAV interventional gene therapy clinical trials .............. 149 Figure 17. Treatment strategy for advanced HCC ..................................................................... 153 Figure 18. Vascular-normalizing therapies can reprogram the immunosuppressive tumor microenvironment .................................................................................................................... 155 Figure 19. HCC classification according to its immunological features and potential response to the combination therapy .......................................................................................................... 157 Figure 20. Summary of the studies comprised in this doctoral thesis ...................................... 159 TABLE OF CONTENTS 7 TABLES Table 1. Mutational signatures included in the COSMIC catalog associated with known or suspected carcinogens in humans. ............................................................................................. 33 Table 2. Recurrent somatic driver alterations in resected HCCs. ................................................ 46 Table 3. Summary of surveillance strategies. ............................................................................. 57 Table 4. Summary of main outcomes and adverse events among systemic therapies approved for advanced HCC. ....................................................................................................................... 60 ANNEX ANNEX A ................................................................................................................................... 181 1. Publications ....................................................................................................................... 183 2. Communications to Scientific Meetings ........................................................................... 185 3. Grants and Awards ........................................................................................................... 187 ANNEX B .................................................................................................................................... 189 1. Supplementary Data Study 1 ............................................................................................ 191 2. Supplementary Data Study 2 ............................................................................................ 230 3. Supplementary Data Study 3 ............................................................................................ 235 ABBREVIATIONS ABBREVIATIONS 11 AAV: Adeno-associated virus AAV2: Adeno-associated virus type 2 AFP: Alpha fetoprotein ALT: Alanine aminotransferase BCLC: Barcelona Clinic Liver Cancer BCP: Basal core promoter CNV: Copy number variation CTLA4: Cytotoxic T-lymphocyte-associated protein 4 CTNNB1: Catenin beta 1 DC: Dendritic cell DC1: Type 1 dendritic cell DDR: DNA damage response DMS: Dimethyl sulfate FC: Fold change FDA: Food and Drug Administration FGFR: Fibroblast growth factor receptor FOXP3: Forkhead box protein 3 GEMM: Genetically engineered mouse model GSEA: Gene set enrichment analysis HBsAg: HBV surface antigen HBV: Hepatitis B virus HCC: Hepatocellular carcinoma HCV: Hepatitis C virus HDV: Hepatitis delta virus HFD: High-fat diet HGDN: High-grade dysplastic nodules ICGC: International Cancer Genome Consortium ICI: Immune checkpoint inhibitor IFN- α: Interferon α IHC: Immunohistochemistry LGDN: Low-grade dysplastic nodule MDSC: Myeloid-derived suppressor cell NAFLD: Non-alcoholic fatty liver disease NASH: Non-alcoholic steatohepatitis NGS: Next generation sequencing ABBREVIATIONS 12 NK: natural killer NMFc: Non-negative matrix factorization NTP: Nearest template prediction ORR: Objective response rate PDX: Patient-derived xenograft PD1: Programmed cell death protein 1 rAAV: Recombinant adeno-associated virus RET: RET proto-oncogene RNA-seq: RNA sequencing SBS: Single base substitution SBSM: SBS Mongolia SNP: Single-nucleotide polymorphism SNV: Single nucleotide variant ssGSEA: Single sample gene set enrichment analysis TCGA: The Cancer Genome Atlas TGF-β: Transforming growth factor β TAM: Tumor-associated macrophage TAN: Tumor-associated neutrophils TIM3: T cell immunoglobulin and mucin domain containing-3 TKI: Tyrosine-kinase inhibitor TMB: Tumor mutational burden Treg: Regulatory T cell WES: Whole-exome sequencing WGS: Whole-genome sequencing LIST OF ARTICLES COMPRISED IN THE THESIS LIST OF ARTICLES COMPRISED IN THE THESIS 15 Thesis in the form of a collection of published articles. The thesis includes 3 articles and 3 aims. Study 1 – Molecular Characterization of Hepatocellular Carcinoma in Mongolia Delineates Unique Genomic Features Torrens L, Puigvehí M, Torres-Martín M, Wang H, Maeda M, Haber PK, Leonel T, García-López M, Leow WQ, Montironi C, Torrecilla S, Varadarajan AR, Taik P, Campreciós G, Enkhbold C, Taivanbaatar E, Yerbolat A, Villanueva A, Pérez-del-Pulgar S, Thung S, Chinburen J, Letouzé E, Zucman-Rossi J, Uzilov A, Neely J, Forns X, Roayaie S, Sia D, Llovet JM. Molecular Characterization of Hepatocellular Carcinoma in Mongolia Delineates Unique Genomic Features. Submitted to Proc Natl Acad Sci U S A. 2021. Impact factor: 11.205, 1st quartile. Subject Areas: Multidisciplinary. Aim: 1. To provide a molecular characterization of Mongolian HCC and identify its unique genomic features compared to Western HCC. Study 2 – Liver Injury Increases the Incidence of HCC following AAV Gene Therapy in Mice Dalwadi D, Torrens L, Abril-Fornaguera J, Pinyol R, Willoughby C, Posey J, Llovet JM, Lanciault C, Russell DW, Grompe M, Naugler WE. Liver Injury Increases the Incidence of HCC following AAV Gene Therapy in Mice. Mol Ther. 2021;29:680–90. Impact factor: 11.454, 1st quartile. Subject Areas: Genetics, Molecular Biology, Molecular Medicine, Drug Discovery, Pharmacology, Medicine (miscellaneous). Aim: 2. To assess whether NAFLD-associated liver damage increase the risk of AAV integration inducing HCC. LIST OF ARTICLES COMPRISED IN THE THESIS 16 Study 3 – Immunomodulatory Effects of Lenvatinib Plus Anti-Programmed Cell Death Protein 1 in Mice and Rationale for Patient Enrichment in Hepatocellular Carcinoma Torrens L, Montironi C, Puigvehí M, Mesropian A, Leslie J, Haber PK, Maeda M, Balaseviciute U, Willoughby CE, Abril-Fornaguera J, Piqué-Gili M, Torres-Martín M, Peix J, Geh D, Ramon-Gil E, Saberi B, Friedman SL, Mann DA, Sia D, Llovet JM. Immunomodulatory Effects of Lenvatinib Plus Anti–Programmed Cell Death Protein 1 in Mice and Rationale for Patient Enrichment in Hepatocellular Carcinoma. Hepatology. 2021;74:2652–69. Impact factor: 17.425, 1st quartile. Subject Areas: Hepatology, Medicine (miscellaneous). Aim: 3. To identify the immunomodulatory effects of lenvatinib in combination with anti-PD1 and provide a mechanistic rationale for this treatment in advanced HCC. THESIS SUMMARY (CATALAN) THESIS SUMMARY (CATALAN) 19 Títol Caracterització Molecular i Noves Estratègies Terapèutiques pel Carcinoma Hepatocel·lular Introducció El carcinoma hepatocel·lular (CHC) representa un important problema de salut pública degut a la seva elevada incidència i mortalitat1,2. La incidència mundial de CHC és heterogènia, sent Mongòlia el país amb més casos per habitant, gairebé 10 vegades per sobre de la mitjana global2. Els principals factors de risc del CHC són la infecció pel virus de l'hepatitis B o pel virus de l'hepatitis C i la malaltia del fetge gras no alcohòlica (NAFLD, de les seves sigles en anglès). També s’han proposat altres agents que podrien estar associats a hepatocarcinogènesi, inclosa la infecció per virus adenoassociat (VAA)3,4, però és necessària més informació per determinar en quines condicions aquest virus pot promoure el CHC. Al voltant del 40% dels pacients amb CHC són diagnosticats en etapes avançades de la malaltia, en les quals són elegibles per a teràpies sistèmiques incloent inhibidors multiquinases (p. ex., lenvatinib) i inhibidors de punts de control immunitaris (ICI; p. ex., anticossos anti-PD1). Tot i que els ICI estan revolucionant el tractament del CHC, només aconsegueixen respostes en el ~15% dels pacients en monoteràpia5,6. Per contra, noves teràpies combinades estan incrementant la taxa de resposta fins al ~30%. Considerant això, és necessari identificar noves teràpies que puguin fer sinergia amb ICI per tal d’augmentar la supervivència dels pacients. Durant els últims anys, la caracterització molecular i immunològica de tumors mitjançant estudis translacionals ha proporcionat una millor comprensió de la patogènesi molecular del CHC5,6. Aquest coneixement té una gran rellevància clínica, ja que, entre altres coses, permet 1) identificar determinants genètics i moleculars que afavoreixen l'hepatocarcinogènesi en poblacions específiques de pacients; i 2) dissenyar nous tractaments incloent teràpies combinades. Hipòtesis Les hipòtesis d’aquesta tesi són les següents: - L’anàlisi exhaustiva de les característiques moleculars i immunològiques del CHC proporcionarà nova informació sobre els determinants genètics i moleculars associats al CHC en poblacions d’alt risc, afavorint l’estudi de noves estratègies terapèutiques. Aquestes dades podrien tenir implicacions fonamentals per a la presa de decisions clíniques en CHC. THESIS SUMMARY (CATALAN) 20 - L'avaluació de les alteracions genòmiques i transcripcionals del CHC a Mongòlia pot proporcionar nova informació que ajudi a identificar factors genètics i ambientals propis associats a l’elevada incidència en aquesta població. - La malaltia hepàtica crònica i NAFLD promouen la integració oncogènica del VAA i el desenvolupament de CHC, la qual cosa podria ser un condicionant per a l'ús de la teràpia gènica mitjançant VAA en pacients. - L'inhibidor multiquinasa lenvatinib té potencial immunomodulador i la seva combinació amb ICI anti-PD1 podria incrementar l’efecte anti-tumoral en models experimentals i en pacients amb CHC. Objectius Els objectius específics d'aquesta tesi doctoral són: - Realitzar una caracterització molecular del CHC a Mongòlia i identificar trets moleculars específics en comparació amb tumors de pacients occidentals. - Analitzar si el NAFLD incrementa el risc d’integració del VAA en hepatòcits, induint així el CHC. - Investigar l'efecte immunomodulador de lenvatinib en combinació amb l'anticòs anti-PD1 per proporcionar una base racional pel seu ús com a tractament pel CHC avançat. Mètodes - Estudi #1 (Torrens et al., en revisió): Un total de 192 mostres de tumors hepàtics CHC de pacients provinents de Mongòlia i 187 mostres de CHC de pacients occidentals (Europa i EEUU) van ser analitzades mitjançant seqüenciació completa d'exoma i d’ARN. Per identificar les característiques moleculars úniques del CHC de Mongòlia, es van comparar les característiques clinico-patològiques dels tumors de pacients mongols i occidentals, així com els seus perfils mutacionals i transcripcionals. - Estudi #2 (Dalwadi et al., Mol Ther 2021): Per tal de determinar si la infecció pel virus VAA dels hepatòcis promou l’hepatocarcinogènesi en condicions de malaltia hepàtica, es va fer servir un model animal murí on ratolins adults i nounats de la soca C57BL/6 van ser infectats amb un vector VAA dirigit contra el locus genètic Rian (VAA-Rian). A continuació, els animals van ser a) alimentats amb una dieta rica en greixos per generar NAFLD o b) sotmesos a hepatectomia parcial per generar dany hepàtic i regeneració. Es va monitoritzar el THESIS SUMMARY (CATALAN) 21 desenvolupament tumoral i els tumors generats es van caracteritzar mitjançant seqüenciació d'ARN. - Estudi #3 (Torrens et al., Hepatology 2021): Per avaluar l'activitat anti-tumoral de la combinació de l’inhibidor multiquinasa lenvatinib amb anti-PD1 es van generar tres models murins singènics de CHC. L’impacte dels tractaments sobre el perfil molecular i immunitari dels tumors van ser analitzats per citometria de flux i perfilats a nivell transcripcional i immunohistoquímic. Finalment, es va explorar el perfil d'expressió gènica de 228 tumors humans per identificar pacients amb CHC que es podrien beneficiar de la teràpia combinada de lenvatinib i anti-PD1. Resultats - Estudi #1 (Torrens et al., en revisió): Els resultats del nostre estudi indiquen que els tumors CHC de pacients de Mongòlia presenten les següents particularitats: 1. Una elevada taxa de mutacions, que gairebé dobla la taxa de mutacions de la cohort occidental (121 davant de 70 mutacions per tumor). 2. Un perfil mutacional diferent, amb una major freqüència de mutacions de TP53, APOB, TSC2 i la família de gens KMT2. 3. La presència d'una nova firma mutacional (SBS Mongòlia) al 25% dels casos, que està associada a la firma del genotòxic dimetil sulfat derivat de la combustió del carbó. 4. A nivell d’expressió gènica, els tumors de la cohort de Mongòlia es classifiquen en tres classes moleculars (MGL1, MGL2 i MGL3), dues de les quals (MGL2 al 26% dels CHC i MGL3 al 30%) i presenten característiques moleculars i clínico-patològiques no observades en mostres de CHC occidental. - Estudi #2 (Dalwadi et al., Mol Ther 2021): 1. Només el 5% dels ratolins adults infectats amb VAA-Rian van desenvolupar CHC. Per contra, l’elevada proliferació cel·lular en nounats va promoure un increment en la integració viral i generació de tumors en el 100% dels animals. 2. Tant la lesió hepàtica per la dieta rica en greix com la derivada de l’hepatectomia parcial van conduir a un increment en l’aparició de CHC en ratolins adults infectats amb VAA-Rian en comparació amb animals infectats però no sotmesos a estrès hepàtic. 3. L'anàlisi transcriptòmica de tumors de CHC murins de ratolins originats degut a la infecció amb VAA-Rian va revelar similituds moleculars amb tumors de CHC humans THESIS SUMMARY (CATALAN) 22 amb sobreexpressió del locus Rian, associats amb proliferació, fenotip agressiu i mal pronòstic. 4. El fetge de ratolins alimentats amb una dieta rica en greixos presentaven un perfil immunològic protumorigènic, que podria ser el motiu de l’increment del risc de CHC induït per la infecció de VAA. - Estudi #3 (Torrens et al., Hepatology 2021): 1. La combinació de lenvatinib i anti-PD1 aconsegueix una major taxa de resposta, un temps de resposta més curt i una reducció de la viabilitat del tumor en comparació amb els tractaments administrats en monoteràpia. 2. Lenvatinib exerceix un potent efecte immunomodulador sobre l'infiltrat tumoral caracteritzat per una reducció en la proporció de cèl·lules T reguladores i inhibició de la senyalització immunosupressora mitjançant TGFß. 3. El tractament anti-PD1 modifica l'infiltrat immunològic del tumor augmentant les cèl·lules T i la subpoblació de cèl·lules dendrítiques de tipus 1 al tumor. 4. La combinació de lenvatinib amb anti-PD1 és l’únic dels tractaments avaluats capaç de generar una resposta immune anti-tumoral activada. 5. El 22% dels pacients amb CHC presenten un perfil transcripcional que podria estar associat a resistència primària a anti-PD1 i resposta al règim combinat. Aquests pacients podrien beneficiar-se de l'efecte de reforç del tractament de combinació. Conclusions - El CHC a Mongòlia presenta característiques moleculars úniques comparat amb els tumors de pacients occidentals. El seu perfil mutacional podria ser degut a exposició a factors ambientals responsables de l'elevada incidència en aquest país. - La infecció amb el virus VAA promou el desenvolupament de CHC en models murins amb fetge gras degut a un augment en la proliferació d’hepatòcits i a la presència d'un infiltrat protumorogènic. Per tant, la teràpia gènica amb VAA podria resultar oncogènica en pacients amb inflamació hepàtica crònica. - La combinació de lenvatinib i anti-PD1 té un efecte immunomodulador al CHC caracteritzat per la generació d’un perfil immunològic activat. Aquest tractament podria maximitzar el benefici clínic en un subgrup de pacients de CHC. - Els estudis translacionals basats en models preclínics i anàlisis òmiques tenen la capacitat de revelar les característiques moleculars del CHC, potencials factors de risc i nous enfocaments terapèutics. INTRODUCTION INTRODUCTION 25 1. Principles of Cancer 1.1. Worldwide Impact of Cancer The term cancer refers to a group of diseases involving uncontrollable cell growth with the potential to spread to other parts of the body. Cancer is one of the leading causes of death worldwide and supposes an important barrier to increasing life expectancy7. In 2020, there was an estimated 19.3 million new cases of cancer and 10 million cancer-related deaths worldwide. The burden of cancer incidence and mortality is rapidly increasing, partly due to aging of the population and changes in the prevalence and distribution of cancer risk factors associated with socioeconomic development (e.g., smoking, unhealthy diet, excess body weight, and physical inactivity)2. Worldwide, an estimated 28.4 million new cancer cases are projected to occur in 2040, corresponding to a 47% increase from the current number of cases2. The most commonly diagnosed cancer types are breast (11.7% of total cases), lung (11.4%), colorectal (10.0%), prostate (7.3%), stomach (5.6%) and liver (4.7%) cancers. In terms of mortality, the most common causes of cancer-related death are lung (18.0%), colorectal (9.4%), liver (8,3%), stomach (7.7%), breast (6.9%) and esophagus (5.5%) cancer2. Notably, the incidence and mortality of cancer are not homogeneous around the world and present regional and gender variations reflecting societal, economic, and lifestyle disparities2. Many efforts have been invested to identify risk factors for cancer development and effective therapeutic strategies to reduce its incidence and mortality. In this context, understanding the biological principles of the disease and its clinical implications is crucial. 1.2. Biological Basis of Cancer 1.2.1. Hallmarks of Cancer Tumorigenesis is a multistep process that drives the progressive transformation of normal human cells into malignant cells8. Observations of human tumors and animal models suggest that all cancers are a manifestation of eight essential hallmark capabilities consisting in alterations in cell physiology that collectively permit malignant growth8,9 (Figure 1): - Sustaining Proliferative Signaling: Cancer cells have the capacity to increase the production and release of growth-promoting signals that instruct progression through INTRODUCTION 26 the cell division cycle, thereby deregulating the homeostasis of cell number and achieving an aberrant tissue architecture and function. These signals are conveyed in large part by growth factors that bind to cell-surface receptors, typically containing intracellular tyrosine kinase domains. - Evading Growth Suppressors: Cancer cells can also circumvent inhibitory signaling that negatively regulates cell proliferation; many of which depend on the actions of tumor suppressor genes. - Resisting cell death: Programmed cell death by apoptosis occurs physiologically to maintain cell population and acts as a natural barrier to cancer development. When cells become old or damaged, they are genetically programmed to die in order to prevent the propagation of DNA errors. Through the introduction of abnormalities in sensors that play a key role in triggering apoptosis, tumor cells can avoid cell death. - Enabling replicative Immortality: Most normal cell lineages can undergo only a limited number of successive cell growth-and-division cycles. Cancer cells, on the other hand, require unlimited replicative potential to generate macroscopic tumors. Multiple lines of evidence indicate that telomeres protecting the ends of chromosomes are centrally involved in this capability. - Inducing angiogenesis: The tumor-associated neovasculature, generated by the process of angiogenesis, provides essential nutrients and oxygen, and evacuates metabolic wastes and carbon dioxide. By synthesizing angiogenic factors, tumors promote the production of aberrant vessels enabling cell proliferation and metastasis. - Activating invasion and metastasis: The capability for invasion enables cancer cells to escape the primary tumor mass and colonize other tissues. In this process, the primary tumor spawn cancer cells that move out, invade adjacent tissues, and travel to distant sites where they may form new colonies. These distant settlements of tumor cells — metastases — are the cause of 90% of human cancer deaths. - Deregulation of cellular energetics: The uncontrolled cell proliferation in cancer requires adjustments of energy metabolism to fuel cell growth and division. Cancer cells can reprogram their glucose metabolism, and thus their energy production, by limiting their energy metabolism largely to glycolysis. INTRODUCTION 27 - Avoiding immune destruction: Tumors have developed several strategies to avoid detection by the immune system, which constantly monitors cells and tissues through a process known as immune surveillance. These processes include escaping immune recognition, suppressing immune reactivity, or preventing T-cell infiltration. 1.2.2. Enabling Characteristics Enabling characteristics are necessary for the acquisition of the eight hallmark capabilities (Figure 1): - Genome instability and mutation: The acquisition of oncogenic traits largely depends on a succession of alterations in the genomes of cancer cells. Certain mutant genotypes confer a selective advantage on subsets of cancer cells, enabling their outgrowth and eventual dominance in a local tissue environment. - Tumor-promoting inflammation: Inflammatory cells infiltrated in the tumor can contribute to multiple hallmark capabilities by supplying bioactive molecules to the microenvironment, including growth factors that sustain proliferative signaling, survival factors that limit cell death, proangiogenic factors, extracellular matrix-modifying enzymes that facilitate invasion, etc. Figure 1. Hallmarks of cancer and enabling characteristics. Representation of the eight hallmark capabilities and two enabling characteristics originally proposed by INTRODUCTION 28 Hanahan and Weinberg in 2000 and further expanded in 2011. Modified from Hanahan D & Weinberg D, Cell 20119 using biorender.com. The following section expands on the implications of these enabling characteristics – genomic mutations and inflammation – in shaping cancer and how the mutation and inflammatory profile of a tumor can provide valuable information about the molecular basis of the disease and optimal therapeutic approaches. 1.3. Genomic Basis of Cancer As stated by the eminent oncologist Bert Vogelstein, “cancer is, in essence, a genetic disease”10, since it originates due to changes in the DNA that are acquired and maintained with each somatic cell replication. These acquired changes, known as somatic alterations, are the result of cell-intrinsic processes (e.g., DNA damage or spontaneous decay of nucleotides throughout time) or extrinsic exposures (e.g., ultraviolet radiation or cigarette smoke). Although less frequent, germline alterations associated with hereditary cancer-predisposition are also a mechanism of cancer development11. 1.3.1. Structural alterations Structural alterations correspond to all the molecular aberrations that involve modification of the underlying DNA sequence. - Mutations: These alterations involve changes in the DNA sequence. Depending on the position and specific change, mutations may alter the protein sequence, known as non- synonymous mutations. Most adult solid tumors (e.g., colon, breast, brain, and liver cancer) present an average of 33 to 66 non-synonymous mutations (Figure 2)12. Conversely, certain tumor types display more mutations than average, reflecting the involvement of a potent mutagen in their pathogenesis. This is the case of tobacco smoking in lung cancer and ultraviolet light in melanoma. About 95% of the non-synonymous mutations in a tumor correspond to single-base substitutions (e.g., C>G), also known as single nucleotide variants (SNV). Out of these non-synonymous SNVs, 90.7% result in missense changes, 7.6% result in nonsense changes, and 1.7% result in alterations of splice sites or untranslated regions immediately adjacent to the start and stop codons12. The remaining 5% of non- INTRODUCTION 29 synonymous mutations comprise deletions and insertions of one or a few bases (e.g., CTT>CT)12. Only a small fraction of non-synonymous mutations (3-8 mutations) participate in the transformation of a normally functioning cell into a cancer cell13. These mutations are known as “drivers”. The remaining alterations – termed “passengers” – occur coincidentally as a byproduct of the operative mutational processes during tumorigenesis, without supposing a functional advantage to the cell. To date, only ~500 of the 20,000 genes in the human genome have been shown to act as driver genes13,14. Figure 2. Number of somatic mutations in representative human cancers. The median number of non-synonymous mutations per tumor in a variety of tumor types. Horizontal bars indicate the 25 and 75% quartiles. MSI, microsatellite instability; SCLC, small cell lung cancers; NSCLC, non–small cell lung cancers; ESCC, esophageal squamous cell carcinomas; MSS, microsatellite stable; EAC, esophageal adenocarcinomas. Data obtained from Vogelstein B et al., Science 201312. - Copy-Number Variations (CNV): These changes refer to an increase (copy-number gain) or decrease (copy-number loss) in the number of gene copies present in the normal diploid genome15. According to its size, a CNV can be classified as broad CNV, spanning the length of a chromosome arm or more; or focal CNV, which affects shorter regions. Adult solid tumors present a median of 11 focal amplifications and 12 focal deletions16. Copy number gains may range from one copy to several extra copies (high-level amplifications)15. In terms of copy-number losses, the deletion might generate the loss of one of the gene alleles (loss of heterozygosity) or both alleles (homozygous deletions). Non−synonymous mutations per tumor 0 50 100 150 200 250 Co lor ec ta l ( M SI ) Lu ng (S CL C) Lu ng (N SC LC ) M ela no m a Es op ha ge al (E SC C) No n− Ho dg kin ly m ph om a Co lor ec ta l ( M SS ) He ad a nd n ec k Es op ha ge al (E AC ) Ga str ic En do m et ria l ( En do m et rio id) Pa nc re at ic ad en oc ar cin om a Ov ar ian (H igh −g ra de se ro us ) Pr os ta te G lio bla sto m a Br ea st En do m et ria l ( se ro us ) Lu ng (n ev er sm ok ed N SC LC ) Ch ro nic ly m ph oc yti c l eu ke m ia Ac ut e m ye loi d leu ke m ia Gl iob las to m a Ne ur ob las to m a Ac ut e lym ph ob las tic le uk em ia M ed ull ob las to m a Rh ab do id m ed ian m ut at ion s + /− on e qu ar tile 500 1000 1500 He pa to ce llu la r Mutagens Adult solid tumors Liquid PediatricMSI INTRODUCTION 30 - Insertional mutagenesis: This is the phenomenon by which an exogenous DNA sequence integrates within the human genome, resulting in the deregulation of neighboring genes that can potentially lead to oncogenic development17. Such insertional mutations can be mediated by viruses. Although not all viral integrations are oncogenic, they can lead to cell transformation if the insertion occurs in an essential gene or a gene that is involved in cellular replication or programmed cell death. In this regard, viruses such as human papilloma virus, Epstein Barr virus, hepatitis B virus (HBV), human T lymphotropic virus 1, and human herpes virus 8, are known to contribute to the genesis of one or more types of cancer15. - Chromosomal rearrangements: Rearrangements originate due to a DNA break that is then rejoined to another DNA segment that was not originally contiguous15, which results in the formation of gene fusions. Fusions can exert tumorigenic potential due to overexpression of one of the involved genes or the generation of a hybrid gene. The first gene fusion described was the so-called “Philadelphia chromosome'' in chronic lymphocytic leukemia, where a rearrangement between BCR (chromosome 22) and ABL (chromosome 9) genes renders constitutive activation of ABL leading to a deregulated cell cycle function15. 1.3.3. Epigenetic alterations Epigenetic alterations are heritable changes in gene expression that are not due to modifications in the DNA sequence. The most widely studied epigenetic modifications in cancer are the following: - DNA methylation: The addition of methyl groups in the DNA molecule is a mechanism of gene expression regulation that occurs normally but can be deregulated in cancer. This modification appears almost exclusively in the context of cytosine-guanine (CpG) dinucleotides, which tend to cluster in regions called CpG islands. In most cases. methylation in CpG islands is associated with gene silencing. Cancer cells are characterized by a massive global loss of DNA methylation (20–60% less overall 5- methyl-cytosine). However, hypermethylation at the CpG islands of certain promoters is also frequent18. - Nucleosome positioning: A nucleosome is a segment of DNA that is wrapped around a core of proteins, which blocks the access of activators and transcription factors. Occlusion of the transcription start site due to aberrant nucleosome positioning leads INTRODUCTION 31 to transcription repression. Nucleosome remodeling complexes SWI/SNF are commonly altered in cancer, although in most cases, the molecular mechanisms underlying their function remain unclear18. - Histone modifications: Histone tails are subject to post-transcriptional modification occur such as acetylation, methylation, phosphorylation, and ubiquitination among others. As a consequence, nucleosome architecture is altered, thus exposing or hiding specific regions of the DNA for transcriptional regulation, which ultimately affect gene expression. A common hallmark of human cancer is the reduction in the acetylation and methylation of histone H4, which is associated with the hypomethylation of DNA repetitive sequences18. 1.3.2. Mutational Signatures Despite most somatic mutations in cancer do not have a functional role, they can be extremely informative for the mutational processes that have been operative over the lifetime of a patient, including exogenous or endogenous mutagens, errors of the DNA replication machinery, and defective DNA repair. Each process generates unique nucleotide substitution patterns on the cancer genome, termed “mutational signatures” (Figure 3). Consequently, the presence of mutational signatures provides relevant information that helps unveil the genetic factors and environmental exposures in cancer19. Figure 3. Active mutational processes over the course of cancer development. Each mutational process leaves a characteristic imprint — a mutational signature INTRODUCTION 32 — in the cancer genome. In this hypothetical cancer genome, arrows indicate the duration and intensity of exposure to a mutational process. The final mutational portrait is the sum of all the different mutational processes (A–D) that have been active in the entire lifetime. Extracted from Helleday T et al., Nat Rev Genet 201419. The link between environmental exposures and cancer has been known since the 18th century, when English physicians observed an increased incidence of nasal polyps among users of snuff and associated scrotal cancer in chimney sweeps with chronic exposure to soot20,21. Subsequent associations between environmental agents and tumorigenesis include tobacco smoking and lung cancer, aniline dyes and bladder cancer, asbestos and mesothelioma, aflatoxin and liver cancer, and benzene and leukemia among many others. Many of these agents cause DNA damage that results in carcinogen-specific mutational signatures (Table 1). For instance, C·G → A·T transversions are related to smoking in lung cancer samples, and C·G → T·A transitions are significantly over-represented in skin cancers related to UV light exposure19 (Figure 4). Furthermore, the trinucleotide context of a mutation (i.e., the bases located 5ʹ and 3ʹ of the mutated nucleotide) is known to affect mutation rates in the genome and must also be taken into consideration when defining a mutational signature. Combining the six possible substitutions (C·G → A·T, C·G → G·C, C·G → T·A, T·A → A·T, T·A → C·G, and T·A → G·C) together with their trinucleotide contexts, all SNVs can be classified into 96 combinations. This classification has been used to extract multiple different mutational signatures in cancer21–23. Collective efforts analyzing the mutational fingerprints in thousands of tumor samples have allowed the identification of more than 40 mutational signatures, comprised in the COSMIC catalog23 (Figure 4). Also, experimental analyses in cancer models have unveiled more than 50 additional signatures associated with exposure to known environmental carcinogens21. The presence of these mutational signatures in cancer serves as an archaeological record of the multiple mutational processes that have been operative, including the exposure to environmental agents24,25. In addition, some mutational signatures are associated with a distinct clinical outcome and are emerging as potential biomarkers for novel targeted therapies24. Figure 4. Example of mutational signature in cancer. COSMIC Signature 4, associated with tobacco smoking, is displayed according to the 96 substitution INTRODUCTION 33 classification defined by the substitution type and trinucleotide context. The probability bars for the six types of substitutions are displayed in different colors. The mutation types are on the horizontal axes, and vertical axes depict the percentage of mutations attributed to a specific mutation type. Modified from Alexandrov LB et al., Nature 201322. Table 1. Mutational signatures included in the COSMIC catalog associated with known or suspected carcinogens in humans. Signature Carcinogen IARC category Main tumor type SBS4 Tobacco smoking 1 Lung, liver, head SBS7a-d Ultraviolet light 1 Skin, head SBS11 Temozolomide (alkylating agent) 1 Central nervous system, pancreas SBS16 Alcohol 1 Liver, head SBS18 Reactive oxygen species 3 Stomach, myeloid, colorectal SBS22 Aristolochic acid 1 Liver, kidney, biliary SBS24 Aflatoxin 1 Liver, biliary SBS29 Tobacco chewing 1 Liver, kidney, bone SBS31 Platinum treatment 2A Liver, myeloid SBS32 Azathioprine 1 Biliary, myeloid SBS, single base substitution; IARC, International Agency for Research on Cancer. Data obtained from Alexandrov LB et al., Nature 202023 and the IARC Monograph on the Identification of Carcinogenic Hazards to Humans26. 1.4. Role of the Immune System in Cancer The immune system is a complex network composed of immune cells and lymphoid organs that defend the body against microorganisms – such as bacteria, viruses, and parasites – or cancer cells. In the cancer setting, mutations that accumulate during tumorigenesis give rise to tumor- specific antigens. The function of the immune system is to detect and eliminate cells producing tumor antigens27. Immunity can be classified into innate and adaptive, which cooperate to detect and eliminate non-self bodies28. Immune cells from both the adaptive and innate immune systems are present in the tumor microenvironment and contribute to the modulation of tumor progression. The innate response is the first line of defense and involves cells ready to face foreign bodies. It exerts a tumor suppressive function, either by directly killing tumor cells or by triggering adaptive immune responses. Innate immune cells include natural killer (NK) cells, eosinophils, basophils, and phagocytic cells such as mast cells, neutrophils, monocytes, macrophages, and INTRODUCTION 34 dendritic cells28,29. Other components of the innate immune system are plasmatic proteins (complement system) and epithelial barriers, with fewer implications in cancer. The adaptive response, on the other hand, is highly specific and capable of discriminating between healthy and cancerous tissues. It is initiated when the innate immunity is not able to clear the non-self bodies (i.e., infection or cancer). This leads to the triggering of warning signals that activate the immune adaptive machinery through antigen presentation by the dendritic cells. The main players in adaptive immunity are lymphocytes, including B cells and T cells. B cells participate in humoral immune responses through the generation of antibodies, whereas T cells are mostly involved in cell-mediated immunity28. Among those, CD8+ cytotoxic T lymphocytes and T helper lymphocytes are the most specialized antitumor cells30. 2. Translational Research in Cancer 2.1. Translational Research Translational medicine is an interdisciplinary branch within the biomedical field that aims to translate discoveries and technologies from basic research to the clinical setting in order to improve the prevention, diagnosis, and treatment of human diseases31,32 (Figure 5). To pursue this goal, the translational medicine field incorporates the use of cutting-edge technologies, implements interdisciplinary approaches, and encourages collaboration between institutions. Translational medicine is based on the feedback loop “from bench to bedside and back again” which allows complementing the search for a cure with the understanding of the biological basis of a disease (Figure 5). Specifically, the key elements of this discipline are33: - Basic science studies defining biological aspects of a disease and effects of therapies. - Human research to define the molecular basis of a disease and provide the rationale for the development of therapeutics for human diseases. - Pre-clinical experimental studies that develop principles for application of therapies to human disease or aimed at advancing the research of bringing therapies to the clinic. - Appropriately designed clinical trials originated from the above-mentioned studies with efficacy or toxicity as an endpoint. INTRODUCTION 35 Figure 5. The translational medicine pipeline. Basic research discoveries are used in the design of clinical trials, which in turn will impact patient management policies. Adapted from Torrens L et al., Overview of Translational Medicine. Handbook of Translational Medicine, 201633. Translational studies in the oncology field have led to the identification of driver genes and targetable molecular alterations which eventually have brought to the development of targeted therapies with proven anti-tumoral efficacy and subsequently improved survival of cancer patients. These discoveries also have implications for the identification of patients who are at high-risk for certain solid tumors or who may benefit from specific treatment approaches34. In this regard, the identification of targetable alterations is paving the way for the use of molecular therapies and, consequently, standard-of-care treatments are moving from a clinical- pathological basis to a targeted approach based on the molecular characteristics of the tumors. Targeted therapies are designed to disrupt specific molecular targets in tumor cells, thus increasing specificity and decreasing toxicity35. For instance, this is the case of FDA-approved gefitinib, a small molecule that specifically blocks the EGF receptor thus providing better progression-free survival than chemotherapy to lung cancer patients with EGFR mutations36. In addition to this, translational research studies have elucidated the mechanisms that regulate the anti-tumor immune response, enabling the development of immunotherapies in oncology. Such therapies, designed to boost the action of the immune system against cancer cells, are showing unprecedented clinical benefits in terms of tumor regression and long-term responses in advanced cases37. The implementation of precision medicine requires the use of biomarkers as tools for the selection of patients who might benefit from therapies (predictive biomarkers) or as indicators of the long-term outcome of the patient (prognostic biomarkers). Biomarkers are molecular or cellular markers that can be measured in tumor tissue or body fluids for a wide variety of materials (e.g., DNA, methylation, RNA, protein, and circulating tumor cells)38. Their use as INTRODUCTION 36 surrogate markers of a pathogenic process or drug efficacy is changing the paradigm of patient management, allowing the implementation of targeted therapeutic approaches. Clear examples of this are the detection of ALK fusion protein in non-small cell lung cancer (targetable by crizotinib) or BRAF mutations in melanoma (targetable by vemurafenib), which enable treating only patients that will benefit from a given therapy33. Overall, by impacting on the diagnosis, prognosis, and patient treatment, translational medicine has enabled the development of more individualized treatments for cancer patients33. Because of this, precision medicine is already a reality in some tumor types such as melanoma, small-cell lung cancer, or colorectal cancer39. 2.2. Commonly Used Tools for Cancer Research 2.2.1. Omic techniques Translational research has revolutionized clinical management in oncology by providing an in- depth understanding of the molecular basis of cancer. This has been largely achieved thanks to the use of “omic techniques”. Omics refers to the collection and analysis of large data sets of biological variables and is regularly applied to genes (genomics), RNA molecules (transcriptomics), proteins (proteomics), and their subunits, such as nucleotides and amino acids (metabolomics)40. These new technologies present three clear advantages: a) they work at a high-throughput level, thus presenting an increased speed for tumor profiling; a) they allow researchers to obtain information of multiple data points of a biological sample in a single analysis; and c) they reduce the cost per analyzed unit40,41. A clear example of the implementation of omics is the establishment of two major international collaborative projects – The Cancer Genome Atlas (TCGA) and the International Cancer Genome Consortium (ICGC)42 – that have achieved the molecular profiling of thousands of human samples from the 38 most common cancer types14. Genomic analysis The study of the cancer genome comprises the characterization of mutations, copy number variations, and chromosomal rearrangements. Before the arrival of Next Generation Sequencing (NGS) technologies, the techniques used to study all these structural genomic aberrations were limited to the characterization of a single genome region per analysis, mainly INTRODUCTION 37 by Sanger sequencing for mutations and FISH karyotyping for CNVs. In contrast, NGS technologies allow for massively parallel sequencing reactions at the same time41. This includes a step of DNA fragmentation, library generation for DNA amplification, and the use of potent sequencer machines able to generate an output consisting of a collection of DNA sequences for each spatially separated cluster of DNA fragments43. Compared to previous methods, NGS allows obtaining multiple readings for the same nucleotide. The number of times that a single position of the genome sequence is read is referred to as depth of sequencing41. According to the region of interest that the sequencing technology covers, we can distinguish different assays, such as whole-genome sequencing (WGS), whole-exome sequencing (WES), and targeted sequencing19. These sequencing approaches yield many thousands of genomic alterations that can be used to study the mutation profile of each patient and to extract mutational signatures. Alternatively to NGS, single-nucleotide polymorphism arrays (SNP arrays) are also a common high-throughput tool for genome-wide CNV characterization44. Contrary to NGS, SNP arrays only provide nucleotide information at specific positions instead of all nucleotides of an aligned region. Hence, arrays offer lower resolution and imprecise delineation of the breakpoints of CNVs, than NGS44. The challenging step of genomic analysis is now to translate this knowledge into clinically- meaningful information, such as the identification of causal biological changes that drive cancer phenotypes or to guide the selection of therapy for patients45. NGS has become the foundational technology for modern diagnostic testing in oncology, with several laboratory-developed tests recently achieving recognition by the US Food and Drug Administration (FDA)45. Transcriptomic analysis Tumor phenotypes are determined by aberrant transcriptional patterns, which are the consequence of somatic mutations and epigenetic alterations46. In contrast to DNA, which is mostly identical across all cells of an organism, the transcribed RNA is highly dynamic, reflecting the diversity of cell types, cellular states, and regulatory mechanisms. Therefore, the transcriptome profile of a tumor sample can be regarded as a snapshot of its underlying biological status46. Transcriptomic studies also draw upon NGS strategies. Using the same technology as genomic approaches, RNA sequencing (RNA-seq) is based on the sequencing of cDNA derived from messenger RNA, total RNA, and microRNAs among others47. RNA-seq can be used to assess the expression level of genes, study their association with a given phenotype and compare the INTRODUCTION 38 expression levels between phenotypes. It also provides information about splicing variants, allelic expression, and RNA editing47. Alternatively, the transcriptomic profile of tumors can be assessed with microarray-based techniques. Microarrays consist of a collection of microbeads containing DNA probes corresponding to known sequences. RNA is isolated from the control and the target samples, undergo reverse transcription, and labeling and then cDNA is hybridized to the array. The abundance of hybridization is quantified by fluorescently labeled probes, which is a readout of the RNA expression levels. Despite microarrays do not allow the detection of gene fusions or non-previously known transcripts, they are a widely used tool due to their lower cost. Gene expression profiling can be used to identify cancer biomarkers and aberrant expression of cancer drivers, as well as to unveil molecular targets for anticancer therapies48. In addition, during the past 20 years, gene expression profiling has focused on the identification of clinically useful molecular signatures. Gene expression signatures are alterations in the expression of sets of genes with an association with disease prognosis, therapeutic benefit, or molecular profile48. Signatures showing activation of specific signaling pathways or cell functions are tools for molecular classification of tumors, such as the well-established five classes of breast cancer, with biological and clinical relevance40. The combination of tumor transcriptomic profiling and novel analysis algorithms allow the assessment of the relative cell composition of a tumor. In this regard, transcriptome profiling can be used to infer the immune-cell composition and to classify tumors according to their immune profile, which could have implications for the prediction of immunotherapy responses49,50. Furthermore, RNA-based prognostic panels are now available and are clinically used for major cancer types, including breast (MammaPrint, Oncotype DX, and Prosigna), lung (GeneFx), prostate (Prolaris), and colon (ColoPrint)46. 2.2.3. Preclinical Models of Cancer Preclinical models aim at understanding the pathogenesis and mechanisms involved in disease initiation and progression and are essential to building the groundwork for the development of clinical therapies51. A significant amount of our current understanding of the pathophysiology, prevention, and treatment of cancer is based on preclinical studies that exhibit some of the clinical or molecular features observed in humans. Unlike the processing of human data, these models allow the generation of mechanistic studies. Among other merits, preclinical models of cancer have achieved the identification of oncogenic mutations and pathways involved in tumorigenesis and have provided important clues regarding targetable alterations for cancer treatment. INTRODUCTION 39 In Vitro Experimental Models In vitro models allow researchers to recapitulate aspects of tumor biology using specific cell types, extracellular matrices, and/or soluble factors outside of a living organism. Cancer cell lines are, by far, the most used model in cancer research and a particularly useful tool for screening of anticancer compounds51. The main advantage of cell lines resides in the fact that they allow a thorough control of experimental conditions and provide good reproducibility of results. However, they fail to recapitulate the characteristics of a whole tumor, including genomic heterogeneity of tumor cells and the presence of a tumor microenvironment. A popular alternative to cell lines are organoids, which refer to three-dimensional tissue culture that recapitulate the tumor architecture and maintain multilineage differentiation. Currently, great efforts are being invested in developing tumor organoids that maintain a functional immune microenvironment. In Vivo Experimental Models Laboratory models typically involve animals in which processes that are present in human diseases are either induced or occur naturally. The most widely used animal in cancer research is the mouse (Mus musculus). Remarkably, 90% of the mouse genome can be linked up with a region of the human genome and 99% of the mouse genes have analogs in humans. Other advantages of the murine model include rapid tumor development, low housing cost, and good- sized litters51. Murine models of cancer encompass a wide spectrum of options. The selection of the optimal model depends on the biological question that needs to be tackled and often represents a compromise between time, complexity, and clinical relevance51. - Xenograft models: The xenograft model is based on the implantation of tumor cell lines onto mice to grow tumors of human origin. this implantation can be produced either orthotopically (i.e., in the same organ as the tumor would develop) or subcutaneously. This model is relatively inexpensive, highly reproducible, and applicable to many cancer types51. Furthermore, subcutaneous models are useful in studying the response to therapy, as tumors are easily accessible, and growth can be monitored with serial tumor measurement. The main limitation of this model is that it requires immunocompromised mice to avoid rejection of the foreign tissue52. INTRODUCTION 40 - Syngeneic models: This model utilizes mouse tumor cells implanted orthotopically or subcutaneously in recipient mice with a fully functional immune system52. Therefore, it can mimic a comprehensive antitumor immune response and recapitulate its implications in tumor development and response to treatment. This is especially relevant when assessing the anti-tumoral effect of drugs targeting the anti-tumor immune response (i.e., immunotherapies). However, it might be less representative of the human disease due to the use of mouse tumors. - Patient-derived xenografts (PDX): The PDX model consists in the implantation of primary human tumor samples into immunocompromised mice. Therefore, they retain the architecture and microenvironment of the primary human tumor and provide an accurate representation of the original molecular landscape and heterogeneity. Similar to the conventional xenograft, its main limitation is the lack of a fully-functional immune response51. - Genetically engineered mouse models (GEMMs): Advances in genetic engineering have led to the genetic program of organ-specific activation of oncogenes or inactivation of tumor suppressor genes to induce tumor formation, recapitulating the genetic lesions found in human cancers. GEMMs develop spontaneous and autochthonous tumors including tumor microenvironment and immune infiltrate51. GEMMs enable the study of tumorigenesis and biological tumor development, but they often involve long latency and a lower tumor incidence52. Hydrodynamic tail-vein injection of cells or DNA plasmids carrying the mutations of interest has opened a new avenue for delivering genetic materials to a mouse model52. - Environmentally-induced cancer models: Tumor development in mouse models can be achieved following various environmental exposures, including carcinogen agents, radiation, pathogenic viruses, or specific diets53. The type and route of administration of the carcinogen determine the location in which the tumor is formed (e.g., topical, intramuscular, or oral administration). Some models mimic the setting of chronic inflammation and organ damage which accompany or lead to tumorigenesis. This is the case of high-fat diet (HFD) models, which induce liver disease and steatosis, that can be easily combined with different methods to induce liver cancer (e.g., GEMM, cell implantation, or additional carcinogens). Overall, the main caveat of environmentally- induced models is that only a restricted subset of tumor types can be developed, and tumor penetrance can be often limited. INTRODUCTION 41 2.3. Current Challenges in Translational Oncology Translational medicine is playing a pivotal role in broadening our knowledge about the molecular pathogenesis of cancer. The emergence of omic techniques and collective initiatives to sequence tumor samples has allowed the identification of key molecular alterations and has provided insight into the biological mechanisms involved in oncogenesis. At the same time, functional studies in preclinical models have enabled a more detailed characterization of tumor tissues. Overall, this has led to the identification of drug targets, enabling the development of novel therapies and personalized treatment approaches. In this regard, the use of biomarkers facilitates the selection of patients who would benefit from each treatment, but only a few biomarkers for solid tumors are routinely tested in the clinical setting at the moment34. In the future, the development of more affordable high-throughput sequencing approaches will allow us to increase the assessment of molecular alterations to all patients and will lead to more tailored therapies. New and less invasive methods for obtaining tumor material such as liquid biopsies will enable more frequent monitoring of tumor response to therapy, leading to an improved therapeutic benefit to toxicity ratio34. Finally, the incorporation of novel biomarkers in clinical trials may also improve the efficacy of experimental drug testing. 3. Hepatocellular Carcinoma 3.1. Epidemiology and Risk Factors Primary liver cancer is the sixth most commonly diagnosed cancer and the third leading cause of cancer-related death worldwide, with approximately 906,000 new cases and 830,000 deaths in 2020. It is estimated that, by 2025, more than 1 million individuals will be affected by liver cancer annually2,3. In most geographical regions, rates of both incidence and mortality are 2 to 3 times higher among men than women (Figure 6A). The highest incidence and mortality rates are observed mainly in transitioning countries from East Asia and Africa, with Mongolia exhibiting the highest incidence of HCC worldwide2 (Figure 6B). Nonetheless, liver cancer incidence and mortality are also increasing in different parts of Europe and the USA54. Hepatocellular carcinoma (HCC) is the most common form of primary liver cancer and accounts for ~90% of cases2,3. HCC typically arises in the setting of chronic liver damage generated by known risk factors2,3. The worldwide prevalence of such factors is highly heterogeneous, which INTRODUCTION 42 determines the global distribution of HCC cases (Figure 6B)2,3,55. Well-established risk factors of HCC are: - Hepatitis B virus infection: HBV is a DNA virus that can integrate into the host genome, leading to oncogene activation. HBV infection is the main etiologic agent, in Asia, Africa, Melanesia, and Polynesia2,3,55. - Hepatitis C virus infection: Unlike HBV, hepatitis C virus (HCV) is an RNA virus that does not integrate into the host genome. Therefore, the risk of HCC in HCV-infected patients is due to the development of cirrhosis or chronic liver damage3. HCV infection is the predominant etiology in the North of Africa, Europe, North America, Japan, and Central Asia2,3,56. - Alcohol abuse: Excessive alcohol intake causes alcoholic liver disease, cirrhosis, and HCC. In addition, there is growing evidence supporting alcohol-specific protumorigenic mechanisms besides cirrhosis3. Alcohol consumption is the second most common risk factor in most European countries and also in North America2,3. - Non-alcoholic fatty liver disease: Non-alcoholic fatty liver disease (NAFLD) is a spectrum of chronic liver diseases normally occurring in patients with diabetes mellitus and obesity. NAFLD ranges from excessive hepatocyte triglyceride accumulation and steatosis (nonalcoholic fatty liver) to hepatic triglyceride accumulation plus inflammation and hepatocyte injury (nonalcoholic steatohepatitis [NASH]), which finally leads to hepatic cirrhosis and HCC57. NAFLD is the fastest growing etiology of HCC, particularly in Western countries 2,3,55. Other emerging factors have been reported to contribute to disease risk, such as: - Hepatitis D virus infection (HDV): HDV is an RNA virus that requires the presence of HBV surface antigens for its replication and infectivity. Despite being associated with a more severe course of liver disease and increased risk of HCC compared with HBV infection alone58, HDV is not currently included in the roster of carcinogenic agents since data supporting this relationship remains scarce26,59. - Adeno-associated virus type 2 (AAV2): Adeno-associated virus (AAV) is a defective DNA virus that causes frequent non-pathogenic infections in the general population60. However, recurrent clonal integrations of certain AAV strains – mostly AAV2 – have been INTRODUCTION 43 identified in a subgroup of HCC (~5%)3,4, and preclinical data points towards HCC development after AAV2 infection in murine models3,4, suggesting that this virus could be hepatocarcinogenic in certain circumstances. Finally, several carcinogenic cofactors can exacerbate the risk of HCC in patients exposed to these factors61. For instance, the fungal compound aflatoxin B1, a common food contaminant in Southeast Asia and Sub-Saharan Africa, has been reported to act synergistically with HBV to induce HCC. Other known cofactors in HCC are aristolochic acid, which is used in Asian traditional medicine, and tobacco smoke3,61. Figure 6. Age-standardized incidence rates in liver cancer. A-B. Age-standardized incidence by sex (A) and geographical area (B). The major etiological factors involved in hepatocarcinogenesis are depicted in figure B. NASH, nonalcoholic A B INTRODUCTION 44 steatohepatitis; ASR, age standardized rate. Modified from Llovet JM et al., Nat Rev Dis Prim 20213 and Sung CA et al., Cancer J Clin 2021 (Global Cancer Statistics 2020)2. 3.1.1. Particularities of HCC in Mongolia Mongolia, a landlocked East Asian country between Russia and China, shows the world’s highest incidence of HCC, with a burden of 85.6 cases per 100,000 inhabitants (106.0 and 68.4 cases per 100,000 inhabitants in males and females, respectively) 62. This incidence far exceeds that of the surrounding countries such as China, with 18.2 cases per 100,000 inhabitants (5-fold higher in Mongolia) and Russia, with 4.2 (>20-fold higher in Mongolia), or any other country worldwide62,63. The extremely high incidence observed in Mongolia has been largely attributed to the unique combination of HCC risk factors, with a historical high prevalence of both HBV (10.6%) and HCV (6.4%) viruses, and alcohol consumption56,64,65. Indeed, 90% of Mongolian HCC cases are positive for HBV, HCV, or both66, and HDV is present in 50-80% of Mongolian HBV- infected individuals67,68. In this regard, despite the implementation of universal infant HBV vaccination in 1991, which achieved a marked decline in HBV prevalence64, and the higher control in HCV blood-based transmission, the burden of HCC in Mongolia is increasing year after year, and the incidence is now ~10 times higher compared to the 1960s69,70. Other reasons such as low screening and treatment rates due to financial barriers have also been proposed, suggested by the fact that 75% of cases are diagnosed at advanced stages and have poor 1-year survival of less than 25%69,71. However, it is likely that earlier diagnosis would even increase the actual burden, as probably many cases remain undiagnosed, especially in remote areas. Besides the abovementioned risk factors, Mongolia has many particularities that might eventually play a role in HCC burden. It is a huge but sparsely populated country (1.9 people per km2), but half of its 3-million population lives in Ulaanbaatar, an overpopulated capital with dismal environmental conditions, whereas the rest is still predominantly nomad. Mongolian individuals are descendants of the Genghis Khan Empire, which spread from East Asia to Europe during the 13th century72. This explains why modern Mongolian populations have a certain degree of genotypic similarity with Siberian, Finn, Chinese Han, and Japanese populations, and are more similar to European than to the majority of Asian populations73. Interestingly, its neighbor countries also show a high prevalence of HCC risk factors (in Russia HBV is 2%, HCV is 2.7%, and alcohol consumption is a major issue; and in China, HBV prevalence is 6.1%), but, as stated, its HCC burden is much lower56,64. Around 60% of the 10 million Mongols live in the Northern region of China called Inner Mongolia, representing 17% of the population in that INTRODUCTION 45 region. Interestingly, this area shows a prevalence of HBV infection below 4% and a low incidence of HCC74,75. Regarding HCC onset, the male/female ratio in Mongolia is 1.5/1, as opposed to that observed in the surrounding countries (2.6/1 in Russia, 3.4/1 in China, and 3/1 in East Asia)76, which could potentially be linked to distinct patterns of exposure to HCC risk factors in Mongolia compared to other populations. Overall, it is unclear whether HCC incidence in Mongolia is completely explained by the unique combination of risk factors, or eventually, other genetic or environmental factors might play a role. 3.2. Molecular Pathogenesis of HCC HCC development is a complex multistep process involving the interplay between genetic factors and environmental exposures77,78. The combination of these factors triggers the transformation of the normal liver to an inflammatory and fibrogenic procarcinogenic field, which constitutes the background for HCC development77. indeed, 70–80% of HCC cases occur in the context of established liver cirrhosis, the last stage of this underlying chronic liver disease79. The natural history of HCC in cirrhosis follows a sequence of events starting with pre- cancerous cirrhotic nodules, called low-grade dysplastic nodules (LGDN) and high-grade dysplastic nodules (HGDN), which can finally transform into HCC80. However, HCC might also develop in the context of chronic liver disease without cirrhosis. For instance, HBV is able to insert into the genome in cancer genes, triggering the formation of HCC without the need for a cirrhotic background78. In normal liver, HCC can also arise from the malignant transformation of hepatocellular adenoma, a rare benign lesion81. 3.2.1. Molecular Landscape Over the last decade, translational genomic studies have provided an overview of the molecular landscape of HCC82–86. HCC tumors present 40-60 non-silent somatic mutations accumulated in coding regions78,83 (Figure 2). The most commonly mutated driver genes include TERT, CTNNB1, and TP53 (~55%, 29%, and 27% of patients, respectively) (Table 2)78,83. CNAs are also recurrent in HCC, with 21-30 focal alterations per tumor87. Characteristic focal CNAs in HCC include gains in 8q24.21 involving MYC (12%) and 11q13.3 affecting CCND1 and FGF19 (6-7%), as well as losses in 9p21.3 affecting CDKN2A (Table 2)83,87. Finally, few recurrent fusion proteins have been described in HCC, including the ABCB11- LRP2 fusion in only 2% of the patients88. INTRODUCTION 46 Each HCC tumor is a unique combination of genetic and epigenetic alterations, underlining the complexity and diversity in HCC. As a result, the major signaling pathways disrupted in HCC progression include the following: Table 2. Recurrent somatic driver alterations in resected HCCs. Altered pathway Altered gene Type of alteration Percentage (range) Mutations Telomere maintenance TERT$ Promoter Activating mutation 55 (44–59) Cell cycle regulation TP53$ Loss of function mutation 27 (18–31) ATM Loss of function mutation 4 (2–5) RB1 Loss of function mutation 4 (3–5) CDKN2A Loss of function mutation 2 (1–3) Wnt / β-catenin signaling CTNNB1$ Activating mutation 29 (23–36) AXIN1 Loss of function mutation 7 (4–10) APC Loss of function mutation 2 (0–3) Chromatin remodeling ARID1A Loss of function mutation 8 (4–12) ARID2 Loss of function mutation 7 (3–10) KMT2A Loss of function mutation 3 (0–4) KMT2C Loss of function mutation 3 (2–5) KMT2B Loss of function mutation 2 (0–4) BAP1 Loss of function mutation 2 (0–5) ARID1B Loss of function mutation 1 (0–3) Ras/PI3K/mTOR RPS6KA3 Unclassified 4 (3–6) PIK3CA# Activating mutation 2 (1–4) KRAS# Activating mutation 1 (0–1) NRAS Activating mutation 0 (0–1) PDGFRA# Mutation 1 (0–4) EGFR# Activating mutation 1 (0–2) PTEN Loss of function mutation 1 (0–2) Oxidative stress NFE2L2 & Activating mutation 4 (2–6) KEAP1& Activating mutation 3 (2–5) Hepatocyte differentiation ALB Mutation 9 (5–13) APOB Mutation 8 (1–10) JAK–STAT IL6ST Mutation 2 (0–3) JAK1$ Mutation 1 (0–3) TGFβ signaling$ ACVR2A Loss of function mutation 4 (1–10) IGF signaling$ IGF2R Mutation 1 (0–2) Copy number alterations Telomere maintenance TERT$ High-level focal amplification 6 (1–9) Cell cycle regulation MYC High-level focal amplification 12 (4–18) CCND1$ High-level focal amplification 7 (5–7) CDKN2A Homozygous deletion 5 (4–6) RB1 Homozygous deletion 5 (4–6) TP53$ Homozygous deletion 2 (0–2) RTK-RAS-PI3K signaling FGF19 # High-level focal amplification 6 (5–6) VEGFA# High-level focal amplification 5 (1–8) Viral insertions Telomere maintenance TERT$ HBV insertion 3 (1 – 5) Cell cycle regulation CCNA2 HBV insertion 5 (1-6) CCNE HBV insertion 3 (1-6) Chromatin remodeling KMT2B HBV insertion 1 HCC, hepatocellular carcinoma; IGF, insulin growth factor; mTOR, mammalian Target of Rapamycin; STAT, signal transducer and activator of transcription; TGFβ, transforming growth factor β. #: targetable by an FDA-approved drug. $: targetable by a drug in testing phases. &: targetable using mTOR inhibitors in testing phases. Adapted from Llovet JM et al., Nat Cancer 202189 and Bayard Q et al., Nat Commun 201890. INTRODUCTION 47 - Telomere maintenance: Approximately 90% of human HCCs harbor increased telomerase expression, the enzyme responsible for the maintenance of telomere length3. Telomerase prevents the erosion of the chromosomes that physiologically occur at each cell division during aging. In HCC, overexpression occurs mainly due to somatic TERT promoter mutations (55%)91,92, HBV insertion in the promoter (3%)93, and copy- number amplification (6%)84. While these alterations are mutually exclusive, TERT promoter mutations are frequently associated with CTNNB1 mutations, suggesting cooperation between telomerase maintenance and the β-catenin pathway in liver tumorigenesis83,91. Remarkably, 19% of HGDN exhibit these mutations, suggesting TERT as a “gatekeeper” during hepatocarcinogenesis91. - WNT/β-catenin signaling: This pathway is implicated in physiologic embryogenesis, zonation, and metabolic control in the liver. CTNNB1, a gene coding for β-catenin, presents activating mutations in 29% of HCC. Inactivating mutations or deletions have also been identified in AXIN1 (7%) and APC (2%)82–86. - Cell cycle control: TP53 – a key tumor suppressor participating in cell cycle regulation – present inactivating mutations in 27% of HCC patients82–86. The only recurrent hotspot identified so far in TP53 is R249S, related to aflatoxin exposure94,95. Additionally, the retinoblastoma pathway that controls progression from G1 to S phase is often inactivated in HCC mainly by homozygous deletions in CDKN2A (5%) or RB1 mutations (4%), both associated with poor prognosis82,83. Finally, recurrent HBV insertions in CCNE1 (5%)93 and amplification of the CCND1/FGF19 locus (6-7%)96,97, two key proteins involved in cell cycle progression, have been reported in HCC. - Epigenetic modifiers: Mutations in epigenetic modifiers from the SWI/SNF chromatin remodeling complex are recurrent in HCC, including inactivating mutations in ARID1A (8%) and ARID2 (7%)82,83,98. Less frequent mutations have been described in the histone methylation writer family, including KMT2A and KMT2C (3%)82,83. Furthermore, KMT2B can be affected by mutations (2%) or recurrent HBV insertions (10%)82,83,93. Altogether, the functional consequences of ARID1A, ARID2, and KMT2 family mutations in hepatocarcinogenesis remain to be further explored. - Oxidative stress pathway: The oxidative stress pathway is deregulated by activating mutations in NFE2L2 – coding for NRF2 – and inactivating mutations in KEAP1 in (3- 4%)82,83,98. Interestingly, NRF2 pathway activation was previously shown to protect INTRODUCTION 48 against liver tumor initiation, but its constitutive activation can drive tumor progression in late-stage HCC99. - Tyrosine kinase receptor-RAS-PI3K signaling: The RAS/RAF/mitogen-activated protein kinase pathway is activated by loss-of-function mutations in RP6SKA3 (4%), coding for the RAS inhibitor RSK298. In contrast to other tumors, activating mutations of genes belonging to the RAS family are rarely observed in HCC (<2%)77. On the other hand, PI3K/AKT/mTOR signaling is stimulated by activating mutations in PIK3CA (2%) and homozygous-deletions in PTEN (1%)82,83, as well as FGF19 focal amplification (6%)97,100. Notably, some HCCs with activation of the PI3K/AKT/MTOR cascade have no genetic alterations in this pathway. In this sense, indirect upstream signaling through the insulin growth factor (IGF) pathway has been proposed as an alternative mechanism of activation of this cascade77,101. Other mutated genes of HCC belong to the including JAK/STAT signaling (IL6ST and JAK1, 1-2%) and TGFβ (ACVR2A, 4%) signaling pathways. Genes related to liver function such as hepatic differentiation ALB and APOB also present recurrent mutations in 8-9% of cases79 (Table 2). 3.2.2. Genomic Alterations Associated with Viral Infections Two DNA viruses – HBV and AAV2 have been reported to induce insertional mutagenesis leading to HCC4,102. On the other hand, HCV is a single-strand DNA that is unable to insert into the host DNA and does not present a clear direct mechanism of carcinogenesis103. - Hepatitis B virus: Oncogenic HBV-mediated insertional mutagenesis can occur within the TERT promoter, leading to an overexpression of telomerase3,102. Other recurrent HBV insertions have been mapped in several oncogenes including the cyclins CCNA2 and CCNE1, which activate in cell cycle progression90 (Table 2). - Adeno-associated virus type 2: AAV2 infection is frequent in the human population, and no specific diseases have yet been associated with natural infection4. This, together with its high cell infectivity, has supported extensive development of AAV2-derived vectors for gene therapy for the last 30 years4. Indeed, AAV2 vectors have been largely considered safe, given that they generally persist episomally with infrequent integration104. The growing popularity of AAV2 viral vectors has resulted in the FDA and EMA approval of two gene therapies for the treatment of spinal muscular atrophy and retinal dystrophy, and ~130 active clinical trials are currently testing AAV-based gene INTRODUCTION 49 therapy for a great variety of human diseases104,105. However, mounting recent evidence supports that AAV2 vectors may have genotoxic potential and AAV2 insertions have been identified in a small set of HCC patients (2-5%)4,60. Preclinical data points towards the development of HCC after AAV transfer in murine models as a result of integration in the chromosome 12 locus that includes the noncoding RNA gene Rian (analogous to the human DLK1-DIO3 locus in 14q32.2)106,107. Furthermore, Molecular profiling of human HCC revealed a subclass of 6-19% HCC with overexpression of this microRNA cluster, associated with an aggressive phenotype and poor prognosis108,109. Similar to HBV, other recurrent AAV2 insertion points also include TERT, CCNA2, and CCNE1 genes4. Finally, considering that AAV2 integration capacity is enhanced in cells undergoing cell cycle progression110, it remains to be studied whether hepatocyte replication in response to liver injury and chronic inflammation could potentially favor oncogenic AAV integration111. Adverse events in these patients could potentially be a concern for the use of AAV gene therapy. 3.2.3. Mutational Signatures in HCC During the development of chronic liver disease and cirrhosis, hepatocytes progressively accumulate genetic mutations, which constitute mutational signatures that can be associated with specific risk factors. In this regard, genome and exome sequencing analyses of HCC have allowed the identification of mutational signatures from the COSMIC catalog that suggest the presence of intrinsic and extrinsic mutational processes23,83,112 (Figure 7). For instance, SBS5 and SBS1, which are among the most common signatures in this tumor type, recapitulate clock-like endogenous mutations that occur in cell division and accumulate with age. Other signatures can be linked to specific HCC risk factors. For instance, SBS16 has been associated with alcohol intake, and signatures SBS4 and SBS29, with tobacco smoking. Interestingly, some of these signatures show clear geographical differences due to distinct exposure to known carcinogens23. This is the case of SBS22 and SBS24, mostly detected in HCC patients from Asia and Africa exposed to aristolochic acid and aflatoxin B1, respectively23,83,113. Finally, some signatures with unknown etiology can be routinely found in the HCC genome, such as SBS12 and SBS40, and future studies will be required to elucidate the mutational mechanisms causing their mutational patterns23. Overall, these observations align with the role of the liver in detoxifying numerous metabolites, which can damage the hepatocyte genome3. In this context, assessing the mutational landscape INTRODUCTION 50 of HCC can provide relevant information that may help unveil the genetic factors and environmental exposures underlying hepatocarcinogenesis. Figure 7. Mutational signatures identified in HCC. Single base mutational signatures from the COSMIC catalog version 3.0 reported in hepatocellular carcinoma (HCC) samples (n = 323). Signatures are sorted by frequency and median mutations per megabase (Mb) explained by the signature in positive samples. The proposed etiology associated with each signature (if any) is indicated. Modified from Alexandrov LB et al., Nature 202023. 3.2.4. Molecular and Immune Classes Integrative efforts involving genomic, transcriptomic, and epigenetic data have established a molecular classification of HCC3,77,114. These molecular classes reflect specific genomic alterations, histopathological fingerprints, and clinical outcomes, with potential implications in patient prognostication and therapy selection. The molecular profile of HCC can be roughly divided into two major molecular types, each accounting for ~50% of patients with this disease – the proliferation class and the non-proliferation class3,80 (Figure 8): - Proliferation class: HCC tumors belonging to this class are associated with mutations in TP53, chromosomal instability, and enrichment in HBV-associated HCC 3,115–117. They also present enrichment of poor prognosis signatures and clinical characteristics of aggressive tumors (e.g., vascular invasion)3,118. The proliferation class can be further subdivided into two subclasses3: 1) The proliferation-progenitor cell group, characterized by the activation of classic cell proliferation pathways (e.g., PI3K–AKT– mTOR, RAS–MAPK or MET) and expression of progenitor cell markers (e.g., EPCAM and Clock-like Clock-like Tobacco chewing Tobacco smoking Alcohol Aristolochic acid exposure Aflatoxin B1 exposure Defective DNA mismatch repair Defective DNA mismatch repair Platinum treatment Platinum treatment Reactive oxygen species Mismatch repair deficiency Defective base excision repair Polymerase activity SBS5 SBS1 SBS12 SBS29 SBS4 SBS40 SBS16 SBS22 SBS24 SBS26 SBS6 SBS31 SBS35 SBS18 SBS19 SBS19 SBS30 SBS28 SBS9 SBS17b SBS17a Proportion of tumors with the signature Median mutations per Mb due to signature (among tumors with the signature) Signature Proposed etiology INTRODUCTION 51 α-fetoprotein); and 2) the proliferation–WNT–TGFβ group, characterized by non- canonical activation of Wnt. Figure 8. Molecular classification of HCC. Hepatocellular carcinoma (HCC) can be classified into two major molecular groups based on transcriptomic features, which present association with genomic, histopathological, and clinical characteristics. FLC, fibrolamellar carcinoma; IHC, immunohistochemistry; FLC, fibrolamellar carcinoma; TCGA, The Cancer Genome Atlas; IHC, immunohistochemistry; AFP; α- fetoprotein; HBV, hepatitis B virus; HCV, hepatitis C virus; NASH, non-alcoholic steatohepatitis; miRNA, microRNA. Obtained from Llovet JM et al., Nat Rev Dis Prim 20213. INTRODUCTION 52 - Non-proliferation class: Tumors from this class lack strong proliferative signaling and retain molecular features resembling normal hepatic physiology (e.g., metabolic functions and protein synthesis). This class is enriched in alcohol-associated and HCV- related HCC and is associated with better outcomes3. Although the non-proliferation class is heterogeneous, it can be divided into 2 main subclasses – one characterized by canonical Wnt signaling activation and CTNNB1 mutations, and another characterized by the activation of IFNα signaling3,100,119. HCCs can also be classified based upon their immune microenvironment profile using transcriptomic data120,121 (Figure 9). Around ~25% of the tumors belong to the immune class, characterized by high levels of immune infiltrate, high cytolytic activity, and expression of PD1/PDL1 immune checkpoints. This class can be further divided into immune-active tumors, presenting an enrichment in cytotoxic T cell infiltrate and signatures of response to immunotherapy; and immune-exhausted tumors, with TGFβ signaling activation driving immune exhaustion. Recently, the concept of “hot tumors” in HCC – i.e., tumors with high immune infiltration – has been further expanded with the inflamed class (~35% of cases), which encompasses both the immune class and the novel immune-like class121. Tumors from the immune-like class are dominated by high interferon-γ signaling coexisting with CTNNB1 mutations. Whether inflamed HCCs or other immune-related biomarkers are associated with response to ICI is currently being investigated. On the other side of the spectrum, the non-inflamed class (~65%) encompasses two further subclasses based on their mechanisms of immune escape121. First, the immune excluded class (~25%) is characterized by low immune cell infiltrate and enrichment in CTNNB1 mutations, which has been associated with resistance to immunotherapy122. Finally, the intermediate class (~45%) presents enrichment in TP53 mutations and frequent deletions in genomic regions harboring genes related to interferon signaling or antigen presentation. 3.3. Tumor Microenvironment The tumor microenvironment of HCC is a complex and spatially structured mixture of tumor cells, immune cells and tumor-associated fibroblasts, and hepatic non-parenchymal resident cells. All these populations dynamically interact and influence the inflammatory profile of the tumor123. INTRODUCTION 53 Approximately 90% of HCC cases are associated with chronic inflammatory processes due to viral hepatitis, alcohol intake, or NAFLD. Consequently, HCC is a prototypical inflammation- associated cancer, and the immune microenvironment plays a pivotal role in hepatocarcinogenesis3,124. Remarkably, immune-related gene expression patterns in the non- tumoral liver parenchyma have been associated with enhanced risk of HCC development in patients with cirrhosis125. In fully-developed HCC, the presence of an intra-tumoral immune infiltrate is associated with good prognosis, likely due to the activation of a more effective antitumor immunity3,126. Figure 9. HCC immune-based classification. Classification of hepatocellular carcinoma (HCC) based on immune-related parameters. TCR, T cell receptor; TIL, tumor-infiltrating lymphocyte; TLS, tertiary lymphoid structures; Treg, regulatory T. Adapted from Llovet JM et al., Nat Rev Clin Oncol 2021121. 3.3.1. Immune Cell Infiltrate Immune cells from the innate and adaptive immune systems interact in the tumor microenvironment to enable or suppress anticancer immune surveillance. The main players in the HCC immune infiltrate are the following121,127,128: INTRODUCTION 54 - CD8+ cytotoxic T cells: CD8+ lymphocytes exert effector anti-tumoral functions by eliciting cytotoxic activity through the release of granzyme B and perforin, and by producing proinflammatory cytokines such as IFN-γ, which inhibit tumor cell growth. Association of CD8+ T cell infiltrate with good overall survival has been widely demonstrated in several tumors including HCC128. However, there is a need to further explore additional markers that define the functional state of the CD8 infiltrate to improve its prognostic value (e.g., expression of immune checkpoints). - CD4+ helper T cells: CD4+ cells encompass different subtypes with opposing effects. TH1 cells and their derived cytokines (e.g., IFNγ) are strongly associated with good clinical outcomes in most cancer types, whereas cytokines generated by TH2 cells (e.g., IL-4 and IL-10) are upregulated in advanced HCC with vascular invasion and metastasis. - Regulatory T cells (Treg): Treg cells are a subset of CD4+ T cells characterized that inhibit immune responses through several mechanisms including suppression of CD8+ T cells via TGF-β and IL-10 signaling. High tumor Treg infiltrate has been proposed as an independent prognostic factor for poor overall survival in HCC. - B cells: B cells are at the center of the humoral adaptive immunity and are responsible for mediating the production of antibodies directed against tumor antigens. Despite they being abundant in the tumor microenvironment, no clear prognostic value has been assigned to this population. - Macrophages: Tumor-associated macrophages (TAMs) arise from two distinct lineages. Tissue-resident macrophages, which self-renew locally, and short-lived monocyte- derived macrophages that infiltrate into the tumor129. Kupffer cells, which constitute up to 90% of tissue-resident liver macrophages, and other TAMs can contribute to hepatocarcinogenesis and immune evasion. An abundance of these TAMs has been associated with a poor prognosis in HCC121. Furthermore, TAMs have been traditionally classified into M1 and M2 macrophages on basis of their functional role. M1 macrophages exert a proinflammatory role by secreting cytokines with anti-tumoral effects, such as IL-12. Conversely, M2 macrophages produce anti-inflammatory cytokines that promote HCC tumor growth, invasion, and metastasis (e.g., IL-5, IL-6, TGF- β). Several studies have shown that a greater proportion of M2 macrophages in the HCC microenvironment results in worse clinical outcomes121,127,128. INTRODUCTION 55 - Dendritic cells (DCs): DCs are antigen-presenting cells that exert pro-immunogenic functions by promoting T cell activation and differentiation. Subsets of DCs with distinct functions and morphology have been identified, including anti-tumoral type 1 DCs, and regulatory type 2 DCs. - Myeloid-derived suppressor cells (MDSCs): MDSCs comprise a heterogeneous population of immature and immunosuppressive myeloid cells with protumoral capacities. They have been reported to suppress adaptive antitumor immunity by impairing CD4+ and CD8+ T cell responses and promoting Treg cell expansion. Increased numbers of MDSCs have been found in tumor tissue and peripheral blood from patients with HCC, and elevated cell counts have been associated with tumor progression. - Neutrophils: Tumor-associated neutrophils (TANs) release a plethora of factors exerting mostly protumoral functions, including promotion of tumor growth and invasion128. An increase TAN count has been associated with poor clinical outcomes in most cancer types. Notably, TANs can be further subdivided into N1 and N2 neutrophils, which represent the extreme of a wide spectrum displaying intermediate phenotypes. N2 neutrophils are likely responsible for the protumoral role of TANs, as opposed to the antitumoral and immunostimulatory N1 phenotype. 3.3.2. Mechanisms of Immune Evasion Dysfunctional tumor-immune system interactions lead to immune evasion through different mechanisms of immune escape29, including: - Secretion or expression of immunosuppressive factors: Tumor cells can secrete immunosuppressive cytokines such as TGF-β and IL-10. This promotes a permissive tumor microenvironment and negatively regulates cytotoxic T lymphocytes effector activity, while also favoring tumor cell proliferation and survival. In addition, tumor cells express co-inhibitory receptors that negatively regulate T cell function, such as cytotoxic T lymphocyte-associated antigen 4 (CTLA4), Programmed cell death protein 1 (PD1), and T cell immunoglobulin and mucin domain containing-3 (TIM3) among others123. - Recruitment of immunosuppressive cells: Infiltration of cells with negative regulatory immune activity to the tumor microenvironment can counteract the effector function INTRODUCTION 56 of T cells27,29. This is the case of regulatory Treg, myeloid-derived suppressor cells, and M2-polarized TAMs. - Antigen loss: Tumor cells can downregulate the expression of tumor antigens, making it harder for the immune cells to identify them as non-self29. - Reduced lymphocytic extravasation: Aberrant tumor vasculature restricts the entry of immune cells to the tumor microenvironment130. Furthermore, tumor blood vessels lack adhesion molecules, thus reducing lymphocytic extravasation and infiltrations29. Any attempt to overcome these barriers to achieve an effective anti-tumor immune activation can potentially represent a therapeutic approach for HCC treatment. For instance, recent studies indicate that VEGF expression by malignant hepatocytes exerts pro-angiogenic effects and generates an immune-tolerant, pro-tumorigenic microenvironment by decreasing cytotoxic T cell and dendritic cell function, and promoting the recruitment of immunosuppressive cells such as Treg, myeloid-derived suppressor cells, and tumor-associated macrophages130–132 (Figure 10). This suggests suggesting that inhibition of the VEGF/VEGFR pathway could be an effective approach to boost the anti-tumoral immune response3,100,130. Figure 10. Direct effects of VEGF pathway activation on tumor-infiltrated immune cells. VEGF produced by tumor and immune cells modulates the functions of innate INTRODUCTION 57 and adaptive immune cells towards immunosuppression. Modified from Fukumura D et al., Nat Rev Clin Oncol 2018130. 3.4. Clinical Management of HCC Patients 3.4.1. Diagnosis, Staging, and Management HCC development is a multistep process with a prolonged subclinical course, and it occurs in the context of a diversity of etiologies and liver disease. In light of this, surveillance protocols have been developed for HCC detection in high-risk patients3,55 (Table 3). Early diagnosis of HCC in patients who have been enrolled in surveillance programs relies on the identification of a liver nodule by abdominal ultrasound and its confirmation using non-invasive radiological approaches or liver biopsy. Nevertheless, diagnosis at symptomatic advanced stages occurs in ~50% of cases globally, particularly in developing countries3. Table 3. Summary of surveillance strategies. Patient population Expected incidence per population Cirrhosis from any etiology, Child-Pugh A or B Hepatitis B cirrhosis 3–8% per year Hepatitis C cirrhosis 3–5% per year Alcohol-related cirrhosis 1.3–3% per year NASH cirrhosis Unknown, but probably 1–2% per year Hemochromatosis and cirrhosis Unknown, but probably >1.5% per year α1 antitrypsin deficiency and cirrhosis Unknown, but probably >1.5% per year Stage 4 primary biliary cirrhosis 3–5% per year Other cirrhosis Unknown Non-cirrhotic hepatitis B Asian male hepatitis B carriers >40 years 0.4–0.6% per year Asian female hepatitis B carriers >50 years 0.3–0.6% per year Hepatitis B carrier with family history of HCC Incidence higher than without family history African Black people with hepatitis B HCC occurs at a younger age (<40 years) Patients with sufficient risk by risk score such as PAGE-B >3% cumulative 5-year incidence if score >10 Other causes Patients with NASH in the absence of cirrhosis <1.5% per year Hepatitis C infection without cirrhosis (including F3) <1.5% per year HCC, hepatocellular carcinoma; NASH, non-alcoholic steatohepatitis. Adapted from Llovet et al., Nat Rev Dis Prim 20213. INTRODUCTION 58 Since most HCC patients suffer from underlying chronic liver damage, the clinical management of the disease is complex and needs to take into account the patient’s overall health status as well as anti-tumor benefits. The Barcelona Clinic Liver Cancer (BCLC) – endorsed by the European Guidelines133 – is a worldwide recognized HCC clinical algorithm for the stratification of patients. It considers clinical variables such as performance status (i.e., ECOG), liver dysfunction (i.e., Child-Pugh), and tumor-related features such as size, number of nodules, or portal invasion80,134. The BCLC staging system defines five prognostic subclasses and allocates specific treatments for each one, based on the levels of evidence defined by the National Cancer Institute (Figure 11)134,135: - BCLC 0 (or very early HCC): Patients at this stage are asymptomatic with well-preserved liver function and present low tumor burden (i.e., single tumor of <2 cm without vascular invasion). These patients are candidates for local curative treatments, including resection and ablation, and present very low recurrence rates. - BCLC A (or early HCC): This stage includes asymptomatic patients with preserved liver function that present a single tumor >2 cm or up to 3 nodules measuring <3 cm. These patients are also considered for curative treatments. Specifically, they can be treated with resection, transplantation, or ablation depending on liver-related variables, can extend median survival beyond 60 months. However, recurrence rates are high (70% in 5 years), and no adjuvant therapies have demonstrated efficacy to date. - BCLC B (or intermediate stage): These patients present multinodular disease with large nodules (>3 cm) or more than 3 nodules of any size. At this stage, patients are asymptomatic and maintain adequate liver function. The established standard of care for this stage is transarterial chemoembolization, which achieves a median survival of 26–30 months. - BCLC C (or advanced stage): At this stage, patients present an advanced disease with macrovascular invasion or extrahepatic spread. These patients are eligible for systemic therapies, including both tyrosine-kinase inhibitors (TKI) and immune checkpoint inhibitors (ICI), which provide a median overall survival of around 19 months in first line. - BCLC D (or terminal stage): This stage includes patients with impaired liver function or relevant tumor-related symptoms, for which best supportive care is recommended. INTRODUCTION 59 Figure 11. The Barcelona Clinic Liver Cancer (BCLC) staging system. The algorithm based on BCLC classifies patients in five stages depending on disease extension, liver function and performance status. AFP, α-fetoprotein; DDLT, deceased-donor liver transplantation; ECOG, Eastern Cooperative Oncology Group; HCC, hepatocellular carcinoma; LDLT, living-donor liver transplantation; M1, distant metastasis; N1, lymph node metastasis; OS, overall survival; RCT, randomized controlled trial; TACE, transarterial chemoembolization. Adapted from Llovet JM et al., Nat Rev Dis Prim 20213. 3.4.2. Systemic Therapies in Advanced HCC Translational research has deeply improved our understanding of the pathophysiology and molecular alterations that drive HCC3,80. It is estimated that ~25% of HCC tumors present actionable mutations, but the low prevalence of such mutations (<10% in most cases) hampers the design of proof-of-concept studies79,83. Furthermore, the most common mutations in HCC are not targetable with existing drugs (e.g., TERT, TP53, and CTNNB1). This knowledge is yet to be translated into clinical practice. However, several molecular therapies and immunotherapy- based approaches have been implemented in the treatment of advanced HCC during the last decades. Tyrosine-kinase inhibitors TKI are compounds that target tyrosine kinase proteins, thus blocking many signaling pathways deregulated in cancer such as proliferation and angiogenesis136. In HCC, the approval of the TKI sorafenib in 2007 paved the way for implementing molecular therapies137. It wasn’t until 10 years later that another TKI, lenvatinib, showed non-inferiority compared to sorafenib and INTRODUCTION 60 received FDA approval for the treatment of advanced HCC in the first-line setting138. The REFLECT trial, a global open-label randomized phase III study established an improved median overall survival for lenvatinib (13.6 months) compared with sorafenib (12.3 months), as well as improved ORR (24.1% versus 9.2%)3,138 (Table 4). The main targets of lenvatinib include VEGF receptors 1-3, FGF receptors 1-4, PDGF receptor α, RET, and KIT, and its difference with sorafenib corresponds to a higher potency blocking VEGF receptors and the FGFR family139. In the second-line setting, three TKI – regorafenib, cabozantinib, and ramucirumab – have been approved for the treatment of HCC patients progressing to sorafenib (Table 4). Importantly, ramucirumab is the only biomarker-guided therapy for HCC, which is indicated for patients with baseline α-fetoprotein levels of ≥400 ng/dl140. Overall, TKI inhibition in second-line confers a median overall survival of 13-15 months3. Table 4. Summary of main outcomes and adverse events among systemic therapies approved for advanced HCC. Treatment Study name Median overall survival (months) Median PFS (months) ORR mRECIST RECIST First-line therapies Atezolizumab + bevacizumab IMbrave150 19.2 6.8 35.4%; 29.8% Sorafenib SHARP (IMbrave150, REFLECT) 10.7–13.4 3.7–4.3 NA 2% Lenvatinib REFLECT 13.6 7.4 24.1% 18.8% Second-line therapies Regorafenib RESORCE 10.6 3.1 11% 7% Cabozantinib CELESTIAL 10.2 5.2 NA 4% Ramucirumab REACH-2 8.5 2.8 NA 5% Second-line therapies based on FDA accelerated approval Pembrolizumab Keynote 240 13.9 3.0 NA 18.3% Ipilimumab + nivolumab Checkmate 040* 22.8 n.a. 34% 32% AST, aspartate aminotransferase; BR, bilirubin; HCC, hepatocellular carcinoma; ORR, overall response rate; PFS, progression-free survival; mRECIST, modified Response Evaluation Criteria In Solid Tumors; RECIST, Response Evaluation Criteria In Solid Tumors. *Data from the Checkmate 404 trial corresponds to phase Ib/II. Adapted from Llovet et al., Nat Rev Dis Prim 20213 and Bruix et al., J Hep 2021141. Immune Checkpoint Inhibitors Immunotherapies with ICIs have emerged as promising treatment options for multiple solid tumors including HCC. ICIs are monoclonal antibodies directed against negative regulators of T cell immune function such as PD1, PDL1, and CTLA4, expressed by tumor or immune cells INTRODUCTION 61 present in the tumor milieu. By blocking these immune checkpoint proteins, ICIs induce an expansion of CD8+ T cells infiltrate, resulting in the activation of the anti-tumor immune response142. In HCC, the combination of atezolizumab (anti-PDL1 antibody) and bevacizumab (anti-VEGF antibody) was the first regimen to improve overall survival compared with sorafenib143, achieving a median survival of 19.2 months and 30% ORR (Table 4)144. As a consequence of these findings, this combination has become the standard of care in first-line therapies for advanced HCC135,141. Based upon promising phase Ib/II studies, two additional therapies – pembrolizumab and nivolumab plus ipilimumab – have received accelerated approval by the FDA in the second- line setting145,146. Despite the promising results of the anti-PD1 monoclonal antibody pembrolizumab, with an ORR of 18% and a median overall survival of 13.9 months, phase III studies failed to demonstrate that it prolongs overall survival compared to sorafenib but showed a non-significant trend with improved long-term survival rates141,146,147. This raised concerns that single-agent ICI may not have sufficient activity to show significant improvements in median OS in an unselected population. The combination of nivolumab (anti-PD1) plus ipilimumab (anti- CTLA4) achieved an ORR of 31% with a median overall survival of 23 months145, thus providing great hopes for treatment combinations in HCC. Emerging Combination Regimens Although ICIs are changing the landscape of HCC treatment, monotherapy approaches elicit responses in only ~15% of patients, while the majority are primarily resistant5,6. Thus, much effort has been invested into identifying existing kinase inhibitors that can effectively synergize with ICI. Indeed, the combination of ICIs with VEGF inhibitors has shown promising activity in many solid tumors121,130, including atezolizumab plus bevacizumab in HCC. The rationale for these combinations is that the VEGFA/VEGFR pathway has a direct immunosuppressive effect on the tumor infiltrate130–132 (Figure 10). It also favors tumor growth, progression, and aberrant vasculature formation130. Therefore, therapies targeting the VEGF pathway likely mitigate the local immunosuppressive effects of VEGF signaling and promote T cell infiltration. In this context, the combination of lenvatinib plus pembrolizumab is currently being tested as first-line therapy in unresectable HCC in a phase III trial. Phase Ib data showed an encouraging objective response rate (ORR) of 46% by mRECIST, with a median overall survival of 22 months and a median progression-free survival of 9.5 month148. Lenvatinib could boost the effects of ICIs on the antitumor immune response by ‘releasing the brake’ on inflammation. However, the INTRODUCTION 62 immunomodulatory capacity of lenvatinib alone or in combination with anti-PD1 still remains poorly characterized. In addition, several trials testing the combinations of a variety of other multi-kinase inhibitors plus ICIs are underway, including cabozantinib plus atezolizumab (NCT03755791). Notably, the inhibition of other tyrosine kinase receptors besides VEGFR, such as FGFR1-4, RET, and PDGF could potentially have immunological and molecular implications. Finally, combinations of different ICIs present another promising strategy, as evidenced by the recent FDA accelerated approval of nivolumab + ipilimumab in second line145. This regimen is now being tested in a phase III trial versus sorafenib or lenvatinib as first-line treatment in patients with advanced-stage HCC (NCT04039607). Furthermore, the combination of the anti- PD-L1 durvalumab with the anti-CTLA4 tremelimumab recently demonstrated superior efficacy compared to sorafenib first-line therapy (NCT03298451)121. 3.4.3. Biomarkers for Patient Selection To date, the translation of predictive biomarkers that guide systemic therapies in HCC is under investigation. Few candidate biomarkers of response to TKI have been reported, with the sole exception of α-fetoprotein serum levels in ramuricumab (>400 ng/ml)3,140. For instance, FGFR4 inhibition elicited promising responses in a subset of patients with tumors overexpressing FGF19 in a phase I trial, but this discovery has not been translated in phase III investigations149. Similarly, MET inhibition with tivantinib did not improve survival in patients with tumor-MET overexpression150. A variety of candidate biomarkers for ICI are under investigation across different solid tumors, including HCC. PD-L1 expression by immunohistochemistry has been approved as companion diagnostics or complementary test of anti-PD-1 treatments in other malignancies, but its predictive role in HCC is unclear3,151. Furthermore, pembrolizumab has been FDA-approved for the treatment of advanced-stage cancers with microsatellite instability or mismatch repair- deficiency irrespectively of tumor type or histology, but this alteration is found in only ~3% of HCC79,152. Tumor lymphocytic infiltration, gene signatures of immune activity, and CTNNB1 mutation status also warrant examination for predictive value in patients treated with ICIs in HCC37. INTRODUCTION 63 Overall, most patients receiving ICIs do not derive benefit, and there is an urgent need to identify and develop predictive biomarkers of response to these therapies, both to enable a precision medicine approach in HCC and to better understand and overcome mechanisms of resistance. HYPOTHESES AND AIMS HYPOTHESES AND AIMS 67 1. Hypotheses Primary liver cancer is the sixth most commonly diagnosed cancer and the second leading cause of cancer-related death worldwide2,3. Around 90% of cases correspond to hepatocellular carcinoma (HCC)2,3, which arises almost unfailingly in the setting of chronic liver diseases. The worldwide incidence of HCC cases is heterogeneous, reflecting the distribution of risk factors, with Mongolia exhibiting the highest HCC burden worldwide2. Well-established HCC risk factors are hepatitis B virus, hepatitis C virus, and non-alcoholic fatty liver disease (NAFLD). Other agents have also been reported to promote hepatocarcinogenesis, including adeno-associated virus (AAV) integration3,4, but more studies are needed to determine in which conditions this virus can induce HCC. Approximately, 50–60% of HCC patients are exposed to systemic therapies in their lifespan, particularly in advanced stages of the disease3. These mostly include multikinase inhibitors such as lenvatinib and immune checkpoint inhibitors (ICI) such as anti-PD1 antibodies. Although ICIs are revolutionizing HCC treatment, monotherapy approaches elicit responses in only 15% of patients, while the majority are primarily resistant5,6. Thus, identifying existing multikinase inhibitors that can effectively synergize with ICI is urgently needed. During the last decade, molecular profiling of tumors using translational approaches has significantly contributed to the understanding of the molecular pathogenesis of HCC3,83. This knowledge has provided important opportunities in clinical oncology by 1) identifying novel genetic and environmental features associated with hepatocarcinogenesis in specific patient populations; and 2) unraveling new treatment strategies. Considering all this, the hypotheses of this thesis are that: 1. Performing a comprehensive analysis of the molecular and immunological features of HCC will provide relevant information about the genetic and molecular determinants associated with HCC in specific populations and will unveil novel potential therapeutic approaches. This could result in fundamental implications for the clinical decision- making in HCC. 2. Assessing the distinct genomic and transcriptomic alterations of Mongolian HCC could provide relevant information that may help unveil the genetic factors and environmental exposures underlying the high incidence in this population. HYPOTHESES AND AIMS 68 3. Hepatocyte replication in response to liver injury and NAFLD could promote oncogenic AAV integration and HCC development, which could be a concern for the use of AAV gene therapy in patients with chronic liver. 4. The multikinase inhibitor lenvatinib has immune-modulating potential and its combination with anti-PD1 checkpoint inhibitors could improve its anti-tumoral effect in HCC. HYPOTHESES AND AIMS 69 2. Aims Considering the background and hypotheses exposed above, the specific aims of this doctoral thesis were the following: 1. To provide a molecular characterization of Mongolian HCC and identify its unique genomic features compared to Western HCC. 2. To assess whether NAFLD-associated liver damage increase the risk of AAV integration inducing HCC. 3. To identify the immunomodulatory effects of lenvatinib in combination with anti-PD1 and provide a mechanistic rationale for this treatment in advanced HCC. RESULTS RESULTS 73 Study 1 – Hepatocellular Carcinoma in Mongolia Delineates Unique Genomic Features Laura Torrens*, Marc Puigvehí*, Miguel Torres-Martín, Huan Wang, Miho Maeda, Philipp K. Haber, Thais Leonel, Mireia García-López, Wei Qiang Leow, Carla Montironi, Sara Torrecilla, Ajay Ramakrishnan Varadarajan, Patricia Taik, Genís Campreciós, Chinbold Enkhbold, Erdenebileg Taivanbaatar, Amankyeldi Yerbolat, Augusto Villanueva, Sofía Pérez-del-Pulgar, Swan Thung, Jigjidsuren Chinburen, Eric Letouzé, Jessica Zucman-Rossi, Andrew Uzilov, Jaclyn Neely, Xavier Forns, Sasan Roayaie, Daniela Sia, Josep M. Llovet * Contributed equally Submitted to Proc Natl Acad Sci U S A (IF: 11.205) Summary Mongolia has the world’s highest incidence of HCC (~100 cases per 100,000 inhabitants)2, far exceeding that of the surrounding countries2. The Mongolian population presents many particularities that might play a role in HCC burden, including specific risk factors, socioeconomic particularities and genetic background72,153. Despite a strikingly high prevalence of HBV (10.6%), HCV (6.4%), and HDV (50-80% of HBV-positive individuals) infections and alcohol consumption56,64,65, it is unclear whether HCC incidence is completely explained by the combination of risk factors, or eventually other unknown factors might play a role. In this context, we hypothesize that assessing the distinct genomic and transcriptomic alterations of Mongolian HCC could provide relevant information that may help unveil the genetic factors and environmental exposures underlying the high incidence in this population. Indeed, large-scale next-generation sequencing studies conducted during the last decade have been key in deciphering the molecular alterations and transcriptomic-based subtypes occurring in HCC77,83. Particularly, the presence of mutational signatures consisting in nucleotide substitution patterns has allowed to track the exposure of endogenous and exogenous agents in HCC samples22. With the aim to provide a molecular characterization of Mongolian HCC and identify its unique genomic features compared to Western HCC, we collected 192 Mongolian HCC and 187 Western HCC from Europe and US. Whole exome and RNA sequencing were conducted, and the mutational landscape, mutational signatures and transcriptomic profiles were evaluated and compared between cohorts. Furthermore, viral characteristics were assessed, including HBV and HDV viral genotypes and presence of HBV pro-oncogenic mutations. RESULTS 74 We were able to identify distinct clinical and molecular features of Mongolian HCC compared to Western cases, including: 1. High prevalence in females (up to 46% of the cohort), consistent with the reported male/female 1.5/1 ratio in Mongolia76, as opposed to that observed in the Western cohort (20% females) and globally3. In addition, Mongolian patients were younger, with less advanced hepatic fibrosis, and higher rate of HBV-HDV co-infection (84% of HBV- infected patients). 2. HBV characteristics associated with low oncogenic potential, including genotype D1 and low prevalence of HBV precore and basal core promoter mutations compared to Western cases154,155. 3. High rate of protein-coding mutations, almost doubling that in the Western in-house cohort (121 vs 70 mutations per tumor) and publicly available datasets82,83,85,153. This could potentially be explained by the presence of intrinsic and/or extrinsic factors promoting mutagenesis in Mongolian HCC. 4. Higher mutation rates in HCC drivers such as TP53, APOB, and the KMT2 gene family. Furthermore, TSC2 mutations were identified as a potential drivers in 9% of Mongolian tumors by in silico tools revealing positive selection of damaging alterations in this gene156,157. 5. Presence of a novel mutational signature (SBS Mongolia) in 25% of cases associated with the carcinogenic dimethyl sulfate (DMS) genotoxic signature. This potentially suggests that long-term exposure to DMS generated from coal combustion could be a risk factor for HCC development in Mongolia. 6. A distinct transcriptomic profile consisting in three molecular clusters (MGL1-3), of which MGL2 (26%) and MGL3 classes (30%) were specific for Mongolian tumors and presented molecular and clinico-pathological features not observed in HCC samples from Western countries, including enrichment in HBV-HDV infection, female gender, and inflamed profile (p < 0.05). In conclusion, Mongolian HCC is characterized by unique molecular classes, high mutational burden, and a distinct mutational signature associated with environmental factors. These RESULTS 75 findings could pave the way for the identification of environmental or genetic factors associated with the increased incidence in this country. RESULTS 77 Hepatocellular Carcinoma in Mongolia Delineates Unique Genomic Features Laura Torrens1,2#, Marc Puigvehí1,3#, Miguel Torres-Martín1,2, Huan Wang4, Miho Maeda1, Philipp K. Haber1, Thais Leonel5, Mireia García-López5, Wei Qiang Leow1,6, Carla Montironi1,2, Sara Torrecilla2, Ajay Ramakrishnan Varadarajan4, Patricia Taik4, Genís Campreciós1,5, Chinbold Enkhbold7, Erdenebileg Taivanbaatar8, Amankyeldi Yerbolat8, Augusto Villanueva1, Sofía Pérez- del-Pulgar5, Swan Thung1, Jigjidsuren Chinburen8, Eric Letouzé9, Jessica Zucman-Rossi9, Andrew Uzilov4,10, Jaclyn Neely11, Xavier Forns5, Sasan Roayaie12, Daniela Sia1, Josep M. Llovet1,2,13* 1Liver Cancer Program, Division of Liver Diseases, Tisch Cancer Institute, Department of Medicine, Icahn School of Medicine at Mount Sinai, New York, New York, USA; 2Translational research in Hepatic Oncology, Liver Unit, Institut d'Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Hospital Clínic, University of Barcelona, Barcelona, Spain; 3Hepatology Section, Gastroenterology Department, Parc de Salut Mar, IMIM (Hospital del Mar Medical Research Institute), Barcelona, Catalonia, Spain4Sema4, Stamford, Connecticut, USA; 5Liver Unit, Hospital Clínic, Institut d'Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), CIBEREHD, University of Barcelona, Barcelona, Spain; 6Department of Anatomical Pathology, Singapore General Hospital, Singapore, Singapore; 7Hepato-Pancreatico-Biliary Surgery Department, National Cancer Center, Ulaanbaatar, Mongolia; 8National Cancer Center, Ulaanbaatar, Mongolia; 9Centre de Recherche des Cordeliers, Sorbonne Université, Inserm, Université de Paris, Université Paris 13, Functional Genomics of Solid Tumors laboratory, F-75006, Paris, France; 10Department of Genetics and Genomic Sciences and Icahn Institute for Data Science and Genomic Technology, Icahn School of Medicine at Mount Sinai, New York, New York 10029, USA; 11 Bristol Myers Squibb, Princeton, New Jersey, USA; 12Department of Surgery, White Plains Hospital, White Plains, New York, USA; 13Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Catalonia, Spain # These authors contributed equally; * Corresponding author Publication RESULTS 78 ABSTRACT Background and aims: Mongolia has the world’s highest incidence of hepatocellular carcinoma (HCC), with ~100 cases/105 inhabitants/year. Here, we aimed to provide a molecular characterization of Mongolian HCC and unveil unique genomic and environmental features compared to Western HCC. Methods: We collected 192 well-annotated paired fresh-frozen HCC/non-tumoral samples from Mongolian patients and compared its molecular profile using whole exome and RNA sequencing with a newly collected and unreported Western cohort (n=187). Mutational calling, mutational signature analysis, tumor mutational burden, and transcriptome analysis were conducted. Viral genotypes were assessed in HBV+ and HDV+ samples by direct sequencing. Results: Mongolian patients, compared to Western, were significantly younger, with higher female predominance, and presented higher rates of HBV-HDV co-infected non-cirrhotic livers (all p<0.001). Mongolian HCCs present three unique molecular features: a) higher rates of protein-coding mutations (121 vs 70 mutations per tumor in Western) with higher mutations rates in known (e.g., APOB) and putative HCC drivers (TSC2); b) a novel mutational signature (SBS Mongolia) identified in 25% of cases (vs 4% in Western samples) that was enriched in a signature of genotoxic exposure to dimethyl sulfate, a byproduct of coal combustion; and c) a distinct transcriptomic profile consisting in three molecular clusters (MGL1-3), of which MGL2 (26%) and MGL3 classes (30%) were specific for Mongolian tumors and were enriched in HBV- HDV infection, female gender, and inflamed profile (p<0.05). Conclusion: Mongolian HCC is characterized by unique molecular classes, high mutational burden, and a unique mutational signature associated with environmental factors. INTRODUCTION Liver cancer is the third leading cause of cancer-related mortality, and its global burden has increased in recent years [1,2]. Hepatocellular carcinoma (HCC) accounts for 90% of liver cancer cases and arises almost unfailingly in the setting of chronic liver diseases. The worldwide incidence of HCC cases is heterogeneous, reflecting the distribution of known liver disease risk factors such as hepatitis B virus (HBV), hepatitis C virus (HCV), alcohol consumption, and non- alcoholic steatohepatitis [2]. Mongolia, a landlocked East Asian country between Russia and China, shows the world’s highest incidence of HCC, with a burden of 86 cases per 100,000 RESULTS 79 inhabitants [1]. This incidence far exceeds that of the surrounding countries (4x and >20x compared to China and Russia, respectively) or any other country worldwide [1], and has been attributed to a historical high prevalence of HBV (10.6%) and HCV (6.4%) viruses, as well as alcohol consumption [3–5]. Indeed, 90% of Mongolian HCC cases are positive for HBV, HCV, or both [6], and co-infection with hepatitis delta virus (HDV), a defective virus that needs HBV for its replication cycle and has been associated with liver fibrosis and HCC development, occurs in 50-80% of HBV-infected individuals [7]. Despite the implementation of universal infant HBV vaccination in 1991 [3] and the improved control in HCV blood-based transmission, the burden of HCC in Mongolia is increasing year after year, and the incidence is now ~10x compared to the 1960s [1]. Overall, the reasons for this extreme incidence have never been thoroughly understood. During the last 10 years, large-scale next-generation sequencing studies have been key in deciphering the transcriptomic-based HCC subtypes and the molecular alterations occurring in HCC [8,9]. The presence of mutational signatures consisting in unique nucleotide substitution patterns has allowed to track the exposure of endogenous and exogenous factors in cancer [10]. For instance, HCC risk factors such as aristolochic acid, aflatoxin, alcohol, or tobacco smoking can be related to specific signatures [8,11]. In this context, assessing the genetic alterations, mutational frequencies and transcriptomic profile of Mongolian HCC could provide relevant information that may help unveil the genetic factors and environmental exposures underlying the high incidence in this population. Previous studies have provided valuable data about the molecular landscape in Mongolian HCC [12]. However, further analyses comparing Mongolian tumors with a Western cohort are needed to understand the relevance of the molecular traits in Mongolian HCC. Additionally, viral characteristics in Mongolian HCC remain underexplored. To identify the unique molecular features of Mongolian HCC, we performed whole exome (WES) and RNA sequencing (RNA-seq) in 379 HCC tumors of Mongolian and Western origin. Herein, we provide a comprehensive characterization of the molecular profile in Mongolian HCC and reveal unique features consisting in an increased number of mutations, as well as specific mutational and transcriptomic patterns. These findings could pave the way for the identification of environmental or genetic factors associated with the increased incidence in this country. RESULTS 80 MATERIALS AND METHODS Study design A total of 219 paired HCC/non-tumoral liver samples from distinct patients undergoing HCC resection were collected at the National Cancer Center, Ulaanbaatar, Mongolia (Fig. 1a). Samples were collected from October 2015 to October 2017, in accordance with Mongolian regulations, the National Cancer Center, and the Ministry of Health of Mongolia. Written informed consent was obtained from all participants. A Western cohort was used as internal control, including tumor and matched non-tumoral liver samples (n = 187) from patients undergoing resection (Fig. 1a). Samples were collected from two institutions of the HCC Genomic Consortium: IRCCS Istituto Nazionale dei Tumori (Milan, Italy; n = 110) and Icahn School of Medicine at Mount Sinai (New York, USA; n = 77). Samples were collected with written informed consent upon Institutional Review Board approval. Whole exome sequencing WES analysis was run in NovaSeq 6000 (Illumina, San Diego, CA) in the New York Genome Center facilities. WES data was used for mutation calling, mutational signature analysis, and tumor mutational burden (TMB) evaluation. TMB was calculated based on protein-coding mutations assuming an average exome size of 30 Mb, in accordance with previously published studies [10]. Mutations were called comparing the tumor with its paired non-tumoral counterpart. Molecular variant calling was performed by Sema4 (Stamford, CT, USA), using the Tigris pipeline (v2.0.1), which carries out modified GATK4 (4.0.11.0) best practices (https://software.broadinstitute.org/gatk/). In addition, WES data from a European (n = 241) [8], Korean (n = 231) [13], TCGA (n = 363) [14] and Mongolian NCI (n = 71) [12] HCC cohorts were used for mutation calling and TMB evaluation. Mutational signature analysis Somatic SNVs in exome region (defined by coding exons in Ensembl GRCh37 built) at allelic frequency cutoff of 0.05 and with gnomAD population frequency or ethnic-specific frequency ≤ 0.5% were selected. Tumor samples with total SNV count ≥ 50 (after aforementioned filtering) were used in downstream mutational signature analyses, resulting in a total of 254 samples (148 from the Mongolian cohort plus 106 from the Western cohort). The remaining samples (n = 9) were considered negative for the signatures. All signature fitting and de novo signature RESULTS 81 extraction analyses were performed using exome region SNVs and trinucleotide frequencies normalized via `exome2genome` approach in the deconstructSigs R package [15]. RNA sequencing RNA data was processed by the RAPiD pipeline at the Mount Sinai Genomic Core Facility. Briefly, Fastq files were aligned using STAR (v 2.7.0f) [16] to hg19 with gencode annotation v19. QoRTs (v1.3.6) [17] was used for QC and obtaining raw counts. Batch correction was performed using RUVSeq [18]. Empirical method (RUVg) with 10.000 low expressed genes and normalization for subsequent analysis was performed using VST method from DESeq2 [19]. Unsupervised clustering analysis of the whole cohort (n = 224) was performed using the Non-negative matrix factorization (NMFc) module from GenePattern and Euclidean distance and Ward’s agglomerative procedure. Clustering of the Mongolian (n = 118) and NCI Mongolian cohort [12] (n = 70) was also performed using NMFc. Gene expression characterization was performed using Nearest Template Prediction (NTP), Gene Set Enrichment Analysis (GSEA), and single sample GSEA (ssGSEA) modules from GenePattern. To this end, Molecular Signature Database (MSigDB, www.broadinstitute.org/msigdb) and previously reported gene sets were used (Supplementary Table 1). Class comparison between molecular clusters was performed using subclass mapping analysis. Finally, the stromal infiltration and relative tumor purity were assessed using the ESTIMATE R package [20]. Statistical analysis Statistical analyses were performed using either SPSS software package (version 24.0; SPSS Inc, Chicago, IL, USA) or scipy (v 1.2.1) and matplotlib (v3.0.3) modules from Python (v3.7.3). Differences between qualitative variables were assessed with the Fisher exact test and corrected for multiple comparisons using false discovery rate. Differences between quantitative variables were analyzed with a two-sided non-parametric test (Mann-Whitney or Kruskal-Wallis, were appropriate), and adjustments for multiple comparison analysis were performed using Dunn’s test. Data Availability Original whole exome sequencing and RNA sequencing data are available at the European Genome-Phenome Archive (EGAS00001005364). The remaining data are available in the Article, Supplementary Information, or upon request. Additional detailed information is provided in the Supplementary Materials and Methods. RESULTS 82 RESULTS Clinico-pathological characteristics of the cohorts HCC patients from the Mongolian cohort were younger (61 vs 66 years old, p < 0.001), with a higher rate of HBV/HDV co-infection (84% of HBV infected vs 7% in the Western cohort, p < 0.001), and lower rate of non-infected cases (15% vs 40%, p < 0.001) (Table 1, Supplementary Table 1). In line with previous data [21]. 54% of the HCC patients in Mongolia were male compared to 80% in the Western cohort (p < 0.001). The rate of advanced hepatic fibrosis (F3- 4) and cirrhosis (F4) was significantly lower in Mongolian compared to Western cases (38% vs 79% and 16% vs 60%, respectively), independently of etiology (Supplementary Fig. 1a). Tumor characteristics were similar in both cohorts, with most tumors within Barcelona Clinic Liver Cancer (BCLC) stages 0-A, and with alpha-fetoprotein (AFP) < 400 (IU/mL). However, tumors in the Mongolian cohort showed a lower differentiation grade (Table 1). No significant survival differences were observed between cohorts. Viral characterization of Mongolian and Western individuals To understand whether the particular clinico-pathological features of the Mongolian HCCs were due to unique viral characteristics, we analyzed the phenotypes of HBV and HDV in infected patients. HBV genomes can be classified into 9 genotypes (A to I), according to differences in nucleotide sequences. HBV genotypes have a characteristic geographical and ethnic distribution, and HBV genotype D is known to be almost universal in Mongolia [22]. In the Western cohort, patients were infected mostly by genotypes C and D (27% and 43%, respectively, p < 0.001) whereas in the Mongolian cohort, all HBV-infected individuals were genotype D, including 2 patients with recombinant forms of genotype C and D (Supplementary Table 2). Interestingly, genotype D has been previously associated with reduced HCC development as compared to genotype C [23]. The most frequent HBV sub-genotypes in Mongolian and Western cases were D1 (77%) and D3 (39%), respectively (Supplementary Fig. 1b). Among Western cohort samples, the most prevalent genotype in European patients was genotype D3 (71%), whereas in patients from the USA it was genotype C (60%) (Supplementary Fig. 1c). Regarding HDV infection, all Mongolian patients were genotype 1 (Supplementary Table 2). We then evaluated the presence of basal core promoter (BCP) and precore HBV mutations, which have been associated with liver disease progression and HCC development [24]. It has been previously suggested that these mutations are more frequent in HBV-infected patients with genotype D, with only 11% of patients with such genotype being wild-type [25]. We detected BCP A1762T, BCP G1764A, and precore G1896A mutations in 14%, 21% and 29% of HBV-infected Mongolian patients, RESULTS 83 respectively (Supplementary Fig. 1d, Supplementary Table 2). Interestingly, the prevalence of these mutations was significantly higher in the Western cohort (60%, 66%, and 46%, respectively), (p < 0.001, p < 0.001, and p = 0.089, respectively, Supplementary Fig. 1d). The percentages observed in our Western patients were similar to those previously reported [26]. In line with previously reported data [27], precore G1896A mutations were less frequent in HBV/HDV co-infected versus HBV mono-infected in Mongolia (20% vs 64%, p = 0.002); however, we found no differences for both BCP A1762T and G1764A mutations (14% vs 14%, and 19% vs 29%, respectively, p = ns). Overall, these data suggest that the rate of BCP and precore HBV pro- oncogenic mutations in Mongolia is particularly low, despite the predominance of genotype D in this population. The interaction between HBV and HDV viruses is complex and not fully understood. It has been proposed that HDV can suppress HBV [28,29], but HBV and HDV levels can fluctuate over time [29]. To further investigate this, we determined HBV-DNA and HDV-RNA levels in tumor- adjacent liver tissue in Mongolian and Western samples. Of note, high HBV-DNA levels in blood have been associated with more aggressive liver disease, including HCC development [30]. First, we analyzed the differences between HDV positive and negative samples. Intrahepatic HBV-DNA load was significantly higher in samples with HBV/HDV co-infection than those with HBV mono- infection (5.0 vs 3.8 log copies/μg total DNA, p = 0.001, Supplementary Fig. 2a). Furthermore, high HBV-DNA levels were associated with advanced liver fibrosis, advanced tumor stage, and worse survival in Mongolian individuals (Supplementary Table 3, Supplementary Fig. 2b). Conversely, patients with high HDV-RNA levels showed significantly higher alanine aminotransferase levels, suggesting greater inflammation (Supplementary Table 4). Finally, no significant differences in HBV-DNA load were found between Mongolian and Western samples (4.85 vs. 4.97 log copies/μg total DNA, respectively p = 0.23, Supplementary Fig. 2c, Supplementary Table 5). Taken together, our analysis suggests that HBV and HDV viral characteristics in Mongolia are highly homogeneous and with low oncogenic potential. Analysis of the genomic landscape in Mongolian HCC To gain further insights into the molecular landscape of Mongolian HCC, we performed mutation and copy number variation (CNV) analysis. The pattern of broad gains and losses in our Western cohort was consistent with previous reports in HCC [8], with 1q and 8q gains and 8p losses being the most common alterations (Supplementary Fig. 3a-b, Supplementary Table 6). When we compared the broad chromosomal variation profiles between Mongolian and Western HCC patients, no difference was observed in terms of overall CNV burden (Supplementary Fig. 3c-e). RESULTS 84 Nonetheless, Mongolian HCCs showed a significantly higher occurrence of 1p gains, 9q gains, as well as fewer 9q losses, 1q losses, and 8p losses (Supplementary Table 6). The average number of mutations per tumor was significantly higher in Mongolian patients compared to Western, with a median of 121 and 70 mutations/tumor, respectively (p < 0.001) (Fig. 1b). Accordingly, the median tumor mutational burden (TMB) in the Mongolian and Western cohorts was 4.0 and 2.3 mutations/Mb, respectively (p < 0.001). No significant differences were observed depending on the origin of Western samples (p = 0.645, Fig. 1c). To rule out the possibility that the observed difference was due to random selection bias, we then compared the median protein-coding mutations with previously published cohorts in Western and Asian countries, applying the same filtering criteria [8,12–14]. The median number of mutations per tumor was 111 in the Mongolian NCI cohort [12], 76 in TCGA [14], 61 in the European [8] cohort and 63 in the Korean cohort [13], corresponding to a TMB of 3.7, 2.5, 2.0, and 2.1, respectively (Fig. 1b, Supplementary Table 7) confirming that the number of mutations in both Mongolian HCC cohorts was significantly higher than in other countries (all 0.001). There was a positive association between the number of mutations and tumor grade in Mongolian cases, with a median of 100, 120, and 171 mutations/tumor in samples with good, moderate or poor differentiation grade (p = 0.004). We did not observe any significant difference in the median of mutations according to etiology in either cohort (Supplementary Fig. 4a-c). Previous studies in HCC have suggested that highly mutated tumors (TMB ≥4 mutations/Mb) are enriched with mutations in DNA damage response (DDR) genes [31]. In our Mongolian cohort, 84 (56%) of samples showed ≥ 4 mutations/Mb, compared to 7 (6.3%) in the Western cohort (p < 0.001). However, no association was observed with presence of mutations in DDR genes (Supplementary Table 8). Mutational profile of Mongolian HCC We then explored whether the higher frequency of mutations in the Mongolian cohort was due to enrichment in specific genes. Among the 250 most frequently mutated genes considering either cohort, 225 (90%) were more mutated in Mongolian HCCs, suggesting that the higher number of mutations was broadly occurring across the whole genome and not concentrated in specific loci (Fig. 1c). Overall, we detected a significant mutational increase in multiple known HCC driver genes in Mongolia compared to the Western cohort, including TP53 (46% vs 32%), APOB (15% vs 5%), TSC2 (9% vs 1%), and NFE2L2 (6% vs 1%), (p < 0.05, Fig. 2, Supplementary Fig. 5, Supplementary Tables 9). The mutation rate of HCC drivers was also assessed in the NCI Mongolian cohort for comparison [12] (Fig. 2a, Supplementary Table 10). Furthermore, genes RESULTS 85 belonging to the KMT2 histone lysine methyltransferase family were significantly more mutated in Mongolian HCC (34% vs 18%, p = 0.005, Fig. 3a). Mongolian HCC with mutations in the KMT2 gene family displayed higher TMB than patients without these mutations in both Mongolian cohorts (median TMB 4.6 and 4.7 vs 3.9, p = 0.005 and p < 0.0001, Fig. 3b). Next, potential drivers in Mongolian HCC were further assessed by OncodriveCLUSTL and dN/dScv algorithms [32,33]. Among the genes significantly more mutated in the Mongolian cohort, 6 were enriched for damaging alterations, suggesting that they could exert a drivel role in Mongolian HCC (q < 0.05; Fig. 3c-d), including TSC2 (9%). Similar TSC2 mutation rates were confirmed in the Mongolian NCI cohort (7%) (Fig. 3c). Finally, differences in the mutation profile depending on etiology were detected in the Mongolian cohort (Supplementary Fig. 4d-e). CDKN2A mutations were enriched in patients with HBV (7.4% vs 0%) and HDV (8.0% vs 0%) infection, while ARID2 mutations were more common in HCV-infected patients (11.9% vs 1.2%) (p < 0.05, Supplementary Fig. 4d). No significant differences between Western samples from Europe and USA were observed (Fig. 2b, Supplementary Table 9). Overall, our results indicate that Mongolian HCC shows a significantly higher tumor mutational burden and, although the mutational and chromosomal spectrum highly resembled that of Western HCC, significant differences were observed for key driver genes including APOB, KMT2 family, and TSC2. Mutational signature analysis We then analyzed the pattern of single base substitutions in the Mongolian and Western cohorts. Notably, Mongolian HCC was characterized by a higher proportion of T>G substitutions compared to Western tumors (Supplementary Fig. 6a-c). This was also observed in the Mongolian NCI cohort (Supplementary Fig. 6d). De novo mutational signature extraction from the Mongolian and Western cohorts revealed four signatures (Fig. 4a), three of which were mapped to COSMICv3 signatures previously found in liver cancer: SBS22, a combination of SBS6-SBS40 and of SBS16-SBS26 (cosine similarity >0.90 in all cases) [34]. The fourth signature did not present strong similarities with any of the COSMIC signatures and was therefore considered novel (Supplementary Table 11-12). We then performed signature fitting using the de novo signature 4 and HCC-specific COSMICv3 signatures by applying a bootstrap approach (exposure cutoff ≥ 0.1). Interestingly, the Mongolian cohort was enriched in the de novo signature 4, henceforth renamed SBS Mongolia (SBSM) (25.2% RESULTS 86 [38/151 vs 4.5% [5/112] in Mongolian vs Western cohorts; p < 0.0001). Notably, Mongolian HCC presenting the SBSM signature showed a distinct substitution profile consisting in a high proportion of T>G substitutions (14% vs 8% in SBSM positive and negative samples, respectively [p < 0.001], and 6% in Western HCC, Fig. 4c-d). Other COSMICv3 signatures previously reported in HCC [34] presented similar prevalence between cohorts (Fig. 4c-d, Supplementary Fig. 6e-f, Supplementary Table 13). To further characterize SBSM positive samples, we assessed the presence of mutational signatures linked to the effects of known or suspected environmental mutagens from the Compendium of Mutational Signatures of Environmental Agents [35]. Samples presenting SBSM were significantly enriched for the mutational signature associated with exposure to dimethyl sulfate (DMS) (71.1% [27/38] vs 26.5% [30/113], p < 0.0001, Fig. 4b). In line with this, DMS was the only environmental-related signature significantly enriched in Mongolian HCC compared to Western (37.7% [57/151] vs 18.8% [21/112], Supplementary Table 14). Patients presenting the DMS signature were older (64.3 vs 59.5 years) and predominantly HCV-positive; Supplementary Fig. 7). No association between SBSM and TMB, etiology, fibrosis, or other clinical and molecular variables were found (Fig. 4b). Finally, we investigated the mutational profile in adjacent matched liver tissue of Mongolia and Western HCCs. Due to the small number of SNVs present in the adjacent tissues, mutational signature fitting was performed on pooled variants from the Mongolian and Western cohorts using HCC-specific COSMICv3 signatures plus SBSM (Supplementary Fig. 8). SBSM was the only dominant signature in adjacent tissue from the Mongolian cohort (Supplementary Fig. 8c-d), suggesting that non-tumoral liver tissue in Mongolia presents the signature before HCC arises. SBS5, associated with age-related clock-like mutations [34], was the main signature in Western non-tumoral tissue. Overall, Mongolian HCC presents a unique substitution profile characterized by a novel mutational signature, SBSM, which is associated with the DMS-related signature. Considering this, DMS exposure warrants further investigation as a potential environmental factor for HCC in Mongolia. Identification of unique gene expression patterns in Mongolian HCC We then investigated the transcriptome profiling of HCC samples using RNA-seq data to define the molecular patterns in Mongolian HCC tumors. Unsupervised clustering analysis of Mongolian and Western samples using non-negative matrix factorization (NMFc) identified two robust RESULTS 87 clusters (Supplementary Fig. 9a-c). Notably, 80% of Western tumors were included in one cluster, while the second cluster showed a strong enrichment of Mongolian HCC samples, suggesting that Mongolian HCC may present a distinct molecular profile. Furthermore, NMFc analysis of the Mongolian HCC samples alone revealed three main clusters -MGL1, MGL2, and MGL3- (Supplementary Fig. 9b-c), which overlapped with the classification of the whole cohort. Unsupervised clustering showed similar results, thus confirming the robustness of our findings (Supplementary Fig. 9d). To elucidate the transcriptomic differences between Mongolian and Western HCC, we performed single-sample Gene Set Enrichment Analysis (ssGSEA) and Nearest Template Prediction (NTP) (Supplementary Fig. 10a). The Mongolian cohort presented an enrichment in Hoshida S1 and Proliferation classes (39% vs 20% and 36% vs 14%, p < 0.01). In addition, Mongolian tumors showed enhanced inflammatory signaling (i.e., IFN and HCC Immune class), response to viral infection, and growth factor-related pathways (all p < 0.05). Comparatively, the Western cohort was enriched in the Hoshida S3 class (26% vs 47%, p < 0.01) and liver-related metabolic activation. We then characterized each one of the MGL clusters. Patients belonging to MGL1 class (44% of the cohort) were more frequently HCV-infected, older, and mostly males (female:male ratio of 1:2), (Supplementary Table 15) and with a molecular profile closer to Western HCCs than the rest of Mongolian HCC (Fig. 5). On the other hand, patients of the MGL2 (26%) and MGL3 clusters (30%) were significantly younger than MGL1 and enriched in HBV/HDV infection and triple infections (HBV/HDV/HCV). Interestingly, while patients of the MGL3 class showed a female:male ratio similar to MGL1 (1:2), the MGL2 class showed a female:male ratio of 2:1 and higher AFP levels. None of the MGL clusters was associated with specific outcomes (Fig. 5, Supplementary Fig. 11, Supplementary Table 15). In terms of molecular features (Fig. 5), the MGL1 cluster was characterized by enrichment in CTNNB1 class and activation of metabolic and liver-specific pathways (all p < 0.05). MGL2 HCC cases displayed a proliferative HCC phenotype (i.e enrichment of Proliferation, G3, and Cluster A gene signatures), higher rates of RB1 mutations, and lower rates of CTNNB1 mutations (Fig. 5). Finally, MGL3 were characterized by lower rates of TP53 mutations and fewer broad chromosomal alterations and were particularly enriched in immune-related features (i.e immune class, interferon and inflammatory pathways, and PD1 signaling). The clinico-pathological and molecular features of each MGL cluster were further validated using RNA-seq data from the Mongolian NCI cohort [12] (Supplementary Fig. 12a, Supplementary RESULTS 88 Table 16). In addition, subclass mapping and NTP analysis in the in-house Mongolian cohort indicated a good overlap of the MGL clusters with the previously published MO classification of Mongolian HCC (Supplementary Fig. 13) [12]. Specifically, MGL1 aligned with the MO1 class, MLG2 with MO4; and MGL3 with MO2 and MO3 (all FDR < 0.05). Overall, our results suggest that three distinct gene expression patterns characterize Mongolian HCC of whom two presents unique clinico-pathological features not observed in HCC samples from Western countries. Immune characterization of the molecular classes of Mongolian HCC We then further explored the inflammatory profile of Mongolian HCC using ESTIMATE [20] and ssGSEA in our in-house Mongolian cohort and the NCI cohort [12]. In both Mongolian cohorts, MGL3 showed a significantly higher immune enrichment score, presence of both an innate and adaptive immune response, and signatures predicting response to immunotherapy (Fig. 6, Supplementary Fig. 12b). A certain degree of inflammation was also observed in MGL2, even if significantly lower than MGL3. Our previously reported HCC immune class was detected in most patients belonging to MGL3 (88% in the Mongolian cohort, 100% in the NCI Mongolian cohort [12], Fig. 6, Supplementary Fig. 12b). Compared to Western HCC, the immune class in Mongolian tumors was larger (42% versus 29%, p = 0.05) with an inverted ratio of Exhausted/Active subtypes (65/35 vs 30/70, Supplementary Fig. 10b), indicating a more prominent immunosuppressive phenotype. Overall, we observed a higher presence of immune signaling in the Mongolia clusters MGL2 and MGL3. Considering that most patients belonging to these clusters were HBV/HDV infected and younger, this suggests that HDV infection could accelerate disease progression through inflammatory mechanisms. DISCUSSION Our study entails a comprehensive characterization of the molecular profile of HCC in Mongolia, the country with the highest global incidence. Mongolia has many particularities that might play a role in HCC burden, including specific risk factors, socioeconomic particularities, and genetic profiles [12,36]. Despite a strikingly high prevalence of HBV (10.6%), HCV (6.4%), and HDV (70% of HBV-positive individuals) infections and alcohol consumption [3–5], it is unclear whether HCC incidence is completely explained by the unique combination of risk factors, or eventually, other unknown factors might be responsible. In addition, a direct comparison with a comprehensive Western cohort is necessary to understand the relevance of newly-identified molecular traits. RESULTS 89 By analyzing WES and RNA-seq data from 379 new Mongolian and Western HCC samples, we identified unique genomic and transcriptomic footprints in Mongolian tumors that suggest a role of specific genetic and environmental factors in the country. The study provides novel information in three major areas: a) Clinical characteristics of Mongolian cases, b) High tumor mutational burden and mutational profile associated with environmental agents, and c) Unique transcriptomic-based molecular classes. Regarding the clinico-pathological particularities of Mongolian HCC patients, we confirmed the high prevalence in females (up to 46% of the cohort), consistent with the reported male/female 1.5/1 ratio [21], as opposed to that observed globally [2] and in the surrounding countries (2.6/1 in Russia, 3.4/1 in China, and 3/1 in East Asia) [21]. Mongolian HCC also occurred in younger patients with milder underlying liver fibrosis (F1-2 stages in >60% of cases) and with a dominant viral-related etiology (85% of either HBV, HBV-HDV, or HCV-positive). In our study, HBV characteristics such as genotype D1 and precore mutations in < 30% of cases revealed traits associated with a low oncogenic potential of the virus in Mongolia [23,25], whereas HBV load was similar to Western samples. Thus, other factors not associated with HBV infection might be responsible for the high HCC incidence rates in Mongolia [23,25]. In this sense, 84% of the HBV-infected Mongolian patients showed co-infection with HDV, which contrasts with the Western data (less than 7% of coinfection) [37]. This unique co-infection profile was associated with two molecular subclasses only identified in Mongolian patients (MGL2 and MGL3), thus pointing towards a potential role of HDV in the oncogenic process, despite it is not currently considered a carcinogenic agent [38]. From the genomic standpoint, Mongolian HCCs had a high rate of protein-coding mutations, which almost doubled that in the Western in-house cohort (121 vs 70 mutations per tumor) and publicly available datasets [8,12–14]. This suggests the presence of intrinsic and/or extrinsic factors promoting mutagenesis in Mongolian HCC. Notably, while several HCC driver genes presented similar mutation rates in the Mongolian cohort compared to the Western tumors and previous studies [8,13,14] (e.g., CTNNB1 and ARID1A), others such as APOB, and the KMT2 gene family were significantly more mutated in Mongolian HCC (e.g., APOB in 15.2% vs 4.5% in Western). For instance, KMT2 family mutations (34% vs 17% in Western) were associated with higher TMB in Mongolian HCC, suggesting that they could be partially responsible for the increased mutational burden. In line with this, loss of function in KMT2 methyltransferases has been proposed to induce DNA damage due to aberrant chromatin remodeling [39]. Finally, TSC2 mutations (9%) were identified as potential drivers in Mongolian tumors. TSC2, a known cancer- RESULTS 90 related gene [8,40] participating in the mTOR oncogenic pathway, has been proposed as an actionable alteration with level 2B evidence, as it could be a predictor of response to the FDA- approved drug everolimus [40]. To understand whether Mongolian HCCs present differential genomic and genotoxic footprints, we assessed the presence of distinct mutational signatures [10,35]. We identified a new mutational signature (SBS Mongolia) with no similarities with previously reported COSMIC signatures. SBS Mongolia was significantly enriched in Mongolian HCC (25% vs 4.5%), indicating distinct substitution patterns characterized by increased T>G substitutions. Overall, this mutational landscape points towards specific exposure to environmental factors in Mongolian patients. In this regard, Mongolian HCC samples presenting the novel SBS Mongolia were significantly enriched for the mutational signature associated with exposure to DMS (71.1% vs 26.5%). DMS has been classified as a probable carcinogenic hazard to humans by the International Agency for Research on Cancer [41] (category 2A carcinogen) and its production has been associated with coal combustion. Most of the Mongolian population is currently exposed to coal combustion, as coal is used to fight against the intense cold weather both in urban and rural areas. Half of the 3-million population of Mongolia lives in Ulaanbaatar, an overpopulated capital with dismal environmental conditions [42], whereas the rest is still predominantly nomad and lives in traditional tents or gers, where coal is used both for cooking and heating. This fact has been recognized by international organizations as a major health threat in this country [42]. Considering this, we hypothesize that long-term exposure to DMS from coal combustion could be a risk factor for HCC development in Mongolia. In this regard, the DMS signature was associated with older patients, potentially due to a longer exposure time. The transcriptomic profile of Mongolian tumors was consistent with known HCC features [2]. Nonetheless, we observed two striking differences a) Mongolian HCC was characterized by an enhanced proliferative and immunological signaling, with a proportion of tumors belonging to proliferative/progenitor HCC classes (39%) doubling the one in Western HCC (20%) and previously published studies [43]; and b) Two out of the three identified molecular classes presented distinct molecular features compared to Western HCC, and thus are deemed unique for Mongolian tumors. Effectively, MGL2 (26%) and MGL3 classes (30%) were specific for Mongolian tumors but not for Western HCC and were enriched in HBV/HDV infection. Interestingly, the MGL2 class was associated with both clinical and molecular features of aggressiveness and showed a female:male ratio of 2:1. The presence of this molecular class could be due to the increased HCC incidence in females in this population [21]. Finally, MGL3 presented an inflamed profile, potentially linked to the response to HBV/HDV infection [7]. RESULTS 91 In conclusion, we provided an exhaustive comparison of the genomic and transcriptomic characteristics of Mongolian HCC with an in-house Western cohort. Mongolian HCC is characterized by high mutational rates, a distinct mutational signature profile, and the presence of two unique molecular subclasses. Finally, environmental factors such as DMS need to be further explored as a potential risk factor in this population. REFERENCES 1. Sung H, Ferlay J, Siegel RL, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71:209–49. 2. Llovet JM, Kelley RK, Villanueva A, et al. Hepatocellular carcinoma. Nat Rev Dis Prim. 2021;7:7. 3. Polaris Observatory Collaborators. Global prevalence, treatment, and prevention of hepatitis B virus infection in 2016: a modelling study. Lancet Gastroenterol Hepatol. 2018;3:383–403. 4. Blach S, Zeuzem S, Manns M, et al. Global prevalence and genotype distribution of hepatitis C virus infection in 2015: a modelling study. Lancet Gastroenterol Hepatol. 2017;2:161–76. 5. Demaio AR, Dugee O, de Courten M, et al. Exploring knowledge, attitudes, and practices related to alcohol in Mongolia: a national population-based survey. BMC Public Health. 2013;13:178. 6. Dondog B, Lise M, Dondov O, et al. Hepatitis B and C virus infections in hepatocellular carcinoma and cirrhosis in Mongolia. Eur J Cancer Prev. 2011;20:33–9. 7. Puigvehí M, Moctezuma-Velázquez C, Villanueva A, et al. The oncogenic role of hepatitis delta virus in hepatocellular carcinoma. JHEP Reports. 2019;1:120–30. 8. Schulze K, Imbeaud S, Letouzé E, et al. Exome sequencing of hepatocellular carcinomas identifies new mutational signatures and potential therapeutic targets. Nat Genet. 2015;47:505–11. 9. Zucman-Rossi J, Villanueva A, Nault JC, et al. Genetic Landscape and Biomarkers of Hepatocellular Carcinoma. Gastroenterology. 2015;149:1226–39. 10. Alexandrov LB, Nik-Zainal S, Wedge DC, et al. Signatures of mutational processes in human cancer. Nature. 2013;500:415–21. 11. Letouzé E, Shinde J, Renault V, et al. Mutational signatures reveal the dynamic interplay of risk factors and cellular processes during liver tumorigenesis. Nat Commun. 2017;8:1–13. RESULTS 92 12. Candia J, Bayarsaikhan E, Tandon M, et al. The genomic landscape of Mongolian hepatocellular carcinoma. Nat Commun. 2020;11. 13. Ahn S-M, Jang SJ, Shim JH, et al. Genomic portrait of resectable hepatocellular carcinomas: implications of RB1 and FGF19 aberrations for patient stratification. Hepatology. 2014;60:1972–82. 14. Ally A, Balasundaram M, Carlsen R, et al. Comprehensive and Integrative Genomic Characterization of Hepatocellular Carcinoma. Cell. 2017;169:1327-1341.e23. 15. Rosenthal R, McGranahan N, Herrero J, et al. deconstructSigs: Delineating mutational processes in single tumors distinguishes DNA repair deficiencies and patterns of carcinoma evolution. Genome Biol. 2016;17. 16. Dobin A, Davis CA, Schlesinger F, et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics. 2013;29:15–21. 17. Hartley SW, Mullikin JC. QoRTs: a comprehensive toolset for quality control and data processing of RNA-Seq experiments. BMC Bioinformatics. 2015;16:224. 18. Risso D, Ngai J, Speed TP, et al. Normalization of RNA-seq data using factor analysis of control genes or samples. Nat Biotechnol. 2014;32:896–902. 19. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA- seq data with DESeq2. Genome Biol. 2014;15:550. 20. Yoshihara K, Shahmoradgoli M, Martínez E, et al. Inferring tumour purity and stromal and immune cell admixture from expression data. Nat Commun. 2013;4. 21. Petrick JL, Braunlin M, Laversanne M, et al. International trends in liver cancer incidence, overall and by histologic subtype, 1978-2007. Int J cancer. 2016;139:1534–45. 22. Baatarkhuu O, Gerelchimeg T, Munkh-Orshikh D, et al. Epidemiology, Genotype Distribution, Prognosis, Control, and Management of Viral Hepatitis B, C, D, and Hepatocellular Carcinoma in Mongolia. Euroasian J Hepato-Gastroenterology. 2018;8:57– 62. 23. Wong GL-H, Chan HL-Y, Yiu KK-L, et al. Meta-analysis: The association of hepatitis B virus genotypes and hepatocellular carcinoma. Aliment Pharmacol Ther. 2013;37:517–26. 24. Wei F, Zheng Q, Li M, et al. The association between hepatitis B mutants and hepatocellular carcinoma: A meta-analysis. Medicine (Baltimore). 2017;96:e6835. 25. Sonneveld MJ, Rijckborst V, Zeuzem S, et al. Presence of precore and core promoter mutants limits the probability of response to peginterferon in hepatitis B e antigen-positive chronic hepatitis B. Hepatology. 2012;56:67–75. 26. Chen B-F, Liu C-J, Jow G-M, et al. High prevalence and mapping of pre-S deletion in hepatitis B virus carriers with progressive liver diseases. Gastroenterology. 2006;130:1153–68. RESULTS 93 27. Jardi R, Rodriguez F, Buti M, et al. Role of hepatitis B, C, and D viruses in dual and triple infection: Influence of viral genotypes and hepatitis B precore and basal core promoter mutations on viral replicative interference. Hepatology. 2001;34:404–10. 28. Giersch K, Homs M, Volz T, et al. Both interferon alpha and lambda can reduce all intrahepatic HDV infection markers in HBV/HDV infected humanized mice. Sci Rep. 2017;7. 29. Schaper M, Rodriguez-Frias F, Jardi R, et al. Quantitative longitudinal evaluations of hepatitis delta virus RNA and hepatitis B virus DNA shows a dynamic, complex replicative profile in chronic hepatitis B and D. J Hepatol. 2010;52:658–64. 30. Chen CJ, Yang HI, Su J, et al. Risk of hepatocellular carcinoma across a biological gradient of serum hepatitis B virus DNA Level. J Am Med Assoc. 2006;295:65–73. 31. Totoki Y, Tatsuno K, Covington KR, et al. Trans-ancestry mutational landscape of hepatocellular carcinoma genomes. Nat Genet. 2014;46:1267–73. 32. Martincorena I, Raine KM, Gerstung M, et al. Universal Patterns of Selection in Cancer and Somatic Tissues. Cell. 2017;171:1029-1041.e21. 33. Arnedo-Pac C, Mularoni L, Muiños F, et al. OncodriveCLUSTL: A sequence-based clustering method to identify cancer drivers. Bioinformatics. 2019;35:4788–90. 34. Alexandrov LB, Kim J, Haradhvala NJ, et al. The repertoire of mutational signatures in human cancer. Nature. 2020;578:94–101. 35. Kucab JE, Zou X, Morganella S, et al. A Compendium of Mutational Signatures of Environmental Agents. Cell. 2019;177:821-836.e16. 36. Bai H, Guo X, Zhang D, et al. The Genome of a Mongolian Individual Reveals the Genetic Imprints of Mongolians on Modern Human Populations. Genome Biol Evol. 2014;6:3122– 36. 37. Tsatsralt-Od B, Takahashi M, Nishizawa T, et al. High prevalence of dual or triple infection of hepatitis B, C, and delta viruses among patients with chronic liver disease in Mongolia. J Med Virol. 2005;77:491–9. 38. Rao RC, Dou Y. Hijacked in cancer: The KMT2 (MLL) family of methyltransferases. Vol. 15, Nature Reviews Cancer. Nature Publishing Group; 2015. p. 334–46. 39. Chakravarty D, Gao J, Phillips S, et al. OncoKB: A Precision Oncology Knowledge Base. JCO Precis Oncol. 2017;2017:1–16. 40. International Agency for Research on Cancer. Agents Classified by the IARC Monographs, Volumes 1–127 – IARC Monographs on the Identification of Carcinogenic Hazards to Humans [Internet]. 2018 [cited 2020 Nov 5]. Available from: https://monographs.iarc.fr/agents-classified-by-the-iarc 41. WHO. Air pollution in Mongolia. Bull World Health Organ. 2019;97:79:80. RESULTS 94 42. Chiang DY, Villanueva A, Hoshida Y, et al. Focal gains of VEGFA and molecular classification of hepatocellular carcinoma. Cancer Res. 2008;68:6779–88. RESULTS 95 FIGURES Figure 1. Flow chart of the study and mutational profile in Mongolian HCC. a A total of 192 HCC samples from Mongolia were used in this study. A Western cohort including 187 HCCs was used as internal control. b Mutations per tumor in the Mongolian (n = 151) and Western (n = 112) cohorts. Mongolian NCI (n = 71), European (n = 241), Korean (n = 231) and TCGA (n = 363) external cohorts are shown RESULTS 96 as reference. c Mutations per tumor in Mongolian HCC (n = 151) compared to Western HCC from Europe (n = 69) and USA (n = 43). Y axis was cut at 300 mutations/tumor to facilitate data interpretation. P-value corresponds to Kruskal- Wallis test. Box plots indicate median (middle line), 25th, 75th percentile (box) and 5th and 95th percentile (whiskers). d Percentage of mutated samples for the top 250 most frequently altered genes in the Mongolian and Western cohort. Y axis was cut at 20% to improve readability. RESULTS 97 Figure 2. Mutational landscape of Mongolian and Western HCC. a Mutations present in HCC samples from the Mongolian cohort (n = 151). The frequency of mutations in the Western and Mongolian NCI cohort are indicated for comparison (left). Genes with significant differences between the Mongolian and Western cohorts are highlighted in green (Fisher p < 0.05). b Mutations present in HCC samples from the Western cohort (n = 112), sorted by sample origin. Overall frequency of mutations and frequencies in Western samples from Europe (n = 69) and USA (n = 43) are shown (left). No significant differences between samples from Europe and USA were found (Fisher test). Top panel shows tumor mutational burden (TMB, mutations/Mb) per sample. Middle panel indicates the presence of mutations per sample (right) and overall percentage (left) in the most frequently mutated genes. Bottom panel details clinico-pathological parameters. Figure 2 b a Mongolian cohort (n = 151) 0 5 10 15 TMB Etiology Gender Age >= 60 Fibrosis stage 02040 % Mutations TP53 CTNNB1 ALB ARID1A APOB AXIN1 TSC2 ATM RB1 KEAP1 ARID2 CDKN2A MAP1B ACVR2A PIK3CA AHCTF1 PTEN Alterations Missense variant Splice acceptor variant Frameshift variant Stop gained Splice donor variant Inframe deletion Etiology HBV HBV/HCV/HDV HBV/HDV HCV Non−infec Gender Female Male Positive Yes No NA M on go lia n NC I M on go lia n 29.6% 22.5% 14.1% 2.8% 12.7% 4.2% 7% 2.8% 9.9% 1.4% 4.2% 0% 7% 1.4% 2.8% 2.8% 4.2% W es te rn 32.1% 42.9% 11.6% 9.8% 4.5% 7.1% 0.9% 6.2% 4.5% 5.4% 3.6% 1.8% 0% 3.6% 3.6% 0% 1.8% 46.4% 36.4% 17.2% 17.2% 15.2% 11.3% 9.3% 8.6% 7.3% 6.6% 6% 4.6% 4.6% 3.3% 2.6% 2.6% 2% Wester cohort (n = 112) 0 5 10 15 TMB Etiology Gender Age >= 60 Fibrosis stage Origin 02040 % Mutations TP53 CTNNB1 ALB ARID1A APOB AXIN1 TSC2 ATM RB1 KEAP1 ARID2 CDKN2A MAP1B ACVR2A PIK3CA AHCTF1 PTEN 32.1% 42.9% 11.6% 9.8% 4.5% 7.1% 0.9% 6.2% 4.5% 5.4% 3.6% 1.8% 0% 3.6% 3.6% 0% 1.8% 26.1% 46.4% 13% 10.1% 1.4% 4.3% 0% 5.8% 4.3% 5.8% 2.9% 1.4% 0% 2.9% 5.8% 0% 0% 41.9% 37.2% 9.3% 9.3% 9.3% 11.6% 2.3% 7% 4.7% 4.7% 4.7% 2.3% 0% 4.7% 0% 0% 4.7% Alterations Missense variant Splice acceptor variant Frameshift variant Stop gained Splice donor variant Inframe deletion Etiology HBV HBV/HDV HCV Non−infec Gender Female Male Origin Europe USA Positive Yes No NA W es te rn E ur op e W es te rn W es te rn U SA n RESULTS 98 Figure 3. Potential driver and KMT2 family mutations in Mongolian and Western cohorts. a Mutations in the KMT2 gene family in the Mongolian (n = 151) and Western cohorts (n = 112). b TMB in samples with KMT2 family mutations and wild type (wt). Box plots indicate median (middle line), 25th, 75th percentile (box) and 5th and 95th percentile (whiskers). P-value corresponds to Kruskal-Wallis test. c Mutations in potential driver genes in the Mongolian and Western cohorts according to enrichment in damaging alterations. Percentage of mutations in external cohorts is indicated in the right panel. Top panel shows tumor mutational burden (TMB, mutations/Mb) per sample. Middle panel indicates the presence of mutations per sample (left) and overall percentage (right) in the Mongolian and Western cohorts. Bottom panel details clinico-pathological characteristics. Significant differences between the Mongolian and Western cohorts are indicated in green (Fisher p < 0.05). d Mutation distribution in the TSC2 gene. 29.6% 7.0% 1.4% 2.8% 7.0% 1.4% 27.8% 22.4% 31.2% 3.3% 4.6% 3.0% 2.8% 2.5% 0% 2.5% 6.2% 2.6% 2.5% 5.8% 5.2% 0.8% 0% 1.7% Mongolian cohort Western cohort 0 5 10 15 TMB Etiology Gender Age >= 60 Fibrosis stage Origin KMT2A KMT2B KMT2C KMT2D KMT2E SETD1A SETD1B Alterations Missense variant Splice acceptor variant Frameshift variant Stop gained Splice donor variant Inframe deletion Etiology HBV HBV/HCV/HDV HBV/HDV HCV Non-infec Gender Female Male Origin Europe USA Positive Yes No NA 46.4% 32.2% 9.3% 0.9% 6.6% 0.9% 6.0% 0.9% 4.6% 0% 4.0% 0% Eu ro pe an Ko re an TC GA M on go lia n NC I W es te rn M on go lia n b c d a W es te rn M on go lia n M on go lia n NC I Mongolian cohort West rn coh rt 0 5 10 15 TMB Etiology Gender Age >= 60 Fibrosis stage Origin TP53 TSC2 NOTCH3 NFE2L2 MAP1B FANCD2 Alterations Missense variant Splice acceptor variant Frameshift variant Stop gained Splice donor variant Inframe deletion Etiology HBV HBV/HCV/HDV HBV/HDV HCV Non-infec Gender Female Male Origin Europe USA Positive Yes No NA 10.6% 2.7% 7.3% 4.5% 8.6% 2.7% 10.6% 6.2% 1.3% 0.9% 0.7% 0.9% 2.6% 0.9% 5.6% 5.6% 5.6% 4.2% 1.4% 1.4% 7.0% Mongolian cohort Western cohort 0 5 10 15 TMB Etiology Gender Age >= 60 Fibrosis stage Origin TP53 TSC2 NOTCH3 NFE2L2 MAP1B FANCD2 Alterations Missense variant Splice acceptor variant Frameshift variant Stop gained Splice donor variant Inframe deletion Etiology HBV HBV/HCV/HDV HBV/HDV HCV Non−infec Gender Female Male Origin Europe USA Positive Yes No NA Mutation rate 0 10 20 30 Mongolian cohort Western cohort 0 5 10 15 TMB Etiology Gender Age >= 60 Fibrosis stage Origin KMT2A KMT2B KMT2C D E SETD1A SETD1B Alterations Missense variant Splice acceptor variant Frameshift variant Stop gained Splice donor variant Inframe deletion Etiology HBV HBV/HCV/HDV HBV/HDV HCV Non-infec Gender Female Male Origin Europe USA Positive Yes No NA Mongolian cohort West rn coh rt 0 5 10 15 TMB Etiology Gender Age >= 60 Fibrosis stage Origin TP53 TSC2 NOTCH3 NFE2L2 MAP1B FANCD2 Alterations Missense variant Splice acceptor variant Frameshift variant Stop gained Splice donor variant Inframe deletion Etiology HBV HBV/HCV/HDV HBV/HDV HCV N n-infec Gender Female Male Origin Europe USA Positive Yes No NA RESULTS 99 Figure 4. Single-base substitution signatures in Mongolian and Western HCC. a De novo single-base substitution (SBS) signatures found in Mongolian (n = 151) and Western HCC (n = 112) samples. For each signature, the 96-substitution classification is displayed, including the substitution type and sequence context. b Distribution of samples harboring SBSM in the Mongolian cohort. Presence of dimethyl sulfate (DMS) mutational signature, main clinical variables and T>G substitution frequency are shown for each sample. FDR-adjusted p-values comparing SBSM positive and negative samples are indicated. Percentages in the lower panel refer to median T>G frequency. c-d Signature fitting results in Mongolian (c) and Western (d) HCC using HCC-specific COSMIC signatures and SBS Mongolia (SBSM). d 0.0 0.3 0.6 0.9 1.2 JM L2 70 −M GL 43 JM L1 39 −M GL 11 1 JM L2 83 −M GL 15 9 JM L1 38 −M GL 10 9 JM L2 34 −M GL 30 3 JM L1 35 −M GL 99 JM L1 28 −M GL 79 JM L1 29 −M GL 85 JM L1 50 −M GL 18 3 JM L2 12 −M GL 08 9 JM L2 47 −M GL 35 5 JM L2 50 −M GL 36 1 JM L2 95 −M GL 31 7 JM L2 03 −M GL 02 9 JM L2 44 −M GL 34 3 JM L2 01 −M GL 00 5 JM L2 98 −M GL 36 5 JM L2 39 −M GL 32 1 JM L2 07 −M GL 06 1 JM L2 15 −M GL 13 7 JM L1 49 −M GL 17 7 JM L2 29 −M GL 27 1 JM L2 66 −M GL 43 7 JM L2 41 −M GL 32 7 JM L2 37 −M GL 31 5 JM L1 47 −M GL 16 3 JM L1 48 −M GL 17 1 JM L1 55 −M GL 20 3 JM L2 96 −M GL 33 7 JM L1 21 −M GL 17 JM L1 51 −M GL 18 5 JM L1 26 −M GL 49 JM L2 08 −M GL 06 3 JM L1 60 −M GL 22 3 JM L2 89 −M GL 22 9 JM L2 49 −M GL 35 9 JM L1 32 −M GL 93 JM L2 90 −M GL 23 3 JM L3 00 −M GL 37 7 JM L2 60 −M GL 41 5 JM L1 30 −M GL 87 JM L1 24 −M GL 41 JM L1 58 −M GL 21 9 JM L1 67 −M GL 24 7 JM L2 05 −M GL 04 7 JM L2 06 −M GL 05 7 JM L2 13 −M GL 10 7 JM L1 41 −M GL 13 3 JM L2 56 −M GL 40 5 JM L1 20 −M GL 15 JM L3 04 −M GL 43 5 JM L1 18 −M GL 9 JM L2 42 −M GL 33 1 JM L2 81 −M GL 13 1 JM L2 62 −M GL 42 1 JM L2 88 −M GL 21 1 JM L2 18 −M GL 14 5 JM L2 25 −M GL 25 1 JM L2 48 −M GL 35 7 JM L2 93 −M GL 28 3 JM L2 10 −M GL 06 9 JM L2 26 −M GL 25 3 JM L2 16 −M GL 14 1 JM L1 64 −M GL 23 9 JM L2 51 −M GL 36 7 JM L2 45 −M GL 34 5 JM L1 22 −M GL 33 JM L2 54 −M GL 39 7 JM L2 94 −M GL 28 5 JM L2 30 −M GL 27 5 JM L2 61 −M GL 41 9 JM L1 17 −M GL 3 JM L1 63 −M GL 23 1 JM L2 63 −M GL 42 7 JM L1 27 −M GL 65 JM L1 31 −M GL 91 JM L2 79 −M GL 12 1 JM L1 61 −M GL 22 5 JM L2 31 −M GL 29 5 JM L1 25 −M GL 45 JM L2 58 −M GL 41 1 JM L1 46 −M GL 16 1 JM L2 92 −M GL 28 1 JM L2 76 −M GL 11 3 JM L3 01 −M GL 38 1 JM L2 64 −M GL 42 9 JM L2 02 −M GL 02 5 JM L2 71 −M GL 51 JM L2 84 −M GL 16 5 JM L2 33 −M GL 30 1 JM L2 67 −M GL 19 JM L2 73 −M GL 71 JM L2 82 −M GL 15 5 JM L1 23 −M GL 39 JM L2 77 −M GL 11 5 JM L1 56 −M GL 20 7 JM L2 04 −M GL 03 7 JM L2 86 −M GL 18 9 JM L2 75 −M GL 75 JM L2 14 −M GL 12 3 JM L2 38 −M GL 31 9 JM L2 36 −M GL 31 1 JM L1 19 −M GL 13 JM L2 40 −M GL 32 3 JM L2 78 −M GL 11 7 JM L1 44 −M GL 14 9 JM L2 87 −M GL 19 3 JM L2 97 −M GL 34 9 JM L1 40 −M GL 12 9 JM L2 65 −M GL 43 1 JM L2 21 −M GL 19 1 JM L2 43 −M GL 33 3 JM L2 11 −M GL 07 7 JM L1 42 −M GL 13 5 JM L2 91 −M GL 27 9 JM L2 55 −M GL 40 1 JM L2 69 −M GL 35 JM L2 20 −M GL 17 9 JM L1 43 −M GL 13 9 JM L1 62 −M GL 22 7 JM L1 57 −M GL 21 7 JM L1 52 −M GL 18 7 JM L1 53 −M GL 19 5 JM L3 03 −M GL 41 7 JM L2 09 −M GL 06 7 JM L3 02 −M GL 38 3 JM L1 65 −M GL 24 1 JM L1 33 −M GL 95 JM L2 27 −M GL 25 5 JM L2 00 −M GL 00 1 JM L2 24 −M GL 23 5 JM L2 53 −M GL 39 5 JM L1 54 −M GL 19 7 JM L2 23 −M GL 21 3 JM L2 57 −M GL 40 7 JM L1 45 −M GL 15 1 JM L2 28 −M GL 25 7 JM L2 85 −M GL 17 5 JM L1 34 −M GL 97 JM L2 17 −M GL 14 3 JM L2 46 −M GL 35 1 JM L2 59 −M GL 41 3 JM L2 22 −M GL 20 5 JM L2 74 −M GL 73 JM L2 80 −M GL 12 7 JM L1 36 −M GL 10 3 JM L1 59 −M GL 22 1 JM L2 35 −M GL 30 5 0 200 400 600 JM L2 70 −M GL 43 JM L1 39 −M GL 11 1 JM L2 83 −M GL 15 9 JM L1 38 −M GL 10 9 JM L2 34 −M GL 30 3 JM L1 35 −M GL 99 JM L1 28 −M GL 79 JM L1 29 −M GL 85 JM L1 50 −M GL 18 3 JM L2 12 −M GL 08 9 JM L2 47 −M GL 35 5 JM L2 50 −M GL 36 1 JM L2 95 −M GL 31 7 JM L2 03 −M GL 02 9 JM L2 44 −M GL 34 3 JM L2 01 −M GL 00 5 JM L2 98 −M GL 36 5 JM L2 39 −M GL 32 1 JM L2 07 −M GL 06 1 JM L2 15 −M GL 13 7 JM L1 49 −M GL 17 7 JM L2 29 −M GL 27 1 JM L2 66 −M GL 43 7 JM L2 41 −M GL 32 7 JM L2 37 −M GL 31 5 JM L1 47 −M GL 16 3 JM L1 48 −M GL 17 1 JM L1 55 −M GL 20 3 JM L2 96 −M GL 33 7 JM L1 21 −M GL 17 JM L1 51 −M GL 18 5 JM L1 26 −M GL 49 JM L2 08 −M GL 06 3 JM L1 60 −M GL 22 3 JM L2 89 −M GL 22 9 JM L2 49 −M GL 35 9 JM L1 32 −M GL 93 JM L2 90 −M GL 23 3 JM L3 00 −M GL 37 7 JM L2 60 −M GL 41 5 JM L1 30 −M GL 87 JM L1 24 −M GL 41 JM L1 58 −M GL 21 9 JM L1 67 −M GL 24 7 JM L2 05 −M GL 04 7 JM L2 06 −M GL 05 7 JM L2 13 −M GL 10 7 JM L1 41 −M GL 13 3 JM L2 56 −M GL 40 5 JM L1 20 −M GL 15 JM L3 04 −M GL 43 5 JM L1 18 −M GL 9 JM L2 42 −M GL 33 1 JM L2 81 −M GL 13 1 JM L2 62 −M GL 42 1 JM L2 88 −M GL 21 1 JM L2 18 −M GL 14 5 JM L2 25 −M GL 25 1 JM L2 48 −M GL 35 7 JM L2 93 −M GL 28 3 JM L2 10 −M GL 06 9 JM L2 26 −M GL 25 3 JM L2 16 −M GL 14 1 JM L1 64 −M GL 23 9 JM L2 51 −M GL 36 7 JM L2 45 −M GL 34 5 JM L1 22 −M GL 33 JM L2 54 −M GL 39 7 JM L2 94 −M GL 28 5 JM L2 30 −M GL 27 5 JM L2 61 −M GL 41 9 JM L1 17 −M GL 3 JM L1 63 −M GL 23 1 JM L2 63 −M GL 42 7 JM L1 27 −M GL 65 JM L1 31 −M GL 91 JM L2 79 −M GL 12 1 JM L1 61 −M GL 22 5 JM L2 31 −M GL 29 5 JM L1 25 −M GL 45 JM L2 58 −M GL 41 1 JM L1 46 −M GL 16 1 JM L2 92 −M GL 28 1 JM L2 76 −M GL 11 3 JM L3 01 −M GL 38 1 JM L2 64 −M GL 42 9 JM L2 02 −M GL 02 5 JM L2 71 −M GL 51 JM L2 84 −M GL 16 5 JM L2 33 −M GL 30 1 JM L2 67 −M GL 19 JM L2 73 −M GL 71 JM L2 82 −M GL 15 5 JM L1 23 −M GL 39 JM L2 77 −M GL 11 5 JM L1 56 −M GL 20 7 JM L2 04 −M GL 03 7 JM L2 86 −M GL 18 9 JM L2 75 −M GL 75 JM L2 14 −M GL 12 3 JM L2 38 −M GL 31 9 JM L2 36 −M GL 31 1 JM L1 19 −M GL 13 JM L2 40 −M GL 32 3 JM L2 78 −M GL 11 7 JM L1 44 −M GL 14 9 JM L2 87 −M GL 19 3 JM L2 97 −M GL 34 9 JM L1 40 −M GL 12 9 JM L2 65 −M GL 43 1 JM L2 21 −M GL 19 1 JM L2 43 −M GL 33 3 JM L2 11 −M GL 07 7 JM L1 42 −M GL 13 5 JM L2 91 −M GL 27 9 JM L2 55 −M GL 40 1 JM L2 69 −M GL 35 JM L2 20 −M GL 17 9 JM L1 43 −M GL 13 9 JM L1 62 −M GL 22 7 JM L1 57 −M GL 21 7 JM L1 52 −M GL 18 7 JM L1 53 −M GL 19 5 JM L3 03 −M GL 41 7 JM L2 09 −M GL 06 7 JM L3 02 −M GL 38 3 JM L1 65 −M GL 24 1 JM L1 33 −M GL 95 JM L2 27 −M GL 25 5 JM L2 00 −M GL 00 1 JM L2 24 −M GL 23 5 JM L2 53 −M GL 39 5 JM L1 54 −M GL 19 7 JM L2 23 −M GL 21 3 JM L2 57 −M GL 40 7 JM L1 45 −M GL 15 1 JM L2 28 −M GL 25 7 JM L2 85 −M GL 17 5 JM L1 34 −M GL 97 JM L2 17 −M GL 14 3 JM L2 46 −M GL 35 1 JM L2 59 −M GL 41 3 JM L2 22 −M GL 20 5 JM L2 74 −M GL 73 JM L2 80 −M GL 12 7 JM L1 36 −M GL 10 3 JM L1 59 −M GL 22 1 JM L2 35 −M GL 30 5 sample mu tC ou nt SBS1 SBS4 SBS5 SBS6 SBS12 SBS16 SBS18 SBS22 SBS26 SBS29 SBS40 SBSN 0.0 0.3 0.6 0.9 1.2 JML27 0−MG L43 JML13 9−MG L111 JML28 3−MG L159 JML13 8−MG L109 JML23 4−MG L303 JML13 5−MG L99 JML12 8−MG L79 JML12 9−MG L85 JML15 0−MG L183 JML21 2−MG L089 JML24 7−MG L355 JML25 0−MG L361 JML29 5−MG L317 JML20 3−MG L029 JML24 4−MG L343 JML20 1−MG L005 JML29 8−MG L365 JML23 9−MG L321 JML20 7−MG L061 JML21 5−MG L137 JML14 9−MG L177 JML22 9−MG L271 JML26 6−MG L437 JML24 1−MG L327 JML23 7−MG L315 JML14 7−MG L163 JML14 8−MG L171 JML15 5−MG L203 JML29 6−MG L337 JML12 1−MG L17 JML15 1−MG L185 JML12 6−MG L49 JML20 8−MG L063 JML16 0−MG L223 JML28 9−MG L229 JML24 9−MG L359 JML13 2−MG L93 JML29 0−MG L233 JML30 0−MG L377 JML26 0−MG L415 JML13 0−MG L87 JML12 4−MG L41 JML15 8−MG L219 JML16 7−MG L247 JML20 5−MG L047 JML20 6−MG L057 JML21 3−MG L107 JML14 1−MG L133 JML25 6−MG L405 JML12 0−MG L15 JML30 4−MG L435 JML11 8−MG L9 JML24 2−MG L331 JML28 1−MG L131 JML26 2−MG L421 JML28 8−MG L211 JML21 8−MG L145 JML22 5−MG L251 JML24 8−MG L357 JML29 3−MG L283 JML21 0−MG L069 JML22 6−MG L253 JML21 6−MG L141 JML16 4−MG L239 JML25 1−MG L367 JML24 5−MG L345 JML12 2−MG L33 JML25 4−MG L397 JML29 4−MG L285 JML23 0−MG L275 JML26 1−MG L419 JML11 7−MG L3 JML16 3−MG L231 JML26 3−MG L427 JML12 7−MG L65 JML13 1−MG L91 JML27 9−MG L121 JML16 1−MG L225 JML23 1−MG L295 JML12 5−MG L45 JML25 8−MG L411 JML14 6−MG L161 JML29 2−MG L281 JML27 6−MG L113 JML30 1−MG L381 JML26 4−MG L429 JML20 2−MG L025 JML27 1−MG L51 JML28 4−MG L165 JML23 3−MG L301 JML26 7−MG L19 JML27 3−MG L71 JML28 2−MG L155 JML12 3−MG L39 JML27 7−MG L115 JML15 6−MG L207 JML20 4−MG L037 JML28 6−MG L189 JML27 5−MG L75 JML21 4−MG L123 JML23 8−MG L319 JML23 6−MG L311 JML11 9−MG L13 JML24 0−MG L323 JML27 8−MG L117 JML14 4−MG L149 JML28 7−MG L193 JML29 7−MG L349 JML14 0−MG L129 JML26 5−MG L431 JML22 1−MG L191 JML24 3−MG L333 JML21 1−MG L077 JML14 2−MG L135 JML29 1−MG L279 JML25 5−MG L401 JML26 9−MG L35 JML22 0−MG L179 JML14 3−MG L139 JML16 2−MG L227 JML15 7−MG L217 JML15 2−MG L187 JML15 3−MG L195 JML30 3−MG L417 JML20 9−MG L067 JML30 2−MG L383 JML16 5−MG L241 JML13 3−MG L95 JML22 7−MG L255 JML20 0−MG L001 JML22 4−MG L235 JML25 3−MG L395 JML15 4−MG L197 JML22 3−MG L213 JML25 7−MG L407 JML14 5−MG L151 JML22 8−MG L257 JML28 5−MG L175 JML13 4−MG L97 JML21 7−MG L143 JML24 6−MG L351 JML25 9−MG L413 JML22 2−MG L205 JML27 4−MG L73 JML28 0−MG L127 JML13 6−MG L103 JML15 9−MG L221 JML23 5−MG L305 0 200 400 600 JML27 0−MG L43 JML13 9−MG L111 JML28 3−MG L159 JML13 8−MG L109 JML23 4−MG L303 JML13 5−MG L99 JML12 8−MG L79 JML12 9−MG L85 JML15 0−MG L183 JML21 2−MG L089 JML24 7−MG L355 JML25 0−MG L361 JML29 5−MG L317 JML20 3−MG L029 JML24 4−MG L343 JML20 1−MG L005 JML29 8−MG L365 JML23 9−MG L321 JML20 7−MG L061 JML21 5−MG L137 JML14 9−MG L177 JML22 9−MG L271 JML26 6−MG L437 JML24 1−MG L327 JML23 7−MG L315 JML14 7−MG L163 JML14 8−MG L171 JML15 5−MG L203 JML29 6−MG L337 JML12 1−MG L17 JML15 1−MG L185 JML12 6−MG L49 JML20 8−MG L063 JML16 0−MG L223 JML28 9−MG L229 JML24 9−MG L359 JML13 2−MG L93 JML29 0−MG L233 JML30 0−MG L377 JML26 0−MG L415 JML13 0−MG L87 JML12 4−MG L41 JML15 8−MG L219 JML16 7−MG L247 JML20 5−MG L047 JML20 6−MG L057 JML21 3−MG L107 JML14 1−MG L133 JML25 6−MG L405 JML12 0−MG L15 JML30 4−MG L435 JML11 8−MG L9 JML24 2−MG L331 JML28 1−MG L131 JML26 2−MG L421 JML28 8−MG L211 JML21 8−MG L145 JML22 5−MG L251 JML24 8−MG L357 JML29 3−MG L283 JML21 0−MG L069 JML22 6−MG L253 JML21 6−MG L141 JML16 4−MG L239 JML25 1−MG L367 JML24 5−MG L345 JML12 2−MG L33 JML25 4−MG L397 JML29 4−MG L285 JML23 0−MG L275 JML26 1−MG L419 JML11 7−MG L3 JML16 3−MG L231 JML26 3−MG L427 JML12 7−MG L65 JML13 1−MG L91 JML27 9−MG L121 JML16 1−MG L225 JML23 1−MG L295 JML12 5−MG L45 JML25 8−MG L411 JML14 6−MG L161 JML29 2−MG L281 JML27 6−MG L113 JML30 1−MG L381 JML26 4−MG L429 JML20 2−MG L025 JML27 1−MG L51 JML28 4−MG L165 JML23 3−MG L301 JML26 7−MG L19 JML27 3−MG L71 JML28 2−MG L155 JML12 3−MG L39 JML27 7−MG L115 JML15 6−MG L207 JML20 4−MG L037 JML28 6−MG L189 JML27 5−MG L75 JML21 4−MG L123 JML23 8−MG L319 JML23 6−MG L311 JML11 9−MG L13 JML24 0−MG L323 JML27 8−MG L117 JML14 4−MG L149 JML28 7−MG L193 JML29 7−MG L349 JML14 0−MG L129 JML26 5−MG L431 JML22 1−MG L191 JML24 3−MG L333 JML21 1−MG L077 JML14 2−MG L135 JML29 1−MG L279 JML25 5−MG L401 JML26 9−MG L35 JML22 0−MG L179 JML14 3−MG L139 JML16 2−MG L227 JML15 7−MG L217 JML15 2−MG L187 JML15 3−MG L195 JML30 3−MG L417 JML20 9−MG L067 JML30 2−MG L383 JML16 5−MG L241 JML13 3−MG L95 JML22 7−MG L255 JML20 0−MG L001 JML22 4−MG L235 JML25 3−MG L395 JML15 4−MG L197 JML22 3−MG L213 JML25 7−MG L407 JML14 5−MG L151 JML22 8−MG L257 JML28 5−MG L175 JML13 4−MG L97 JML21 7−MG L143 JML24 6−MG L351 JML25 9−MG L413 JML22 2−MG L205 JML27 4−MG L73 JML28 0−MG L127 JML13 6−MG L103 JML15 9−MG L221 JML23 5−MG L305 sample mutCo unt SBS1 SBS4 SBS5 SBS6 SBS12 SBS16 SBS18 SBS22 SBS26 SBS29 SBS40 SBSN20 40 60 SN V co un t Mongolian HCC 0.00 0.25 0.50 0.75 1.00 JML01 1−M37 9 JML02 0−M50 5 JML08 8−NY1 651 JML01 0−M37 3 JML01 7−M47 9 JML03 0−M57 7 JML08 0−NY1 627 JML09 0−NY1 655 JML06 3−M77 3 JML01 3−M41 1 JML01 6−M47 7 JML05 6−M73 3 JML09 3−NY1 661 JML03 7−M62 9 JML02 6−M54 7 JML09 6−NY1 667 JML09 7−NY1 669 JML11 6−NY1 717 JML11 4−NY1 713 JML07 4−NY1 613 JML02 9−M57 3 JML08 2−NY1 631 JML05 8−M74 7 JML07 2−NY1 609 JML06 8−M79 7 JML00 9−M35 7 JML07 7−NY1 619 JML00 4−M24 3 JML10 7−NY1 699 JML07 0−M80 5 JML04 7−M67 7 JML11 2−NY1 709 JML02 5−M53 9 JML03 8−M63 3 JML04 2−M65 7 JML04 0−M64 7 JML03 1−M58 5 JML11 1−NY1 707 JML04 1−M65 1 JML03 4−M60 7 JML04 6−M67 3 JML10 4−NY1 693 JML01 2−M38 5 JML00 6−M25 5 JML04 9−M68 1 JML07 6−NY1 617 JML00 3−M23 3 JML10 6−NY1 697 JML06 7−M79 5 JML09 2−NY1 659 JML03 5−M61 9 JML06 5−M78 7 JML07 3−NY1 611 JML02 2−M52 3 JML01 4−M42 1 JML07 1−M80 7 JML05 4−M71 3 JML05 1−M68 7 JML04 5−M66 9 JML09 9−NY1 673 JML01 5−M43 5 JML01 9−M49 9 JML10 1−NY1 677 JML06 2−M76 3 JML06 4−M78 1 JML01 8−M49 3 JML10 2−NY1 679 JML07 5−NY1 615 JML09 5−NY1 665 JML08 7−NY1 649 JML11 5−NY1 715 JML00 1−M21 3 JML03 6−M62 1 JML00 7−M34 5 JML02 3−M53 3 JML07 9−NY1 623 JML03 3−M59 3 JML02 4−M53 7 JML05 7−M74 5 JML09 1−NY1 657 JML10 9−NY1 703 JML06 6−M78 9 JML08 4−NY1 643 JML10 5−NY1 695 JML05 0−M68 5 JML05 9−M75 1 JML00 2−M22 3 JML04 8−M67 9 JML10 0−NY1 675 JML11 0−NY1 705 JML08 3−NY1 635 JML08 9−NY1 653 JML03 2−M58 7 JML06 9−M79 9 JML05 2−M69 1 JML00 8−M35 3 JML03 9−M63 5 JML05 5−M71 9 JML02 8−M55 1 JML08 6−NY1 647 JML02 7−M54 9 JML09 8−NY1 671 JML10 8−NY1 701 JML08 1−NY1 629 JML06 1−M75 7 JML10 3−NY1 691 JML04 3−M66 1 0 100 200 300 JML01 1−M37 9 JML02 0−M50 5 JML08 8−NY1 651 JML01 0−M37 3 JML01 7−M47 9 JML03 0−M57 7 JML08 0−NY1 627 JML09 0−NY1 655 JML06 3−M77 3 JML01 3−M41 1 JML01 6−M47 7 JML05 6−M73 3 JML09 3−NY1 661 JML03 7−M62 9 JML02 6−M54 7 JML09 6−NY1 667 JML09 7−NY1 669 JML11 6−NY1 717 JML11 4−NY1 713 JML07 4−NY1 613 JML02 9−M57 3 JML08 2−NY1 631 JML05 8−M74 7 JML07 2−NY1 609 JML06 8−M79 7 JML00 9−M35 7 JML07 7−NY1 619 JML00 4−M24 3 JML10 7−NY1 699 JML07 0−M80 5 JML04 7−M67 7 JML11 2−NY1 709 JML02 5−M53 9 JML03 8−M63 3 JML04 2−M65 7 JML04 0−M64 7 JML03 1−M58 5 JML11 1−NY1 707 JML04 1−M65 1 JML03 4−M60 7 JML04 6−M67 3 JML10 4−NY1 693 JML01 2−M38 5 JML00 6−M25 5 JML04 9−M68 1 JML07 6−NY1 617 JML00 3−M23 3 JML10 6−NY1 697 JML06 7−M79 5 JML09 2−NY1 659 JML03 5−M61 9 JML06 5−M78 7 JML07 3−NY1 611 JML02 2−M52 3 JML01 4−M42 1 JML07 1−M80 7 JML05 4−M71 3 JML05 1−M68 7 JML04 5−M66 9 JML09 9−NY1 673 JML01 5−M43 5 JML01 9−M49 9 JML10 1−NY1 677 JML06 2−M76 3 JML06 4−M78 1 JML01 8−M49 3 JML10 2−NY1 679 JML07 5−NY1 615 JML09 5−NY1 665 JML08 7−NY1 649 JML11 5−NY1 715 JML00 1−M21 3 JML03 6−M62 1 JML00 7−M34 5 JML02 3−M53 3 JML07 9−NY1 623 JML03 3−M59 3 JML02 4−M53 7 JML05 7−M74 5 JML09 1−NY1 657 JML10 9−NY1 703 JML06 6−M78 9 JML08 4−NY1 643 JML10 5−NY1 695 JML05 0−M68 5 JML05 9−M75 1 JML00 2−M22 3 JML04 8−M67 9 JML10 0−NY1 675 JML11 0−NY1 705 JML08 3−NY1 635 JML08 9−NY1 653 JML03 2−M58 7 JML06 9−M79 9 JML05 2−M69 1 JML00 8−M35 3 JML03 9−M63 5 JML05 5−M71 9 JML02 8−M55 1 JML08 6−NY1 647 JML02 7−M54 9 JML09 8−NY1 671 JML10 8−NY1 701 JML08 1−NY1 629 JML06 1−M75 7 JML10 3−NY1 691 JML04 3−M66 1 sample mutCo unt SBS1 SBS4 SBS5 SBS6 SBS12 SBS16 SBS18 SBS22 SBS26 SBS29 SBS40 SBSN 0.0 0 0.2 5 0.5 0 0.7 5 1.0 0 JM L0 11 −M 37 9 JM L0 20 −M 50 5 JM L0 88 −N Y1 65 1 JM L0 10 −M 37 3 JM L0 17 −M 47 9 JM L0 30 −M 57 7 JM L0 80 −N Y1 62 7 JM L0 90 −N Y1 65 5 JM L0 63 −M 77 3 JM L0 13 −M 41 1 JM L0 16 −M 47 7 JM L0 56 −M 73 3 JM L0 93 −N Y1 66 1 JM L0 37 −M 62 9 JM L0 26 −M 54 7 JM L0 96 −N Y1 66 7 JM L0 97 −N Y1 66 9 JM L1 16 −N Y1 71 7 JM L1 14 −N Y1 71 3 JM L0 74 −N Y1 61 3 JM L0 29 −M 57 3 JM L0 82 −N Y1 63 1 JM L0 58 −M 74 7 JM L0 72 −N Y1 60 9 JM L0 68 −M 79 7 JM L0 09 −M 35 7 JM L0 77 −N Y1 61 9 JM L0 04 −M 24 3 JM L1 07 −N Y1 69 9 JM L0 70 −M 80 5 JM L0 47 −M 67 7 JM L1 12 −N Y1 70 9 JM L0 25 −M 53 9 JM L0 38 −M 63 3 JM L0 42 −M 65 7 JM L0 40 −M 64 7 JM L0 31 −M 58 5 JM L1 11 −N Y1 70 7 JM L0 41 −M 65 1 JM L0 34 −M 60 7 JM L0 46 −M 67 3 JM L1 04 −N Y1 69 3 JM L0 12 −M 38 5 JM L0 06 −M 25 5 JM L0 49 −M 68 1 JM L0 76 −N Y1 61 7 JM L0 03 −M 23 3 JM L1 06 −N Y1 69 7 JM L0 67 −M 79 5 JM L0 92 −N Y1 65 9 JM L0 35 −M 61 9 JM L0 65 −M 78 7 JM L0 73 −N Y1 61 1 JM L0 22 −M 52 3 JM L0 14 −M 42 1 JM L0 71 −M 80 7 JM L0 54 −M 71 3 JM L0 51 −M 68 7 JM L0 45 −M 66 9 JM L0 99 −N Y1 67 3 JM L0 15 −M 43 5 JM L0 19 −M 49 9 JM L1 01 −N Y1 67 7 JM L0 62 −M 76 3 JM L0 64 −M 78 1 JM L0 18 −M 49 3 JM L1 02 −N Y1 67 9 JM L0 75 −N Y1 61 5 JM L0 95 −N Y1 66 5 JM L0 87 −N Y1 64 9 JM L1 15 −N Y1 71 5 JM L0 01 −M 21 3 JM L0 36 −M 62 1 JM L0 07 −M 34 5 JM L0 23 −M 53 3 JM L0 79 −N Y1 62 3 JM L0 33 −M 59 3 JM L0 24 −M 53 7 JM L0 57 −M 74 5 JM L0 91 −N Y1 65 7 JM L1 09 −N Y1 70 3 JM L0 66 −M 78 9 JM L0 84 −N Y1 64 3 JM L1 05 −N Y1 69 5 JM L0 50 −M 68 5 JM L0 59 −M 75 1 JM L0 02 −M 22 3 JM L0 48 −M 67 9 JM L1 00 −N Y1 67 5 JM L1 10 −N Y1 70 5 JM L0 83 −N Y1 63 5 JM L0 89 −N Y1 65 3 JM L0 32 −M 58 7 JM L0 69 −M 79 9 JM L0 52 −M 69 1 JM L0 08 −M 35 3 JM L0 39 −M 63 5 JM L0 55 −M 71 9 JM L0 28 −M 55 1 JM L0 86 −N Y1 64 7 JM L0 27 −M 54 9 JM L0 98 −N Y1 67 1 JM L1 08 −N Y1 70 1 JM L0 81 −N Y1 62 9 JM L0 61 −M 75 7 JM L1 03 −N Y1 69 1 JM L0 43 −M 66 1 0 100 200 300 JM L0 11 −M 37 9 JM L0 20 −M 50 5 JM L0 88 −N Y1 65 1 JM L0 10 −M 37 3 JM L0 17 −M 47 9 JM L0 30 −M 57 7 JM L0 80 −N Y1 62 7 JM L0 90 −N Y1 65 5 JM L0 63 −M 77 3 JM L0 13 −M 41 1 JM L0 16 −M 47 7 JM L0 56 −M 73 3 JM L0 93 −N Y1 66 1 JM L0 37 −M 62 9 JM L0 26 −M 54 7 JM L0 96 −N Y1 66 7 JM L0 97 −N Y1 66 9 JM L1 16 −N Y1 71 7 JM L1 14 −N Y1 71 3 JM L0 74 −N Y1 61 3 JM L0 29 −M 57 3 JM L0 82 −N Y1 63 1 JM L0 58 −M 74 7 JM L0 72 −N Y1 60 9 JM L0 68 −M 79 7 JM L0 09 −M 35 7 JM L0 77 −N Y1 61 9 JM L0 04 −M 24 3 JM L1 07 −N Y1 69 9 JM L0 70 −M 80 5 JM L0 47 −M 67 7 JM L1 12 −N Y1 70 9 JM L0 25 −M 53 9 JM L0 38 −M 63 3 JM L0 42 −M 65 7 JM L0 40 −M 64 7 JM L0 31 −M 58 5 JM L1 11 −N Y1 70 7 JM L0 41 −M 65 1 JM L0 34 −M 60 7 JM L0 46 −M 67 3 JM L1 04 −N Y1 69 3 JM L0 12 −M 38 5 JM L0 06 −M 25 5 JM L0 49 −M 68 1 JM L0 76 −N Y1 61 7 JM L0 03 −M 23 3 JM L1 06 −N Y1 69 7 JM L0 67 −M 79 5 JM L0 92 −N Y1 65 9 JM L0 35 −M 61 9 JM L0 65 −M 78 7 JM L0 73 −N Y1 61 1 JM L0 22 −M 52 3 JM L0 14 −M 42 1 JM L0 71 −M 80 7 JM L0 54 −M 71 3 JM L0 51 −M 68 7 JM L0 45 −M 66 9 JM L0 99 −N Y1 67 3 JM L0 15 −M 43 5 JM L0 19 −M 49 9 JM L1 01 −N Y1 67 7 JM L0 62 −M 76 3 JM L0 64 −M 78 1 JM L0 18 −M 49 3 JM L1 02 −N Y1 67 9 JM L0 75 −N Y1 61 5 JM L0 95 −N Y1 66 5 JM L0 87 −N Y1 64 9 JM L1 15 −N Y1 71 5 JM L0 01 −M 21 3 JM L0 36 −M 62 1 JM L0 07 −M 34 5 JM L0 23 −M 53 3 JM L0 79 −N Y1 62 3 JM L0 33 −M 59 3 JM L0 24 −M 53 7 JM L0 57 −M 74 5 JM L0 91 −N Y1 65 7 JM L1 09 −N Y1 70 3 JM L0 66 −M 78 9 JM L0 84 −N Y1 64 3 JM L1 05 −N Y1 69 5 JM L0 50 −M 68 5 JM L0 59 −M 75 1 JM L0 02 −M 22 3 JM L0 48 −M 67 9 JM L1 00 −N Y1 67 5 JM L1 10 −N Y1 70 5 JM L0 83 −N Y1 63 5 JM L0 89 −N Y1 65 3 JM L0 32 −M 58 7 JM L0 69 −M 79 9 JM L0 52 −M 69 1 JM L0 08 −M 35 3 JM L0 39 −M 63 5 JM L0 55 −M 71 9 JM L0 28 −M 55 1 JM L0 86 −N Y1 64 7 JM L0 27 −M 54 9 JM L0 98 −N Y1 67 1 JM L1 08 −N Y1 70 1 JM L0 81 −N Y1 62 9 JM L0 61 −M 75 7 JM L1 03 −N Y1 69 1 JM L0 43 −M 66 1 sample mu tC ou nt SBS1 SBS4 SBS5 SBS6 SBS12 SBS16 SBS18 SBS22 SBS26 SBS29 SBS40 SBSN 0 10 200 300 SN V co un t Western HCC c C>A C>G C>T T>A T>C T>G denovoSig_1 denovoSig_2 denovoSig_3 denovoSig_4 A. A A. C A. G A. T C. A C. C C. G C. T G. A G. C G. G G. T T.A T.C T.G T. T A. A A. C A. G A. T C. A C. C C. G C. T G. A G. C G. G G. T T.A T.C T.G T. T A. A A. C A. G A. T C. A C. C C. G C. T G. A G. C G. G G. T T.A T.C T.G T. T A. A A. C A. G A. T C. A C. C C. G C. T G. A G. C G. G G. T T.A T.C T.G T. T A. A A. C A. G A. T C. A C. C C. G C. T G. A G. C G. G G. T T.A T.C T.G T. T A. A A. C A. G A. T C. A C. C C. G C. T G. A G. C G. G G. T T.A T.C T.G T. T 0.0 0.1 0.2 0.0 0.1 0.2 0.0 0.1 0.2 0.0 0.1 0.2 context Re lat ive co nt rib ut ion SBS22 SBS6 + SBS40 SBSM (De novo signature 4) SBS16 + SBS26 Mongolian HCC (n=151) SBS Mongolia DMS signature Etiology Gender Age >= 60 Fibrosis stage 0 0.1 0.2T>G frequency Signature Absent Present Etiology HBV HBV/HCV/HDV HBV/HDV HCV Non−infec Gender Female Male Positive Yes No NA p value 0.0001 0.610 0.686 0.738 0.950 <0.000114% 8% a b RESULTS 100 Figure 5. Molecular classification of Mongolian HCC. Consensus-clustered classification of Mongolian HCC samples using Non-negative matrix factorization. In the heatmap, clinico-pathological characteristics, nearest template prediction and gene set enrichment in each sample are shown. MGL1, 45% MGL2, 25% MGL3, 30% MONGOLIAN COHORT (n=106) NMFc-based MGL clusters Etiology Age Gender Fibrosis F3-4 Chiang 5 cls Hoshida 3 cls MO classification G3 signature Cluster A signature TGFB late signature MET signature NOTCH signature HCC Immune class TMB high CTNNB1 mut TP53 mut RB1 mut RB1 signature Broad chromosomal alterations 1111111111111111111111111111111111111111111111122222222222222222222222222233333333333333333333333333333333 > = > = < 6 < 6 < 6 < 6 > = > = > = > = > = > = > = > = < 6 > = > = > = < 6 > = < 6 < 6 < 6 < 6 > = > = < 6 > = < 6 > = > = > = > = > = > = > = > = > = < 6 > = < 6 < 6 < 6 < 6 > = < 6 > = > = < 6 < 6 < 6 < 6 < 6 < 6 < 6 < 6 > = < 6 > = < 6 < 6 < 6 < 6 < 6 < 6 < 6 > = > = < 6 < 6 < 6 < 6 > = < 6 < 6 < 6 > = > = < 6 > = < 6 < 6 > = > = > = < 6 < 6 < 6 > = < 6 < 6 < 6 > = < 6 < 6 < 6 > = < 6 > = < 6 < 6 < 6 # N # N # N # N # N # N # N # N # N # N # N # N # N # ¡ # ¡ # ¡ N/A N/A MGL1 HBV HCV <60 2 Male CTNNB1 Poly 7 S1 MO1 MO4 MGL2 HBV/HDV Non-infected > 60 1 Female Proliferation 5 Unan. S2 MO2 MGL3 HBV/HCV/HDV IFN S3 MO3 Hoshida classes MO classificationViral etiologyMGL clusters Age Gender Chiang 5 classes Immune Significant Present Rest Non-significant Absent N/A Immune class Gene signature Other features Chromosomal alterations max min RESULTS 101 Figure 6. Inflammatory profile of Mongolian HCC. Characterization of inflammatory profile in the MGL clusters assessed by ESTIMATE analysis and single sample gene set enrichment analyses capturing distinct immune populations. Th1, Type 1 helper; Th2, type 2 helper; TFH, T follicular helper; Treg, regulatory T; Tem, effector memory T, Tcm central memory T. Figure 6 Gene expression Low High MGL1 MGL2 MGL3 NMFc-based MGL clusters MO classification HCC Immune class Immune enrichment score Stroma enrichment score PD1 signaling Exhaustion signature B cells CD8 T cells Cytotoxic T cells T helper cells Th1 cells Th2 cells TFH cells Treg cells Tem cells Tcm cells iDC cells Macrophages NK cells Eosinophils Neutrophils Mast cells Im m une signatures Adaptive im m une response Innate im m une response MGL1 MO1 MO4 Immune Significant MGL2 MO2 Rest Non-significant MGL3 MO3 MO classification Immune class Gene signatureMGL clusters RESULTS 102 TABLES Table 1. Baseline characteristics of the Mongolian and Western cohorts Mongolian Cohort (n = 192) Western Cohort (n = 187) p value Age (years) < 60 years (n, %) < 50 years (n, %) 61 (18-80) 82 (44.6) 20 (10.9) 66 (29-91) 32 (18.6) 7 (4.1) <0.001 <0.001 0.017 Gender (male, %) 98 (53.6) 137 (79.7) <0.001 Etiology HBV+ (n, %) HBV/HDV+ (n, %) HBV/HCV/HDV+ (n, %) HCV+ (n, %) HBV/HCV (n, %) Non-infected (n, %) 15 (7.8) 77 (40.1) 12 (6.3) 57 (29.7) 2 (1) 29 (15) 41 (21.9) 3 (1.6) 0 (0) 69 (36.9) 0 (0) 74 (39.6) <0.001 Bilirubin (mg/dL) 0.6 (0.1-3.7) 0.9 (0.3-3.8) <0.001 Albumin (g/L) 41 (29-49) 40 (22-54) ns Platelets (109/L) < 150x109/L (n, %) 181 (76-574) 50 (28.6) 160 (27-493) 81 (47.4) <0.001 <0.001 AFP (IU/mL) > 400 IU/mL (n, %) 22 (1-121000) 32 (20.9) 12 (1-311190) 26 (16.4) ns ns Tumor size > 5 cm (n, %) 5 (0.8-20) 93 (52.8) 4.2 (1-20) 67 (41.4) 0.001 0.039 BCLC stage (0-A, %) 132 (78.1) 129 (79.6) ns Multinodular disease (n, %) 26 (15.4) 42 (25.8) 0.021 Advanced liver fibrosis (F3-4, %) Cirrhosis (F4, %) 64 (38.1) 27 (16.1) 106 (78.5) 81 (60) <0.001 <0.001 Microvascular invasion (yes, %)# 72 (47.7) 80 (46.5) ns Tumor grade (G3-4, %) 11 (10.6) 41 (28.7) <0.001 HBV, hepatitis B virus; HCV, hepatitis C virus; HDV, hepatitis delta virus; AFP, alfa- fetoprotein; BCLC, Barcelona Clinic Liver Cancer The following variables have missing values for the Mongolian and Western cohorts, respectively: Age: 8 and 15 patients. Gender: 9 and 15 patients. Bilirubin, albumin, and platelets: 19 and 17 patients. AFP: 39 and 28 patients. Tumor size: 16 and 25 patients. BCLC stage: 23 and 25 patients. Tumor number: 23 and 24 patients. Liver fibrosis in 24 and 52 patients. Microvascular invasion in 41 and 15 patients. Tumor grade in 88 and 44 patients. RESULTS 103 Financial support This study was partially supported by Bristol-Myers Squibb. MP received a scholarship grant from Asociación Española para el Estudio del Hígado (AEEH). PKH is the recipient of a grant from the German Research Foundation (DFG, HA 8754/1-1). MGL is supported by the i-PFIS program (fellowship IFI18/00006) of the ISCIII, co-funded by the European Social Fund (ESF). AV is supported by the US Department of Defense grant (CA150272P3) and the Tisch Cancer Institute (Cancer Center grant P30 CA196521). SPP is supported by the Instituto de Salud Carlos III (ISCIII) through the Plan Estatal de Investigación Científica y Técnica y de Innovación 2013-2016 and 2017-2020 co-funded by the European Regional Development Fund (ERDF) (PI16/00111 and PI19/00036). JZR’s team is supported by Inserm, Labex OncoImmunology Investissement d’Avenir and is “Equipe labellisée par la Ligue Nationale Contre le Cancer». XF is supported by the Instituto de Salud Carlos III (ISCIII) through the Plan Estatal de Investigación Científica y Técnica y de Innovación 2013-2016 and 2017-2020 co-funded by the European Regional Development Fund (ERDF) (PI15/00151 and PI18/00079), by the Secretaria d’Universitats i Recerca del Departament d’Economia i Coneixement (grant 2017_SGR_1753) and by CERCA Programme/Generalitat de Catalunya. DS is supported by the Gilead Sciences Research Scholar Program in Liver Disease. JML is supported by grants from the Samuel Waxman Cancer Research Foundation, the Spanish National Health Institute (MICINN, PID2019-105378RB-I00), NIH (R01 DK128289-01), HUNTER (Ref. C9380/A26813) through a partnership between Cancer Research UK, Fondazione AIRC and Fundación Científica de la Asociacion Española Contra el Cáncer and by the Generalitat de Catalunya (AGAUR, SGR-1358). Conflict of interest HW, ARV and PT are employed by Sema4. AV has received consulting fees from Guidepoint, Fujifilm, Boehringer Ingelheim, FirstWord, and MHLife Sciences; advisory board fees from Exact Sciences, Nucleix, Natera, Gilead and NGM Pharmaceuticals; and research support from Eisai. AU reports employment and stock ownership from Sema4. JN is employed by Bristol-Myers Squibb. JML is receiving research support from Bayer HealthCare Pharmaceuticals, Eisai Inc, Bristol-Myers Squibb, Boehringer-Ingelheim and Ipsen, and consulting fees from Eli Lilly, Bayer HealthCare Pharmaceuticals, Bristol-Myers Squibb, Eisai Inc, Celsion Corporation, Exelixis, Merck, Ipsen, Genentech, Roche, Glycotest, Nucleix, Sirtex, Mina Alpha Ltd and AstraZeneca. The remaining authors declare no competing interests. RESULTS 104 Author contributions L.T. and M.P. contributed equally to this work. M.P., D.S., and J.M.L. developed the study concept and design. L.T., M.P., M.T.M., H.W., M.M.., T.L., M.G.L., W.Q.L., C.M., S.T., A.R., P.T., C.E., E.T., A.Y., G.C., S.P.P., S.T., and D.S. acquired the experimental data. L.T., M.P., M.T.M., H.W., M.M., P.K.H., T.L., M.G.L., W.Q.L., C.M., S.T., A.R.V., P.T., G.C., S.P.P., A.V., E.L., J.Z.R., A.U., J.N., X.F., S.R., D.S., and J.M.L. conducted the analysis and interpretation of the data. L.T., M.P., D.S., M.T.M., H.W., and A.U. performed statistical analyses. L.T., M.P., M.T.M., P.K.H., D.S., and J.M.L. drafted the manuscript. L.T., M.P., M.T.M., P.K.H., S.P.P., A.V., S.T., J.C., E.L., J.Z.R., A.U., J.N., X.F., S.R., D.S., and J.M.L. provided critical revision of the manucript. D.S. and J.M.L. supervised the study. Acknowledgmenents We thank Clara Rossi, Fellow at Mount Sinai, for her help. RESULTS 105 Study 2 – Liver Injury Increases the Incidence of HCC following AAV Gene Therapy in Mice Dhwanil A Dalwadi, Laura Torrens, Jordi Abril-Fornaguera, Roser Pinyol, Catherine Willoughby, Jeffrey Posey, Josep M Llovet, Christian Lanciault, David W Russell, Markus Grompe, Willscott E Naugler Molecular Therapy. 2021 Feb 3;29:680–690. Epub 2020 Oct 22 (IF: 11.454) Summary Recombinant AAV is a widely used platform for gene replacement, silencing, and editing which holds great promise for the implementation of gene therapies in the clinical setting. In the past few years, two AAV-based gene therapies have been approved by regulatory agencies. Furthermore, ~130 active clinical trials are currently ongoing and could result in groundbreaking therapeutic advances for a great variety of medical conditions4,105. However, compelling studies provide evidence that AAV infection could induce hepatocarcinogenesis due to integration into oncogenic genomic sites, which poses serious safety concerns105,106. In this regard, AAV2 insertions have been identified in a small set of HCC patients (2-5%)4,60 and murine models, often occurring in the Rian locus, a cluster of oncogenic microRNA located at the murine chromosome 12106,107,158, which is analogous to the human DLK1-DIO3 locus in chromosome 14. Furthermore, overexpression of the DLK1-DIO3 locus has been identified in a subclass of 6-19% HCC tumors, associated with an aggressive phenotype and poor prognosis108,109. Common causes of chronic liver disease such as NAFLD could potentially favor AAV integration in the genome due to increased liver damage, inflammation, and regenerative proliferation of hepatocytes110. Considering the high prevalence of NAFLD, affecting up to 30% of the U.S. population159, adverse events in these patients could be a concern for the use of AAV gene therapy. Herein we aimed to assess whether NAFLD-associated liver damage increase the risk of AAV integration inducing HCC. To determine whether hepatocyte proliferation impacts AAV-induced oncogenesis, wildtype C57BL/6 mice were infected with an AAV editing vector targeting the Rian locus (AAV-Rian). Specifically, the AAV-Rian vector consisted of a strong promoter flanked by regions homologous to the Rian gene, which allow recombination and promoter insertion in that genomic region158. AAV infection was performed both at neonatal and adult mice. Furthermore, animals were treated either with HFD to induce NAFLD-like liver injury, or with partial hepatectomy to RESULTS 106 promote hepatocyte proliferation. The molecular and inflammatory profiles in HCC and background liver samples from each group were characterized by RNA sequencing. Our results revealed that: 1. AAV targeting the Rian locus (AAV-Rian) led to HCC in all male mice infected as neonates, likely due to growth-related hepatocyte proliferation in young mice. Conversely, only 5% of untreated male mice infected with AAV as adults developed HCC. 2. Liver injury through HFD and partial hepatectomy led to increased HCC development in adult AAV-Rian infected mice compared to untreated animals (100% vs 5%). 3. Transcriptomic analysis of murine tumors from AAV-Rian infected mice revealed remarkable molecular similarities with human HCC tumors overexpressing the Rian locus, associated with proliferation, aggressive phenotype, and poor prognosis. 4. NAFLD mice infected with a non-targeted control AAV developed HCC, though only half as frequently as those exposed to the targeted AAV-Rian (50% vs 100% in adults). These tumors presented increased expression of genes located in the Rian locus as well as a similar gene expression profile to AAV-Rian tumors. This suggests random integrations into the Rian locus, in line with results from previous studies158. 5. Transcriptomic analysis of background liver samples from our murine model showed that HFD elicited a protumorigenic immune cancer field as well as activated lipid metabolism pathways. This could contribute to the AAV-induced in this model. 6. Female mice were less susceptible to develop AAV-Rian-induced HCC compared to males (29% vs 100% female and male neonates, respectively). Furthermore, livers from male mice with NAFLD treated with estrogen exhibited a reduction in aberrant immune exhaustion signaling compared to untreated males. In conclusion, this study shows that AAV gene therapy increase HCC development in murine models of NAFLD, likely due to enhanced hepatocyte proliferation and the generation of a protumorigenic immune cancer field effect in the liver. This study raises concerns about the risks of AAV gene therapy causing HCC, particularly in patients with chronic liver injury such as NAFLD. RESULTS 107 Publication Original Article Liver Injury Increases the Incidence of HCC following AAV Gene Therapy in Mice Dhwanil A. Dalwadi,1,2 Laura Torrens,3 Jordi Abril-Fornaguera,3 Roser Pinyol,3 Catherine Willoughby,3 Jeffrey Posey,2 Josep M. Llovet,3,4,5 Christian Lanciault,6 David W. Russell,7,8 Markus Grompe,2 and Willscott E. Naugler1 1Department of Medicine, Division of Gastroenterology and Hepatology, Oregon Health and Science University, Portland, OR 97239, USA; 2Papé Family, Pediatric Research Institute, Department of Pediatrics, Oregon Health and Science University, Portland, OR 97239, USA; 3Translational Research in Hepatic Oncology, Liver Unit, IDIBAPS-Hospital Clínic, University of Barcelona, Catalonia, Spain; 4Mount Sinai Liver Cancer Program, Divisions of Liver Diseases, Tisch Cancer Institute, Icahn School of Medicine at Mount Sinai, NY, NY, USA; 5Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Catalonia, Spain; 6Department of Pathology, Oregon Health and Science University, Portland, OR 97239, USA; 7Department of Medicine, University of Washington, Seattle, WA 98195, USA; 8Department of Biochemistry, University of Washington, Seattle, WA 98195, USA Adeno-associated virus (AAV) integrates into host genomes at low frequency, but when integration occurs in oncogenic hotspots it can cause hepatocellular carcinoma (HCC). Given the possibility of recombinant AAV (rAAV) integra- tion leading to HCC, common causes of liver inflammation like non-alcoholic fatty liver disease (NAFLD) may increase the risk of rAAV-induced HCC. A rAAV targeting the oncogenic mouse Rian locus was used, and as expected led to HCC in all mice infected as neonates, likely due to growth-related hepatocyte proliferation in young mice. Mice infected with rAAV as adults did not develop HCC unless they were fed a diet leading to NAFLD, with increased inflammation and hepatocyte proliferation. Female mice were less susceptible to rAAV-induced HCC, and male mice with NAFLD treated with estrogen exhibited less inflammation and immune exhaustion associated with oncogenesis compared to those without estrogen. Adult NAFLD mice infected with a non-targeted control rAAV also developed HCC, though only half as frequently as those exposed to the Rian targeted rAAV. This study shows that adult mice exposed to rAAV gene therapy in the context of chronic liver disease developed HCC at high frequency, and thus warrants further study in humans given the high prevalence of NAFLD in the population. INTRODUCTION Recombinant adeno-associated virus (rAAV) is a promising gene therapy tool and is currently widely used for this purpose in the laboratory and increasingly in the clinic.1 Key advantages of rAAV as a gene therapy vector include its apparent lack of path- ogenicity, ability to infect both dividing and quiescent cells, and a very mild immune response, though this is to some extent depen- dent on the dose.2 There are more than a hundred ongoing rAAV clinical trials along with at least two FDA approved therapies in the US. Although rAAV is a promising gene therapy vector, concern remains regarding potential side-effects, notably inser- tional mutagenesis in oncogenes or tumor suppressor genes with resultant carcinogenesis. While random rAAV insertions might result in cancer in any tissue, both human and mouse data suggest that hepatocellular carcinoma (HCC) is the most likely malignancy.3,4 The majority of experimental evidence to date suggests that rAAV gene therapy is safe, but there are some compelling rodent studies that provide evidence of potential genotoxicity of rAAV vectors.5 For example, when neonatal mice were infected with a rAAV gene therapy vector, the mice that received the vector were four to seven times more prone to developing HCC than control mice, and the oncogenic vector integration site was mapped to the Rian locus on chromosome 12.6,7 When a gene-editing rAAV vector was used to insert a strong promoter specifically into the Rian locus, 100% of neonatal mice infected with the vec- tor developed HCC.4 However, to date no evidence shows that adult mice develop HCC when treated with rAAV vectors.8,9 It is hypothesized that proliferating hepatocytes in the neonatal liver are responsible for the high frequency of homologous and non- homologous rAAV integrations. It follows, then, that the absence of proliferating hepatocytes in the healthy adult liver would diminish rAAV integration and resultant oncogenesis. While quiescent under healthy homeostatic conditions, however, hepatocytes readily proliferate in the adult liver in response to injury.10–13 The clearest example of such proliferation occurs after partial hepatectomy, wherein hepatocytes in the remnant liver have a striking response to the injury—marked hepatocyte DNA Received 19 June 2020; accepted 19 October 2020; https://doi.org/10.1016/j.ymthe.2020.10.018. Correspondence: Willscott E. Naugler, Department of Medicine, Division of Gastroenterology and Hepatology, Oregon Health and Science University, Port- land, OR 97239, USA. E-mail: nauglers@ohsu.edu 680 Molecular Therapy Vol. 29 No 2 February 2021 ª 2020 The American Society of Gene and Cell Therapy. RESULTS 108 synthesis and proliferation.10,14 More commonly seen (especially in humans) are conditions of chronic liver inflammation and injury, which are themselves associated with compensatory prolif- eration as the body seeks to repair the damaged liver.11,12 Multi- ple conditions of chronic liver inflammation/injury/proliferation make up the bulk of human liver disease and include entities such as alcoholic hepatitis, hepatitis B and C infections, and various auto-immune conditions. The most common cause of chronic liver inflammation in the US and Western countries is non-alcoholic fatty liver disease (NAFLD), affecting up to 30% of the US population.15 The high prevalence of obesity and NAFLD makes it likely to be present in a significant number of patients who could benefit from gene therapy. In this study we assessed whether hepatocyte replication and liver injury increase HCC formation in adult mice who receive rAAV gene therapy. Given the high prevalence of inflammatory liver conditions such as NAFLD, adverse events in these patients could be a concern for the use of rAAV gene therapy. To test this hy- pothesis, we infected wild-type C57BL/6 mice with a rAAV edit- ing vector targeting the Rian locus by homologous recombina- tion.4 Partial hepatectomy was used to induce hepatocyte proliferation and a high-fat diet (HFD) to induce liver injury.10,16 The results of this study demonstrate that both partial hepatec- tomy and HFD-induced liver injury are sufficient to cause HCC if a strong promoter integration occurs in an oncogenic locus. This study raises concerns about the risks of rAAV gene therapy causing HCC in patients with chronic liver injury, particularly patients suffering from non-alcoholic steatohepatitis (NASH) or other inflammatory liver conditions. RESULTS Neonatal mice infected with AAV-Rian-CMV (AAV-Rian), a gene editing vector that inserts a strong CAG promoter in the Rian locus, all developed HCC as a result of a Rian-specific rAAV integration.4 The development of HCC in neonates was attributed to the hepato- cyte proliferation expected in the neonatal liver. However, there are currently no reports of adult mice developing HCC when infected with rAAV. In this study we report for the first time that rAAV can cause HCC in adult mice under conditions causing hepatocyte prolif- eration. Two models were utilized, a neonatal model where the mice were infected with rAAV followed by initiation of HFD or regular diet (RD) and an adult model wheremice were first started onHFD or RD, followed by rAAV infection. HFD-Induced Chronic Liver Injury Does Not Increase Hepatocarcinogenesis in rAAV-Infected Neonates 1-day-old neonatal mice were injected with 3! 1010 vg of AAV-Rian or AAV-CAG-tdTomato (AAV-tdTomato) (Figure 1) via the tempo- ral vein. At 3 weeks old, they were either started on a HFD or RD, and at 6 months of age the mice were sacrificed and assessed for tumors (Figure 2A). Gross inspection of livers revealed multiple nodules in vector-infected mice (Figure 2D), whereas mice that received vehicle injections (No rAAV, Dulbecco's PBS [DPBS], n = 6, 3 males, 3 females) did not develop tumors regardless of diet (data not shown). Mice that received the tdTomato virus and were on the RD did not develop tumors (n = 6, 3 males, 3 females); however, 20% of males and 25% of females that were on the HFD developed tumors (n = 9, 4 males, 5 females), though this difference was not statistically significant compared to RD by chi-square test (p = 0.54 for females and p = 0.41 for males; Figure 2B). All males that received the Rian vi- rus developed tumors, regardless of diet (n = 21, 14 HFD, 7 RD), whereas only 29% of females developed tumors when on the RD, and 20% developed tumors when on the HFD (Figure 2B, n = 13, 5 HFD, 8 RD). The average number of tumors per mouse in mice that received the tdTomato virus on the HFD was 0.25 for male (n = 4) and 0.2 for females (n = 5; Figure 2B). In mice that received the Rian virus, the tumor burden was higher in males than females with an average of 9 tumors per mouse when on the RD (n = 7), and 10 when on the HFD (n = 14; Figure 2C). Tumor burden in female mice was lower, 0.25 tumors per mouse for mice on RD (n = 8) and 5.4 for mice on HFD (n = 5), however, the variance was such that the difference was not statistically significant (p > 0.05; Figure 2B). Histological analysis showed that the tumors had characteristic fea- tures of HCC such as cytologic atypia with bizarre, enlarged hepato- cytes and wide trabeculae, and pseudoacinar formation (Figure 2E). At the chosen time points inflammatory liver injury did not change AAV-induced HCC initiation or progression in the neonatal model. Hepatocyte Proliferation Caused by Chronic Liver Injury or Partial Hepatectomy Leads to rAAV-Induced HCC in Adult Mice To determine whether liver injury and regeneration affects rAAV- induced HCC development in adult mice, we started 3-week-old male mice on HFD or RD and infected them at full adult maturity (10 weeks old) with either AAV-Rian or an AAV-tdTomato control virus. To stimulate hepatocyte proliferation, a subset of rAAV-in- fected mice on the RD underwent a 2/3 partial hepatectomy 1 week after viral infection. 6 months after viral infection, the mice were euthanized and assessed for tumors (Figures 3A and 3D). None of Figure 1. Schematic of Rian Targeting Vector and a Control tdTomato Vector The AAV-Rian vector consists of a CAG promoter flanked by arms of homology to Rian. The green arrows indicate the positions of primers, one inside the CAG promoter and one outside of the arm of homology, used to detect CAG inte- gration by homologous recombination. AAV-tdTomato is a non-editing vector that expresses the tdTomato transgene driven by the CAG promoter. www.moleculartherapy.org Molecular Therapy Vol. 29 No 2 February 2021 681 RESULTS 109 the mice on the RD that received the tdTomato virus developed tumors (n = 5), but 50% of the mice on the HFD (n = 10) developed tumors, with an average tumor burden of approximately 3 tumor nodules per mouse (Figures 3B and 3C). However, the tumor inci- dence (Figure 3B, chi-square test p = 0.10) and burden (Figure 3C, one-way ANOVA p > 0.05) were not statistically different. Only 5% of themice on the RD that received the Rian vector developed tumors, with an average tumor burden of 0.05 (n = 20), whereas 100% of the mice on the HFD developed tumors with a tumor burden of 6 (n = 10, p < 0.001; Figures 3B and 3C). Similarly, all of the mice on the RD that underwent a 2/3 partial hepatectomy developed tumors, with an Figure 2. HFD-Induced Liver Injury Does Not Exacerbate AAV-Induced HCC in the Neonatal Mouse Model (A) Outline of experimental design (n = 3–14 per group). (B) Tumor incidence in neonatal mice infected with AAV-Rian or AAV-tdTomato in presence or absence of diet induced liver injury. Tumor incidence was not statistically significant between diet groups by chi-square analysis (p > 0.05). Numbers above the bars represent the sample size. (C) Tumor burden at 6 months. No statistical significance was achieved between diet groups by one-way ANOVA fol- lowed by Bonferroni post hoc (p > 0.05). Numbers above the bars represent the sample size. (D) Images of gross liver specimens from 6-month-old mice. (E) Representative H&E staining of livers from mice on RD and HFD. (F) Representative gel of CAG insertion in Rian tumors by integration PCR, where one primer targets the CAG region and the other primer targets a region outside of the ho- mology arm. Four tumors, two from female and two from male tumors, were selected. No integration was detected in mice on RD infected with AAV-tdTomato. average tumor burden of 16 (n = 5, Figures 3B and 3C). Both HFD and partial hepatectomy had a significant effect on tumor incidence and burden compared to mice on RD that received the Rian virus (p < 0.01). Histologic findings of HCC were similar to the neonatal livers (Figure 3E). CAG Integration in the Rian Locus Activates Oncogenic Pathways Several tumors and background liver were dissected and assessed for integration of the CAG promoter into the Rian locus by homolo- gous recombination. Primers were designed such that one primer was in the 30 end of the CAG promoter and another in the flanking gDNA outside the homology arm. All of the Rian tumors from the neonatal experiment and the adult mice experiment (AAV-Rian tumors from mice on HFD and those that underwent PH) had a targeted CAG integrations (Figures 2F and 3F), which were confirmed by sequencing. Bands seen in the RD/tdTomato and RD/Rian group were sequenced and did not contain Rian or CAG sequences, suggest- ing nonspecific amplification. In order to determine whether the tumors from this study presented a Rian gene-expression profile similar to prior studies, a gene signature (AAV-Rian signature) recapitulating the top differentially expressed genes was generated.4 Single-sample gene set enrichment analysis (ssGSEA) and nearest template prediction (NTP) analysis were per- formed to assesswhether the tumorshad theAAV-Rian signature using non-tumor samples as controls (Figure 4A). ssGSEA analysis showed Molecular Therapy 682 Molecular Therapy Vol. 29 No 2 February 2021 RESULTS 110 Figure 3. HFD-Induced Liver Injury and Partial Hepatectomy Exacerbates AAV-Induced HCC in the Adult Mouse Model (A) Outline of experimental design (n = 5–20 per group). (B) Tumor incidence in adult mice infected with AAV. HFD increased Rian induced tumor incidence (chi-square test, p < 0.05). Numbers above the bars represent the sample size. (C) Tumor burden in adult mice infected with AAV. Both HFD and partial hepatectomy increased tumor burden in AAV-Rian-CMV infected mice (one-way ANOVA, Bonferroni post hoc, p < 0.01). Numbers above the bars represent the sample size. (D) Images of gross liver specimens (legend continued on next page) www.moleculartherapy.org Molecular Therapy Vol. 29 No 2 February 2021 683 RESULTS 111 an enrichment of the AAV-Rian signature in the tumor samples. The “AAV-Rian-UP” set of genes was significantly enriched (p < 0.0001) in tumor samples and the “AAV-Rian-DOWN” set of genes were significantly enriched in the non-tumoral samples (p < 0.0001). The NTP approach also demonstrated a similar AAV-Rian signature in the tumors. Interestingly, the tumor from the RD-Rian-Female mouse did not have the Rian signature, in line with the different clustering of female tumors (n = 2) revealed by non-negative matrix factorization (NMF) analysis (Figure S1) but did have a discernibly reduced expres- sion of the main HCC-related oncogenic pathways. Nevertheless, the RD-Rian-Female tumor did have an increase in gene expression in/ near the Rian locus where CAG integrated compared to background liver (Figure 4A). In contrast to the female tumors, the HFD-Tomato tumor (n = 1) clustered with the AAV-Rian tumors (Figure S1) and had a similar AAV-Rian signature (Figure 4A; Figure S2). The gene- expression pattern of the tomato tumor strongly suggests random inte- gration into the Rian locus, given that no other known genomic targets would produce the same pattern. Overall, these results indicate that the gene-expression profile of these tumors highly resemble the gene expression profile of the Rian tumors reported previously.4 Gene-expression profiles of the tumors was assessed to determine whether they recapitulated any of the molecular classes that have been described for human HCC. Interestingly, the Rian tumors ex- pressed a molecular profile similar to a subclass of human HCC asso- ciated with proliferative and progenitor like features and a poor prog- nosis (Figure 4B).17–19 Additional HCC signaling pathways involved enrichment of a gene signature associated with poor outcome (i.e., poor-survival signature), including positivity for the Met signature (associated with MET activation), EpCAM-positivity, AKT signaling, IGFR1 positivity, WNT/transforming growth factor b (TGF-b) acti- vation, and NOTCH1 signaling.20–25 The above described molecular profile of the Rian tumors parallels a subtype of human HCCs with overexpression of a cluster of microRNAs (miRNAs) located in the MEG8 locus.26 Moreover this group of human HCC is associated with molecular signatures of poor prognosis similar to that seen in Rian tumors.27 Enrichment of EpCAM and NOTCH1 signaling sig- natures is probably due to certain biological differences between mu- rine tumors and human HCC. Murine tumors tend to be very aggres- sive, which is linked to more deregulated pathways than human tumors. Estrogen Partially Ameliorates HFD-Induced Inflammation Results of this study confirmed a delay in initiation of HCC in females compared to males as seen in other studies.4,28 Since estrogen has anti-inflammatory properties, it was hypothesized that estrogen may be involved in delaying the initiation of HCC. To test this, 3- week-old male mice were started on HFD or RD, received tri-weekly intraperitoneal (i.p.) injections of 80 mg/kg estrogen for 1 month, and were then sacrificed for analysis. H&E staining showed a small but noticeable reduction in liver damage and fat in male mice on HFD that received estrogen (Figure 5B). Further, estrogen treatment re- sulted in a small (!4%) but significant reduction in hepatocyte pro- liferation (Figure 6). This reduction in proliferation was not captured in the RNA sequencing (RNA-seq) analysis, likely due to the overall small proportion of proliferating hepatocytes in the liver. RNA-seq of background livers from mice on RD and HFD, with or without estrogen did show differences in immune-related pathways, oncogenic pathways, lipid metabolism, and liver function. ssGSEA indicated that background livers from mice fed the HFD presented an increased immune cancer field, recently defined by our group in adjacent liver tissue from HCC patients, which is associated with in- flammatory signaling and higher risk of HCC development.29 This is in line with the increased inflammation of the liver observed by the enrichment in inflammatory-related signaling pathways such as hall- marks of inflammatory response, interferon-a (IFN-a) or IFN-g response, signatures of T cell exhaustion or presence of T regs (Fig- ure 5C). The gene expression pathways in HFD mice treated with E2 are similar in most respect in HFD mice not treated with E2. This likely indicates that estrogen has little role in ameliorating the NASH condition at the dose tested. However, with the estrogen treat- ment we observed an amelioration of the immune exhaustion fea- tures, including a decrease in TNF-a, TGF-b, and Wnt/b-catenin signaling, an increase of antigen processing and presentation, Toll- like receptor signaling, and activation of the adaptive immune system including general activation of B and T cells. Livers from mice fed the HFD also presented upregulation of some oncogenic pathways (Myc, MTORC, or EGFR) and increased lipid metabolism, and downregu- lation of other metabolic pathways (i.e., bile acid or drugmetabolism). DISCUSSION Here we show for the first time that adult mice that received a rAAV gene targeting vector developed HCC where hepatocyte proliferation is increased due to NAFLD and partial hepatectomy. Adult mice typi- cally do not develop rAAV-induced liver cancer in the absence of injury.8,9 We hypothesized that the age-dependence of gene therapy induced cancer is due to the rate of hepatocyte proliferation, which is high in neonates due to natural liver growth. To determine whether the proliferation rate impacts rAAV induced oncogenesis, we evalu- ated two injury regimens. Hepatocyte cell division is increased after both partial hepatectomy and fat-induced liver injury. Both condi- tions led to the development of HCC in adult mice, which raises con- cerns for the use of rAAV therapy in the general human population, where significant numbers have chronic inflammatory liver diseases. In the US, for example, 0.34% of the population have HBV infection, from 9-month-old mice. (E) Representative H&E staining of livers frommice on RD and HFD. (F) Representative gel of CAG insertion in Rian tumors by integration PCR. Target PCR product was not observed inmice on RD infected with either Rian or tdTomato virus but was observed inmouse on HFD infectedwith Rian and inmice that had received a partial hepatectomy. The bands seen in the RD/tdTomato and RD/Rian lanes were sequenced and did not show CAG or Rian sequences, suggesting that these bands are artifacts. Molecular Therapy 684 Molecular Therapy Vol. 29 No 2 February 2021 RESULTS 112 (legend on next page) www.moleculartherapy.org Molecular Therapy Vol. 29 No 2 February 2021 685 RESULTS 113 1.7% have HCV infection, 4.3% have alcoholic liver disease, and up to 30% have fatty liver disease.15,30–32 These inflammatory conditions alone are risk factors for HCC development, thus adding a rAAV- associated oncogenic risk may significantly diminish the possible therapeutic benefits of gene therapy. Inflammation has long been associated with carcinogenesis, and hep- atocarcinogenesis is no exception.33 Most human HCCs arise in the background of chronic liver diseases characterized by injury and inflammation. Inflammation itself may promote hepatocarcinogene- sis through several mechanisms, such as DNA damage from reactive oxygen species, changes in the immune system milieu, and an in- crease in hepatocyte proliferation.34 This study demonstrates an increase in the risk of rAAV-induced HCC in adult mice fed with HFD to approximate NAFLD, a common chronic liver disease in humans. While the exact mechanism for this increase is unclear, the most likely reason is an increase in hepatocyte proliferation. AAV is more likely to integrate into proliferating hepatocytes as Figure 4. Tumors Exhibit a Prototypical Rian Signature and Have a Similar Transcriptomic Profile to Human HCC Subclassifications (A) Heatmap showing the AAV-Rian signature and the “AAV-Rian UP” and “AAV-Rian Down” gene sets generated from theWang et al.4 study, and analysis of the expression of the genes in the Rian Locus (GRCm38/mm10, chr12:108860000–110418000). (B) HCC classification, survival signatures, oncogenic pathways, and lipid metabolism activation in the tumor samples of the study. p Values were calculated comparing diets within male Rian tumor samples. Figure 5. Estrogen Partially Ameliorates HFD-Induced Liver Injury (A) Outline of experimental design. (B) Representative H&E liver images of mice on RD or HFD for 1month or 6months. Females have slightly lower HFD-induced injury at both time points, and male mice on HFD+E2 also had less injury than males on HFD that didn’t receive E2. (C) Heatmap showing the expression of immune-related pathways, oncogenic pathways, lipid metabolism, and liver function in the background liver samples. Molecular Therapy 686 Molecular Therapy Vol. 29 No 2 February 2021 RESULTS 114 seen in the adult HFD model.13,35,36 These findings raise concern regarding rAAV gene therapy in humans, where resultant HCC has not been detected. In patients with chronic liver disease like NAFLD, and with other viral infections like hepatitis B and C, the risk of random rAAV integration is greater due to the proliferating hepato- cytes associated with chronic liver injury. The correlation between AAV gene therapy and other viral infection is not extensively studied and is an area that would benefit from further investigation. One caveat to generalizing these mouse data to humans, however, is the possibility that mice may have a lower threshold for HCC develop- ment than humans. Males have long been known to have a significantly higher incidence of HCC compared to females, and the mechanism is likely related to a relative increase in hepatic inflammation for a given injury.28,37 Here we confirm that female mice are less susceptible to rAAV-induced HCC compared to males. We show that estrogen treatment of male mice fed a HFD partially ameliorates inflammation and proliferation. Interestingly, estrogen also altered the immune milieu in males with NAFLD, decreasing immune exhaustion associated with activation of oncogenic pathways. CD4+ lymphocyte loss in patients with NAFLD was found to promote hepatocarcinogenesis, and the link here found between estrogen and liver inflammation suggests a further connec- tion explaining gender disparity seen in hepatocarcinogenesis.38 We observed a high incidence of HCC in adult mice on a HFD that received the control tdTomato AAV. Unlike the Rian AAV, the tdTomato virus does not have the capability for targeted integration into the genome. Nonetheless, we found 5 of 10 mice on the HFD that received the tdTomato AAV developed HCC, some of which had the same tumor signature as the AAV-Rian-induced HCC. There are numerous reports that show that AAV randomly integrates in host genomes and that these sites are scattered throughout the genome.3,39–42 It is likely that the virus integrated in other sites as well, some of which could have been in oncogenic loci; however, we are not aware of any other oncogenic loci that have a similar onco- genic signature as Rian. Although there is not sufficient data in this report talk about the frequency of integration in oncogenic loci, the Rian locus appears to be a hotspot for random integration since the tomato tumors analyzed had a similar, but slightly shifted gene acti- vation profile as the tumors derived from a targeted insertion (Fig- ure S2). The fact that tumors developed only in those on the HFD points toward the same factors noted above, namely that inflamma- tion and increased hepatocyte proliferation increase HCC incidence. Whether the human liver is similarly susceptible to random rAAV in- tegrations in the MEG8 locus (analogous to mouse Rian) leading to HCC is unclear but should be studied. We did not observe a difference in initiation or progression of HCC in mice infected with AAV as neonates, though this may be a result of assessing the mice at a time point where all of the control mice already had cancer. Indeed, a small number of mice on a HFD who had neonatal AAV-infection died before the pre-determined 6-month period (data not shown, but necropsied mice all had liver tumors). We suspect that acceleration of hepatocarcinogenesis occurred in the AAV-infected neonates on HFD, but that our pre-determined experimental end (6 months) was too late to document changes such as the earlier appearance of HCC. We chose the 6-month time point in the neonates based on prior studies, which led to 100% of mice developing HCC.4 To understand if HCC developed earlier in AAV-infected neonates on a HFD, earlier time points would need to be assessed, prior to 100% of mice developing HCC. Indeed, there Figure 6. Estrogen Treatment Partially Suppressed Hepatocyte Proliferation (A) BrdU staining (brown dots) identifies dividing hepatocytes. (B) Morphometric analysis of BrdU hepatocytes. Estrogen partially reversed HFD-induced injury (Student’s t test, p < 0.05). www.moleculartherapy.org Molecular Therapy Vol. 29 No 2 February 2021 687 RESULTS 115 was a trend toward higher tumor incidence and burden in both gen- ders and ages of AAV-infected mice fed the HFD but did not reach statistical significance. Given the many conditions that were tested, the sample size per group became relatively small, which could explain why some of the trends did not reach significance. Viral load is another parameter that could have influenced hepato- carcinogenesis in adult mice. It is possible that a higher AAV dose could result in higher HCC incidence, however, we did not assess whether there might be a dose response effect of the virus on hepatocarcinogenesis. In conclusion, this study demonstrates that adult mice infected with both targeted and non-targeted rAAV develop HCC when there is also a stimulus for hepatocyte proliferation such as fatty liver induced by a HFD. Given the high prevalence of inflammatory liver conditions in the general population, this should raise an alarm about the use of rAAV, an otherwise promising vector used for diverse gene therapy application. Female mice are less susceptible to rAAV-induced HCC compared to males, part of which is due to a more favorable im- mune milieu related to estrogen and likely applies to humans given the documented gender disparity. More studies are needed to assess the risks of rAAV-induced hepatocarcinogenesis given the promise and widespread use of rAAV for gene therapies. MATERIALS AND METHODS Vector Production The AAV-Rian-CMV plasmid was obtained from David W. Russell (University of Washington).4 The vector was packaged in the AAV- DJ serotype by cotransfection of HEK293 cells with the vector plasmid pAAV-Rian-CMV, pDJ (capsid plasmid), and pHelper.43 Briefly, 7! 106 HEK293T cells were plated in 10-cm dishes, cultured in DMEM complete (10% FBS, 100 U/mL penicillin, 100 mg/mL streptomycin, 2 mM GlutaMax, 1 mM sodium pyruvate). After 16 to 20 h, media was replaced with reduced serumDMEM (2% FBS) fol- lowed by polyethylenimine (PEI)-based triple transfection. Equi- molar quantities of each plasmid was combined with a total mass of 24 mg in OptiMEM (GIBCO, Grand Island, NY) media and 3 times the plasmid weight of PEI (72 mg from 1 mg/mL stock solution, pH 7.1, Polysciences, Warrington, PA) were combined, with a final vol- ume of 1.5 mL per 10-cm dish. The transfection mixture was incu- bated at room temperature for 30 min followed by addition to cells. 5 days post transfection the media was collected and centrifuged twice at 1,900 ! g for 30 min. The supernatant was treated with benzonase (Sigma, St. Louis, MO) at 37"C for 1 h and then centrifuged at 1,900! g for 30 min. The supernatant was filtered through 0.45 mm PES vacuum filter, precipitated with PEG #8,000 (Spectrum Chemical, Gardena, CA; 8% w/v PEG-8000, 0.5 M NaCl) overnight, and then centrifuged at 6,000 ! g for 1 h. The viral pellet was resuspended in DPBS. The viral titer was determined by a dot-blot assay as described previously.44 Animal Care C57BL/6J neonatal mice were bred at OHSU and 3-week-old C57BL/ 6J male mice were purchased from Jackson Laboratory. All animal experiments were approved by the Oregon Health & Science Univer- sity Institutional Animal Care and Use Committee (Portland, OR) and performed in accordance with the approved protocols. Effect of Diet-Induced Liver Injury on AAV Induced HCC in Neonates 1-day-old mice received either 3 ! 1010 viral genomes of AAV-Rian- CMV (AAV-Rian) or AAV-CAG-tdTomato (AAV-tdTomato; tdTo- mato reporter gene driven by the CAG promoter) vectors in 16 mL DPBS or DPBS alone via superficial temporal vein injections. This dose has been shown to cause to HCC in neonatally infected mice.4 At 3 weeks of age, mice were either placed on a RD (5LOD) or a HFD (Envigo, Madison, WI). Table S1 summarizes the different groups and sample size. Mice were sacrificed at 6 months of age and their livers were harvested. Tumor burden was assessed by count- ing visible nodules and calculating the average number per mouse. Sections of the liver were fixed in 4% PFA (v/v) or frozen at #80"C for histological and molecular analysis. Effect of Age- and Diet-Induced Liver Injury on AAV-Induced HCC 3-week-old mice (purchased from Jackson Laboratory) were placed on either RD or HFD. When the mice were 10 weeks old, they received 8 ! 1011 AAV-Rian or AAV-tdTomato vectors in 100 mL DPBS via retro-orbital injection. This dose reflects a weight-based calculation of virus analogous to the dose given to the neonatal mice. A week later (11 weeks), a subset of mice on the regular diet received a 2/3 partial hepatectomy (PH). The mice were sacrificed and their livers were harvested at 9.5 months of age. Table S2 summa- rizes the different groups and sample size. Tumor burden assessment and liver sample processing was performed as described in the previ- ous section. Effect of Estrogen on HFD-Induced Liver Injury 3-week-old mice (purchased from Jackson Laboratory) were either on a HFD or a RD and received either estrogen (80 mg/kg dissolved in corn oil, Sigma, St. Louis, MO) or vehicle (corn oil) every other day for 1 month (5 mice per group). 2 h before tissue harvest, mice were injected with 100 mg/kg bromodeoxyuridine (BrdU; Alfa Aesar, Ward Hill, MA) to label actively dividing hepatocytes. At the end of treatment period, livers were harvested and sections were fixed in 4% PFA or frozen for histology and molecular analysis. Histology Tissues were fixed in 4% PFA (v/v) and then transferred to 70% (v/v) EtOH. H&E staining was performed by the Digestive Diseases Center at Texas Children’s Hospital. DNA Isolation and Vector Integration Analysis 1 mg of tissue was used for DNA isolation using the MasterPure Complete DNA and RNA Purification kit (Lucigen, Middleton, WI) per manufacturer’s protocol, followed by phenol/chloroform extraction and ethanol precipitation. The DNA pellet was resus- pended in TE buffer and the concentration was measured using the Molecular Therapy 688 Molecular Therapy Vol. 29 No 2 February 2021 RESULTS 116 Qubit Fluorometer. Integration of AAV-Rian was assessed by PCR and sequencing. Primers were designed such that the forward primer (50-GCTCCTGGGCAACGTGCTGGT-30) was complementary to the CAG promoter and the reverse primer (50-TGGAAGAGCCGG- GAAGCCTTTGA-30) was outside of the AAV-Rian homology arm (Figure 1A). PCR was performed using the 2! PrimeStar polymerase mix (Takara, Mountain View, CA) per manufacturer’s protocol. Briefly, 100 ng of gDNA and 0.2 mM of each primer was used in a 50 mL reaction. Thermocycler condition was as follows: denatured 98"C for 10 s, followed by extension at 68"C for 45 s, for 30 cycles; the expected product size is 951 base pairs. PCR products were analyzed on a 0.8% agarose gel and verified by sequencing (GeneWiz, South Plainfield, NJ). RNA-Seq and Analysis 1 mg of tissue was homogenized in RNAzol RT (MRC, Cincinnati, OH) and RNA was extracted per manufacturer’s protocol. RNA was resuspended in nuclease free water and sent to GeneWiz for library prep and sequencing. Three independent Rian tumors from the neonatal model on HFD and RD were used for library prepara- tion. RNA from three individual livers were pooled to prepare one library for RD/tdTomato and HFD/tdTomato background liver. Only a single nodule was used to prepare libraries for HFD/ tdTomato male tumor and HFD/tdTomato female tumor. RNA from liver sections of three individual mice were pooled for each of the following groups from the estrogen experiment: RD/vehicle male, RD/vehicle female, HFD/vehicle male, RD/estrogen male, and HFD/estrogen male (a single library was prepared for each of the five groups). The data was standardized to fragments per kilobase of transcript per million mapped reads (FPKM). See Supplemental Materials and Methods sections for a detailed descrip- tion of data analysis. Statistical Analysis All data presented as mean ± SEM and statistical significance was as- sessed by one-way ANOVA followed by Bonferroni post hoc test or Student’s t test. Chi-square test was used to assess significance of tumor incidence. For RNA-seq data, correlations for categorical variables were analyzed by Fisher exact test. In all cases, p < 0.05 (two-tailed) or FDR < 0.05 was considered significant. Methods used in the statistical analysis of RNA-seq data are noted in the Sup- plemental Materials and Methods section. SUPPLEMENTAL INFORMATION Supplemental Information can be found online at https://doi.org/10. 1016/j.ymthe.2020.10.018. AUTHOR CONTRIBUTIONS D.A.D.: Designed and performed experiments, analyzed data, and wrote the manuscript. D.W.R., M.G., and W.E.N.: Conceived the project, helped design experiments, and helped write the manuscript. J.P.: Quantified AAV titers. C.L.: Analyzed HCC histology. L.T., J.A.-F., R.P., C.W., and J.M.L.: Performed bioinformatics analysis and helped review the manuscript. CONFLICTS OF INTEREST J.M.L. is receiving research support from Bayer HealthCare Pharma- ceuticals, Eisai, Bristol-Myers Squibb, Boehringer-Ingelheim, and Ipsen and consulting fees from Bayer HealthCare Pharmaceuticals, Merck, Eisai, Bristol-Myers Squibb, Celsion Corporation, Eli Lilly, Roche, Genentech, Glycotest, Nucleix, AstraZeneca, and Exelixis. ACKNOWLEDGMENTS We thank Amita Tiyaboonchai for her assistance with liver harvest- ing. This research was funded by the National Institutes of Health grant R01CA190144 to W.E.N. and M.G. J.M.L. is supported by European Commission (EC)/Horizon 2020 Program (HEPCAR, Ref. 667273-2), EIT Health (CRISH2, reference number 18053), Accelerator Award (CRUK, AECC, AIRC; HUNTER, reference num- ber C9380/A26813), National Cancer Institute (P30-CA196521), U.S. Department of Defense (CA150272P3), Samuel Waxman Cancer Research Foundation, Spanish National Health Institute (SAF2016- 76390), and the Generalitat de Catalunya/AGAUR (SGR-1358). L.T. is supported by an Accelerator Award (CRUCK, AECC, AIRC; HUNTER, C9380/A26813). REFERENCES 1. Wang, D., Tai, P.W.L., and Gao, G. (2019). Adeno-associated virus vector as a plat- form for gene therapy delivery. Nat. Rev. Drug Discov. 18, 358–378. 2. Hinderer, C., Katz, N., Buza, E.L., Dyer, C., Goode, T., Bell, P., Richman, L.K., and Wilson, J.M. (2018). Severe Toxicity in Nonhuman Primates and Piglets Following High-Dose Intravenous Administration of an Adeno-Associated Virus Vector Expressing Human SMN. Hum. Gene Ther. 29, 285–298. 3. Nault, J.C., Datta, S., Imbeaud, S., Franconi, A., Mallet, M., Couchy, G., Letouzé, E., Pilati, C., Verret, B., Blanc, J.F., et al. (2015). Recurrent AAV2-related insertional mutagenesis in human hepatocellular carcinomas. Nat. Genet. 47, 1187–1193. 4. Wang, P.R., Xu, M., Toffanin, S., Li, Y., Llovet, J.M., and Russell, D.W. (2012). Induction of hepatocellular carcinoma by in vivo gene targeting. Proc. Natl. Acad. Sci. USA 109, 11264–11269. 5. Chandler, R.J., LaFave, M.C., Varshney, G.K., Trivedi, N.S., Carrillo-Carrasco, N., Senac, J.S., Wu, W., Hoffmann, V., Elkahloun, A.G., Burgess, S.M., and Venditti, C.P. (2015). Vector design influences hepatic genotoxicity after adeno-associated vi- rus gene therapy. J. Clin. Invest. 125, 870–880. 6. Donsante, A., Miller, D.G., Li, Y., Vogler, C., Brunt, E.M., Russell, D.W., and Sands, M.S. (2007). AAV vector integration sites in mouse hepatocellular carcinoma. Science 317, 477. 7. Donsante, A., Vogler, C., Muzyczka, N., Crawford, J.M., Barker, J., Flotte, T., Campbell-Thompson, M., Daly, T., and Sands, M.S. (2001). Observed incidence of tumorigenesis in long-term rodent studies of rAAV vectors. Gene Ther. 8, 1343– 1346. 8. Bell, P., Wang, L., Lebherz, C., Flieder, D.B., Bove, M.S., Wu, D., Gao, G.P., Wilson, J.M., and Wivel, N.A. (2005). No evidence for tumorigenesis of AAV vectors in a large-scale study in mice. Mol. Ther. 12, 299–306. 9. Li, H., Malani, N., Hamilton, S.R., Schlachterman, A., Bussadori, G., Edmonson, S.E., Shah, R., Arruda, V.R., Mingozzi, F., Wright, J.F., et al. (2011). Assessing the potential for AAV vector genotoxicity in a murine model. Blood 117, 3311–3319. 10. Taub, R. (2004). Liver regeneration: from myth to mechanism. Nat. Rev. Mol. Cell Biol. 5, 836–847. 11. Michalopoulos, G.K. (2017). Hepatostat: Liver regeneration and normal liver tissue maintenance. Hepatology 65, 1384–1392. 12. Forbes, S.J., and Newsome, P.N. (2016). Liver regeneration - mechanisms and models to clinical application. Nat. Rev. Gastroenterol. Hepatol. 13, 473–485. www.moleculartherapy.org Molecular Therapy Vol. 29 No 2 February 2021 689 RESULTS 117 13. Cast, A., Kumbaji, M., D’Souza, A., Rodriguez, K., Gupta, A., Karns, R., Timchenko, L., and Timchenko, N. (2019). Liver Proliferation Is an Essential Driver of Fibrosis in Mouse Models of Nonalcoholic Fatty Liver Disease. Hepatol. Commun. 3, 1036– 1049. 14. Michalopoulos, G.K. (2010). Liver regeneration after partial hepatectomy: critical analysis of mechanistic dilemmas. Am. J. Pathol. 176, 2–13. 15. Cotter, T.G., and Rinella, M. (2020). Nonalcoholic Fatty Liver Disease 2020: The State of the Disease. Gastroenterology. 158, 1851–1864. 16. Tsuchida, T., Lee, Y.A., Fujiwara, N., Ybanez, M., Allen, B., Martins, S., Fiel, M.I., Goossens, N., Chou, H.I., Hoshida, Y., and Friedman, S.L. (2018). A simple diet- and chemical-induced murine NASH model with rapid progression of steatohepati- tis, fibrosis and liver cancer. J. Hepatol. 69, 385–395. 17. Boyault, S., Rickman, D.S., de Reyniès, A., Balabaud, C., Rebouissou, S., Jeannot, E., Hérault, A., Saric, J., Belghiti, J., Franco, D., et al. (2007). Transcriptome classification of HCC is related to gene alterations and to new therapeutic targets. Hepatology 45, 42–52. 18. Hoshida, Y., Nijman, S.M., Kobayashi, M., Chan, J.A., Brunet, J.P., Chiang, D.Y., Villanueva, A., Newell, P., Ikeda, K., Hashimoto, M., et al. (2009). Integrative tran- scriptome analysis reveals common molecular subclasses of human hepatocellular carcinoma. Cancer Res. 69, 7385–7392. 19. Llovet, J.M., Montal, R., Sia, D., and Finn, R.S. (2018). Molecular therapies and pre- cision medicine for hepatocellular carcinoma. Nat. Rev. Clin. Oncol. 15, 599–616. 20. Kaposi-Novak, P., Lee, J.S., Gòmez-Quiroz, L., Coulouarn, C., Factor, V.M., and Thorgeirsson, S.S. (2006). Met-regulated expression signature defines a subset of hu- man hepatocellular carcinomas with poor prognosis and aggressive phenotype. J. Clin. Invest. 116, 1582–1595. 21. Yamashita, T., Forgues, M., Wang, W., Kim, J.W., Ye, Q., Jia, H., Budhu, A., Zanetti, K.A., Chen, Y., Qin, L.X., et al. (2008). EpCAM and alpha-fetoprotein expression de- fines novel prognostic subtypes of hepatocellular carcinoma. Cancer Res. 68, 1451– 1461. 22. Majumder, P.K., Febbo, P.G., Bikoff, R., Berger, R., Xue, Q., McMahon, L.M., Manola, J., Brugarolas, J., McDonnell, T.J., Golub, T.R., et al. (2004). mTOR inhibition reverses Akt-dependent prostate intraepithelial neoplasia through regulation of apoptotic and HIF-1-dependent pathways. Nat. Med. 10, 594–601. 23. Tovar, V., Cornella, H., Moeini, A., Vidal, S., Hoshida, Y., Sia, D., Peix, J., Cabellos, L., Alsinet, C., Torrecilla, S., et al. (2017). Tumour initiating cells and IGF/FGF signalling contribute to sorafenib resistance in hepatocellular carcinoma. Gut 66, 530–540. 24. Lachenmayer, A., Alsinet, C., Savic, R., Cabellos, L., Toffanin, S., Hoshida, Y., Villanueva, A., Minguez, B., Newell, P., Tsai, H.W., et al. (2012). Wnt-pathway acti- vation in two molecular classes of hepatocellular carcinoma and experimental mod- ulation by sorafenib. Clin. Cancer Res. 18, 4997–5007. 25. Villanueva, A., Alsinet, C., Yanger, K., Hoshida, Y., Zong, Y., Toffanin, S., Rodriguez- Carunchio, L., Sole, M., Thung, S., Stanger, B.Z., et al. (2012). Notch signaling is acti- vated in human hepatocellular carcinoma and induces tumor formation in mice. Gastroenterology 143, 1660–1669. 26. Toffanin, S., Hoshida, Y., Lachenmayer, A., Villanueva, A., Cabellos, L., Minguez, B., Savic, R., Ward, S.C., Thung, S., Chiang, D.Y., et al. (2011). MicroRNA-based classi- fication of hepatocellular carcinoma and oncogenic role of miR-517a. Gastroenterology 140, 1618–1628. 27. Lee, J.S., Chu, I.S., Heo, J., Calvisi, D.F., Sun, Z., Roskams, T., Durnez, A., Demetris, A.J., and Thorgeirsson, S.S. (2004). Classification and prediction of survival in hepa- tocellular carcinoma by gene expression profiling. Hepatology 40, 667–676. 28. Naugler, W.E., Sakurai, T., Kim, S., Maeda, S., Kim, K., Elsharkawy, A.M., and Karin, M. (2007). Gender disparity in liver cancer due to sex differences in MyD88-depen- dent IL-6 production. Science 317, 121–124. 29. Moeini, A., Torrecilla, S., Tovar, V., Montironi, C., Andreu-Oller, C., Peix, J., Higuera, M., Pfister, D., Ramadori, P., Pinyol, R., et al. (2019). An Immune Gene Expression Signature Associated With Development of Human Hepatocellular Carcinoma Identifies Mice That Respond to Chemopreventive Agents. Gastroenterology 157, 1383–1397. 30. Kim, H.S., Rotundo, L., Yang, J.D., Kim, D., Kothari, N., Feurdean, M., Ruhl, C., and Unalp-Arida, A. (2017). Racial/ethnic disparities in the prevalence and awareness of Hepatitis B virus infection and immunity in the United States. J. Viral Hepat. 24, 1052–1066. 31. Rosenberg, E.S., Hall, E.W., Sullivan, P.S., Sanchez, T.H., Workowski, K.A., Ward, J.W., and Holtzman, D. (2017). Estimation of State-Level Prevalence of Hepatitis C Virus Infection, US States and District of Columbia, 2010. Clin. Infect. Dis. 64, 1573–1581. 32. Wong, T., Dang, K., Ladhani, S., Singal, A.K., and Wong, R.J. (2019). Prevalence of Alcoholic Fatty Liver Disease Among Adults in the United States, 2001-2016. JAMA 321, 1723–1725. 33. Kasai, Y., Takeda, S., and Takagi, H. (1996). Pathogenesis of hepatocellular carci- noma: a review from the viewpoint of molecular analysis. Semin. Surg. Oncol. 12, 155–159. 34. Yang, Y.M., Kim, S.Y., and Seki, E. (2019). Inflammation and Liver Cancer: Molecular Mechanisms and Therapeutic Targets. Semin. Liver Dis. 39, 26–42. 35. Trobridge, G., Hirata, R.K., and Russell, D.W. (2005). Gene targeting by adeno-asso- ciated virus vectors is cell-cycle dependent. Hum. Gene Ther. 16, 522–526. 36. Vansaun, M.N., Mendonsa, A.M., and Lee Gorden, D. (2013). Hepatocellular prolif- eration correlates with inflammatory cell and cytokine changes in a murine model of nonalchoholic fatty liver disease. PLoS ONE 8, e73054. 37. Fattovich, G., Stroffolini, T., Zagni, I., and Donato, F. (2004). Hepatocellular carci- noma in cirrhosis: incidence and risk factors. Gastroenterology 127 (5, Suppl 1), S35–S50. 38. Ma, C., Kesarwala, A.H., Eggert, T., Medina-Echeverz, J., Kleiner, D.E., Jin, P., Stroncek, D.F., Terabe, M., Kapoor, V., ElGindi, M., et al. (2016). NAFLD causes se- lective CD4(+) T lymphocyte loss and promotes hepatocarcinogenesis. Nature 531, 253–257. 39. Miller, D.G., Trobridge, G.D., Petek, L.M., Jacobs, M.A., Kaul, R., and Russell, D.W. (2005). Large-scale analysis of adeno-associated virus vector integration sites in normal human cells. J. Virol. 79, 11434–11442. 40. Nakai, H., Wu, X., Fuess, S., Storm, T.A., Munroe, D., Montini, E., Burgess, S.M., Grompe, M., and Kay, M.A. (2005). Large-scale molecular characterization of ad- eno-associated virus vector integration in mouse liver. J. Virol. 79, 3606–3614. 41. Chandler, R.J., Sands, M.S., and Venditti, C.P. (2017). Recombinant Adeno- Associated Viral Integration and Genotoxicity: Insights from Animal Models. Hum. Gene Ther. 28, 314–322. 42. La Bella, T., Imbeaud, S., Peneau, C., Mami, I., Datta, S., Bayard, Q., Caruso, S., Hirsch, T.Z., Calderaro, J., Morcrette, G., et al. (2020). Adeno-associated virus in the liver: natural history and consequences in tumour development. Gut 69, 737–747. 43. Grimm, D., Lee, J.S., Wang, L., Desai, T., Akache, B., Storm, T.A., and Kay, M.A. (2008). In vitro and in vivo gene therapy vector evolution via multispecies inter- breeding and retargeting of adeno-associated viruses. J. Virol. 82, 5887–5911. 44. Powers, J.M., Chang, X.L., Song, Z., and Nakai, H. (2018). A Quantitative Dot Blot Assay for AAV Titration and Its Use for Functional Assessment of the Adeno-asso- ciated Virus Assembly-activating Proteins. J. Vis. Exp. 12, 56766. Molecular Therapy 690 Molecular Therapy Vol. 29 No 2 February 2021 RESULTS 119 Study 3 – Immunomodulatory Effects of Lenvatinib Plus Anti- Programmed Cell Death Protein 1 in Mice and Rationale for Patient Enrichment in Hepatocellular Carcinoma Laura Torrens, Carla Montironi, Marc Puigvehí, Agavni Mesropian, Jack Leslie, Philipp K Haber, Miho Maeda, Ugne Balaseviciute, Catherine E Willoughby, Jordi Abril-Fornaguera, Marta Piqué- Gili, Miguel Torres-Martín, Judit Peix, Daniel Geh, Erik Ramon-Gil, Behnam Saberi, Scott L Friedman, Derek A Mann, Daniela Sia, Josep M Llovet Hepatology. 2021 Jun 22;74:2652–2669 (IF: 17.425) Summary Approximately, 50–60% of patients with HCC are estimated to be exposed to systemic therapies in their lifespan, particularly in advanced stages of the disease78. Recently, the tyrosine kinase inhibitor lenvatinib showed efficacy in an open-label randomized phase III trial and became the first new FDA-approved drug for advanced-stage HCC in the first-line setting in over 10 years138. On the other hand, immunotherapies with immune checkpoint inhibitors (ICIs) are emerging as promising treatment options, but responses in HCC are only observed in 15-20% of patients, while the majority are primarily resistant. Much effort has been invested into identifying existing kinase inhibitors that can effectively synergize with ICI. The combination of lenvatinib plus the anti-PD1 ICI pembrolizumab has shown unprecedented phase Ib results (objective response rate of 46% and median survival of 22 months)148 and is currently being assessed in a phase III trial. This strategy is based on the hypothesis that lenvatinib could boost the antitumor immune response and improve the clinical benefit of anti-PD1. However, a deeper understanding of the immunomodulatory capacity of these treatments is still needed. Considering this, we aimed to identify the immunomodulatory effects of lenvatinib in combination with anti-PD1 and provide a mechanistic rationale for this treatment in advanced HCC. To fill this gap, we generated three murine immunocompetent syngeneic models of HCC – two subcutaneous and an orthotopic – and assessed the anti-tumoral activity of lenvatinib alone or in combination with anti-PD1 and its effects on the systemic and tumor-infiltrating immune cells. By performing flow cytometry, transcriptomic, and immunohistochemistry analyses, we explored the immunomodulatory effect of each treatment and unveiled unique effects of the combination. Next, we explored the gene expression analysis of murine and human tumors to RESULTS 120 identify HCC patients who are likely to exhibit primary resistance to single agents but could potentially be rescued with the combination treatment. Our results showed that: 1. Anti-PD1 and lenvatinib in monotherapy were able to improve survival and reduce tumor growth in our murine models. However, the combination treatment achieved a higher response rate, shorter time to response, and a reduction of tumor viability. 2. Lenvatinib exerted a potent immunomodulatory effect on the tumor infiltrate by reducing the Treg proportion and altering their histological distribution by eliciting Treg exclusion from the intratumoral region. In addition, lenvatinib blocked immunosuppressive signaling including TGFß pathway inhibition. 3. Anti-PD1 treatment was also able to modify the immune infiltrate by increasing the T cells and type 1 dendritic cells in the tumor. However, transcriptomic profiling of the tumors revealed that the treatment elicited an immune exhausted phenotype in the infiltrate. 4. The combined effect of lenvatinib plus anti-PD1 treatments induced an increase in the proinflammatory component in the tumor. Consequently, only the combination treatment generated a specific activated antitumoral immune response. 5. No significant alterations were detected in circulating immune cells, suggesting that the effect of the treatments may be tumor-specific. 6. Transcriptomic analysis in a human HCC cohort revealed that 22% of tumors presented down-regulation of genes associated with the molecular effect of the combination, along with reduced pro-inflammatory signaling, high Treg levels, and VEGF pathway activation. Thus, these patients could harbor primary resistance to anti-PD1 and potentially benefit from the booster effect of the combination treatment. In conclusion, these findings provide a comprehensive understanding of the immune- remodeling capacities of the combination of lenvatinib plus anti-PD1, with important implications for patient selection to ultimately maximize the clinical benefit from this treatment regimen. RESULTS 121 Publication 2652 HEPATOLOGY, VOL. 74, NO. 5, 2021 Immunomodulatory Effects of Lenvatinib Plus Anti– Programmed Cell Death Protein 1 in Mice and Rationale for Patient Enrichment in Hepatocellular Carcinoma Laura Torrens ,1,2 Carla Montironi ,1,2 Marc Puigvehí ,1,3 Agavni Mesropian ,2 Jack Leslie ,4 Philipp K. Haber,1 Miho Maeda,1 Ugne Balaseviciute ,2 Catherine E. Willoughby ,2 Jordi Abril- Fornaguera ,2 Marta Piqué- Gili ,2 Miguel Torres- Martín ,1,2 Judit Peix ,2 Daniel Geh ,4 Erik Ramon- Gil ,4 Behnam Saberi,1 Scott L. Friedman ,1 Derek A. Mann ,4 Daniela Sia ,1 and Josep M. Llovet 1,2,5 BACKGROUND AND AIMS: Lenvatinib is an effective drug in advanced HCC. Its combination with the anti- PD1 (programmed cell death protein 1) immune checkpoint inhibi- tor, pembrolizumab, has generated encouraging results in phase Ib and is currently being tested in phase III trials. Here, we aimed to explore the molecular and immunomodulatory ef- fects of lenvatinib alone or in combination with anti- PD1. APPROACH AND RESULTS: We generated three synge- neic models of HCC in C57BL/6J mice (subcutaneous and orthotopic) and randomized animals to receive placebo, len- vatinib, anti- PD1, or combination treatment. Flow cytom- etry, transcriptomic, and immunohistochemistry analyses were performed in tumor and blood samples. A gene signature, capturing molecular features associated with the combina- tion therapy, was used to identify a subset of candidates in a cohort of 228 HCC patients who might respond beyond what is expected for monotherapies. In mice, the combina- tion treatment resulted in tumor regression and shorter time to response compared to monotherapies (P  <  0.001). Single- agent anti- PD1 induced dendritic and T- cell infiltrates, and lenvatinib reduced the regulatory T cell (Treg) proportion. However, only the combination treatment significantly inhib- ited immune suppressive signaling, which was associated with the TGFß pathway and induced an immune- active microen- vironment (P  <  0.05 vs. other therapies). Based on immune- related genomic profiles in human HCC, 22% of patients were identified as potential responders beyond single- agent therapies, with tumors characterized by Treg cell infiltrates, low inflammatory signaling, and VEGFR pathway activation. CONCLUSIONS: Lenvatinib plus anti- PD1 exerted unique immunomodulatory effects through activation of immune pathways, reduction of Treg cell infiltrate, and inhibition of TGFß signaling. A gene signature enabled the identification of ~20% of human HCCs that, although nonresponding to single agents, could benefit from the proposed combination. (Hepatology 2021;74:2652-2669). Liver cancer is the second- leading cause of cancer- related death and a major health problem globally.(1) HCC is the most com- mon form of liver cancer, accounting for >90% of Abbreviations: CTLA4, cytotoxic T- lymphocyte- associated protein 4; CTNNB1, catenin beta 1; DC, dendritic cell; DC1, type 1 dendritic cell; FC, fold change; FGFR, f ibroblast growth factor receptor; FOXP3, forkhead box protein 3; GSEA, gene set enrichment analysis; ICI, immune checkpoint inhibitor; IHC, immunohistochemistry; MDSC, myeloid- derived suppressor cell; ORR, objective response rate; PD1, programmed cell death protein 1; RET, RET proto- oncogene; TGFβ, transforming growth factorβ; Treg, regulatory T cell. Received November 25, 2020; accepted June 13, 2021. Additional Supporting Information may be found at onlinelibrary.wiley.com/doi/10.1002/hep.32023/suppinfo. Supported by a grant from Eisai Inc. C.M. is supported by a Rio Hortega grant from Instituto de Salud Carlos III (ISCIII), Fondo Social Europeo, ID code CM19/00039. M.P. received a Juan Rodés scholarship grant from Asociación Española para el Estudio del Hígado (AEEH). P.K.H. is supported by the fellowship grant of the German Research Foundation (DFG; HA 8754/1- 1). U.B. is supported by a Juan Rodés Ph.D. student fellowship from the European Association for the Study of the Liver (EASL). C.E.W. is supported by a Sara Borrell fellowship (CD19/00109) from the ISCIII and Fondo Social Europeo. J.A.F. is supported by a doctoral training grant from the University of Barcelona (PREDOCS- UB) and by a mobility grant from Fundació Universitària Agustí Pedro i Pons. S.L.F. is supported by NIH RO1- DK56621, NIH R01- DK128289- 01, and the U.S. Department of Defense (CA150272P1). D.S. is supported by the Gilead Sciences Research Scholar Program in Liver Disease. D.A.M. is supported by CRUK grants C18342/A23390 and C9380/A26813. J.M.L. is supported by grants from the Samuel Waxman Cancer Research RESULTS 122 HEPATOLOGY, Vol. 74, No. 5, 2021 TORRENS ET AL. 2653 cases.(2) Around 40% of HCC patients are diag- nosed at advanced stages of the disease, in which the tyrosine kinase inhibitor (TKI) sorafenib, has been the only approved treatment for >10  years.(3) Only recently, lenvatinib has shown noninferiority com- pared to sorafenib and received U.S. Food and Drug Administration (FDA) approval for the treatment of advanced HCC.(4) On the other hand, immunother- apies with immune checkpoint inhibitors (ICIs) are emerging as promising treatment options. For exam- ple, the new combination of the ICI, atezolizumab, plus bevacizumab (VEGFA inhibitor) has been FDA approved as first- line therapy(5) and pembrolizumab and nivolumab plus ipilimumab as second- line treatments.(6- 8) Although ICIs are changing the landscape of can- cer medicine, responses in HCC are only observed in 15%- 20% of patients, whereas the majority are pri- marily resistant. Thus, much effort has been invested into identifying existing kinase inhibitors that can effectively synergize with ICIs. In this context, lenva- tinib plus the anti- PD1 ICI, pembrolizumab, is cur- rently being tested in unresectable HCC in phase Ib and III trials, with an encouraging objective response rate (ORR) of 46%, 22- month median survival, and 9.5- month median progression- free survival.(9) This strategy is based on the hypothesis that lenvatinib could inhibit the immunosuppressive and proangio- genic effect of the VEGFA- VEGFR pathway on the tumor microenvironment,(10,11) thus boosting anti- tumor immune response and improving the clini- cal benefit of anti- PD1. In this regard, experimental studies conducted in HCC models have recently sug- gested a link between lenvatinib and inflammation as well as a greater antitumoral effect of the combination treatment.(12,13) However, a deeper understanding of the immunomodulatory capacity of these treatments is still needed. To fill this gap, we generated three murine immu- nocompetent models of HCC— two subcutaneous and an orthotopic— and assessed the antitumoral activity of lenvatinib alone or in combination with anti- PD1 and its effects on systemic and tumor- infiltrating immune cells. By performing flow cytom- etry, transcriptomic, and immunohistochemistry (IHC) analyses, we explored the immunomodulatory Foundation, NIH R01 DK128289- 01, the Spanish National Health Institute (MICINN; PID2019- 105378RB- I00), the Generalitat de Catalunya (AGAUR, SGR- 1358), and through a partnership between Cancer Research UK, Fondazione AIRC, and Fundación Científ ica de la Asociacion Española Contra el Cáncer (HUNTER, Ref. C9380/A26813). © 2021 by the American Association for the Study of Liver Diseases. View this article online at wileyonlinelibrary.com. DOI 10.1002/hep.32023 Potential conflict of interest: Dr. Leslie owns stock in Fibrof ind. Dr. Llovet consults for and received grants from Bayer, Eisai, Boehringer Ingelheim, Bristol- Myers Squibb, and Ipsen. He consults for Celsion, Eli Lilly, Merck, Roche, Genentech, Glycotest, Nucleix, Can- Fite, Sirtex, AstraZeneca, and Mina Alpha. Dr. Friedman consults for, received grants from, and owns stock in Morphic Therapeutics and Galmed. He consults for and owns stock in Blade, Escient, Glympse, North Sea, Scholar Rock, and Surrozen. He consults for 89 Bio, Amgen, Axcella, Bristol- Myers Squibb, Can- Fite, ChemomAb, Forbion, Gordion, Glycotest, In sitro, Novartis, Ono, and Pf izer. He received grants from Novo Nordisk and Abalone. He owns stock in Galectin, Genf it, Lifemax, Metacrine, Nimbus, Intercept, Madrigal, and Group K. Dr. Mann is employed by and owns stock in Fibrof ind. He received grants from GlaxoSmithKline. ARTICLE INFORMATION: From the 1 Mount Sinai Liver Cancer Program,  Division of Liver Diseases,  Tisch Cancer Institute,  Icahn School of Medicine at Mount Sinai, New York, NY; 2 Translational Research in Hepatic Oncology,  Liver Unit,  Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS)- Hospital Clínic,  Universitat De Barcelona, Barcelona, Spain; 3 Hepatology Section,  Gastroenterology Department,  Parc de Salut Mar,  IMIM (Hospital del Mar Medical Research Institute), Barcelona, Spain; 4 Newcastle Fibrosis Research Group,  Biosciences Institute,  Newcastle University, Newcastle upon Tyne, United Kingdom; 5 Institució Catalana De Recerca i Estudis Avançats, Barcelona, Spain. ADDRESS CORRESPONDENCE AND REPRINT REQUESTS TO: Josep M. Llovet, M.D. Liver Cancer Translational Research Laboratory, Liver Unit IDIBAPS- Hospital Clinic, Faculty of Medicine University of Barcelona Rosselló 153 08036, Barcelona, Catalonia, Spain E- mail: jmllovet@clinic.cat Tel.: 0034- 932- 279- 155 RESULTS 123 HEPATOLOGY, November 2021TORRENS ET AL. 2654 effect of each treatment and unveiled noteworthy effects of the combination. Gene expression analysis of murine and human tumors allowed the identifi- cation of a subset of HCC patients (~22%) who are likely to exhibit primary resistance to single agents, but could potentially be rescued with the combina- tion treatment. Overall, these findings provide a comprehensive understanding of the immune- remodeling capacities of the combination of lenvatinib plus anti- PD1, with important implications for patient selection to ulti- mately maximize the clinical benefit from this treat- ment regimen. Materials and Methods SUBCUTANEOUS SYNGENEIC MOUSE MODELS Two subcutaneous syngeneic HCC models were generated by injecting 5  ×  106 Hepa1- 6 cells (ATCC, Manassas, VA) in 100 μL of PBS in 5- to 6- week- old female C57BL/6J mice (n  =  59; Charles River Laboratories, Wilmington, MA) and 5  ×  106 Hep53.4 cells (CLS, Eppelheim, Germany) in 5- to 6- week- old male C57BL/6J mice (n = 40).(14) Animals were weighed, and tumor volume was assessed three times per week. Once tumors reached 200 mm3, animals were randomly assigned to receive lenvatinib (Eisai, Ibaraki, Japan), anti- PD1 (anti- murine PD- 1 monoclonal antibody clone J43 BioXCell, San Diego, CA, BE0033- 2), combination therapy (lenvatinib plus anti- PD1), or placebo (drug vehicle plus polyclonal IgG, BioXCell BE0091; Fig. 1A). Mice from the Hepa1- 6 model were euthanized at day 13 postrandomization (early time point, n = 20), once a tumor volume of 1,000 mm3 was reached or at study termination (late time point, n  =  39). The Hep53.4 model was used as validation, and all animals were euthanized at day 13 postrandomization. Tumor and blood samples were collected and processed for subsequent analyses (Figs. 1A and 2A). Assessment of the tumorigenic and histological features in an addi- tional set of male and female mice (n = 30) revealed no sex differences in our model (Supporting Fig. S1). Studies were performed in compliance with guidelines for the use of animals established by the institution ethical committee and the Guide for the Care and Use of Laboratory Animals. ORTHOTOPIC SYNGENEIC MOUSE MODEL To generate the orthotopic model, 5  ×  106 luciferase- transfected Hep53.4 cells were implanted in liver of 5- to 6- week- old male C57BL/6J mice. Tumor growth was serially assessed using the In Vivo Imaging System once per week. Only mice with biolumines- cence values >107 counts before randomization were included in the study (n = 33). Animals were random- ized 2 weeks after implantation to receive lenvatinib, anti- PD1, combination therapy, or placebo. Mice were euthanized at day 13 postrandomization (Fig. 2A), and tumor burden was assessed ex vivo in liver samples. MULTICOLOR FLOW CYTOMETRY ANALYSIS Flow cytometry analysis was performed on tumor and blood samples from the Hepa1- 6 model col- lected at the early time point. After sample process- ing, ~1 × 106 freshly prepared cells were stained with fluorochrome- coupled antibodies targeting cell mark- ers. Three antibody panels were designed to detect lymphocyte and myeloid cell populations of interest (Supporting Table S1; Supporting Fig. S2). Cells were stained according to standard flow cytometer proto- cols (Supporting Tables S2 and S3). Fluorescence data from 50,000 events per sample were collected on a LSRII cytometer (BD Biosciences, Franklin Lakes, NJ), available at Icahn School of Medicine at Mount Sinai (New York, NY) facilities and analyzed using FlowJo Flow Cytometry analysis software. HISTOLOGICAL AND IHC ANALYSIS OF TUMOR SAMPLES Tumor samples were fixed in buffered 4% parafor- maldehyde for 24 hours and underwent tissue process- ing and embedding in paraffin to create formalin- fixed, paraffin- embedded blocks. In the Hepa1- 6 model, only tumors with sufficient tumoral material were processed for histological analysis (n = 14 and 17 in the early and late time points, respectively). Twenty- four tumors from the Hep53.4 subcutaneous model were used for validation. Tumor viability, defined as the proportion of tumor- presenting viable cells in a sample (i.e., excluding necrotic regions or granulation tissue), was assessed on RESULTS 124 HEPATOLOGY, Vol. 74, No. 5, 2021 TORRENS ET AL. 2655 C um S ur vi va l 1.00 0.75 0.50 0.25 0.00 0 20 40 60 80 100 120 Follow-up (days)Number at risk Placebo Lenvatinib Anti-PD1 Combination 10 10 10 9 0 9 20 2 9 40 2 9 8 60 80 100 120 2 8 9 8 1 7 9 8 1 6 9 8 Follow-up days Number at risk Follow-up days Treatment Placebo Lenvatinib Anti-PD1 Combination Follow-up (days) 15 15 14 15 14 13 12 10 150 5 20 25 15 13 8 1 10 7 2 0 10 5 1 0 10 4 1 0 A B DC E Hepa1-6 subcutaneous model (n = 59) Survival (time to 1000 mm³) Time to objective response Lenvatinib (10mg/kg, daily) Anti-PD1 (10mg/kg, q3d5) Day 0 3 6 9 12 Injection Hep1-6 Tumor (200 mm³) Early timepoint - Flow cytometry - Gene expression array - Histological analysis Late timepoint - Response to treatment - Histological analysis 1000 mm³ tumor volume Tumor growth Tu m or v ol um e (m m 3) Days post randomization 800 600 400 200 0 0 5 10 15 ns p<0.01 O R P ro ba bi lit y ns 1.00 0.75 0.50 0.25 0.00 0 5 10 15 20 25 Treatment Placebo Lenvatinib Anti-PD1 Combination % C ha ng e in tu rm or v ol um e 100 80 60 40 20 0 −20 −40 −60 −80 −100 PD SD OR Response to treatment (early timepoint) Tumor response rate Placebo Lenvatinib Anti-PD1 Combination PD SD OR 11 4 0 3 9 3 1 0 4 0 10 14 159 9 9 9 9 9 Placebo Lenvatinib Anti-PD1 Combination Treatment Placebo Lenvatinib Anti-PD1 Combination RESULTS 125 HEPATOLOGY, November 2021TORRENS ET AL. 2656 hematoxylin and eosin slides. Further histopatholog- ical examination was performed by IHC (Supporting Table S3). All analyses were performed by an expert pathologist blinded to the treatment arms. TRANSCRIPTOMIC ANALYSIS OF TUMOR SAMPLES Tumor samples from the Hepa1- 6 model col- lected at the early time point or from mice reaching the survival endpoint at day 13 postrandomization (n  =  21) were processed for transcriptome analysis. Gene expression microarray studies were conducted using the Clariom S Mouse Array (GSE15 3203; Affymetrix, Santa Clara, CA). The combination rescue signature was generated by selecting the top differentially expressed genes between tumors from the combination and placebo arms (Bonferroni, P  <  0.05; fold change [FC], >3 or <0.33). Genes significantly enriched in the monother- apy groups compared to placebo according to the same criteria were eliminated from the signature to capture only the combination- specific transcriptomic effect. IDENTIFICATION OF POTENTIAL RESPONDERS TO COMBINATION TREATMENT We analyzed gene expression data from a cohort of 228 surgically resected fresh- frozen HCC samples (Heptromic data set, GSE63898) previously collected in the setting of the HCC Genomic Consortium.(15,16) A gene signature was generated to identify candidates that might respond to the combination beyond what is expected for single treatment effect alone. STATISTICAL ANALYSES Statistical analysis was conducted using R (ver- sion 3.6.2; R Foundation for Statistical Computing, Vienna, Austria) or GraphPad Prism software (ver- sion 5.01; GraphPad Software Inc., San Diego, CA). Comparison of continuous variables was performed using Kruskal- Wallis and Dunn’s tests for nonpara- metric distributions or ANOVA and Tukey tests for parametric distributions. Correlations for categorical variables were analyzed by Fisher’s exact test. Survival and time to response were assessed with Kaplan- Meier estimates and the log- rank test. Additional detailed information is provided in the Supporting Materials and Methods. Results ANTITUMOR ACTIVITY OF LENVATINIB AND ANTI- PD1 COMBINATION IN A SYNGENEIC MURINE MODEL To evaluate the antitumor activity of lenvatinib, anti- PD1 immunotherapy, and its combination, we generated three syngeneic murine HCC models(14) (Figs. 1A and 2A). Median tumor volume at random- ization was equal in all treatment arms, and no signif- icant differences in body weight or other toxicity signs were observed, indicating that all treatments were well tolerated (Supporting Fig. S3A,B). In the Hepa1- 6 model, mice treated with lenvatinib, anti- PD1, or its combination exhibited a significant reduction in tumor growth and improved survival compared to the placebo arm (median survival of 29  days in placebo and not reached [NR] in the remaining groups; P < 0.0001; Fig. 1B,C). Both anti- PD1 and combination treatments showed higher antitumor efficacy than lenvatinib, but no significant differences were observed between the anti- PD1 and combination arms. However, mice receiving combination treatment required a shorter time to objective responses compared to anti- PD1, lenvatinib, and placebo (median 5, 11, and 20 days and NR, respectively; P < 0.0001; Fig. 1D). Notably, the combination treatment achieved a sig- nificantly higher ORR compared to both lenvatinib and anti- PD1 at the early time point (P  <  0.05; Fig. 1E). At the late time point, the best ORR was also achieved by the combination treatment (P < 0.001 vs. FIG. 1. Antitumoral effect of lenvatinib plus anti- PD1 in the Hepa1- 6 model. (A) Timeline of the study. (B) Tumor growth, (C) survival, and (D) time to objective response of treated mice. (E) Response to treatment at early time point (n = 59). Upper part indicates differences in progressive disease rate, and lower part the differences in objective response rate. Right table shows the number of mice per group and response. *P < 0.05; **P < 0.01; ***P < 0.001 versus placebo (unless indicated); #P < 0.05; ##P < 0.01; ###P < 0.001 versus lenvatinib. Abbreviations: Cum, cumulative; OR, objective response; PD, progressive disease; SD, stable disease. RESULTS 126 HEPATOLOGY, Vol. 74, No. 5, 2021 TORRENS ET AL. 2657 1XPEHUDWULVN 3ODFHER /HQYDWLQLE $QWL3' &RPELQDWLRQ                             )ROORZXSGD\V    7UHDWPHQW 3ODFHER /HQYDWLQLE $QWL3' &RPELQDWLRQ        )ROORZXS GD\V 2 5 3 UR ED EL OLW \          7UHDWPHQW 3ODFHER /HQYDWLQLE $QWL3' &RPELQDWLRQ A B C D E F 7UHDWPHQW 3ODFHER /HQYDWLQLE $QWL3' &RPELQDWLRQ 7LPHWRREMHFWLYHUHVSRQVH 7XPRUJURZWK 7XPRUJURZWK +HS 6& 1  /HQYDWLQLE PJNJGDLO\ $QWL3' PJNJTG 'D\ ,QMHFWLRQ +HS 7XPRU PPñ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ovember 2021TORRENS ET AL. 2658 placebo; Supporting Fig. S3C). All treatments induced a reduction in the progressive disease rate compared to placebo at both time points. The subcutaneous Hep53.4 model confirmed the significant enhanced antitumoral effect, fastest response, and higher ORR in the combi- nation treatment arm (Fig. 2B,C and Supporting Fig. S4A). Similarly, in the orthotopic Hep53.4 model, the combination treatment achieved the greatest reduction in tumor growth, although differences compared to anti- PD1 were nonsignificant (mean FC, 2.7 and 1.6 in anti- PD1 and combination, respectively; P = 0.156; Fig, 2E,F and Supporting Fig. S4B). Finally, tumor viability was assessed to further under- stand the antitumoral effect of the treatments. Only the combination treatment significantly reduced tumor via- bility at both time points of the Hepa1- 6 model (P < 0.01 vs. placebo; Supporting Fig. S3D- F) and in the subcuta- neous Hep53.4 model (P < 0.001; Fig. 2D). Tumor viabil- ity was also significantly lower than in the anti- PD1 arm at the Hep53.4 model and Hepa1- 6 late endpoint (all P < 0.05). The combination treatment reduced cell pro- liferation compared to placebo and anti- PD1 (P < 0.05). Therefore, although no differences in tumor growth were detected between the two treatments, tumors from the combination arm were less viable and contained higher proportions of necrotic and granulation tissue. Overall, all treatments presented antitumoral activity, but the combination of lenvatinib plus anti- PD1 reduced the time to treatment response and tumor viability, and improved ORR compared to the monotherapies. MECHANISM OF ACTION OF LENVATINIB AND ANTI- PD1 COMBINATION The Combination Treatment Elicits an Immune- Activating Lymphocytic Infiltrate The presence of the main lymphocytic popula- tions in tumor samples from each treatment arm in the Hepa1- 6 model was assessed by flow cytometry. Tumors from animals treated with anti- PD1 and com- bination therapies contained a significantly increased intratumoral T- cell infiltrate and a reduction in the PD1+CD8+ T- cell proportion, indicating that the anti- PD1 treatment reached the tumor. Both lenva- tinib and combination treatments reduced the intra- tumoral regulatory T cell (Treg) infiltrate (P  <  0.05 vs. placebo; Fig. 3A; Supporting Table S4). Only the combination treatment was able to enhance the CD8+ T- cell to Treg ratio and proportion of proliferating CD8 T cells, ultimately indicating an increase in the tumoral proinflammatory immune cell component. No differences in the CD4+ T- cell and CD8+ T- cell or B- cell populations were detected between treat- ments. Finally, the analysis of the immune cell pop- ulations in blood samples did not reveal differences in systemic immune response among treatment arms (Supporting Fig. S5A; Supporting Table S4). To study the mid- and long- term effects of the treat- ments, CD8, CD4 T cells, and Treg cells were analyzed by immunostaining in tumor samples collected at the early and late time points. Interestingly, CD4 staining was exclusively located in the intratumoral area in the combination arm compared to a mainly peritumoral location in the placebo and lenvatinib arms (P < 0.05; Supporting Fig. S6), which could potentially induce a more effective antitumoral immune response.(17) CD8 lymphocytes were mostly intratumoral, with no significant differences between treatment arms. Next, forkhead box protein 3 (FOXP3) staining indicated a significant Treg decrease in the combination group at the late time point and Treg exclusion from the intra- tumoral region in the combination arm compared to placebo (P < 0.05; Supporting Fig. S7A- C). Together with the flow cytometry data, this indicates that len- vatinib and combination treatments might alter the proportion of Treg cells in the lymphocytic infiltrate and its location. No clear association between Treg cell localization and tumor vasculature was observed (Supporting Fig. S7D,E). FIG. 2. Antitumoral effect of lenvatinib plus anti- PD1 in the subcutaneous and orthotopic Hep53.4 models. (A) Timeline of the subcutaneous and orthotopic studies. (B) Tumor growth and (C) time to objective response of treated mice from the subcutaneous model. (D) Tumor viability assessed in H&E slides from the subcutaneous model. Representative images captured at 20×. (E) Tumor growth in the orthotopic model, measured as changes in bioluminescence compared to the start of the treatment. (F) Tumor volume measured ex vivo in liver samples. Box plots indicate median and quartiles. *P < 0.05; **P < 0.01; ***P < 0.001 versus placebo (unless indicated); #P < 0.05; ##P < 0.01; ###P < 0.001 versus lenvatinib; +++P < 0.001 versus anti- PD1. Abbreviations: FC, fold change; OR, objective response; SC, subcutaneous. RESULTS 128 HEPATOLOGY, Vol. 74, No. 5, 2021 TORRENS ET AL. 2659 FIG. 3. Immune cell populations in tumor samples detected by flow cytometry analysis. (A) Lymphoid and (B) myeloid immune cell populations from tumor samples collected at the early time point. Results for each treatment arm are shown (n = 5 samples per arm). Box plots indicate median and quartiles. *P < 0.05; **P < 0.01. Abbreviations: MDSC, myeloid- derived suppressor cells; Treg, regulatory T cell. Tumor lymphocyte panelA B Tumor myeloid panel CD4+ T cellsT cells Regulatory T cells sllecT+8DCgnitarefilorPsllecT+8DC Proliferating CD8+ T cells oitargerT/llecT+8DCsllecBsllecT+8DC+1DP segahporcaMCSDM Type 1 dendritic cells % C D4 5 % C D4 5 % C D4 5 % C D4 5 % C D4 5 % C D8 % C D4 5 % C D4 5 % C D4 5 % C D4 5 % C D4 5 % C D4 5 100 75 50 25 0 25 20 15 10 5 0 6 4 2 0 60 40 20 0 30 20 10 0 60 40 20 0 30 20 10 0 40 30 20 10 0 40 30 20 10 0 30 20 10 0 10.0 7.5 5.0 2.5 0.0 6 4 2 0 Pla ce bo Le nv ati nib An ti-P D1 Co mb ina tio n Pla ce bo Le nv ati nib An ti-P D1 Co mb ina tio n Pla ce bo Le nv ati nib An ti-P D1 Co mb ina tio n Pla ce bo Le nv ati nib An ti-P D1 Co mb ina tio n Pla ce bo Le nv ati nib An ti-P D1 Co mb ina tio n Pla ce bo Le nv ati nib An ti-P D1 Co mb ina tio n Pla ce bo Le nv ati nib An ti-P D1 Co mb ina tio n Pla ce bo Le nv ati nib An ti-P D1 Co mb ina tio n Pla ce bo Le nv ati nib An ti-P D1 Co mb ina tio n Pla ce bo Le nv ati nib An ti-P D1 Co mb ina tio n Pla ce bo Le nv ati nib An ti-P D1 Co mb ina tio n Pla ce bo Le nv ati nib An ti-P D1 Co mb ina tio n n.s. n.s. n.s. n.s. n.s. n.s. RESULTS 129 HEPATOLOGY, November 2021TORRENS ET AL. 2660 This histological profile was validated in the Hep53.4 subcutaneous model, which confirmed a reduction in FOXP3 staining in both lenvati- nib and combination arms compared to placebo (P < 0.01) and shift toward a peritumoral localiza- tion in the combination group (P < 0.05 vs. placebo; Fig. 4). Despite that no significant differences in CD3, CD8, and CD4 staining were detected, there was an increase in CD8 intratumoral location in tumors from the anti- PD1 and combination arms (Supporting Fig. S6D- F). This immunological pro- file aligns with a more active intratumoral infiltrate in the combination group. Taken together, the treatments were able to mod- ify the lymphoid infiltrate in the tumor. Interestingly, anti- PD1 increased the T- cell infiltrate, and lenvati- nib reduced the proportion and intratumoral location of Treg cells, associated with immune suppression.(18) Consequently, the combination treatment achieved a greater tumoral proinflammatory immune cell component. Impact of Treatments on Intratumoral and Systemic Myeloid Populations The analysis of the intratumoral myeloid populations revealed that anti- PD1 and combination treatments increased the type 1 dendritic cell (DC1) infiltrate compared to placebo (P  <  0.01; Fig. 3B; Supporting Table S4). No significant differences were observed in the proportion of infiltrating macrophages and myeloid- derived suppressor cells (MDSCs). However, further IHC analysis revealed that anti- PD1 increased the percentage of M2 macrophages at the early time point (P < 0.05 vs. placebo), whereas the combination treatment did not alter this immunosuppressive pop- ulation compared to placebo (Supporting Fig. S8A). This effect was not observed at the late time point. The circulating myeloid populations were not significantly modified by the treatments (Supporting Fig. S5B). Overall, we observed that anti- PD1 and combination treatments increased tumor infiltrating DC1, the main stimulator of T- cell function.(19) The immune- suppressive FIG. 4. Histological analysis of Treg tumor infiltrate in the Hep53.4 subcutaneous model. (A) Percentage of positive cells for FOXP3 staining in tumor samples from treated animals. (B) Percentage of samples with intratumoral or peripheral FOXP3 staining. (C) Representative images of FOXP3 staining captured with 40× magnification. Box plots indicate median and quartiles. *P < 0.05; **P < 0.01; ***P < 0.001. A B C noitanibmoC1DP-itnAbinitavneLobecalP Pla ce bo Le nv ati nib An ti-P D1 Co mb ina tio n Pla ce bo Le nv ati nib An ti-P D1 Co mb ina tio n 100 75 50 25 0 10 5 0 FOXP3 staining location Periphery Intratumoral Location % s am pl es FOXP3 % p os iti ve RESULTS 130 HEPATOLOGY, Vol. 74, No. 5, 2021 TORRENS ET AL. 2661 M2 macrophage component was increased by anti- PD1 treatment only at the early time point, indicating that this effect could be lost once treatment is suspended. Combination Treatment Significantly Alters the Tumor Inflammatory and Proliferative Gene Expression Profile We next sought to investigate the molecular pro- file of Hepa1- 6 tumors from each treatment arm. Principal component analysis showed that murine tumors clustered together with HCC human tumors from a cohort of 228 surgically resected fresh- frozen HCCs (Heptromic data set),(16) indi- cating that the syngeneic model was able to reca- pitulate the transcriptomic characteristics of human HCC (Supporting Fig. S9A). In mice treated with the combination, there were 1,265, 1,621, and 30 significantly differentially expressed genes com- pared to placebo, lenvatinib, and anti- PD1, respec- tively (Supporting Fig. S9B; Supporting Table S5). Notably, 68.8% (870 of 1,265) of the differentially expressed genes in the combination arm compared to placebo were not altered by the monotherapies. This noteworthy molecular effect of the combination treatment could be attributable to a synergistic effect at the transcriptional level. Pathway analysis revealed that the genes solely deregulated by the combina- tion treatment were associated with activation of proinflammatory pathways (i.e., T- cell, B- cell, and chemokine signaling; Supporting Table S6). By comparing the expression profile of each treatment arm, we observed that tumors from the combination arm had fewer differentially expressed genes associated with proliferation and cell- cycle progression (P  <  0.05 vs. placebo; Fig. 5A). In addition, in both anti- PD1 and combination treat- ments, there was enhanced inflammatory signal- ing (e.g., T- cell receptor and chemokine signaling, inflammatory response, and dendritic cell [DC] chemotaxis). Lenvatinib also showed a more mod- est but significant enrichment of proinflammatory gene sets (P  <  0.05 vs. placebo). Of note, lenvati- nib was able to inhibit signaling pathways down- stream of its main targets, including VEGFR, RET proto- oncogene (RET), and fibroblast growth factor receptor (FGFR) 2 (Supporting Fig. S9C). Combination Therapy Elicits Immune Activation and Down- Regulation of TGFß Signaling We then assessed the immunological profile of our samples using the recently described HCC immune class.(15,20) Interestingly, only the tumors from the combination arm were enriched in the immune- active class (4 of 5 vs. 0 of 5 in anti- PD1; P < 0.05), asso- ciated with an adaptive T- cell response activation, whereas all anti- PD1 tumors showed an immune- exhausted profile (5 of 5 vs. 1 of 5 in combination; P  <  0.05), which is characterized by activation of immunosuppressive signaling hampering the antitu- moral immune response (Fig. 5A). Samples from the placebo and lenvatinib groups were nonimmune, con- sistent with the lower T- cell infiltration observed by flow cytometry. Subclass mapping analysis confirmed that tumors from the combination group showed genetic similarity to human HCCs belonging only to the immune- active class, whereas anti- PD1- treated tumors showed a similarity with exhausted tumors as well (Fig. 5B). Gene expression data were then used to character- ize the impact of the treatments on the tumor compo- sition. Anti- PD1 and combination treatments reduced the malignant component of the tumors and increased the immune infiltrate (Fig. 6A,B), whereas tumors from the placebo and lenvatinib arms had higher estimated tumor purity. In line with flow cytometry and IHC results, tumors from the anti- PD1 and combination arms contained an enrichment of gene sets associated with T- cell and DC infiltrates (Fig. 6A). The mac- rophage expression profile revealed an increased M1 proportion in the combination arm whereas anti- PD1 tumors were enriched in M2. The analysis of immu- nosuppressive signaling in tumors revealed that the combination treatment inhibited transforming growth factor β (TGFß) signaling and reduced the expression of Wnt ligands, likely impairing Wnt/β- catenin sig- naling in the microenvironment (Fig. 5A). Lenvatinib alone also displayed an immunomodulatory effect by reducing TGFß signaling (P < 0.05 vs. placebo). The reduced TGFß signaling in these treatment arms could be linked to the decreased proportion of Tregs.(18) In addition, gene set enrichment analysis (GSEA) con- firmed an increase in proinflammatory signaling by RESULTS 131 HEPATOLOGY, November 2021TORRENS ET AL. 2662 all treatments (false discovery rate [FDR], <0.05 vs. placebo). Compared to anti- PD1, the combination induced down- regulation of gene sets associated with resistance to ICI and TGFß signaling and increased proinflammatory signaling (FDR < 0.05; Supporting Fig. S9D- G; Supporting Table S7). To investigate the impact of the treatments on the tumoral expression of checkpoint inhibitors, the expres- sion of PD1, programmed death ligand 1 (PDL1), and cytotoxic T- lymphocyte- associated protein 4 (CTLA4) was assessed by IHC (Supporting Fig. S8B- D). CTLA4 positivity was increased in tumors from the anti- PD1 and combination groups at the early time point (P < 0.05 in combination vs. placebo; n.s. in anti- PD1 vs. placebo). At the late time point, increased CTLA4 expression was maintained only in the anti- PD1 arm (P < 0.001 vs. placebo). No significant differences were observed in PD1 and PDL1 expression between groups. FIG. 5. Gene expression profile of treated tumors. (A) Transcriptomic and immunological profile of tumors from each treatment arm. (B) Subclass mapping analysis showing the transcriptomic similarity between tumors from each treatment arm and human HCC classified according to the HCC immune class. *P < 0.05; **P < 0.01; ***P < 0.001. Abbreviations: C, combination; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; L, lenvatinib; NA, not available; P, placebo; PD1, anti- PD1. A B Cell proliferation Inflammation TGFß pathway Wnt/CTNNB1 pathway Chiang classification Hallmark G2M checkpoint Biocarta cell cycle pathway Cell proliferation GO 0008283 Cyclin D1 KE V1 Hallmark inflammatory response KEGG cytokine receptor interaction KEGG chemokine signaling pathway GO positive regulation of antigen processing and presentation KEGG T cell receptor signaling pathway GO regulation of T cell migration GO dendritic cell chemotaxis HCC immune class KEGG TGFß signaling pathway GO positive regulation of SMAD protein import into nucleus GO regulation of SMAD protein import into nucleus Fibroblast response to TGFß T cell response to TGFß Hallmark Wnt ß-catenin signaling GO positive regulation of Wnt signaling pathway GO regulation of canonical Wnt signaling pathway ß-catenin 100 UP V1 UP WNT3 WNT6 WNT10A Placebo Lenva Anti-PD1 Combo p-value C-P C-L C-PD1 ns ns ns ns ns nsns ns ns ns nsns nsns ns ns ns ns ns ns M ou se tr ea tm en t Placebo Lenvatinib Anti-PD1 Combination R es t Ac tiv e Ex ha us te d P>0.01 P>0.01 P>0.01 Bonferroni p 1 0.8 0.6 0.4 0.2 0 Heptromic immune class High Low Positive Negative Immune Sia Active Intermediate Exhausted Excluded Chiang classification Wnt/CTNNB1 IFN Proliferation Poly 7 Unknown/ NA RESULTS 132 HEPATOLOGY, Vol. 74, No. 5, 2021 TORRENS ET AL. 2663 Overall, though all treatments were able to impact the molecular and immunological profile of tumors, only the combination treatment induced an immuno- modulatory active profile associated with proinflamma- tory signaling and TGFß inhibition (Fig. 6C). Lenvatinib Suppresses Aberrant Angiogenesis in Tumors Given that lenvatinib is a known antiangiogenic agent, the capacity of the treatments to reduce aberrant vasculature in tumors was investigated. In the GSEA analysis, lenvatinib and combination groups showed a reduction in the expression of endothelial- related genes, which could be indicative of decreased angio- genesis (Supporting Fig. S9D,F). IHC analysis showed a reduction in the amount of vessels encapsulating tumor clusters (VETCs) in the lenvatinib and com- bination arms, reaching significance at the late time point (P  <  0.05 vs. placebo; Supporting Fig. S8E,F). VETCs have been proposed as a predictor of aggressive HCC(21) and could be linked to the antitumoral effect of lenvatinib. CD31 staining showed similar results (Supporting Fig. 8G). At the late time point, only the combination group maintained a significant reduction in CD31 expression (P  <  0.05 vs. placebo). Globally, our results indicate that lenvatinib and combination treatments exerted an antiangiogenic effect (Fig. 6C). FIG. 6. Characterization of tumor composition based on gene expression data. (A) Tumor composition assessed by ESTIMATE analysis or ssGSEA capturing distinct cell populations. (B) Immune score and estimated tumor purity in tumors from each treatment arm measured by ESTIMATE analysis. Box plots indicate median and quartiles. (C) Summary of the molecular and immunological effects of lenvatinib, anti- PD1, and combination treatment on the tumor. *P < 0.05; **P < 0.01; ***P < 0.001. Abbreviations: C, combination; ESTIMATE, estimation of stromal and immune cells in malignant tumor tissues using expression data; iDC, immature dendritic cell; L, lenvatinib; NK, natural killer; P, placebo; PD1; anti- PD1; ssGSEA, single- sample gene set enrichment analysis; Th, T helper, Treg, regulatory T cell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ȕ 7FHOOLQILOWUDWH '&LQILOWUDWH &RPELQDWLRQ ,PPXQHUHVSRQVH DFWLYDWLRQ (VWLPDWHGWXPRUSXULW\,PPXQHVFRUH S  U HO DW LY H WX P RU S XU LW\     3O DF HE R /H QY DW LQL E $Q WL 3'  &R P ELQ DW LRQ 3O DF HE R /H QY DW LQL E $Q WL 3'  &R P ELQ DW LRQ     ( QU LF KP HQ WV FR UH RESULTS 133 HEPATOLOGY, November 2021TORRENS ET AL. 2664 ENRICHMENT OF POTENTIAL RESPONDERS TO COMBINATION TREATMENT IN A HUMAN COHORT OF HCC A Gene Signature Capturing the Molecular Effects of the Combination Treatment Identifies Potential Responders Previous studies from our group suggested that the HCC immune class could be able to predict response to ICIs.(15) Based on our results defining the molec- ular impact of the combination therapy, we sought a potential biomarker capable of identifying patients that are not likely to respond to anti- PD1 alone, but could benefit from the combination treatment. To this end, the combination rescue signature was gen- erated by selecting the top differentially expressed genes in the combination group compared to pla- cebo, but not altered by stand- alone monotherapies (Supporting Table S8). We hypothesize that human tumors with a similar gene expression profile to the combination- treated murine tumors according to our FIG. 7. Identification of HCC human tumors expressing the combination rescue signature. (A) Transcriptomic profile of HCC samples classified as HCC immune class (potential responders to ICIs)(15) or combination- only responder class. P values reflect the comparison of the combination- only responder class (green) and nonimmune tumors (white). (B) Estimated proportion of immune cells (CIBERSORT) in human tumors classified according to its potential response to therapies. Box plots indicate median and quartiles. *P < 0.05; **P < 0.01; ***P < 0.001. Abbreviations: GO, Gene Ontolog; ICI, immune checkpoint inhibitors; KEGG, Kyoto Encyclopedia of Genes and Genomes; NA, not available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ol. 74, No. 5, 2021 TORRENS ET AL. 2665 signature (i.e., those with high immune infiltration and activation) do not need the booster effect of the combination treatment because they could potentially respond to anti- PD1 or lenvatinib monotherapies.(15) On the other hand, samples with down- regulation of genes associated with the molecular effect of the combination therapy could benefit from the immune- activating effect of the combination treatment. Human tumors resembling the combination- treated murine tumors according to our signature were enriched in the HCC immune class (33 of 55 immune vs. 15 of 173 nonimmune; P  <  0.0001; Fig. 7A). The HCC immune class accounted for 24.1% of the cohort (55 of 228) as reported.(15) Conversely, 22.8% of the tumors (52 of 228) were classified as combination- only responders according to the combination rescue signature, indicat- ing that they are likely to be resistant to ICIs, but could be rescued with the combination treatment. There was no overlap of tumors expressing the HCC immune class and the combination- only responder class. Even if just patients in the immune class (24.1%) and combination- only responder class (22.8%) were considered, 46.9% (107 of 228) of the patients could potentially benefit from the combination treatment. Characterization of Tumors From the Combination- Only Responder Class Compared to the other nonimmune tumors, the combination- only responder class presented an enrich- ment of the S2 and G1 molecular classes,(22,23) asso- ciated with proliferation and progenitor- like features. There was no significant overlap with the immune excluded or catenin beta 1 (CTNNB1) classes,(20,24) indicating that the combination rescue signature is not recapitulating Wnt/ß- catenin activation (Figs. 7A and 8). On the other hand, the combination- only responder class displayed a significant activation of endothelial cell proliferation and VEGFR path- way. Thus, VEGF inhibition with lenvatinib could potentially boost the inflammatory status in these tumors.(10) These samples also presented activation of FGFR1- 4 and RET downstream pathways, which are also targeted by lenvatinib, although no significant differences compared to other nonimmune samples were found (Supporting Fig. S10). The combination- only responder class was also characterized by a significant decrease in proinflamma- tory signaling (e.g., T- cell receptor [TCR] activation, cytokine production, and antigen presentation) and in gene sets associated with response to ICIs and T- cell infiltrate. Interestingly, we observed an enrichment of gene sets associated with Treg cells, which further highlights the relevance of this cell population in the immune- remodeling effect of the combination treat- ment. Cell- type abundance analysis by digital cytome- try (CIBERSORT) confirmed the differences in Treg infiltrate between the combination- only responder class and the other nonimmune samples (Fig. 7B). A reduc- tion of T- cell and macrophage infiltrates compared to the immune class was also observed. According to our data, the increased VEGF and Treg signaling in this group could be inhibited by the combination treat- ment, whereas other nonimmune tumors may present different mechanisms of immune suppression, which cannot be corrected with this therapeutic approach. Of note, the combination- only responder class was not associated with survival or other clinicopathological variables (Supporting Table S9). Overall, our signature identified ~22% of human HCC patients with gene expression deregulations that could potentially be restored by the combination therapy. These samples were characterized by reduced proinflam- matory signaling, high Treg levels, and VEGF signaling. Discussion Over the past several years, checkpoint- inhibitor– based immunotherapies have achieved unprecedented success for cancer treatment. However, responses are only observed in ~15%- 20% of HCC patients, high- lighting the need to identify either biomarkers of response or combination treatments that could increase their clinical benefit. Here, we generated three synge- neic models to assess the antitumoral, immunological, and molecular effects of combining anti- PD1 ICIs with the TKI, lenvatinib, currently approved for first- line treatment of advanced HCC.(9) We demonstrated that lenvatinib exerts an immunomodulatory effect, which, together with anti- PD1, could induce antitu- moral immune response activation, thus providing a mechanistic rationale for this combination. We also identified a group of human HCC tumors that dis- play features of primary resistance to ICIs, but could potentially benefit from the combination treatment. In the quest to identify combination strategies for cancer treatment, antiangiogenic agents and RESULTS 135 HEPATOLOGY, November 2021TORRENS ET AL. 2666 multikinase inhibitors are among the most notable candidates because of their immunomodulatory capac- ities.(10,25) The lenvatinib plus pembrolizumab combi- nation is a promising approach for HCC, with phase Ib data showing an unprecedented ORR of 46% and median survival of 22 months.(9) This combination is currently being assessed in a phase III trial compared to lenvatinib alone (NCT03713593). In our models, anti- PD1 and lenvatinib alone improved survival and decreased tumor growth; however, the combination treatment achieved a higher response rate, shorter time to response, and a reduction of tumor viability in accordance with a previous publication.(13) The VEGFR family is one of the main targets of lenvatinib. VEGFA- VEGFR pathway activation favors tumor growth, progression, and aberrant vas- culature formation.(10) More important, the VEGF pathway has a direct immunosuppressive effect on the tumor infiltrate by decreasing cytotoxic T- cell and DC function and promoting the recruitment of immu- nosuppressive cells, such as Tregs, M2 macrophages, and MDSCs.(10,26,27) In addition, the inhibition of other lenvatinib targets, such as FGFR1- 4, RET, and platelet- derived growth factor, could potentially have other immunological and molecular implications. Considering this, lenvatinib could boost the effects of ICIs on antitumor immune response by “releasing the brake” on inflammation. However, the immunomodu- latory capacity of lenvatinib alone or in combination with anti- PD1 still remains poorly characterized. Here, we demonstrate that lenvatinib exerts a potent effect on the immune infiltrate by reducing the Treg proportion and altering their intratumoral location. On the other hand, anti- PD1 induced an increase of T- cell and DC1 infiltrate. This is in accordance with human studies reporting T- cell recruitment fol- lowing PD1 blockade and association between DC1 and T- cell function.(19,28) The combined effect of the two therapeutic approaches induced an increase in the proinflammatory component in the tumor, associated with enhanced antitumor immunity.(29) Data from treated cancer patients indicate that intratumoral Tregs might limit anti- PD1 efficacy.(30,31) Therefore, the effect of lenvatinib in reducing this immunosup- pressive population could be responsible for the greater antitumoral capacity of the combination therapy. Previous experimental studies have suggested a link between lenvatinib and antitumor immune responses.(12,13) The combination treatment was able to increase the CD8 T- cell infiltrate and decrease the monocyte/macrophage component in an HCC murine model,(13) in line with previous studies assessing the effect of lenvatinib alone.(12) The study also reported a DC decrease, suggesting that although anti- PD1 promotes DC1 recruitment according to our data, this may not happen in all DC subpopulations. Our data indicate that the combination regimen elicited a reduction of the protumorigenic M2 phenotype compared to the anti- PD1 arm. This could be linked to the decreased Treg proportion in the combination arm, which reportedly promotes the M2 phenotype in tumor- infiltrated macrophages.(32) The reprogram- ming of the macrophage phenotype corresponds to in vivo data linking lenvatinib treatment with a reduc- tion in M2 macrophages in colon cancer.(12) Besides its effects in tumor, it has been proposed that VEGF can cause systemic immunosuppression.(10) However, no significant alterations were detected in circulating immune cells, suggesting that the effect of the treat- ments may be tumor specific. Transcriptomic analysis from tumor samples allowed us to assess the molecular and inflammatory changes induced by the treatments. Tumors from animals receiving the combination treatment showed a reduc- tion of pathways associated with proliferation, in line with its greater antitumoral capacity and histological analysis. Interestingly, lenvatinib blocked immunosup- pressive signaling through TGFß pathway inhibition, likely by decreasing the Treg proportion given that TGFß signaling is one of its key immunosuppres- sive mechanisms.(18) On the other hand, anti- PD1 increased the T- cell infiltrate and elicited an immune exhausted phenotype in tumor characterized by expres- sion of immune- suppressive pathways and Treg infil- trate,(18) in accordance with human studies reporting an increase in the exhausted T- cell component following PD1 blockade.(28) Therefore, the combined effect of lenvatinib plus anti- PD1 generated both an enhanced T- cell infiltrate and a reduction of immunosuppres- sive signaling (i.e., Treg infiltrate and TGFß pathway). Consequently, only the combination treatment gener- ated a specific activated antitumoral immune response. Anti- PD1 treatment also induced a deregulation in expression of checkpoint inhibitors. CTLA4 expression was increased in tumors treated with anti- PD1 and combination treatment, probably attributable to feed- back mechanisms induced by TCR stimulation.(33,34) Increased CTLA4 expression has also been associated RESULTS 136 HEPATOLOGY, Vol. 74, No. 5, 2021 TORRENS ET AL. 2667 with T- cell exhaustion.(18) Overall intratumoral PD1 positivity was not altered by any treatment. Of note, anti- PD1 and the combination treatment reduced the CD8+PD1+ T- cell infiltrate. This population has been recently proposed to limit the efficacy of ICIs in NASH- related HCC.(35) Therefore, the decrease of this population could be an important mechanism promoting response to these treatments. On the other hand, this result could also be influenced by the block- ade of the PD1 epitope by anti- PD1 treatment, rather than a reduction of this cell population in the infiltrate. The identification of the molecular and immunologi- cal effects induced by lenvatinib plus anti- PD1 is a step toward in identifying patients with primary resistance to ICI monotherapies who could benefit from the combi- nation treatment. In a previous study, we defined a gene- expression– based immune classifier able to identify 24% of HCC patients (immune class) with markers of T- cell infiltrate and molecular features similar to melanoma tumors most responsive to ICIs.(15) Using gene expres- sion data from our murine model, we here generated a signature capturing the transcriptomic modifications induced by the combination therapy, but not by mono- therapies as stand- alone therapies. The assessment of our signature in a human cohort of 228 samples showed that 22% of patients presented down- regulation of genes associated with the molecular effect of the combination, along with reduced proinflammatory signaling, high Treg levels, and VEGF signaling. Thus, tumors shar- ing these hallmarks could harbor primary resistance to anti- PD1, but potentially benefit from the booster effect of the combination treatment. Altogether, around half of patients could respond to combination therapies (Fig. 8). Considering that the gene signature has been gen- erated from on- treatment murine tumor samples, fur- ther studies extrapolating these findings to pretreatment HCC profiles will be required. Therefore, the predictive capacity of response of the signature will need to be val- idated in a prospective human cohort of HCC patients receiving lenvatinib plus an anti- PD1 ICI. In summary, our study provides a characterization of the antitumor, immunological, and molecular effects of lenvatinib, anti- PD1, and its combination in murine HCC models. We demonstrated that lenvatinib exerts an immunomodulatory effect on the tumor infiltrate associ- ated with a reduction in Tregs and inhibition of immune- suppressive pathways. Its combination with anti- PD1 favored the generation of an activated immune profile and faster response to treatment. Finally, the identifica- tion of the mechanisms underlying a beneficial effect of the combination therapy led to the generation of a gene signature present in ~20% of human HCC that cor- relates with high Treg infiltrate, VEGFR pathway acti- vation, and low inflammatory signaling. This signature might recognize patients likely to benefit most from this combination. Thus, further investigations are warranted to confirm whether the signature can be a tool to iden- tify HCC patients who may respond to the combination therapy beyond their responses to single- agent therapies. FIG. 8. HCC classification according to its immunological features and potential response to the combination therapy. Diagram summarizing the HCC immune classification and potential response to ICIs or combination treatment beyond single agents according to the combination rescue signature. Percentage of HCCs belonging to each class is shown in brackets. Abbreviations: ICI, Immune checkpoint inhibitors; Treg, regulatory T cell. Immune infiltrate Proinflammatory signaling Treg infiltrate VEGFR signaling Proinflammatory signaling Predicted response Molecular features Immune class Active Exhausted Responders to ICI Intermediate class Excluded class Responders to combination beyond single agents Total combination responders ~ 46%(~24%) (~22%) RESULTS 137 HEPATOLOGY, November 2021TORRENS ET AL. 2668 Acknowledgment: We thank Alice Kamphorst for her help and valuable inputs on experimental design and flow cytometry data analysis. We acknowledge the per- sonnel at the Icahn School of Medicine Flow Cytometry core for their help on experimental design and technical support. We acknowledge the technical assistance pro- vided by Jordi Farre, Samir Luli, and Rainie Cameron. Figures 2A and 6C were created with BioRender.com. Author Contributions: L.T., D.S., and J.M.L. were in- volved in the study conceptualization. L.T., C.M., M.P., A.M., J.L., P.K.H., M.M., U.B., C.E.W., J.A.F., M.P.G., and J.P. contributed to the investigation and methodol- ogy. C.M., A.M., and C.E.W. contributed to pathologi- cal characterization of tumor samples. L.T. and M.T.M. contributed to data curation and software analysis. L.T., C.M., M.P., P.K.H., A.M., C.E.W., J.A.F., M.T.M., and D.S. contributed to formal analysis. C.E.W., M.T.M., B.S., S.L.F., D.A.M., D.S., and J.M.L. provided super- vision. J.M.L. contributed to funding acquisition. L.T., D.S., and J.M.L. wrote the manuscript. All authors were involved in the critical review and editing of the manuscript. REFERENCES 1) Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global cancer statistics 2020: GLOBOCAN esti- mates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2021;71:209- 249. 2) Llovet JM, Kelley RK, Villanueva A, Singal AG, Pikarsky E, Roayaie S, et al. Hepatocellular carcinoma. Nat Rev Dis Primers 2021;7:6. 3) Llovet JM, Ricci S, Mazzaferro V, Hilgard P, Gane E, Blanc JF, et al. Sorafenib in advanced hepatocellular carcinoma. N Engl J Med 2008;359:378- 390. 4) Cheng AL, Finn RS, Qin S, Han KH, Ikeda K, Piscaglia F, et al. Phase III trial of lenvatinib (LEN) vs sorafenib (SOR) in first- line treatment of patients (pts) with unresectable hepatocellu- lar carcinoma (uHCC). J Clin Oncol 2017;35(Suppl.):Abstract 4001. 5) Finn RS, Qin S, Ikeda M, Galle PR, Ducreux M, Kim TY, et al. Atezolizumab plus bevacizumab in unresectable hepatocellular carcinoma. N Engl J Med 2020;382:1894- 1905. 6) Llovet JM, Villanueva A, Marrero JA, Schwartz M, Meyer T, Galle PR, et al. Trial design and endpoints in hepatocellular carcinoma: AASLD Consensus Conference. Hepatology 2021;73:158- 191. 7) Finn RS, Ryoo BY, Merle P, Kudo M, Bouattour M, Lim HY, et al. Results of KEYNOTE- 240: phase 3 study of pembrolizumab (Pembro) vs best supportive care (BSC) for second line ther- apy in advanced hepatocellular carcinoma (HCC). J Clin Oncol 2019;37(Suppl. 15):Abstact 4004. 8) Yau T, Kang YK, Kim TY, El- Khoueiry AB, Santoro A, Sangro B, et al. Efficacy and safety of nivolumab plus ipilimumab in patients with advanced hepatocellular carcinoma previously treated with sorafenib: the CheckMate 040 Randomized Clinical Trial. JAMA Oncol 2020;6:e204564. 9) Finn RS, Ikeda M, Zhu AX, Sung MW, Baron AD, Kudo M, et al. Phase ib study of lenvatinib plus pembrolizumab in pa- tients with unresectable hepatocellular carcinoma. J Clin Oncol 2020;38:2960- 2970. 10) Fukumura D, Kloepper J, Amoozgar Z, Duda DG, Jain RK. Enhancing cancer immunotherapy using antiangiogenics: oppor- tunities and challenges. Nat Rev Clin Oncol 2018;15:325- 340. 11) Longo V, Gnoni A, Gardini AC, Pisconti S, Licchetta A, Scartozzi M, et al. Immunotherapeutic approaches for hepatocel- lular carcinoma. Oncotarget 2017;8:33897- 33910. 12) Kato YU, Tabata K, Kimura T, Yachie- Kinoshita A, Ozawa Y, Yamada K, et al. Lenvatinib plus anti- PD- 1 antibody combination treatment activates CD8+ T cells through reduction of tumor- associated macrophage and activation of the interferon pathway. PLoS One 2019;14:e0212513. 13) Kimura T, Kato YU, Ozawa Y, Kodama K, Ito J, Ichikawa K, et al. Immunomodulatory activity of lenvatinib contributes to anti- tumor activity in the Hepa1- 6 hepatocellular carcinoma model. Cancer Sci 2018;109:3993- 4002. 14) Brown ZJ, Heinrich B, Greten TF. Mouse models of hepatocel- lular carcinoma: an overview and highlights for immunotherapy research. Nat Rev Gastroenterol Hepatol 2018;15:536- 554. 15) Sia D, Jiao Y, Martinez- Quetglas I, Kuchuk O, Villacorta- Martin C, Castro de Moura M, et al. Identification of an immune- specific class of hepatocellular carcinoma, based on molecular features. Gastroenterology 2017;153:812- 826. 16) Villanueva A, Portela A, Sayols S, Battiston C, Hoshida Y, Méndez- González J, et al. DNA methylation- based progno- sis and epidrivers in hepatocellular carcinoma. Hepatology 2015;61:1945- 1956. 17) Galon J, Costes A, Sanchez- Cabo F, Kirilovsky A, Mlecnik B, Lagorce- Pagès C, et al. Type, density, and location of immune cells within human colorectal tumors predict clinical outcome. Science 2006;313:1960- 1964. 18) Wherry EJ, Kurachi M. Molecular and cellular insights into T cell exhaustion. Nat Rev Immunol 2015;15:486- 499. 19) Böttcher JP, Reis e Sousa C. The role of type 1 conventional den- dritic cells in cancer immunity. Trends Cancer 2018;4:784- 792. 20) Pinyol R, Sia D, Llovet JM. Immune exclusion- WNT/CTNNB1 class predicts resistance to immunotherapies in HCC. Clin Cancer Res 2019;25:2021- 2023. 21) Renne SL, Woo HY, Allegra S, Rudini N, Yano H, Donadon M, et al. Vessels encapsulating tumor clusters (VETC) is a powerful predictor of aggressive hepatocellular carcinoma. Hepatology 2020;71:183- 195. 22) Hoshida Y, Nijman SMB, Kobayashi M, Chan JA, Brunet JP, Chiang DY, et al. Integrative transcriptome analysis reveals com- mon molecular subclasses of human hepatocellular carcinoma. Cancer Res 2009;69:7385- 7392. 23) Boyault S, Rickman DS, de Reyniès A, Balabaud C, Rebouissou S, Jeannot E, et al. Transcriptome classification of HCC is related to gene alterations and to new therapeutic targets. Hepatology 2007;45:42- 52. 24) Chiang DY, Villanueva A, Hoshida Y, Peix J, Newell P, Minguez B, et al. Focal gains of VEGFA and molecular classification of he- patocellular carcinoma. Cancer Res 2008;68:6779- 6788. 25) Khan KA, Kerbel RS. Improving immunotherapy outcomes with anti- angiogenic treatments and vice versa. Nat Rev Clin Oncol 2018;15:310- 324. 26) Gavalas NG, Tsiatas M, Tsitsilonis O, Politi E, Ioannou K, Ziogas AC, et al. VEGF directly suppresses activation of T cells from as- cites secondary to ovarian cancer via VEGF receptor type 2. Br J Cancer 2012;107:1869- 1875. RESULTS 138 HEPATOLOGY, Vol. 74, No. 5, 2021 TORRENS ET AL. 2669 27) Gabrilovich DI, Chen HL, Girgis KR, Cunningham HT, Meny GM, Nadaf S, et al. Production of vascular endothelial growth fac- tor by human tumors inhibits the functional maturation of den- dritic cells. Nat Med 1996;2:1096- 1103. 28) Yost KE, Satpathy AT, Wells DK, Qi Y, Wang C, Kageyama R, et al. Clonal replacement of tumor- specific T cells following PD- 1 blockade. Nat Med 2019;25:1251- 1259. 29) Arce Vargas F, Furness AJS, Litchfield K, Joshi K, Rosenthal R, Ghorani E, et al. Fc effector function contributes to the activity of human anti- CTLA- 4 antibodies. Cancer Cell 2018;33:649- 663.e4. 30) Vignali DAA, Collison LW, Workman CJ. How regulatory T cells work. Nat Rev Immunol 2008;8:523- 532. 31) Sakaguchi S, Yamaguchi T, Nomura T, Ono M. Regulatory T cells and immune tolerance. Cell 2008;133:775- 787. 32) Havel JJ, Chowell D, Chan TA. The evolving landscape of bio- markers for checkpoint inhibitor immunotherapy. Nat Rev Cancer 2019;19:133- 150. 33) Kamphorst AO, Pillai RN, Yang S, Nasti TH, Akondy RS, Wieland A, et al. Proliferation of PD- 1+ CD8 T cells in peripheral blood after PD- 1- targeted therapy in lung cancer patients. Proc Natl Acad Sci U S A 2017;114:4993- 4998. 34) Lindsten T, Lee KP, Harris ES, Petryniak B, Craighead N, Reynolds PJ, et al. Characterization of CTLA- 4 struc- ture and  expression on human T cells. J Immunol 1993;151: 3489- 3499. 35) Pfister D, Núñez NG, Pinyol R, Govaere O, Pinter M, Szydlowska M, et al. NASH limits anti- tumour surveil- lance in immunotherapy- treated HCC. Nature 2021;592: 450- 456. Author names in bold designate shared co-first authorship. Supporting Information Additional Supporting Information may be found at onlinelibrary.wiley.com/doi/10.1002/hep.32023/suppinfo. DISCUSSION DISCUSSION 141 Molecular profiling of tumors using translational approaches has significantly contributed to the understanding of the molecular pathogenesis of HCC3,83. The three articles presented in this doctoral thesis provide a comprehensive analysis of the molecular and immunological features of HCC based on multi-omic techniques and animal models of cancer. Specifically, they provide a characterization of genetic and molecular determinants associated with HCC in specific populations or groups of patients, and mechanistic rationale for novel combination therapies for this devastating disease. 1. Molecular Features of HCC in Mongolia Liver cancer presents large geographical variations in incidence in relation to the prevalence of risk factors for liver diseases such as HBV and HCV infection, alcohol consumption and NASH2,3 (Figure 6). Considering this, the biological characterization of cancer patients from regions with different incidences using next-generation sequencing technologies provides an opportunity to study the heterogeneity among populations and unveil distinct molecular particularities or and presence of putative risk factors. An example of this was provided by the TIGER-LC initiative, which unraveled previously unknown biological features of liver cancer in Thailand. Following this approach, in the 1st study of this doctoral thesis, we provide a comprehensive characterization of the molecular profile of HCC in Mongolia, the country the highest incidence worldwide (85.6 cases per 100,000 inhabitants)62 compared to Western HCC. In Mongolia, HCC is the most frequently diagnosed cancer in both sexes and in all regions of the country63 (Figure 12). Despite a strikingly high prevalence of HBV (10.6%), HCV (6.4%), and HDV (70% of HBV-positive individuals) infections and alcohol consumption56,64,65, it is unclear whether HCC incidence is completely explained by the unique combination of risk factors, or eventually, other unknown factors might be responsible. By analyzing WES and RNA-seq data from two new cohorts consisting of 192 Mongolian HCC and 187 Western HCC (European and American), we identified distinct genomic and transcriptomic footprints in Mongolian tumors that suggest the presence of specific genetic factors in the country that could contribute to the higher HCC incidence in this country. Notably, a previous study provided valuable data about the molecular landscape in Mongolian HCC153, but a comparison with an in-house Western cohort and in-depth analysis of potential environmental agents based on mutational fingerprints and virological characterization were still required. Thus, further analyses comparing Mongolian HCC with Western tumors were still needed to better understand potential associations between DISCUSSION 142 molecular traits and high HCC prevalence in this country. Herein, we were able to unveil novel clinical and virological characteristics, mutation profile, and transcriptomic-based molecular classes in Mongolian HCC patients. Figure 12. Cancer incidence in Mongolia. Mean annual population and cumulative risk (0–74 years, %) of most common cancer sites, by sex and region in Mongolia (2008–2012). Extracted from Chimed T et al., Int J Cancer 201763. HCC in Mongolia has a strikingly high prevalence among females, with a male to female ratio of 1.5/176. This contrasts with the strong male predominance observed in HCC patients globally (male to female ratio of 2–3:1), likely related to differential exposition to risk factors as well as differences in sex hormones3. Consistently, our Mongolian cohort presented 46% of females, as opposed to 20% observed in the Western cohort3. Other remarkable clinical characteristics in the Mongolian cohort were younger age, earlier liver fibrosis stages. and higher viral infection rates, all in accordance with previous studies66,153. In this regard, the prevalence of HBV and HCV in Mongolia is strikingly high, with about 20% of the population being infected by at least one or more types of viruses, and many of them not being aware of their status63. Mongolia also has the highest prevalence of HBV-HDV coinfection in the world59. In our study, we observed a higher rate of HBV-HDV co-infection (84% of HBV-infected patients) in Mongolian patients compared to less than 7% of coinfection in the Western cohort. HDV infection has been associated with a more severe course of liver disease and increased risk of HCC compared to HBV infection alone and could contribute to hepatocarcinogenesis through oncogenic mechanisms independent from HBV122. Finally, regarding the potential oncogenic role of AAV in this population, a previous study reported a very low prevalence of oncogenic AAV integration DISCUSSION 143 in Mongolian and Thai HCC patients (<1%), likely due to epidemiological differences between Asian and European patients60. We further evaluated viral characteristics of the Mongolian and Western datasets to assess whether HBV infection in Mongolia presented features associated with elevated oncogenic potential. Notably, HBV-infected individuals in our Mongolian cohort were genotype D, which is known to be almost universal in Mongolia160, while Western patients presented both genotypes C and D. Genotype D has been previously associated with reduced HCC development as compared to genotype C, suggesting lower oncogenic potential in HBV from Mongolia154. Another relevant characteristic of HBV infection is the presence of basal core promoter (BCP) and precore HBV mutations, which have been associated with liver disease progression and a higher risk of HCC development161. The prevalence of these mutations was significantly lower in the Mongolian cohort compared to Western, with mutational frequencies similar to those previously reported162. These data show that the rate of BCP and precore HBV pro-oncogenic mutations in Mongolia is particularly low despite the predominance of genotype D in this population, which has been associated with a higher rate of HBV mutations155. Overall, HBV viral characteristics in Mongolia are highly homogeneous and with low oncogenic potential, suggesting that other features besides HBV infections – such as HDV or other agents – might be contributing to the high HCC incidence rates in this country154,155. From the genomic standpoint, we unveiled a higher rate of protein-coding mutations in Mongolian HCC, almost doubling that in the Western in-house cohort (121 vs 70 mutations per tumor) and publicly available datasets82,83,85,153, which could be due to intrinsic and/or extrinsic factors promoting mutagenesis in Mongolian HCC. The most commonly mutated genes in both cohorts aligned with previous studies in HCC83 (e.g., mutations in CTNNB1, TP53, ARID1A, ALB…) (Table 2) indicating that the mutational spectrum of Mongolian HCC resembled that of Western HCC. Nevertheless, several HCC drivers were significantly more mutated in Mongolian HCC, including APOB (15% vs 5% in Western HCC), TSC2 (9% vs 1%), and NFE2L2 (6% vs 1%), and the KMT2 gene family (34% vs 17%) (Figure 13). Notably, we detected an enrichment in damaging mutations affecting the TSC2 gene in Mongolian tumors. This pattern of mutations suggests a positive selection of this alteration and a potential driver role156. TSC2 is a known cancer-related gene participating in the mTOR oncogenic pathway83,163 and it has been proposed as an actionable alteration with level 2B evidence, as it could be a predictor of response to the FDA- approved drug everolimus163. Thus, these patients could likely benefit from everolimus treatment, despite this drug was not effective in all-comer HCC patients according to phase III DISCUSSION 144 data164. Considering that systemic therapies for HCC treatment are not routinely used in Mongolia to this day160, the potential implementation of a targeted approach using everolimus in this group of patients faces many challenges. Overall, the mutation profile in HCC driver genes aligned with data from a previously described cohort of 71 Mongolian HCC patients153. Compared to this study, herein we were able to 1) identify a higher mutational burden in Mongolian HCC, and 2) confirm statistically significant differences in the mutation rate of HCC drivers in Mongolia compared to an in-house Western cohort. Figure 13. Somatic mutation in HCC-driving signaling pathways. Mutated genes in HCC grouped by the main deregulated signaling pathways proposed by Ally A et al., Cell 201782. Molecular interactions between genes are represented. The percentage of mutations in the Mongolian cohort, Western cohort, and previous bibliography85,89 are indicated for each gene. Genes with significant differences between Mongolian and Western HCC are highlighted. Original figure. To investigate whether Mongolian HCC presents distinct genomic footprints, we assessed the presence of mutational signatures21,22. This approach has been used in other cancers to propose Chromatin modifiers RTK/RAS/PI3K pathway 0% 0% 0% MDM2 9% 6% 4% ATM 46% 32% 27% TP53* Apoptosis, senescence 4% 4% 0% RPS6KA3 1% 0% 2% CDKN2A Cell cycle 7% 4% 4% RB1 Cell cycle progression 0% 0% 0% CCND1 1% 0% 0% CCNE1 11% 7% 7% AXIN1 3% 4% 2% APC Wnt pathway 36% 43% 30% CTNNB1 0% 0% 0% MYC Cell proliferation 52% NA 55% TERT Telomerase 1% 3% 1% MET 0% 1% 0% FGFR1 0% 0% 0% VEGFA 1% 1% 1% KRAS 3% 4% 2% PIK3CA 2% 2% 1% 1% 1% 0% AKT1/2 9% 1% 3% TSC2* Cell growth, angiogenesis 34% 18% NA KMT2 fam* 17% 10% 8% ARID1A 5% 4% 1% ARID1B 2% 3% 2% BAP1 1% 0% 1% IDH2 5% 1% 1% SMARCA4 % Altered cases Activation Inhibition * p < 0.05 (Mongolia vs Western) % Mongolia % Western % Bibliography Gene Adenoma genes 3% 1% 1% GNAS 1% 1% 3% IL6ST 2% 1% 1% STAT3 Inflammatory response 0% 0% 2% HNF1A Hepatocyte differenciation 15% 5% 8% APOB* Hepatocyte differenciation Oxidative stress 6% 1% 4% NFE2L2* PTEN Inactivating Activating DISCUSSION 145 potential exogenous exposure capable of explaining differences in incidence. For instance, a recent study performed mutational signature analyses in squamous cell carcinoma samples from eight countries with varying incidence, which unveiled very similar exposure profiles between countries, including tobacco, alcohol, and opium165. In our study, de novo signature analysis revealed the presence of a new mutational signature (SBS Mongolia) significantly enriched in Mongolian HCC (25% vs 4.5% in Western HCC), indicating unique substitution patterns characterized by increased frequency of T>G substitutions. Other mutational signatures previously associated with HCC such as SBS5, SBS40, and SBS22 were also detected, with no differences between cohorts. The mutational signature profile reported in a previous study of Mongolian HCC also aligned with these results, including signatures associated with tobacco, alcohol consumption, or aristolochic acid153. However, the presence of de novo signatures was not assessed, and the study did not include a comparison of the mutational landscape with a Western HCC cohort. Interestingly, Mongolian HCC samples from our cohort presenting SBS Mongolia were significantly enriched in a mutational signature associated with exposure to dimethyl sulfate (DMS, 71.1% in positive tumors for SBS Mongolia vs 26.5% in negative tumors). The International Agency for Research on Cancer classifies DMS as a probable carcinogenic hazard to humans (category 2A carcinogen)26, which is a byproduct of coal combustion. The DMS signature was also more common in Mongolian HCC compared to Western, potentially indicating higher exposure to DMS in this country. In this regard, most of the Mongolian population is currently exposed to coal combustion, which is used to fight against the intense cold weather both in urban and rural areas. Half of the 3-million population of Mongolia lives in Ulaanbaatar, an overpopulated capital with one of the highest levels of air pollution in the world166. The rest is still predominantly nomad and lives in traditional tents or gers, where coal is used both for cooking and heating. This fact has been recognized by international organizations as a major health threat in this country166,167. In fact, air pollution from coal combustion has been reported to account for 40% of lung cancer deaths in Ulaanbaatar, corresponding to almost 10% of total deaths in the city168. Considering all this, our results suggest that long-term exposure to DMS from coal combustion could also be a risk factor for HCC development in Mongolia. In this regard, the DMS signature was associated with older patients, potentially due to a longer exposure time. Further studies will be required to gain a mechanistic understanding of how these signatures arise. Specifically, two major streams of investigation are required to confirm the association DISCUSSION 146 between a mutational signature and a mutagen22. First, the presence of the signatures needs to be assessed in model systems exposed to the genotoxic. For instance, it should be confirmed whether DMS and/or byproducts of coal combustion in animal or in vitro models originate the DMS signature, SBS Mongolia or both. In second place, epidemiological studies assessing the onset of tumors presenting such mutational footprints in a population exposed to the mutagen would be required. In this case, a causal relationship between DMS and HCC in Mongolia would have to be supported by a positive association between 1) history of exposure to coal combustion and 2) HCC tumors with a strong contribution of the DMS and SBS Mongolia signatures. This has been the case for previous signatures of exposure to HCC risk factors such as aristolochic acid signature (COSMIC signature SBS22), for which the association between the signature identified in human cancers and the genotoxic was validated in vitro169, and exposure of patients presenting the signature was confirmed epidemiologically113,170. Finally, our transcriptomic analysis revealed that Mongolian tumors presented a distinct transcriptomic profile that did not fit into the classical proliferation and non-proliferation subgroups reported in HCC3,100. Specifically, Mongolian HCC was characterized by an enhanced proliferative and immunological signaling, with a proportion of tumors belonging to proliferative HCC classes (39%) doubling the one in Western HCC and previously published studies (~20%)100. Mongolian HCC clustered into three molecular clusters (MGL1-3). Two of these classes (MGL2 in 26% of the patients and MGL3 in 30%) presented distinct molecular and clinical features compared to Western HCC and thus were deemed unique for Mongolian tumors. Both MGL2 and MGL3 classes were enriched in HBV/HDV infection and younger patients. Interestingly, the MGL2 class was associated with clinical and molecular features of aggressiveness and showed a male to female ratio of 1:2, while MGL3 presented an inflamed profile, potentially due to an immunological response to HBV/HDV infection59. The increased HCC incidence among females in Mongolia76 may be due to higher exposure to risk factors compared to males (e.g., environmental agents or viral hepatitis), thus compensating the conventional gender imbalance found in this tumor type. However, the reason why females are enriched in the aggressive HCC cluster needs to be further understood. The molecular profile of the identified MGL1-3 classes aligned with previously-proposed transcriptomic-based clusters in Mongolian HCC153. Overall, in this study, we were able to identify transcriptomic differences compared to Western HCC and novel clinic-pathological characteristics associated with our classification. DISCUSSION 147 Figure 14. Molecular classification of Mongolian HCC. Mongolian HCC can be classified into three molecular clusters with distinct clinico-pathological and molecular features. HCV, hepatitis C virus; HBV, hepatitis B virus; HDV, hepatitis D virus; CNA, copy number alterations. Original figure. In conclusion, we provided an exhaustive comparison of the genomic and transcriptomic characteristics of Mongolian HCC with an in-house Western cohort. We were able to identify novel features of Mongolian tumors, including 1) virological traits associated with low oncogenic potential; 2) high mutational rates; 3) a distinct mutational signature associated with environmental agents; and 4) a transcriptomic profile characterized by two molecular classes not present in Western HCC. Based on our results, environmental factors such as DMS need to be further explored as a potential risk factor in this population. 2. Risk of AAV Integration in NAFLD Patients Undergoing Gene Therapy Chronic viral is one of the main risk factors leading to HCC3, which occurs mainly through the induction of chronic liver disease and cirrhosis due to persistent inflammation and oxidative stress103 (Figure 15). This mechanism explains most HCV-related HCC, which do not present a clear genetic mechanism of carcinogenesis. In contrast, direct oncogenic effects associated with HCC development have been linked to HBV infections and, more recently, to AAV23,4. Both viruses have the capacity to integrate into the human genome, thus giving rise to insertional mutagenesis and subsequent deregulation of neighboring genes leading to HCC17 (Figure 15). Despite mounting evidence linking AAV insertional mutagenesis with HCC, this virus is currently considered non-pathogenic in humans103. Recombinant AAV (rAAV) – and especially the AAV2 serotype – is a widely used vector for gene replacement, silencing, and editing which holds great promise for the implementation of gene therapies in the clinical setting. In the past few years, DISCUSSION 148 two AAV-based gene therapies have been approved by regulatory agencies for the treatment of spinal muscular atrophy and retinal dystrophy105. Furthermore, ~140 active clinical trials are currently ongoing and could result in groundbreaking therapeutic advances for a great variety of medical conditions including genetic disorders, neurological diseases, and cancer (Figure 16). However, compelling studies linking AAV2 infection with HCC development are posing serious safety concerns4,105. Figure 15. Viral mechanisms of liver carcinogenesis. Direct and indirect mechanisms of viral related liver carcinogenesis are represented for hepatitis C virus (HCV), hepatitis B virus (HBV), and adeno-associated virus type 2 (AAV2). Indirect mechanisms are related to the development of cirrhosis triggered by chronic inflammation and oxidative stress induced by chronic viral hepatitis. Direct DISCUSSION 149 oncogenic mechanisms are mainly due to action of viral oncoproteins (Hbx in HBV), chromosomal instability induced by HBV integration and insertional mutagenesis (HBV and AAV2) with aberrant regulation of gene expression. The genes targeted by clonal viral integrations are represented in the blue box. Adapted from Schulze K et al. J Hep 2016103 using biorender.com. Figure 16. Overview of recombinant AAV interventional gene therapy clinical trials. Clinical trials registered in ClinicalTrials.gov, accessed on 13 November 2018 (n = 145). Trials are categorized based on adeno-associated virus (AAV) capsid serotype (a), primary tissue target for gene delivery (b), and clinical trial phase (c). Extracted from Wang D et al., Nat Rev Drug Discov 2019105. AAV2 integration capacity has been reported to be enhanced in cells undergoing cell cycle progression110. While adult hepatocytes are quiescent under homeostatic conditions, they undergo proliferation in response to liver injury111, thus potentially favoring such integration. The clearest example of this phenomenon occurs after partial hepatectomy. Importantly, conditions of chronic liver injury and inflammation are also associated with compensatory proliferation as the body seeks to repair the damaged liver. Considering this, in the 2nd study of this doctoral thesis, we assessed whether common causes of chronic liver disease such as NAFLD could potentially favor AAV2 integration in the genome due to increased liver damage, inflammation, and regenerative proliferation of hepatocytes111. To do so, neonatal and adult mice were infected with an AAV editing vector targeting the oncogenic Rian locus (AAV-Rian). Animals were treated with HFD to induce NAFLD-like liver injury, or partial hepatectomy was performed to promote hepatocyte proliferation. Herein, we showed that hepatocyte proliferation and NAFLD-associated liver damage increased HCC formation in a murine model treated with rAAV gene targeting. Previous studies by our group158 and others106 showed that AAV2 was able to induce HCC in neonatal mice through random vector integration within the oncogenic Rian locus in the murine chromosome 12. The integration site corresponded to a cluster of oncogenic microRNA genes, DISCUSSION 150 which become activated by the promoter contained within the recombinant virus158,171. HCC tumors generated by AAV integration into the Rian locus were molecularly and histologically similar to the C3 subclass of human HCCs present in 6-19% of HCC patients108,109. This miRNA- based subclass is characterized by overexpression of the human ortholog of the Rian cluster (DLK-DIO3 locus), as well as an aggressive phenotype158. To mimic AAV integration in this locus, a rAAV targeting the murine Rian locus was used in this study. Interestingly, RNA-sequencing analysis of murine tumors from our model revealed remarkable molecular similarities with this HCC subclass, associated with proliferation, aggressive phenotype, and poor prognosis. Furthermore, deep sequencing analyses of HCC tumors have previously revealed AAV integration in known cancer driver genes in 2-5% of HCC cases, thus providing further evidence of the oncogenic potential of oncogenic AAV integration in the human genome4,60. Further analyses will be required to determine whether sequence motifs, chromatin states, and vector characteristics influence integration preferences in the Rian locus and other oncogenic sites106. According to previous data, adult mice typically do not develop AAV-induced liver cancer in the absence of injury172,173. To determine whether hepatocyte proliferation impacts rAAV-related oncogenesis, two injury regimens were evaluated: NAFLD induced by the administration of a HFD and partial hepatectomy. Both conditions led to HCC development in 100% of the rAAV- infected adult mice, compared to only 5% in untreated infected animals. Conversely, HCC rates in neonates were significantly higher than in adults irrespectively of whether they received high- fat or normal diets (i.e., 100% incidence in all male groups). The age-dependence of rAAV- induced cancer is likely linked to the rate of hepatocyte proliferation, which is high in neonates due to natural liver growth. This enhanced hepatocyte proliferation would explain the higher HCC incidence in models with liver injury. Overall, these results raise concerns for the use of rAAV therapy in the general human population, where chronic inflammatory liver diseases are very prevalent and could promote rAAV integration. For instance, up to 30% of the population in the US have fatty liver disease, 0.34% have HBV infection, 1.7% have HCV infection, and 4.3% have alcoholic liver disease159. These inflammatory conditions alone are risk factors for HCC development, thus adding a rAAV-associated oncogenic risk may significantly diminish the possible therapeutic benefits of gene therapy. Most human HCCs arise in the background of chronic liver diseases characterized by injury and inflammation3,9. Inflammation itself may promote hepatocarcinogenesis through several mechanisms such as DNA damage from reactive oxygen species, changes in the immune system milieu, and an increase in hepatocyte proliferation. This is especially true in NAFLD, which has DISCUSSION 151 been linked to alterations in the liver immune infiltrate, leading to hepatocyte damage and HCC3,174. In line with this, transcriptomic analysis of background liver samples from our murine model showed that HFD promoted a protumorigenic immune cancer field as well as activated lipid metabolism pathways in the background liver125. This aberrant immune field has been associated with activation of pro-oncogenic immunosuppressive signaling (e.g., TGFβ signaling and T cell exhaustion) and a higher risk of HCC develoment125. Therefore, this immunosuppressive milieu induced by HFD and NAFLD could have a key role in the increased risk of rAAV-induced HCC in our model. Notably, previous studies from our group revealed that NASH-HCC showed a significantly higher prevalence of the immunosuppressive cancer field compared to other etiologies175. As previously discussed in this doctoral thesis, HCC has a strong male predominance, which can be partially attributed to anti-inflammatory properties from estrogens in females3. In our study, we confirmed that female mice are less susceptible to AAV-induced HCC compared to males. We showed that estrogen treatment in male mice fed with a HFD reduced liver damage and fat deposition. This aligns with human studies showing lower NASH prevalence and less severe NAFLD stages in female individuals compared to males and postmenopausal women due to a protective role of estrogens176. In our model, estrogen also altered the immune milieu in males with NAFLD, decreasing pro-oncogenic immune exhaustion signaling (e.g., TGF-β and Wnt/β- catenin pathways) and promoting an adaptive immune response. No information regarding sex differences in oncogenic AAV integration has been currently reported in humans, likely due to the low availability of AAV-related HCC samples. We also investigated whether AAV infection could result in spontaneous viral integration in the Rian locus as indicated by previous studies171. We observed a high incidence of HCC in adult mice receiving HFD who were infected with the control AAV (tdTomato AAV) which, unlike the Rian AAV, does not have the capability for targeted integration into the genome. Specifically, 5 out of 10 mice on the HFD who received the control AAV developed HCC. These tumors presented increased expression of genes located in the Rian locus in line with results from AAV-Rian- induced HCC, as well as a similar gene expression profile and pathway activation, indicating potential integration in the Rian locus. This suggests that random integrations into the Rian locus are common and highly carcinogenic in mice. The fact that only control AAV-infected mice receiving HFD developed tumors (as opposed to control AAV-infected mice receiving normal diet) further suggests that inflammation and hepatocyte proliferation increase HCC incidence. Whether the human liver is similarly susceptible to random rAAV leading to HCC needs to be DISCUSSION 152 further investigated. In this regard, the reasons for the difference between the high rate of AAV2 infection in humans (40-80% seropositivity in the general population) and the low rate of AAV2 related HCC is still unclear103. Overall, this study demonstrates that adult mice infected with both targeted and non-targeted rAAV present increased development HCC in the presence of liver injury such as fatty liver induced by HFD. This was likely due to increased hepatocyte proliferation and the generation of a protumorigenic immune cancer field effect in the liver. Given the high prevalence of inflammatory liver conditions in the general population such as NAFLD and NASH, this should raise an alarm about the use of rAAV, an otherwise promising vector used for diverse gene therapy applications. Female mice were less susceptible to rAAV-induced HCC compared to males, part of which is due to a more favorable immune milieu related to estrogen, and likely applies to humans given the documented gender differences. Given the promise and widespread use of rAAV for gene therapies, more studies are needed to assess the risks of rAAV- induced hepatocarcinogenesis, including the monitoring of patients treated with rAAV vectors in search of viral integration and cancer development104. 3. Immunomodulatory Effects of Lenvatinib Plus Anti-PD1 Combination During the last decade, major breakthroughs have dramatically improved the landscape of advanced HCC treatment thanks to the approval of novel systemic therapies including TKI and ICI. Despite this, survival benefits observed in clinical trials for single agents are modest, and responses for ICI monotherapy are only observed in ~15-20% of patients3. Therefore, there is an urgent need to identify existing therapies that can effectively synergize with ICI. In this regard, antiangiogenic drugs constitute the basis of the standard of care treatment in advanced HCC (Figure 17) and have been proposed as ideal candidates for combination with ICI due to their immunomodulatory capacities130,177. Effectively, the new combination of the ICI atezolizumab plus bevacizumab (VEGFA inhibitor) demonstrated objective responses in 36% of patients and has been FDA-approved as first-line therapy143. Similarly, the combination of lenvatinib plus pembrolizumab is a promising approach for HCC, with phase Ib data showing an unprecedented ORR of 46% and median survival of 22 months148. This combination is currently being assessed in a phase III trial compared to lenvatinib alone (LEAP-002 trial, NCT03713593). DISCUSSION 153 Figure 17. Treatment strategy for advanced HCC. Green boxes designate drugs with positive results from phase III trials with a superiority design. Yellow boxes designate drugs with positive results from phase III trials with a non-inferiority design. Drugs in red boxes have received accelerated approval from the FDA following promising efficacy results in phase II trials. Antiangiogenic agents are indicated with a *. AFP, alpha-fetoprotein; BCLC, Barcelona Clinic Liver Cancer; ECOG PS, Eastern Cooperative Oncology Group performance status; EHS, extrahepatic spread; HCC, hepatocellular carcinoma; HR, Hazard Ratio; PD, Progressive Disease. Adapted from Llovet JM et al., Nat Cancer 2021 (in press)121. In the 3rd study of this doctoral thesis, we explored the anti-tumoral and immunomodulatory effects of lenvatinib in combination with anti-PD1 ICI. To achieve this, we generated three syngeneic models of HCC and assessed the anti-tumoral activity of lenvatinib alone or in combination with anti-PD1 and its effects on the systemic and tumor-infiltrating immune cells. In our models, anti-PD1 and lenvatinib in monotherapy improved survival and decreased tumor growth; however, the combination treatment achieved a higher response rate, shorter time to response, and a reduction of tumor viability. This is in accordance with a previous publication showing an enhanced anti-tumoral potential of the combination compared to monotherapies in vivo178. Furthermore, we demonstrated that lenvatinib exerts an immunomodulatory effect, which together with anti-PD1, elicits an anti-tumoral immune response activation, thus providing a mechanistic rationale for this combination. The immunomodulatory capacity of lenvatinib alone or in combination with anti-PD1 remains poorly characterized. Notably, the main targets of this TKI include VEGFR and FGFRs. The VEGFA- * * * * * * DISCUSSION 154 VEGFR pathway activation has been reported to favor tumor growth, progression, and aberrant vasculature formation130. More importantly, the VEGF pathway has a direct immunosuppressive effect on the tumor infiltrate by decreasing cytotoxic T cell and DC function and promoting the recruitment of immunosuppressive cells such as Treg, M2 macrophages, and MDSC (Figure 10).130–132. Additionally, recent studies propose that FGFR4 inhibition by lenvatinib could also promote immune activation in HCC models179,180. Our study revealed that lenvatinib exerts a potent effect on the immune infiltrate by reducing the Treg proportion and altering their intra- tumoral location, which is in accordance with the potential effects of VEGF inhibition (Figure 10). Considering this, lenvatinib could boost the effects of ICIs on the antitumor immune response by targeting the VEGFR and FGFR pathways. In addition, our results indicate that lenvatinib treatment reduced the aberrant vasculature in the tumors, in line with the anti- angiogenic nature of this drug. The abnormal tumor vasculature contributes to immunosuppression through several direct and indirect mechanisms130, and thus vascular normalization by lenvatinib could also contribute to an immunomodulatory capacity of lenvatinib (Figure 18). The role of anti-PD1 inhibition activating the anti-tumor immune response is well known. In our models, anti-PD1 treatment induced an increase of T cell and type 1 dendritic cell (DC1) infiltrate. This is in accordance with human studies reporting T cell recruitment following PD1 blockade and association between DC1 and T cell function181,182. Importantly, the combined effect of the two therapeutic approaches induced an increase in the pro-inflammatory component in the tumor as shown by the high CD8 to Treg ratio, associated with enhanced antitumor immunity183. Data from treated cancer patients indicate that intra-tumoral Treg cells might limit anti-PD1 efficacy184,185. Therefore, the effect of lenvatinib in reducing this immunosuppressive population could be responsible for the greater anti-tumoral capacity of the combination therapy. Other recent experimental studies have suggested a link between the lenvatinib plus anti-PD1 combination and antitumor immune responses178,179,186. For instance, the combination treatment increased the CD8 T cell infiltrate and decreased the monocyte/macrophage and DC component in an HCC murine model178, in line with previous studies assessing the effect of lenvatinib as monotherapy186. Therefore, while anti-PD1 promotes DC1 recruitment according to our study, this may not be the case for all DC subpopulations. Furthermore, our data indicate that the combination elicited a reduction of the protumorigenic M2 phenotype compared to the anti-PD1 arm. Considering that Treg cells promote the M2 phenotype in tumor-infiltrated DISCUSSION 155 macrophages37, this could be linked to the decreased Treg proportion in the combination arm. The reprogramming of the macrophage phenotype agrees with in vivo data linking lenvatinib treatment with a reduction in M2 macrophages in colon cancer186. Considering the complexity of the myeloid component and different functional states of each sub-population128, the effect of the treatments on additional myeloid subtypes remains to be investigated. Finally, despite VEGF signaling has been proposed to elicit systemic immunosuppression130, no significant alterations were detected in the circulating immune cells, suggesting that the effect of the treatments may be tumor-specific. Figure 18. Vascular-normalizing therapies can reprogram the immunosuppressive tumor microenvironment. The structural and functional abnormalities of tumor blood vessels lead to impaired blood flow thus resulting in a hypoxic tumor microenvironment (TME). Hypoxic conditions in tumors exert numerous immune- suppressive effects and limit the delivery and effectiveness of therapies. TME, tumor microenvironment. Modified from Fukumura D et al., Nat Rev Clin Oncol 2018130. In addition to the described changes in the tumoral infiltrate, the treatments also elicited changes in the immune-related molecular signaling. Notably, lenvatinib blocked immunosuppressive signaling through TGFß pathway inhibition. Considering that TGFß signaling is one of its key immunosuppressive mechanisms of Treg cells187, this is likely a consequence of the decrease in Treg proportion in this treatment arm. Conversely, despite anti-PD1 enhanced the intratumoral T cell infiltrate, this presented an immune exhausted phenotype characterized DISCUSSION 156 by expression of immune-suppressive signaling187, in accordance with human studies reporting an increase in the exhausted T cell component following PD1 blockade181. The combined effect of lenvatinib plus anti-PD1 generated both an increase in the T cell infiltrate and a suppression of immunosuppressive signaling (i.e., Treg infiltrate and TGFß pathway). Consequently, only the combination treatment generated a specific activated anti-tumoral immune response. The identification of the molecular and immunomodulatory effects induced by lenvatinib plus anti-PD1 can provide relevant information to identify which patients are likely to benefit from this treatment based on their immunological profile. Using gene expression data from our murine model, we generated a molecular signature capable of identifying 22% of human HCC patients with downregulation of genes associated with the molecular effect of the combination but not by monotherapies as stand-alone therapies (Figure 19). Tumors from these patients presented reduced pro-inflammatory signaling, high Treg levels, and VEGF signaling. We hypothesize that patients with these characteristics are not likely to respond to anti-PD1 alone but could be responders to the combination treatment. On the other hand, a previous study from our group identified 24% of HCC patients belonging to the HCC immune class, which presented markers of T cell infiltrate and molecular features indicative of good response to ICI monotherapy120. Considering both groups altogether, we hypothesize that about half of HCC patients could respond to combination therapies. These numbers align with the ORR of 46% observed in patients receiving lenvatinib plus anti-PD1 in clinical trials148. Notably, our gene signature has been generated from on-treatment murine tumor samples, and therefore further studies extrapolating these findings to pretreatment HCC profiles will be required. The predictive capacity of response of the signature will need to be validated in a prospective human cohort of HCC patients receiving lenvatinib plus an anti-PD1 ICI. In conclusion, the current study revealed that lenvatinib exerts an immunomodulatory effect on the tumor infiltrate associated with a reduction in the intratumoral Treg infiltrate and inhibition of immune-suppressive pathways. Its combination with anti-PD1 favored the generation of an activated immune profile and a faster response to treatment. Furthermore, we generated a molecular signature present in ~20% HCC patients that correlates with high Treg infiltrate, VEGFR pathway activation, and low inflammatory signaling, and might recognize patients likely to benefit most from this combination. Further investigations are warranted to confirm if the signature can be a tool to identify HCC patients who may respond to the combination therapy beyond their responses to single-agent therapies. DISCUSSION 157 Figure 19. HCC classification according to its immunological features and potential response to the combination therapy. Diagram summarizing the HCC immune classification and potential response to ICIs or combination treatment beyond single agents. Features observed in Study #3 are highlighted in blue and red. Additional features of each subgroup are depicted in grey ICI, Immune checkpoint inhibitors; Treg, regulatory T cell. Original figure. Data extracted from Torrens L et al., Hepatology 2021188 and Llovet JM et al., Nat Cancer 2021121. 4. Improving the Understanding of HCC Pathogenesis with Translational Approaches Using translational approaches based on multi-omic analysis and preclinical models of HCC, the three studies comprised in this thesis provide relevant information in several areas and tackle unmet needs of HCC (Figure 20). First of all, this thesis investigates the molecular heterogeneity DISCUSSION 158 between HCC patients from two geographical regions with different HCC burdens and risk factors. Specifically, distinct virological, genomic, and transcriptomic features of Mongolian HCC patients compared to Western were revealed. Furthermore, our studies elucidate potential risk factors of HCC in specific populations. We uncovered mutational fingerprints in the genome of Mongolian HCC constituting a novel mutational signature (SBS Mongolia) associated with the signature of exposure to carcinogenic DMS. This suggests exposure to DMS from coal combustion in this population, which could contribute to the increased HCC incidence. In addition, we provide evidence suggesting a high risk of AAV2 integration in patients with fatty liver disease, thus promoting HCC development. Finally, we propose novel therapeutic approaches for HCC patients. In this regard, our data shows enhanced anti-tumoral and immune-modulatory effects of the combination of lenvatinib plus anti-PD1 compared to monotherapies. In addition, the identification of targetable drivers in Mongolia such as TSC2 suggests that targeted therapeutic approaches could benefit a subgroup of patients. The abovementioned findings could have implications in advancing the scientific knowledge and clinical management in HCC (Figure 20). Specifically: 1) The improved understanding of clinical, virological, and molecular pathogenesis of HCC in Mongolia provides novel data regarding the heterogeneous distribution of HCC burden in the world. 2) The identification of potential risk factors of HCC could have an impact on the management of the disease. For instance, our data suggest a need to prevent exposure to risk factors in Mongolia, including viral infections and potentially coal combustion. In addition, our second study raises concerns for the use of gene therapy in patients with NAFLD and chronic liver inflammation. 3) We provide a mechanistic rationale for the use of lenvatinib plus pembrolizumab treatment in advanced HCC, which is currently being assessed in phase III trials148, and propose a subgroup of HCC patients which could benefit from this approach. Furthermore, the understanding of the molecular profile of Mongolian HCC could also have implications for the treatment of these patients in the future. DISCUSSION 159 Figure 20. Summary of the studies comprised in this doctoral thesis. This doctoral thesis aims at providing relevant information to advance current unmet needs in hepatocellular carcinoma (HCC). With the use of translational approaches, the three studies herein discussed have potential and clinical impact in the field. AAV2, adeno-associated virus type 2; rAAV, recombinant adeno-associated virus; NAFLD, nonalcoholic fatty liver disease; DMS, dimethyl sulfate. Original figure. Overall, translational studies such as the ones included in the current thesis are key to advancing our knowledge of this devastating disease. As discussed in prior sections of this discussion and summarized in Figure 20, future research will be required to validate the results herein presented and translate this knowledge into actual clinical applications that result in survival benefits for the patients. A joint multidisciplinary effort involving basic, translational, and clinical research will be necessary to keep advancing towards this common goal. CONCLUSIONS CONCLUSIONS 163 The main conclusions arising from the work presented in this thesis are the following: - HCC in Mongolia presents unique genomic and transcriptomic footprints consisting in an increased number of mutations, as well as specific mutational and transcriptomic patterns. This includes the presence of a newly identified mutational signature (SBS Mongolia) in 25% of Mongolian HCC cases, which is associated with a signature of genotoxic DMS exposure. These molecular features suggest a role of environmental factors that might explain the high HCC burden in this country. - NAFLD-related chronic inflammation and liver injury promoted the development of rAAV- induced HCC in mice due to integration in an oncogenic locus. This raises concerns about the risks of rAAV gene therapy causing HCC in patients with chronic liver diseases. - The combination of lenvatinib plus anti-PD1 in HCC elicits immune-modulatory capacities characterized by an activated immune profile. Our gene signature captured ~20% of human HCC which might benefit from the combination treatment. This could have implications for patient selection to ultimately maximize the clinical benefit from this treatment. - Novel molecular features of HCC, potential risk factors, and therapeutic approaches can be unraveled through translational studies based on preclinical models and omics analysis of human tumors. This approach has the potential to impact the clinical management in HCC. REFERENCES REFERENCES 167 1. GBD 2015 Mortality and Causes of Death Collaborators. Global, regional, and national life expectancy, all-cause mortality, and cause-specific mortality for 249 causes of death, 1980–2015: a systematic analysis for the Global Burden of Disease Study 2015. Lancet. 2016;388:1459–544. 2. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71:209–49. 3. Llovet JM, Kelley RK, Villanueva A, Singal AG, Pikarsky E, Roayaie S, et al. Hepatocellular carcinoma. Nat Rev Dis Prim. 2021;7:7. 4. Nault J, Datta S, Imbeaud S, Franconi A, Mallet M, Couchy G, et al. Recurrent AAV2- related insertional mutagenesis in human hepatocellular carcinomas. Nat Genet. 2015;47:1187–93. 5. Finn RS, Ryoo B-Y, Merle P, Kudo M, Bouattour M, Lim H-Y, et al. Results of KEYNOTE- 240: phase 3 study of pembrolizumab (Pembro) vs best supportive care (BSC) for second line therapy in advanced hepatocellular carcinoma (HCC). J Clin Oncol. 2019;37:Abstact 4004 (Suppl 15). 6. El-Khoueiry AB, Sangro B, Yau T, Crocenzi TS, Kudo M, Hsu C, et al. Nivolumab in patients with advanced hepatocellular carcinoma (CheckMate 040): an open-label, non- comparative, phase 1/2 dose escalation and expansion trial. Lancet. 2017;389:2492–502. 7. Bray F, Laversanne M, Weiderpass E, Soerjomataram I. The ever-increasing importance of cancer as a leading cause of premature death worldwide. Cancer. 2021;127:3029–30. 8. Hanahan D, Weinberg RA. The hallmarks of cancer. Cell. 2000;100:57–70. 9. Hanahan D, Weinberg RA. Hallmarks of Cancer: The Next Generation. Cell. 2011;144:646–74. 10. Vogelstein B, Kinzler K. Cancer genes and the pathways they control. Nat Med. 2004;10:789–99. 11. Weir B, Zhao X, Meyerson M. Somatic alterations in the human cancer genome. Cancer Cell. 2004;6:433–8. 12. Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz LA, Jr., et al. Cancer Genome Landscapes. Science. 2013;339:1546–58. 13. Vogelstein B, Kinzler KW. The Path to Cancer — Three Strikes and You’re Out. N Engl J Med. 2015;373:1895–8. 14. Martínez-Jiménez F, Muiños F, Sentís I, Deu-Pons J, Reyes-Salazar I, Arnedo-Pac C, et al. A compendium of mutational cancer driver genes. Nat Rev Cancer. 2020;20:555–72. REFERENCES 168 15. Stratton MR, Campbell PJ, Futreal PA. The cancer genome. Nature. 2009;458:719–24. 16. Zack TI, Schumacher SE, Carter SL, Cherniack AD, Saksena G, Tabak B, et al. Pan-cancer patterns of somatic copy number alteration. Nat Genet. 2013;45:1134–40. 17. Talbot S, Crawford D. Viruses and tumours--an update. Eur J Cancer. 2004;40:1998–2005. 18. Portela A, Esteller M. Epigenetic modifications and human disease. Nat Biotechnol. 2010;28:1057–68. 19. Helleday T, Eshtad S, Nik-Zainal S. Mechanisms underlying mutational signatures in human cancers. Nat Rev Genet. 2014;15:585–98. 20. Brown J, Thornton J. Percivall Pott (1714-1788) and chimney sweepers’ cancer of the scrotum. Br J Ind Med. 1957;14:68–70. 21. Kucab JE, Zou X, Morganella S, Joel M, Nanda AS, Nagy E, et al. A Compendium of Mutational Signatures of Environmental Agents. Cell. 2019;177:821-836.e16. 22. Alexandrov LB, Nik-Zainal S, Wedge DC, Aparicio SAJRJR, Behjati S, Biankin A V., et al. Signatures of mutational processes in human cancer. Nature. 2013;500:415–21. 23. Alexandrov LB, Kim J, Haradhvala NJ, Huang MN, Tian Ng AW, Wu Y, et al. The repertoire of mutational signatures in human cancer. Nature. 2020;578:94–101. 24. Alexandrov LB, Nik-Zainal S, Wedge DC, Campbell PJ, Stratton MR. Deciphering Signatures of Mutational Processes Operative in Human Cancer. Cell Rep. 2013;3:246. 25. Koh G, Degasperi A, Zou X, Momen S, Nik-Zainal S. Mutational signatures: emerging concepts, caveats and clinical applications. . Nat Rev Cancer. 2021;21:619–37. 26. International Agency for Research on Cancer. Agents Classified by the IARC Monographs, Volumes 1–127 – IARC Monographs on the Identification of Carcinogenic Hazards to Humans [Internet]. 2018 [cited 2020 Nov 5]. Available from: https://monographs.iarc.fr/agents-classified-by-the-iarc 27. Tovar V, Llovet JM. Molecular Basis of Cancer. In: Llovet JM, editor. Handbook of Translational Medicine. 2016. p. 79–101. 28. Zhang Y, Zhang Z. The history and advances in cancer immunotherapy: understanding the characteristics of tumor-infiltrating immune cells and their therapeutic implications. Cell Mol Immunol. 2020;17:807–21. 29. Greten FR, Grivennikov SI. Inflammation and Cancer: Triggers, Mechanisms, and Consequences. Immunity. 2019;51:27–41. 30. Grivennikov SI, Greten FR, Karin M. Immunity, Inflammation, and Cancer. Cell. 2010;140:883–99. 31. Sánchez-Serrano I. Success in translational research: lessons from the development of bortezomib. Nat Rev Drug Discov. 2006;5:107–14. REFERENCES 169 32. Woolf SH. The Meaning of Translational Research and Why It Matters. JAMA. 2008;299:211–3. 33. Torrens L, Pinyol R, Jimenez W, Llovet J. Overview of Translational Medicine. In: Llovet J, editor. Handbook of Translational Medicine. 2016. p. 21–30. 34. McPake M. ESMO: Handbook of Principles of Translational Research. Vol. 19, European journal of cancer care. 2010. e6 p. 35. Gharwan H, Groninger H. Kinase inhibitors and monoclonal antibodies in oncology: clinical implications. Nat Rev Clin Oncol. 2016;13:209–27. 36. Maemondo M, Inoue A, Kobayashi K, Sugawara S, Oizumi S, Gemma A, et al. Gefitinib or Chemotherapy for Non-Small-Cell Lung Cancer with Mutated EGFR. N Engl J Med. 2010;362: 2380–8. 37. Havel JJ, Chowell D, Chan TA. The evolving landscape of biomarkers for checkpoint inhibitor immunotherapy. Nat Rev Cancer. 2019;19:133–50. 38. Martin KJ, Fournier M V., Reddy GPV, Pardee AB. A Need for Basic Research on Fluid- Based Early Detection Biomarkers. Cancer Res. 2010;70:5203–6. 39. Kurnit KC, Dumbrava EEI, Litzenburger B, Khotskaya YB, Johnson AM, Yap TA, et al. Precision Oncology Decision Support: Current Approaches and Strategies for the Future. Clin Cancer Res. 2018;24:2719–31. 40. Roychowdhury S, Chinnaiyan AM. Translating cancer genomes and transcriptomes for precision oncology. CA Cancer J Clin. 2016;66:75–88. 41. Meyerson M, Gabriel S, Getz G. Advances in understanding cancer genomes through second-generation sequencing. Nat Rev Genet. 2010;11:685–96. 42. Campbell PJ, Getz G, Korbel JO, Stuart JM, Jennings JL, Stein LD, et al. Pan-cancer analysis of whole genomes. Nature. 2020;578:82–93. 43. Shendure J, Ji H. Next-generation DNA sequencing. Nat Biotechnol. 2008;26:1135–45. 44. Carter NP. Methods and strategies for analyzing copy number variation using DNA microarrays. Nat Genet. 2007;39:S16-21. 45. Donoghue MTA, Schram AM, Hyman DM, Taylor BS. Discovery through clinical sequencing in oncology. Nat Cancer. 2020;1:774–83. 46. Cieślik M, Chinnaiyan AM. Cancer transcriptome profiling at the juncture of clinical translation. Nat Rev Genet. 2017;19:93–109. 47. Wang Z, Gerstein M, Snyder M. RNA-Seq: a revolutionary tool for transcriptomics. Nat Rev Genet. 2009;10:57–63. 48. Supplitt S, Karpinski P, Sasiadek M, Laczmanska I. Current achievements and applications of transcriptomics in personalized cancer medicine. Int J Mol Sci. 2021; 22:1422. REFERENCES 170 49. Rooney MS, Shukla SA, Wu CJ, Getz G, Hacohen N. Molecular and genetic properties of tumors associated with local immune cytolytic activity. Cell. 2015;160:48–61. 50. Moffitt RA, Marayati R, Flate EL, Volmar KE, Loeza SGH, Hoadley KA, et al. Virtual microdissection identifies distinct tumor- and stroma-specific subtypes of pancreatic ductal adenocarcinoma. Nat Genet. 2015;47:1168–78. 51. Soucek L, Torrens L, Pujades C, Clària J. Experimental Models. In: Llovet J, editor. Handbook of Translational Medicine. 2016. p. 118–26. 52. Brown ZJ, Heinrich B, Greten TF. Mouse models of hepatocellular carcinoma: an overview and highlights for immunotherapy research. Nat Rev Gastroenterol Hepatol. 2018;15:536–54. 53. Frese K, Tuveson D. Maximizing mouse cancer models. Nat Rev Cancer. 2007;7:645–58. 54. McGlynn K, Petrick J, London W. Global epidemiology of hepatocellular carcinoma: an emphasis on demographic and regional variability. Clin Liver Dis. 2015;19:223–38. 55. Yang JD, Hainaut P, Gores GJ, Amadou A, Plymoth A, Roberts LR. A global view of hepatocellular carcinoma: trends, risk, prevention and management. Nat Rev Gastroenterol Hepatol. 2019;16:589–604. 56. Blach S, Zeuzem S, Manns M, Altraif I, Duberg AS, Muljono DH, et al. Global prevalence and genotype distribution of hepatitis C virus infection in 2015: a modelling study. Lancet Gastroenterol Hepatol. 2017;2:161–76. 57. Anstee QM, Reeves HL, Kotsiliti E, Govaere O, Heikenwalder M. From NASH to HCC: current concepts and future challenges. Nat Rev Gastroenterol Hepatol. 2019;16:411– 28. 58. Wedemeyer H, Manns MP. Epidemiology, pathogenesis and management of hepatitis D: Update and challenges ahead. Nat Rev Gastroenterol Hepatol. 2010;7:31–40. 59. Puigvehí M, Moctezuma-Velázquez C, Villanueva A, Llovet JM. The oncogenic role of hepatitis delta virus in hepatocellular carcinoma. JHEP Reports. 2019;1:120–30. 60. La Bella T, Imbeaud S, Peneau C, Mami I, Datta S, Bayard Q, et al. Adeno-associated virus in the liver: natural history and consequences in tumour development. Gut. 2020;69:737–47. 61. European Association For The Study Of The Liver, European Organisation For Research And Treatment Of Cancer. EASL–EORTC Clinical Practice Guidelines: Management of hepatocellular carcinoma. J Hepatol. 2012;56:908–43. 62. Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. Global Cancer Statistics 2018: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. Cancer J Clin. 2018;68:394–424. REFERENCES 171 63. Chimed T, Sandagdorj T, Znaor A, Laversanne M, Tseveen B, Genden P, et al. Cancer incidence and cancer control in Mongolia: Results from the National Cancer Registry 2008-12. Int J cancer. 2017;140:302–9. 64. Polaris Observatory Collaborators. Global prevalence, treatment, and prevention of hepatitis B virus infection in 2016: a modelling study. Lancet Gastroenterol Hepatol. 2018;3:383–403. 65. Demaio AR, Dugee O, de Courten M, Bygbjerg IC, Enkhtuya P, Meyrowitsch DW. Exploring knowledge, attitudes, and practices related to alcohol in Mongolia: a national population-based survey. BMC Public Health. 2013;13:178. 66. Dondog B, Lise M, Dondov O, Baldandorj B, Franceschi S. Hepatitis B and C virus infections in hepatocellular carcinoma and cirrhosis in Mongolia. Eur J Cancer Prev. 2011;20:33–9. 67. Oyunsuren T, Kurbanov F, Tanaka Y, Elkady A, Sanduijav R, Khajidsuren O, et al. High frequency of hepatocellular carcinoma in Mongolia; association with mono-, or co- infection with hepatitis C, B, and delta viruses. J Med Virol. 2006;78:1688–95. 68. Tsatsralt-Od B, Takahashi M, Nishizawa T, Endo K, Inoue J, Okamoto H. High prevalence of dual or triple infection of hepatitis B, C, and delta viruses among patients with chronic liver disease in Mongolia. J Med Virol. 2005;77:491–9. 69. Jazag A, Puntsagdulam N, Chinburen J. Status Quo of Chronic Liver Diseases, Including Hepatocellular Carcinoma, in Mongolia. Korean J Intern Med. 2012;27:121. 70. Davaalkham D, Ojima T, Uehara R, Watanabe M, Oki I, Wiersma S, et al. Impact of the universal hepatitis B immunization program in Mongolia: achievements and challenges. J Epidemiol. 2007;17:69–75. 71. Kim YA, Estevez J, Le A, Israelski D, Baatarkhuu O, Sarantuya T, et al. Screening and management of viral hepatitis and hepatocellular carcinoma in Mongolia: results from a survey of Mongolian physicians from all major provinces of Mongolia. BMJ Open Gastroenterol. 2016;3:e000119. 72. Bai H, Guo X, Zhang D, Narisu N, Bu J, Jirimutu J, et al. The Genome of a Mongolian Individual Reveals the Genetic Imprints of Mongolians on Modern Human Populations. Genome Biol Evol. 2014;6:3122–36. 73. Bai H, Guo X, Narisu N, Lan T, Wu Q, Xing Y, et al. Whole-genome sequencing of 175 Mongolians uncovers population-specific genetic architecture and gene flow throughout North and East Asia. Nat Genet. 2018;50:1696–704. 74. Yin J, Zhang H, He Y, Xie J, Liu S, Chang W, et al. Distribution and hepatocellular carcinoma-related viral properties of hepatitis B virus genotypes in Mainland China: a community-based study. Cancer Epidemiol Biomarkers Prev. 2010;19:777–86. REFERENCES 172 75. Chen JG, Zhang SW. Liver cancer epidemic in China: past, present and future. Semin Cancer Biol. 2011;21:59–69. 76. Petrick JL, Braunlin M, Laversanne M, Valery PC, Bray F, McGlynn KA. International trends in liver cancer incidence, overall and by histologic subtype, 1978-2007. Int J cancer. 2016;139:1534–45. 77. Zucman-Rossi J, Villanueva A, Nault JC, Llovet JM. Genetic Landscape and Biomarkers of Hepatocellular Carcinoma. Gastroenterology. 2015;149:1226–39. 78. Llovet JM, Zucman-Rossi J, Pikarsky E, Sangro B, Schwartz M, Sherman M, et al. Hepatocellular carcinoma. Nat Rev Dis Prim. 2016;2:16018. 79. Llovet JM, Montal R, Sia D, Finn RS. Molecular therapies and precision medicine for hepatocellular carcinoma. Vol. 15, Nature Reviews Clinical Oncology. 2018. p. 599–616. 80. Villanueva A. Hepatocellular Carcinoma. Longo DL, editor. N Engl J Med. 2019;380:1450– 62. 81. Nault J, Couchy G, Balabaud C, Morcrette G, Caruso S, Blanc J, et al. Molecular Classification of Hepatocellular Adenoma Associates With Risk Factors, Bleeding, and Malignant Transformation. Gastroenterology. 2017;152:880-894.e6. 82. Ally A, Balasundaram M, Carlsen R, Chuah E, Clarke A, Dhalla N, et al. Comprehensive and Integrative Genomic Characterization of Hepatocellular Carcinoma. Cell. 2017;169:1327- 1341.e23. 83. Schulze K, Imbeaud S, Letouzé E, Alexandrov LBL, Calderaro J, Rebouissou S, et al. Exome sequencing of hepatocellular carcinomas identifies new mutational signatures and potential therapeutic targets. Nat Genet. 2015;47:505–11. 84. Totoki Y, Tatsuno K, Covington KR, Ueda H, Creighton CJ, Kato M, et al. Trans-ancestry mutational landscape of hepatocellular carcinoma genomes. Nat Genet. 2014;46:1267– 73. 85. Ahn S-M, Jang SJ, Shim JH, Kim D, Hong S-M, Sung CO, et al. Genomic portrait of resectable hepatocellular carcinomas: implications of RB1 and FGF19 aberrations for patient stratification. Hepatology. 2014;60:1972–82. 86. Kan Z, Zheng H, Liu X, Li S, Barber TD, Gong Z, et al. Whole-genome sequencing identifies recurrent mutations in hepatocellular carcinoma. Genome Res. 2013;23:1422–33. 87. Bassaganyas L, Pinyol R, Esteban-Fabro R, Torrens L, Torrecilla S, Willoughby CE, et al. Copy-Number Alteration Burden Differentially Impacts Immune Profiles and Molecular Features of Hepatocellular Carcinoma. Clin Cancer Res. 2020;26:6350–61. 88. Fernandez-Banet J, Lee NP, Chan KT, Gao H, Liu X, Sung WK, et al. Decoding complex patterns of genomic rearrangement in hepatocellular carcinoma. Genomics. REFERENCES 173 2014;103:189–203. 89. Llovet J, Pinyol R, Kelley R, El-Khoueiry A, Reeves H, Wang X, et al. Molecular pathogenesis and systemic therapies for hepatocellular carcinoma. Nat Cancer. 2021;In press. 90. Bayard Q, Meunier L, Peneau C, Renault V, Shinde J, Nault J, et al. Cyclin A2/E1 activation defines a hepatocellular carcinoma subclass with a rearrangement signature of replication stress. Nat Commun. 2018;9:5235. 91. Nault JC, Mallet M, Pilati C, Calderaro J, Bioulac-Sage P, Laurent C, et al. High frequency of telomerase reverse-transcriptase promoter somatic mutations in hepatocellular carcinoma and preneoplastic lesions. Nat Commun. 2013;4:2218. 92. Nault JC, Calderaro J, Tommaso L Di, Balabaud C, Zafrani ES, Bioulac-Sage P, et al. TERT promoter mutation is an early somatic genetic alteration in the transformation of premalignant nodules in hepatocellular carcinoma on cirrhosis. Hepatology. 2014;60:6589–604. 93. Sung W-K, Zheng H, Li S, Chen R, Liu X, Li Y, et al. Genome-wide survey of recurrent HBV integration in hepatocellular carcinoma. Nat Genet. 2012;44:765–9. 94. Hsu IC, Metcalf RA, Sun T, Welsh JA, Wang NJ, Harris CC. Mutational hot spot in the p53 gene in human hepatocellular carcinomas. Nature. 1991;350:427–8. 95. Bressac B, Kew M, Wands J, Ozturk M. Selective G to T mutations of p53 gene in hepatocellular carcinoma from southern Africa. Nature. 1991;350:429–31. 96. Wang K, Lim HY, Shi S, Lee J, Deng S, Xie T, et al. Genomic landscape of copy number aberrations enables the identification of oncogenic drivers in hepatocellular carcinoma. Hepatology. 2013;58:706–17. 97. Sawey ET, Chanrion M, Cai C, Wu G, Zhang J, Zender L, et al. Identification of a therapeutic strategy targeting amplified FGF19 in liver cancer by Oncogenomic screening. Cancer Cell. 2011;19:347–58. 98. Guichard C, Amaddeo G, Imbeaud S, Ladeiro Y, Pelletier L, Maad I Ben, et al. Integrated analysis of somatic mutations and focal copy-number changes identifies key genes and pathways in hepatocellular carcinoma. Nat Genet. 2012;44:694–8. 99. Sporn MB, Liby KT. NRF2 and cancer: the good, the bad and the importance of context. Nat Rev Cancer. 2012;12:564–71. 100. Chiang DY, Villanueva A, Hoshida Y, Peix J, Newell P, Minguez B, et al. Focal gains of VEGFA and molecular classification of hepatocellular carcinoma. Cancer Res. 2008;68:6779–88. 101. Martinez-Quetglas I, Pinyol R, Dauch D, Torrecilla S, Tovar V, Moeini A, et al. IGF2 Is Up- regulated by Epigenetic Mechanisms in Hepatocellular Carcinomas and Is an Actionable REFERENCES 174 Oncogene Product in Experimental Models. Gastroenterology. 2016;151:1192–205. 102. Paterlini-Bréchot P, Saigo K, Murakami, Y, Chami M, Gozuacik D, Mugnier C, et al. Hepatitis B virus-related insertional mutagenesis occurs frequently in human liver cancers and recurrently targets human telomerase gene. Oncogene. 2003;22:3911–6. 103. Schulze K, Nault JC, Villanueva A. Genetic profiling of hepatocellular carcinoma using next-generation sequencing. J Hepatol. 2016;65:1031–42. 104. Davé U, Cornetta K. AAV Joins the Rank of Genotoxic Vectors. Mol Ther. 2021;29:418–9. 105. Wang D, Tai P, Gao G. Adeno-associated virus vector as a platform for gene therapy delivery. Nat Rev Drug Discov. 2019;18:358–78. 106. Chandler R, Sands M, Venditti C. Recombinant Adeno-Associated Viral Integration and Genotoxicity: Insights from Animal Models. Hum Gene Ther. 2017;28:314–22. 107. Donsante A, Vogler C, Muzyczka N, Crawford J, Barker J, Flotte T, et al. Observed incidence of tumorigenesis in long-term rodent studies of rAAV vectors. Gene Ther. 2001;8:1343–6. 108. Toffanin S, Hoshida Y, Lachenmayer A, Villanueva A, Cabellos L, Minguez B, et al. MicroRNA-based classification of hepatocellular carcinoma and oncogenic role of miR- 517a. Gastroenterology. 2011;140:1618-28.e16. 109. Luk J, Burchard J, Zhang C, Liu A, Wong K, Shek F, et al. DLK1-DIO3 genomic imprinted microRNA cluster at 14q32.2 defines a stemlike subtype of hepatocellular carcinoma associated with poor survival. J Biol Chem. 2011;286:30706–13. 110. Trobridge G, Hirata RRK, Russell DWD. Gene Targeting by Adeno-Associated Virus Vectors Is Cell-Cycle Dependent. Hum Gene Ther. 2005;16:522–6. 111. Cast A, Kumbaji M, D’Souza A, Rodriguez K, Gupta A, Karns R, et al. Liver Proliferation Is an Essential Driver of Fibrosis in Mouse Models of Nonalcoholic Fatty Liver Disease. Hepatol Commun. 2019;3:1036–49. 112. Letouzé E, Shinde J, Renault V, Couchy G, Blanc JF, Tubacher E, et al. Mutational signatures reveal the dynamic interplay of risk factors and cellular processes during liver tumorigenesis. Nat Commun. 2017;8:1315. 113. Ng A, Poon S, Huang S, Lim J, Boot A, Yu W, et al. Aristolochic acids and their derivatives are widely implicated in liver cancers in Taiwan and throughout Asia. Sci Transl Med. 2017;9:eaan6446. 114. Rebouissou S, Nault J. Advances in molecular classification and precision oncology in hepatocellular carcinoma. J Hepatol. 2020;72:215–29. 115. Villanueva A, Chiang DY, Newell P, Peix J, Thung S, Alsinet C, et al. Pivotal role of mTOR signaling in hepatocellular carcinoma. Gastroenterology. 2008;135:1972–83. REFERENCES 175 116. Kaposi-Novak P, Lee J-S, Gòmez-Quiroz L, Coulouarn C, Factor VM, Thorgeirsson SS. Met- regulated expression signature defines a subset of human hepatocellular carcinomas with poor prognosis and aggressive phenotype. J Clin Invest. 2006;116:1582–95. 117. Newell P, Toffanin S, Villanueva A, Chiang DY, Minguez B, Cabellos L, et al. Ras pathway activation in hepatocellular carcinoma and anti-tumoral effect of combined sorafenib and rapamycin in vivo. J Hepatol. 2009;51:725–33. 118. Villanueva A, Hoshida Y, Battiston C, Tovar V, Sia D, Alsinet C, et al. Combining clinical, pathology, and gene expression data to predict recurrence of hepatocellular carcinoma. Gastroenterology. 2011;140:1501-1512.e2. 119. Lachenmayer A, Alsinet C, Savic R, Cabellos L, Toffanin S, Hoshida Y, et al. Wnt-pathway activation in two molecular classes of hepatocellular carcinoma and experimental modulation by sorafenib. Clin Cancer Res. 2012;18:4997–5007. 120. Sia D, Jiao Y, Martinez-Quetglas I, Kuchuk O, Villacorta-Martin C, Castro de Moura M, et al. Identification of an Immune-specific Class of Hepatocellular Carcinoma, Based on Molecular Features. Gastroenterology. 2017;153:812–26. 121. Llovet JMM, Castet F, Heikenwalder M, Maini MKK, Mazzaferro V, Pinato DJJ, et al. Immunotherapies for hepatocellular carcinoma. Nat Rev Clin Oncol. 2021. 122. Ruiz de Galarreta M, Bresnahan E, Molina-Sánchez P, Lindblad KE, Maier B, Sia D, et al. β-Catenin Activation Promotes Immune Escape and Resistance to Anti–PD-1 Therapy in Hepatocellular Carcinoma. Cancer Discov. 2019;9:1124–41. 123. Sangro B, Sarobe P, Hervás-Stubbs S, Melero I. Advances in immunotherapy for hepatocellular carcinoma. Nat Rev Gastroenterol Hepatol 2021 188. 2021;18:525–43. 124. Ringelhan M, Pfister O, O’Connor T, Pikarsky E, Heikenwalder M. The immunology of hepatocellular carcinoma. Nat Immunol. 2018;19:222–32. 125. Moeini A, Torrecilla S, Tovar V, Montironi C, Andreu-Oller C, Peix J, et al. An Immune Gene Expression Signature Associated With Development of Human Hepatocellular Carcinoma Identifies Mice That Respond to Chemopreventive Agents. Gastroenterology. 2019; 157:1383-1397.e11. 126. Wada Y, Nakashima O, Kutami R, Yamamoto O, Kojiro M. Clinicopathological study on hepatocellular carcinoma with lymphocytic infiltration. Hepatology. 1998;27:407–14. 127. Prieto J, Melero I, Sangro B. Immunological landscape and immunotherapy of hepatocellular carcinoma. Nat Rev Gastroenterol Hepatol. 2015;12:681–700. 128. Bruni D, Angell HK, Galon J. The immune contexture and Immunoscore in cancer prognosis and therapeutic efficacy. Nat Rev Cancer. 2020;20:662–80. 129. Lavin Y, Mortha A, Rahman A, Merad M. Regulation of macrophage development and REFERENCES 176 function in peripheral tissues. Nat Rev Immunol. 2015;15:731–44. 130. Fukumura D, Kloepper J, Amoozgar Z, Duda DG, Jain RK. Enhancing cancer immunotherapy using antiangiogenics: opportunities and challenges. Nat Rev Clin Oncol. 2018;15:325–40. 131. Gavalas NG, Tsiatas M, Tsitsilonis O, Politi E, Ioannou K, Ziogas AC, et al. VEGF directly suppresses activation of T cells from ascites secondary to ovarian cancer via VEGF receptor type 2. Br J Cancer. 2012;107:1869–75. 132. Gabrilovich DI, Chen HL, Girgis KR, Cunningham HT, Meny GM, Nadaf S, et al. Production of vascular endothelial growth factor by human tumors inhibits the functional maturation of dendritic cells. Nat Med. 1996;2:1096–103. 133. Galle PR, Forner A, Llovet JM, Mazzaferro V, Piscaglia F, Raoul JL, et al. EASL Clinical Practice Guidelines: Management of hepatocellular carcinoma. J Hepatol. 2018;69:182– 236. 134. Llovet J, Brú C, Bruix J. Prognosis of Hepatocellular Carcinoma: The BCLC Staging Classification. Semin Liver Dis. 1999;19:329–38. 135. Llovet JM, Villanueva A, Marrero JA, Schwartz M, Meyer T, Galle PR, et al. Trial Design and Endpoints in hepatocellular carcinoma: AASLD Consensus Conference. Hepatology. 2021;73:158–91. 136. Ferguson F, Gray N. Kinase inhibitors: the road ahead. Nat Rev Drug Discov. 2018;17:353– 76. 137. Llovet J, Ricci S, Mazzaferro V, Hilgard P, Gane E, Blanc J-FJJ, et al. Sorafenib in advanced hepatocellular carcinoma. N Engl J Med. 2008;359:378–90. 138. Cheng A-L, Finn RS, Qin S, Han K-H, Ikeda K, Piscaglia F, et al. Phase III trial of lenvatinib (LEN) vs sorafenib (SOR) in first-line treatment of patients (pts) with unresectable hepatocellular carcinoma (uHCC). J Clin Oncol. 2017;35:Abstract 4001 (Suppl). 139. Matsui J, Yamamoto Y, Funahashi Y, Tsuruoka A, Watanabe T, Wakabayashi T, et al. E7080, a novel inhibitor that targets multiple kinases, has potent antitumor activities against stem cell factor producing human small cell lung cancer H146, based on angiogenesis inhibition. Int J Cancer. 2008;122:664–71. 140. Zhu AX, Kang Y-K, Yen C-J, Finn RS, Galle PR, Llovet JM, et al. Ramucirumab after sorafenib in patients with advanced hepatocellular carcinoma and increased α-fetoprotein concentrations (REACH-2): a randomised, double-blind, placebo-controlled, phase 3 trial. Lancet Oncol. 2019;20:282–96. 141. Bruix J, Chan SL, Galle PR, Rimassa L, Sangro B. Systemic treatment of hepatocellular carcinoma: An EASL position paper. J Hepatol. 2021;75:960–74. REFERENCES 177 142. Rizvi NA, Hellmann MD, Snyder A, Kvistborg P, Makarov V, Havel JJ, et al. Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer. Science. 2015;348:124–8. 143. Finn RS, Qin S, Ikeda M, Galle PR, Ducreux M, Kim T-Y, et al. Atezolizumab plus Bevacizumab in Unresectable Hepatocellular Carcinoma. N Engl J Med. 2020;382:1894– 905. 144. Cheng A, Qin S, Ikeda M, Galle P, Ducreux M, Zhu A, et al. IMbrave150: Efficacy and safety results from a ph III study evaluating atezolizumab (atezo) + bevacizumab (bev) vs sorafenib (Sor) as first treatment (tx) for patients (pts) with unresectable hepatocellular carcinoma (HCC). Ann Oncol. 2019;30 (Sup 9). 145. Yau T, Kang YK, Kim TY, El-Khoueiry AB, Santoro A, Sangro B, et al. Efficacy and Safety of Nivolumab plus Ipilimumab in Patients with Advanced Hepatocellular Carcinoma Previously Treated with Sorafenib: The CheckMate 040 Randomized Clinical Trial. JAMA Oncol. 2020;6: e204564. 146. Zhu AX, Finn RS, Edeline J, Cattan S, Ogasawara S, Palmer D, et al. Pembrolizumab in patients with advanced hepatocellular carcinoma previously treated with sorafenib (KEYNOTE-224): a non-randomised, open-label phase 2 trial. Lancet Oncol. 2018;19:940– 52. 147. Finn RS, Ryoo BY, Merle P, Kudo M, Bouattour M, Lim HY, et al. Pembrolizumab As Second-Line Therapy in Patients With Advanced Hepatocellular Carcinoma in KEYNOTE- 240: A Randomized, Double-Blind, Phase III Trial. J Clin Oncol. 2020;38:193–202. 148. Finn RS, Ikeda M, Zhu AX, Sung MW, Baron AD, Kudo M, et al. Phase Ib Study of Lenvatinib Plus Pembrolizumab in Patients With Unresectable Hepatocellular Carcinoma. J Clin Oncol. 2020;38:JCO.20.00808. 149. Kim RD, Sarker D, Meyer T, Yau T, Macarulla T, Park J-W, et al. First-in-Human Phase I Study of Fisogatinib (BLU-554) Validates Aberrant Fibroblast Growth Factor 19 Signaling as a Driver Event in Hepatocellular Carcinoma. Cancer Discov. 2019;CD-19-0555. 150. Rimassa L, Assenat E, Peck-Radosavljevic M, Pracht M, Zagonel V, Mathurin P, et al. Tivantinib for second-line treatment of MET-high, advanced hepatocellular carcinoma (METIV-HCC): a final analysis of a phase 3, randomised, placebo-controlled study. Lancet Oncol. 2018;19:682–93. 151. Masucci G V., Cesano A, Hawtin R, Janetzki S, Zhang J, Kirsch I, et al. Validation of biomarkers to predict response to immunotherapy in cancer: Volume I — pre-analytical and analytical validation. J Immunother Cancer. 2016;4:76. 152. Le DT, Durham JN, Smith KN, Wang H, Bartlett BR, Aulakh LK, et al. Mismatch-repair REFERENCES 178 deficiency predicts response of solid tumors to PD-1 blockade. Science. 2017;357:409. 153. Candia J, Bayarsaikhan E, Tandon M, Budhu A, Forgues M, Tovuu LO, et al. The genomic landscape of Mongolian hepatocellular carcinoma. Nat Commun. 2020;11:4383. 154. Wong GL-H, Chan HL-Y, Yiu KK-L, Lai JW-Y, Chan VK-K, Cheung KK-C, et al. Meta-analysis: The association of hepatitis B virus genotypes and hepatocellular carcinoma. Aliment Pharmacol Ther. 2013;37:517–26. 155. Sonneveld MJ, Rijckborst V, Zeuzem S, Heathcote EJ, Simon K, Senturk H, et al. Presence of precore and core promoter mutants limits the probability of response to peginterferon in hepatitis B e antigen-positive chronic hepatitis B. Hepatology. 2012;56:67–75. 156. Martincorena I, Raine KM, Gerstung M, Dawson KJ, Haase K, Van Loo P, et al. Universal Patterns of Selection in Cancer and Somatic Tissues. Cell. 2017;171:1029-1041.e21. 157. Arnedo-Pac C, Mularoni L, Muiños F, Gonzalez-Perez A, Lopez-Bigas N, Schwartz R. OncodriveCLUSTL: A sequence-based clustering method to identify cancer drivers. Bioinformatics. 2019;35:4788–90. 158. Wang PR, Xu M, Toffanin S, Li Y, Llovet JM, Russell DW. Induction of hepatocellular carcinoma by in vivo gene targeting. Proc Natl Acad Sci U S A. 2012;109:11264–9. 159. Cotter T, Rinella M. Nonalcoholic Fatty Liver Disease 2020: The State of the Disease. Gastroenterology. 2020;158:1851–64. 160. Baatarkhuu O, Gerelchimeg T, Munkh-Orshikh D, Batsukh B, Sarangua G, Amarsanaa J. Epidemiology, Genotype Distribution, Prognosis, Control, and Management of Viral Hepatitis B, C, D, and Hepatocellular Carcinoma in Mongolia. Euroasian J Hepato- Gastroenterology. 2018;8:57–62. 161. Wei F, Zheng Q, Li M, Wu M. The association between hepatitis B mutants and hepatocellular carcinoma: A meta-analysis. Medicine (Baltimore). 2017;96:e6835. 162. Chen B-F, Liu C-J, Jow G-M, Chen P-J, Kao J-H, Chen D-S. High prevalence and mapping of pre-S deletion in hepatitis B virus carriers with progressive liver diseases. Gastroenterology. 2006;130:1153–68. 163. Chakravarty D, Gao J, Phillips S, Kundra R, Zhang H, Wang J, et al. OncoKB: A Precision Oncology Knowledge Base. JCO Precis Oncol. 2017;PO.17.00011. 164. Zhu AX, Kudo M, Assenat E, Cattan S, Kang Y-KK, Lim HY, et al. Effect of Everolimus on Survival in Advanced Hepatocellular Carcinoma After Failure of Sorafenib. JAMA. 2014;312:57–67. 165. Moody S, Senkin S, Islam SMA, Wang J, Nasrollahzadeh D, Cortez Cardoso Penha R, et al. Mutational signatures in esophageal squamous cell carcinoma from eight countries with varying incidence. Nat Genet. 2021;53:1553–63. REFERENCES 179 166. WHO. Air pollution in Mongolia. Bull World Health Organ. 2019;97:79:80. 167. UNICEF. Environment & air pollution | UNICEF Mongolia [Internet]. [cited 2020 Nov 23]. Available from: https://www.unicef.org/mongolia/environment-air-pollution 168. World Bank. Air quality analysis of Ulaanbaatar : improving air quality to reduce health impacts. 2011. 1–128 p. 169. Nik-Zainal S, Kucab JE, Morganella S, Glodzik D, Alexandrov LB, Arlt VM, et al. The genome as a record of environmental exposure. Mutagenesis. 2015;30:763. 170. Zhang W, He H, Zang M, Wu Q, Zhao H, Lu L ling, et al. Genetic Features of Aflatoxin- Associated Hepatocellular Carcinoma. Gastroenterology. 2017;153:249-262.e2. 171. Donsante A, Miller DGG, Li Y, Vogler C, Brunt EMM, Russell DWW, et al. AAV Vector Integration Sites in Mouse Hepatocellular Carcinoma. Science. 2007;317:477. 172. Bell P, Wang L, Lebherz C, Flieder DB, Bove MS, Wu D, et al. No Evidence for Tumorigenesis of AAV Vectors in a Large-Scale Study in Mice. Mol Ther. 2005;12:299– 306. 173. Li H, Malani N, Hamilton SRR, Schlachterman A, Bussadori G, Edmonson SEE, et al. Assessing the potential for AAV vector genotoxicity in a murine model. Blood. 2011;117:3311–9. 174. Ma C, Kesarwala AH, Eggert T, Medina-Echeverz J, Kleiner DE, Jin P, et al. NAFLD causes selective CD4+T lymphocyte loss and promotes hepatocarcinogenesis. Nature. 2016;531:253–7. 175. Pinyol R, Torrecilla S, Wang H, Montironi C, Piqué-Gili M, Torres-Martin M, et al. Molecular characterisation of hepatocellular carcinoma in patients with non-alcoholic steatohepatitis. J Hepatol. 2021;75:865–78. 176. Lefebvre P, Staels B. Hepatic sexual dimorphism — implications for non-alcoholic fatty liver disease. Nat Rev Endocrinol. 2021;17:662–70. 177. Khan KA, Kerbel RS. Improving immunotherapy outcomes with anti-angiogenic treatments and vice versa. Nat Rev Clin Oncol. 2018;15:310–24. 178. Kimura T, Kato Y, Ozawa Y, Kodama K, Ito J, Ichikawa K, et al. Immunomodulatory activity of lenvatinib contributes to antitumor activity in the Hepa1-6 hepatocellular carcinoma model. Cancer Sci. 2018;109:3993–4002. 179. Yi C, Chen L, Lin Z, Liu L, Shao W, Zhang R, et al. Lenvatinib Targets FGF Receptor 4 to Enhance Antitumor Immune Response of Anti–Programmed Cell Death-1 in HCC. Hepatology. 2021;74:2544–60. 180. Deng H, Kan A, Lyu N, Mu L, Han Y, Liu L, et al. Dual Vascular Endothelial Growth Factor Receptor and Fibroblast Growth Factor Receptor Inhibition Elicits Antitumor Immunity REFERENCES 180 and Enhances Programmed Cell Death-1 Checkpoint Blockade in Hepatocellular Carcinoma. Liver cancer. 2020;9:338–57. 181. Yost KE, Satpathy AT, Wells DK, Qi Y, Wang C, Kageyama R, et al. Clonal replacement of tumor-specific T cells following PD-1 blockade. Nat Med. 2019;25:1251–9. 182. Böttcher JP, Reis e Sousa C. The Role of Type 1 Conventional Dendritic Cells in Cancer Immunity. Trends in Cancer. 2018;4:784–92. 183. Arce Vargas F, Furness AJS, Litchfield K, Joshi K, Rosenthal R, Ghorani E, et al. Fc Effector Function Contributes to the Activity of Human Anti-CTLA-4 Antibodies. Cancer Cell. 2018;33:649-663. 184. Vignali DAA, Collison LW, Workman CJ. How regulatory T cells work. Nat Rev Immunol. 2008;8:523–32. 185. Sakaguchi S, Yamaguchi T, Nomura T, Ono M. Regulatory T Cells and Immune Tolerance. Cell. 2008;133:775–87. 186. Kato Y, Tabata K, Kimura T, Yachie-Kinoshita A, Ozawa Y, Yamada K, et al. Lenvatinib plus anti-PD-1 antibody combination treatment activates CD8 + T cells through reduction of tumor-associated macrophage and activation of the interferon pathway. PLoS One. 2019;14:e0212513. 187. Wherry EJ, Kurachi M. Molecular and cellular insights into T cell exhaustion. Nat Rev Immunol. 2015;15:486–99. 188. Torrens L, Montironi C, Puigvehí M, Mesropian A, Leslie J, Haber PK, et al. Immunomodulatory Effects of Lenvatinib Plus Anti–Programmed Cell Death Protein 1 in Mice and Rationale for Patient Enrichment in Hepatocellular Carcinoma. Hepatology. 2021;74:2652–69. ANNEX A ANNEX A 183 1. Publications List of scientific articles and book chapters written during the period of the doctoral thesis (2016- 2021) Published articles • Torrens L, Montironi C, Puigvehí M, Mesropian A, Leslie J, Haber PK, et al. Immunomodulatory Effects of Lenvatinib Plus Anti–Programmed Cell Death Protein 1 in Mice and Rationale for Patient Enrichment in Hepatocellular Carcinoma. Hepatology. 2021 Jun;74:2652–2669. • Dalwadi DA, Torrens L, Abril-Fornaguera J, Pinyol R, Willoughby C, Posey J, et al. Liver Injury Increases the Incidence of HCC following AAV Gene Therapy in Mice. Mol Ther. 2021 Feb;29(2):680-690. • Carrillo-Reixach J, Torrens L, Simon-Coma M, Royo L, Domingo-Sàbat M, Abril-Fornaguera J, et al. Epigenetic footprint enables molecular risk stratification of hepatoblastoma with clinical implications. J Hepatol. 2020 Aug;73(2):328-341. • Bassaganyas L, Pinyol R, Esteban-Fabró R, Torrens L, Torrecilla S, Willoughby C, et al. Copy- Number Alteration Burden Differentially Impacts Immune Profiles and Molecular Features of Hepatocellular Carcinoma. Clin Cancer Res. 2020 Dec 26(23)6350-6361. • Moeini A, Sia D, Zhang Z, Camprecios G, Stueck A, Dong H, Montal R, Torrens L, et al. Mixed hepatocellular cholangiocarcinoma tumors: Cholangiolocellular carcinoma is a distinct molecular entity. J Hepatol. 2017 May;66(5):952-961. Submitted articles • Torrens L*, Puigvehí M*, Torres-Martín M, Wang H, Maeda M, Haber PK, et al. Hepatocellular Carcinoma in Mongolia Delineates Unique Genomic Features Associated with Environmental Agents. Submitted to Proc Natl Acad Sci U S A. 2021. *Contributed equally. • Torrens L*, Abril-Fornaguera J*, Carrillo-Reixach J, Balaseviciute U, Rialdi A, Del Río-Álvarez A, et al. Identification of IGF2 as Genomic Driver and Actionable Therapeutic Target in Hepatoblastoma. Submitted to Cancer Res. 2021. *Contributed equally. ANNEX A 184 • Montironi C*, Castet F*, Haber PK*, Pinyol R, Torres-Martin M, Torrens L, et al. Inflamed and non-inflamed classes of HCC: a revised immunogenomic classification. Submitted to Gut. 2021. *Contributed equally. • Esteban-Fabró R*, Willoughby CE*, Piqué-Gili, M, Montironi C, Abril-Fornaguera J, Judit Peix, Torrens L, et al. Cabozantinib enhances anti-PD1 activity and elicits a neutrophil-based immune response in hepatocellular carcinoma. Submitted to Clin Cancer Res. 2021. *Contributed equally. Book chapters • Torrens L, Pinyol R, Jimenez W, Llovet JM (2016). Overview of Translational Medicine. In Llovet JM (Ed.), Handbook of Translational Medicine (pp 21-28). Barcelona, Spain: Edicions de la Universitat de Barcelona. ISBN: 978-84-475-4030-3. • Soucek L, Torrens L, Pujades C, Clària J (2016). Experimental Models. In Llovet JM (Ed.), Handbook of Translational Medicine (pp 118-126). Barcelona, Spain: Edicions de la Universitat de Barcelona. ISBN: 978-84-475-4030-3. ANNEX A 185 2. Communications to Scientific Meetings Participation in national and international meetings with poster or oral communications. Presenter name is underlined. • Torrens L, Puigvehí M, Torres-Martín M, Wang H, Maeda M, Haber P, et al. Molecular Characterization of HCC in Mongolia Delineates Unique Genomic Features. EASL International Liver Congress 2021. Poster presentation. • Torrens L, Abril-Fornaguera J, Carrillo J, Balaseviciute U, Rialdi A, Haber P, et al. Identification of IGF2 as Genomic Driver and Actionable Therapeutic Target in Hepatoblastoma. ILCA 16th Annual Conference. 2021 Virtual Conference. Poster presentation. • Torrens L, Puigvehí M, Torres-Martín M, Wang H, Maeda M, Haber P, et al. Molecular Characterization of HCC in Mongolia Delineates Unique Genomic Features. ILCA 16th Annual Conference. 2021 Virtual Conference. Poster presentation. • Torrens L, Abril-Fornaguera J, Carrillo J, Balaseviciute U, Rialdi A, Haber P, et al. Identification of IGF2 as Genomic Driver and Actionable Therapeutic Target in Hepatoblastoma. EASL International Liver Congress 2021. Oral presentation. • Esteban-Fabró R, Willoughby CE, Piqué-Gili M, Montironi C, Abril-Fornaguera J, Peix J, Torrens L, et al. Cabozantinib enhances anti-pd1 efficacy and elicits a neutrophil-based immune response in murine models: implications for human HCC. ILCA 16th Annual Conference. 2021 Virtual Conference. Oral presentation. • Torrens L, Abril-Fornaguera J, Carrillo-Reixach J, Balaseviciute U, Rialdi A, Haber P et al. Identificación de IGF2 como diana terapéutica en Hepatoblastoma. 46 Congreso AEEH. 2021. Madrid, Spain. Poster presentation. • Torrens L, Montironi C, Mesropian A, Haber PK, Maeda M, Puigvehí M, et al. Immune- Remodeling Effects of Lenvatinib Plus Anti-PD1 in a Murine Model of Hepatocellular Carcinoma. ILCA 15th Annual Conference. 2020 Virtual Conference. Poster presentation (Best Basic / Translational poster session). • Esteban-Fabró R, Willoughby CE, Piqué-Gili M, Peix J, Montironi C, Abril-Fornaguera J, Torrens L, et al. Cabozantinib enhances the efficacy and immune activity of anti-PD1 ANNEX A 186 therapy in a murine model of hepatocellular carcinoma. ILCA 15th Annual Conference. 2020 Virtual Conference. • Sia D, Puigvehi M, Torrens L, Wang H, Torres-Martin M, Maeda M, et al. Molecular Characterization of Hepatocellular Carcinoma in Mongolia Delineates Unique Genomic Features. AASLD: The Liver Meeting. 2020 Virtual Conference. Poster presentation. • Torrens L, Montironi C, Haber PK, Maeda M, Puigvehi M, Kamphorst A, et al. Efecto de la combinación de lenvatinib y anti-PD1 sobre el sistema inmune en un modelo experimental de carcinoma hepatocelular. 45 Congreso AEEH. 2020. Madrid, Spain. Oral poster presentation. • Esteban-Fabró R, Willoughby CE, Piqué-Gili M, Peix J, Montironi C, Abril-Fornaguera J, Torrens L, et al. Cabozantinib enhances the efficacy and immune modulatory activity of anti-PD1 therapy in a syngeneic mouse model of hepatocellular carcinoma. EASL International Liver Congress 2020. Oral presentation. • Esteban-Fabró R, Willoughby CE, Piqué-Gili M, Peix J, Montironi C, Abril-Fornaguera J, Torrens L, et al. Cabozantinib aumenta la eficacia y actividad inmunomoduladora de la terapia con anti-PD1 en un modelo murino singénico de carcinoma hepatocelular. 45 Congreso AEEH. 2020. Madrid, Spain. Oral communication. • Torrens L, Montironi C, Haber P, Kuchuk O, Akers N, Simon-Coma M, et al. Identification of IGF2 as Genomic Driver and Therapeutic Target in Hepatoblastoma. AASLD: The Liver Meeting 2019. Boston, USA. Poster presentation. • Torrens L, Montironi C, Haber P, Kuchuk O, Akers N, Simon-Coma M, et al. Identification of IGF2 as Genomic Driver and Therapeutic Target in Hepatoblastoma. ILCA 13th Annual Conference. 2019. Chicago, USA. Poster presentation. ANNEX A 187 3. Grants and Awards Fellowship grants • Research Stay Fellowship for International Doctorates. University of Barcelona 2018 Scientific awards • Top-rated basic-translational poster session. ILCA 2020 • Best Poster Presentation – HUNTER Liver Workshop. Cancer Research UK, Beatson Institute, Glasgow (UK) 2020. Competitive project grants (co-investigator) • National Health Institute, Spain. I+D Program (Grant number: PID2019-105378RB-100). “Molecular characterization of obesity / diabetes / NASH-related hepatocellular carcinoma”. PI: JM Llovet. 2020 – 2023. • CRUK, AECC, AIRC, Accelerator Award (C9380/A26813). “HUNTER - Hepatocellular Carcinoma Expediter Network”. PI: JM Llovet. 2019 – Present. • Horizon 2020 – European Commission (Call H2020-PHC-2015, number 667273). Title: “HEP- CAR - Mechanisms underlying hepatocellular carcinoma pathogenesis and impact of co- morbidities”. P: JM Llovet. 2016 – 2019. • National Health Institute, Spain. I+D Program (Grant number: SAF2016-76390-R). “Mechanisms of resistance to TKIs in hepatocellular carcinoma”. PI: JM Llovet. Team member. 2016 – 2019. Investigator initiated sponsored studies (co-investigator) • Bayer Pharmaceuticals. “Discovery of biomarkers predictors of response and/or resistance to anti-PD1-based immune checkpoint inhibitors in advanced HCC”. PI: JM Llovet. 2018 - 2021 • Boehringer-Ingelheim. “Role of Xentuzumab for the treatment of hepatoblastoma overexpressing IGF2”. PI: JM Llovet. 2019 – 2020. ANNEX A 188 • Eisai Inc. “Impact of lenvatinib alone or in combination with anti-PD1 on the immune system in HCC and assessment of the synergistic anti-tumoral effect in experimental models”. PI: JM Llovet. 2018 - 2019. • Ipsen. “Impact of cabozantinib alone or in combination with anti-PD1 on the immune system in HCC and assessment of the synergistic anti-tumoral effect in experimental models”. PI: JM Llovet. 2018 – 2019. ANNEX B ANNEX B 191 Supplementary Data Study 1 SUPPLEMENTARY MATERIALS AND METHODS Clinical and histological data All samples in the study were fresh-frozen. Tissue samples were coded previous to storage using consecutive numbering. The code did not include any patient identifier, and the research team at Mount Sinai received already de-identified samples. The diagnosis of HCC was confirmed after a first evaluation made by 3 independently working expert pathologists in Mount Sinai (WQL, CM, and ST), and those samples with >50% necrotic tissue (n=18), tumors other than HCC (n=7), or repeated (n=2) were excluded. Thus, a final number of 192 patients were included for further evaluation. Baseline clinico-pathological characteristics were collected for both cohorts (Table 1). All histological evaluations were performed by 2 expert pathologists, blinded to clinical data. Fibrosis stage was scored according to the METAVIR Scale [1]. All the above-mentioned variables were also collected for the Western cohort except for BMI, alanine aminotransferase (ALT) values, tumor size and number, region of origin, and presence of steatosis/steatohepatitis. All data were stored in a database containing de- identified information, and electronic files were stored according to Mount Sinai IRB protocols with encryption and password protection. Viral hepatitis evaluation The presence of viral infections (HBV, HCV and HDV) was assessed in the non-tumor tissue of all Mongolian samples, and was compared to the data obtained from Mongolian charts (HBV surface antigen -HBsAg- and HCV antibodies; HDV was not routinely evaluated in Mongolia). Intrahepatic HBV and HDV status were assessed by quantitative PCR (qPCR). HBV-DNA was assessed by Taqman qPCR (ID Pa03453406 s1, ABI, Thermo Fisher) using the ViiA7 Real Time PCR System (ABI) as previously described [2]. The calibration curve was prepared using ten-fold serial dilutions of a plasmid containing an HBV monomer (pHBV-EcoR1). Total HDV-RNA was determined by one-step RT-qPCR as previously reported [3,4]. For absolute quantification, serial dilution of an HDV-RNA standard (WHO 1st International Standard, Paul-Ehrlich-Institut) was included in each assay [5]. All samples positive for HDV were considered HBV/HDV positive. HCV status was determined by conventional PCR. Specifically, HCV RNA was retrotranscribed to cDNA with EcoDry Premix (Double Primed) (Takara cat# 639549) and HCV-specific sequences were amplified under standard conditions using the following primer pair: Fw CACGCAGAAAGCGTCTAG, HCV; Rv TTGATCCAAGAAAGGACCC [6]. PCR products were run on an agarose gel, purified using PureLink Quick Gel Extraction Kit (Invitrogen cat# K210012) and sequenced by Sanger (Macrogen, USA). HBV and HDV genotyping HBV and HDV genotypes were determined by direct sequencing and phylogenetic analysis of a 1100 bp fragment of the HBV retrotranscriptase [7] and a fragment of 370bp encompassing approximately 85% of the large HDV antigen (HDAg) [8], respectively. Multiple alignments were performed with ClustalW [9] and maximum likelihood trees were obtained with MEGA X software [10]. Analysis of HBV mutations HBV mutations associated with HCC development were assessed by nested PCR (GoTaq Flexi DNA Polymerase - Promega). Specifically, precore region was screened for nucleotide substitution G1896A and the basal core promoter (BCP) region was checked for the presence of 2 nucleotides substitutions (A1762T and G1764A). A DNA segment composing of the BCP, precore, and partial C regions was amplified by nested PCR and analyzed by direct sequencing [11]. ANNEX B 192 Whole exome sequencing mutational variant calling in in-house cohorts Mutational variant calling was performed following the Tigris pipeline (v2.0.1). BWA 0.7.17 was used for alignment, followed by base quality score recalibration via BQSR, read deduplication via Picard MarkDuplicates, germline molecular variant (SNV and small indel) calling via HaplotypeCaller, and somatic molecular variant calling via Mutect2, which calls variants using local de novo assembly and then does a two-pass filter using heuristics (further details can be found in the MuTect2 whitepaper from its GitHub repo at https://github.com/broadinstitute/gatk/tree/master/docs/mutect) . After applying these filters in MuTect2, the twice-filtered MuTect2 output was then filtered for 'PASS' variants only with allele frequency >= 5% for downstream analysis. Tigris computes depth-based and other NGS library QC metrics using GATK3 DepthOfCoverage and CallableLoci, as well as Picard. Lastly, somatic copy number variants (sCNV) were called using tumor/normal SAAS-CNV (v0.3.4) workflow that models allele balance to determine balanced versus unbalanced somatic gains and losses, as well as determine somatic copy- neutral loss of heterozygosity [12]. SAAS-CNV output were further processed using GISTIC2.0 for somatic CNV analysis. Analysis of gene mutations and filtering in previously published cohorts We used whole exome sequenxing (WES data to assess the mutation profile in the European [13], Korean [14], TCGA [15] and Mongolian NCI [16] HCC cohorts. In the TCGA cohort, only variants with filter PASS were considered and “3_prime_UTR_variant”, “5_prime_UTR_variant”, “intron_variant”,” synonymous_variant” were filtered from the cohort. For the Korean and European cohorts, only were accepted the following types of mutations, filtering the rest of the annotated subtypes: Missense_Mutation, “Nonsense_Mutation”, “Splice_Site”, “Translation_Start_Site”, “Frame_Shift_Ins”, “In_Frame_Ins”, “Frame_Shift_Del”, “In_Frame_Del”, “3'Flank”, “5'Flank” and “Nonstop_Mutation”. For all cohorts, only VAF ≥ 0.05 was accepted for further analysis. Tumor mutational burden (TMB) from these external cohorts was calculated as previously indicated. Other TMB calculation approaches are provided in Supplementary Table 7 for comparison. Somatic copy number variations (SNVs) analysis HaplotypeCaller [17] was used to generate germline VCF files as input for SAAS-CNV (v0.3.4) [12], which in turn generated segmentation file as input for GISTIC 2.0 run [18]. The “log2ratio.Median.adj” column from saasCNV output was used for GISTIC 2.0 run, with the following parameter flags -genegistic 1 - smallmem 1 -broad 1 -brlen 0.98 -conf 0.99 -armpeel 0 -savegene 1 -gcm extreme -qvt 0.1 -cap 2.0 -ta 0.85 -td 0.74. Identification of potential driver genes OncodriveCLUSTL and dN/dScv algorithms were used to identify genes harboring significantly more mutations than expected by chance [19,20] among the genes significantly more mutated in the Mongolian cohort compared to the Western cohort. Genes predicted to have an enrichment for damaging alterations by OncodriveCLUSTL or dN/dScv were selected (q<0.05). The selected genes were filtered for cancer- related genes according to the OncoKB Cancer Gene List or previously reported studies in HCC [13,21]. TERT promoter mutations detection The promoter region of TERT in Mongolian samples was amplified by PCR and sequenced using Sanger sequencing as previously described [22]. The number of TERT promoter mutations was compared to the reported percentages in Western cohorts (55-60%) [13]. Identification of de novo mutational signatures in Mongolian tumors R package MutationalPatterns [23] was used to perform de novo mutational signature extraction. Extracted signatures were mapped against COSMICv3. De novo signatures were mapped to single signatures and linear combinations of two if the cosine similarity was > 0.9. One novel signature “SBS ANNEX B 193 Mongolia” was revealed with cosine similarity below the threshold for all comparisons (maximum observed cosine similarity of 0.818). Mutational signature fitting was performed using the quadprog R package [24], using HCC specific COSMICv3 mutational signatures plus SBS Mongolia. To select HCC specific signatures, COSMICv3 signatures were assessed in 493 HCC samples from the Mongolian (n=151), Western (n=112) and TCGA (n=230) cohorts. Signatures occurring in ≥ 40 HCC samples (Supplementary Table 17) or signatures that were revealed via de novo mutational signature extraction and able to be mapped to COSMICv3 reference were selected (i.e., SBS1, SBS4, SBS5, SBS6, SBS12SBS16, SBS18, SBS22, SBS26, SBS29, SBS40). In order to assess the confidence of signature assignment across our samples in signature fitting, a previously reported bootstrap approach was adopted [25]. At each bootstrap, we randomly selected the same number of mutations with replacement from the original observed mutational profile of a given tumor sample (classified by the 96 trinucleotide mutation types) and performed signature fitting to estimate signature weight (quadprog R package), resulting in a distribution of signature weights for each signature from all bootstraps (N = 500) in a given tumor. Based on the signature weight distribution, for any given sample, we were able to estimate confidence level. At p value = 0.1 (one sided), the 10% quantile of signature weights would mean we were 90% confident that the signature weight was above that 10% quantile value. Finally, samples were considered positive for a mutational signature when the bootstrap exposure cutoff was ≥ 0.1. Identification of de novo mutational signatures in Mongolian non-tumoral liver samples Mutational signature analysis was used to assess signatures in the adjacent non-tumoral samples. First, for variant calling in the adjacent non-tumoral liver tissue, we subtracted the mutations in tumors from the mutation in non-tumoral tissue using MuTect2. Next, only samples with total SNV count ≥ 10 (for variants in exome region only at allelic frequency cutoff of 0.05) were selected for subsequent mutational signature analysis, resulting in a total of 78 samples (64 Mongolian cohort plus 14 Western cohort). Due to the small number of unique SNVs in adjacent non-tumoral samples, the analysis was performed on pooled variants from each cohort. The mutational signature fitting analysis was performed using all HCC specific COSMICv3 signatures (Supplementary Table 17) plus SBS Mongolia. Analysis of environmental signatures Signature fitting analysis was performed using signatures from the Compendium of Mutational Signatures of Environmental Agent [26]. Specifically, all the 52 signatures included in the Compendium from agents generating significantly different substitution profiles compared to untreated controls were used [26]. The weights of each mutational signature contributing to an individual tumor sample were obtained using the deconstructSigs R package (https://github.com/raerose01/deconstructSigs). The trinucleotide count for each sample was normalized by multiplying it by a ratio of its occurence in the genome to its occurence in the exome (exome2genome method), following recommendations for WES data. Signature contributions with a weight <0.25 were discarded from the analysis. A signature was considered present in an individual tumor sample when the weight threshold was ≥ 0.1. (Supplementary Table 18). SUPPLEMENTARY REFERENCES 1. Bedossa P, Poynard T. An algorithm for the grading of activity in chronic hepatitis C. The METAVIR Cooperative Study Group. Hepatology. 1996;24:289–93. 2. Malmström S, Larsson SB, Hannoun C, et al. Hepatitis B viral DNA decline at loss of HBeAg is mainly explained by reduced cccdna load - down-regulated transcription of PgRNA has limited impact. PLoS One. 2012;7. 3. Ferns RB, Nastouli E, Garson JA. Quantitation of hepatitis delta virus using a single-step internally controlled real-time RT-qPCR and a full-length genomic RNA calibration standard. J Virol Methods. 2012;179:189–94. ANNEX B 194 4. Giersch K, Homs M, Volz T, et al. Both interferon alpha and lambda can reduce all intrahepatic HDV infection markers in HBV/HDV infected humanized mice. Sci Rep. 2017;7. 5. Le Gal F, Brichler S, Sahli R, et al. First international external quality assessment for hepatitis delta virus RNA quantification in plasma. Hepatology. 2016;64:1483–94. 6. Balart LA, Perrillo R, Roddenberry J, et al. Hepatitis C RNA in liver of chronic hepatitis C patients before and after interferon alfa treatment. Gastroenterology. 1993;104:1472–7. 7. Tong Y, Liu B, Liu H, et al. New universal primers for genotyping and resistance detection of low HBV DNA levels. Med. 2016;95:e4618. 8. Le Gal F, Brichler S, Drugan T, et al. Genetic diversity and worldwide distribution of the deltavirus genus: A study of 2,152 clinical strains. Hepatology. 2017;66:1826–41. 9. Larkin MA, Blackshields G, Brown NP, et al. Clustal W and Clustal X version 2.0. Bioinformatics. 2007;23:2947–8. 10. Kumar S, Stecher G, Li M, et al. MEGA X: Molecular evolutionary genetics analysis across computing platforms. Mol Biol Evol. 2018;35:1547–9. 11. Chen CH, Changchien CS, Lee CM, et al. Combined mutations in pre-S/surface and core promoter/precore regions of hepatitis B virus increase the risk of hepatocellular carcinoma: A case- control study. J Infect Dis. 2008;198:1634–42. 12. Zhang Z, Hao K. SAAS-CNV: A Joint Segmentation Approach on Aggregated and Allele Specific Signals for the Identification of Somatic Copy Number Alterations with Next-Generation Sequencing Data. Wang E, editor. PLOS Comput Biol. 2015;11:e1004618. 13. Schulze K, Imbeaud S, Letouzé E, et al. Exome sequencing of hepatocellular carcinomas identifies new mutational signatures and potential therapeutic targets. Nat Genet. 2015;47:505–11. 14. Ahn S-M, Jang SJ, Shim JH, et al. Genomic portrait of resectable hepatocellular carcinomas: implications of RB1 and FGF19 aberrations for patient stratification. Hepatology. 2014;60:1972–82. 15. Ally A, Balasundaram M, Carlsen R, et al. Comprehensive and Integrative Genomic Characterization of Hepatocellular Carcinoma. Cell. 2017;169:1327-1341.e23. 16. Candia J, Bayarsaikhan E, Tandon M, et al. The genomic landscape of Mongolian hepatocellular carcinoma. Nat Commun. 2020;11:4383. 17. Poplin R, Ruano-Rubio V, DePristo MA, et al. Scaling accurate genetic variant discovery to tens of thousands of samples. Biorxiv 201178. 2018; 18. Mermel CH, Schumacher SE, Hill B, et al. GISTIC2.0 facilitates sensitive and confident localization of the targets of focal somatic copy-number alteration in human cancers. Genome Biol. 2011;12:R41. 19. Martincorena I, Raine KM, Gerstung M, et al. Universal Patterns of Selection in Cancer and Somatic Tissues. Cell. 2017;171:1029-1041.e21. 20. Arnedo-Pac C, Mularoni L, Muiños F, et al. OncodriveCLUSTL: A sequence-based clustering method to identify cancer drivers. Bioinformatics. 2019;35:4788–90. 21. Chakravarty D, Gao J, Phillips S, et al. OncoKB: A Precision Oncology Knowledge Base. JCO Precis Oncol. 2017;2017:1–16. 22. Nault JC, Mallet M, Pilati C, et al. High frequency of telomerase reverse-transcriptase promoter somatic mutations in hepatocellular carcinoma and preneoplastic lesions. Nat Commun. 2013;4:2218. 23. Blokzijl F, Janssen R, van Boxtel R, et al. MutationalPatterns: Comprehensive genome-wide analysis of mutational processes. Genome Med. 2018;10. 24. Lynch AG. Decomposition of mutational context signatures using quadratic programming methods. F1000Research. 2016;5:1253. 25. Huang X, Wojtowicz D, Przytycka TM. Detecting presence of mutational signatures in cancer with confidence. Bioinformatics. 2018;34:330–7. 26. Kucab JE, Zou X, Morganella S, et al. A Compendium of Mutational Signatures of Environmental Agents. Cell. 2019;177:821-836.e16. ANNEX B 195 SUPPLEMENTARY FIGURES Supplementary Figure 1. Clinical and viral characterization of the study cohort. a Rate of patients within each fibrosis grade in the Mongolian and Western cohorts. p values show differences in the rate of F3-4 patients in each subgroup (Fisher exact test). b-c Distribution of HBV genotypes (b) in the Mongolian (n = 106) and Western (n = 44) cohorts, and (c) in the Western cohort divided by USA (n = 20) and Europe (n = 24). Sub-genotypes are indicated in color. d Rate of HBV basal core promoter and pre-core mutations in the Mongolian and Western cohorts. 0% 20% 40% 60% 80% 100% BCP_A1762T BCP_G1764A PreC_G1 896A % o f p at ie nt s h ar bo rin g m ut at io ns Mongolia (n=75) Western (n=35) p < 0.001 p < 0.001 p = 0.089 c d a b 0% 20% 40% 60% 80% 100% HBV HBV/HDV HCV Non-infected Pa tie nt s w ith F 3- F4 s ta ge (% ) Mongolian cohort Western cohort ANNEX B 196 Supplementary Figure 2. HBV DNA load in liver samples. a HBV-DNA levels in liver samples from HDV positive and negative patients in the Mongolian cohort (log copies/ug total DNA). b Comparison of survival according to levels of HBV-DNA in the Mongolian cohort. The thresholds of HBV-DNA are established according to the percentile 25 to define low and high HBV-DNA load (log copies/μg total DNA). c HBV DNA levels in (c) all HBV-positive samples from the Mongolian and Western cohorts. Box plots indicate median (middle line), 25th, 75th percentile (box) and 5th and 95th percentile (whiskers). Low HBV-DNA load 0 10 20 30 40 50 Log Rank p=0.025 Follow-up (months) 1.0 0.8 0.6 0.4 0.2 0.0 Ov er al l s ur vi va l Patients at risk High HBV-DNA load High HBV DNA load Low HBV DNA load 29 27 26 15 4 0 70 59 51 44 29 0 a b c p = 0.23 p = 0.001 HBV DNA load in HBV positive samples HBV DNA load in Mongolian samples ANNEX B 197 Supplementary Figure 3. Broad copy number variaton profiles in Mongolian and Western HCCs. a Frequency of broad CNV along the genome in the Mongolian and b Western cohorts. The right panel indicates chromosomal regions and the bottom axis indicates the frequency of broad changes (gains and losses). c-d Total broad CNV (c), broad gains (d), and broad losses (e) in Mongolian and Western HCC. Box plots indicate median (middle line), 25th, 75th percentile (box) and 5th and 95th percentile (whiskers). a 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 2122 −1 −0.5 0 0.5 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 −1 −0.5 0 0.5 1 b c d e ANNEX B 198 Supplementary Figure 4. Mutational profile of Mongolian and Western HCC by etiology. a-c Mutations per tumor in HCC samples from patients positive and negative for HBV (a), HDV (b), and HCV (c) infections. Data from the in-house Mongolian (n = 151), Mongolian NCI (n = 71) and Western (n = 112) cohorts are shown. Y axis was cut at 300 mutations/tumor to facilitate data interpretation. Box plots indicate median (middle line), 25th, 75th percentile (box) and 5th and 95th percentile (whiskers). d-e Mutational landscape in the Mongolian (d; n=151) and Western cohorts (e; n=112) sorted by viral status. Genes with significant differences between Mongolian and Western cohorts are shown in bold green (Fisher p < 0.05). Top panel shows tumor mutational burden (TMB, mutations/Mb) per sample. Middle panel indicates the presence of mutations per sample (right) and overall percentage (left) in the most frequently mutated HCC drivers. Percentage of mutations by viral status is also shown, and genes with significant differences are highlighted in green. Bottom panel details viral etiology, gender, and age. d e a b c Mongolian cohort (n = 151) 0 5 10 15 TMB Gender Age >= 60 Fibrosis stage Etiology 02040 % Mutations TP53 CTNNB1 ALB ARID1A APOB AXIN1 TSC2 ATM RB1 KEAP1 ARID2 CDKN2A MAP1B ACVR2A PIK3CA AHCTF1 PTEN 46.4% 36.4% 17.2% 17.2% 15.2% 11.3% 9.3% 8.6% 7.3% 6.6% 6% 4.6% 4.6% 3.3% 2.6% 2.6% 2% Alterations Missense variant Splice acceptor variant Frameshift variant Stop gained Splice donor variant Inframe deletion Gender Female Male Age >= 60 <60 >=60 Positive Yes No NA M on go lia n HB V po sit ive HB V ne ga tiv e HD V po sit ive HD V ne ga tiv e HC V po sit ive HC V ne ga tiv e 46.3% 35.8% 20.9% 10.4% 13.4% 14.9% 10.4% 7.5% 6% 6% 11.9% 3% 4.5% 6% 0% 4.5% 3% 46.4% 36.9% 14.3% 22.6% 16.7% 8.3% 8.3% 9.5% 8.3% 7.1% 1.2% 6% 4.8% 1.2% 4.8% 1.2% 1.2% 47.4% 38.6% 21.1% 10.5% 15.8% 15.8% 8.8% 8.8% 5.3% 5.3% 14% 0% 3.5% 3.5% 0% 5.3% 1.8% 45.7% 35.1% 14.9% 21.3% 14.9% 8.5% 9.6% 8.5% 8.5% 7.4% 1.1% 7.4% 5.3% 3.2% 4.3% 1.1% 2.1% 46% 32.2% 13.8% 19.5% 14.9% 9.2% 9.2% 9.2% 6.9% 6.9% 1.1% 8% 5.7% 3.4% 4.6% 1.1% 1.1% 46.9% 42.2% 21.9% 14.1% 15.6% 14.1% 9.4% 7.8% 7.8% 6.2% 12.5% 0% 3.1% 3.1% 0% 4.7% 3.1% Western cohort (n = 112) 0 5 10 15 TMB Gender Age >= 60 Fibrosis stage Origin Etiology 02040 % Mutations TP53 CTNNB1 ALB ARID1A APOB AXIN1 TSC2 ATM RB1 KEAP1 ARID2 CDKN2A MAP1B ACVR2A PIK3CA AHCTF1 PTEN 32.1% 42.9% 11.6% 9.8% 4.5% 7.1% 0.9% 6.2% 4.5% 5.4% 3.6% 1.8% 0% 3.6% 3.6% 0% 1.8% 42.4% 42.4% 3% 6.1% 6.1% 12.1% 3% 9.1% 6.1% 9.1% 3% 3% 0% 3% 0% 0% 6.1% 27.8% 43% 15.2% 11.4% 3.8% 5.1% 0% 5.1% 3.8% 3.8% 3.8% 1.3% 0% 3.8% 5.1% 0% 0% 27.7% 44.7% 17% 6.4% 6.4% 6.4% 0% 8.5% 0% 6.4% 6.4% 0% 0% 0% 4.3% 0% 0% 35.4% 41.5% 7.7% 12.3% 3.1% 7.7% 1.5% 4.6% 7.7% 4.6% 1.5% 3.1% 0% 6.2% 3.1% 0% 3.1% Alterations Missense variant Splice acceptor variant Frameshift variant Stop gained Splice donor variant Inframe deletion Gender Female Male Age >= 60 <60 >=60 Etiology HBV HBV/HDV HCV Non−infec Positive Yes No NA HB V po sit ive HB V ne ga tiv e HC V po sit ive HC V ne ga tiv e W es te rn ANNEX B 199 Supplementary Figure 5. Somatic mutation in HCC-driving signaling pathways. Known mutated genes in HCC grouped by signaling pathway. Molecular interactions between them are represented. The percentage of mutations in the Mongolian (left box; n=151) and Western (right box; n=112) cohorts is indicated for each gene. * p < 0.05 in Mongolian vs Western HCC. ANNEX B 200 Supplementary Figure 6. Substitution profile of Mongolian and Western HCCs. a-b Trinucleotide substitution frequency in Mongolian (a) and Western (b) HCC. c-d Differences in trinucleotide substitution frequency between the in-house Mongolian and Western cohorts (c) and Mongolian NCI and Western cohorts (d). Bars indicate the median values for each substitution group. Significant substitutions differences in both in-house and NCI Mongolian cohorts are highlighted in brighter colors. * p < 0.05 (Kruskal-Wallis test). e-f Signature fitting results in Mongolian (e, n = 151) and Western HCC (f, n = 112) using HCC-specific COSMIC signatures and SBS Mongolia. Middle panel indicates the proportion of SNVs assigned to each signature per sample (relative weight). Upper panel indicates the TMB (Mutations/Mb) for each sample, and lower panels represent clinical variables. Mongolian HCC Western HCC a b c Substitution differences in Mongolian vs. Western HCC e f Substitution differences in Mongolian NCI vs. Western HCC d Mongolian HCC Etiology 0 5 10 15 TM B 0 0.5 1 Re lat ive w eig ht Fibrosis F3-4 Age >= 60 Gender Etiology HBV HBV/HDV HCV Non−infec HBV/HCV/HDV Gender Female Male Positive Yes No NA Signatures SBS1 SBS4 SBS5 SBS6 SBS12 SBS16 SBS18 SBS22 SBS26 SBS29 SBS40 SBSM Western HCC Etiology 0 5 10 15 TM B 0 0.5 1 Re lat ive w eig ht Fibrosis F3-4 Age >= 60 Gender Origin Etiology HBV HBV/HDV HCV Non−infec Gender Female Male Origin Europe USA Positive Yes No NA Signatures SBS1 SBS4 SBS5 SBS6 SBS12 SBS16 SBS18 SBS22 SBS26 SBS29 SBS40 SBSM ANNEX B 201 Supplementary Figure 7. Characterization of Mongolian samples presenting the DMS signature. a Clinico-pathological and molecular features of Mongolian HCC samples positive and negative for the DMS signature. P values refer to FDR-adjusted Fisher tests (categorical) and Kruskal-Wallis test (continuous). b Kaplan-Meier estimates of overall survival in Mongolian HCC patients according to the presence of the DMS signature. Mongolian HCC (n=151) DMS signature Etiology HBV Status HCV Status Gender Age >= 60 Age Fibrosis stage Microvascular invasion Tumor grade 3−4 TMB >4 TERT TP53 CTNNB1 NFE2L2 APOB TSC2 NOTCH3 SBSMongolia T>G subs MGL clusters Immune class Broad CNV CNV gains CNV losses DMS signature Absent Present Etiology HBV HBV/HCV HBV/HCV/HDV HBV/HDV HCV Non−infec Gender Female Male Age 40 50 60 70 80 MGL clusters MGL1 MGL2 MGL3 Immune class Immune Rest Positive Yes No NA CNV Min Max 0.027 0.009 0.016 0.825 0.059 0.01 1 1 0.736 0.825 1 0.045 0.825 0.825 0.429 0.429 0.570 <0.0001 <0.0001 0.570 0.476 0.782 0.846 0.782 p value a b ANNEX B 202 Supplementary Figure 8. Single-base substitution signature analysis in adjacent non-tumor samples. a Adjacent non-tumor samples selected for SBS analysis in the Mongolian (n=64) and Western (n=14) cohorts (exome-region SNV count >= 10 at allelic frequency cutoff of 0.05). b Trinucleotide mutational profiles for pooled Mongolian (total variants = 1283) and Western (total variants =215) adjacent non- tumor samples. c Absolute signature fitting results for individual adjacent non-tumor samples. d Mutational signature fitting results of pooled non-tumor samples from the Western and Mongolian cohort. Values in red indicate signatures with weight ≥ 0.1. ANNEX B 203 Supplementary Figure 9. a Non-negative Matrix Factorization of the whole study cohort and b Mongolian cohort. Bottom panels indicate cophenetic coefficients of each clusterization. c Heatmap integrating the whole-cohort and Mongolian clusters and clinic-pathological chatacteristics of the samples. d Hierarchical clustering of the whole study cohort using Euclidean distance and Ward’s agglomerative procedure. d c NMFc-based clusters Country of origin NMFc-based MGL clusters Etiology Age Gender Origin of Western samples Ward hierarchical clusters Country of origin NMFc-based MGL clusters Etiology Age Gender Origin of Western samples a Whole study cohort, n=224 Whole study cohort, n=224 b 11111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111222222222222222222222222222222222222222222222222222222222222222222222222222222222222222 EEEEEEEEEEEEE U S U S U S U S U S U S U S U S U S U S U S U S U S U S U S U SEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE U S U S U S U S U S U S U S U S U S U S U SEEEEEEEEEEEEEEEEEE U S U S U S U S U S M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G U SE U S U SE U SEEEE U SEE U S U S U SEEEE U S U SEE # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N1111311111111111111111111111111111111111111111112222222222222222222222222223333333333333333333333333333333 # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N > = > = > = > = < 6 < 6 > = > = > = < 6 > = > = < 6 < 6 < 6 > = > = > = < 6 < 6 < 6 < 6 > = > = > = > = > = > = > = > = > = > = > = > = > = > = > = > = > = > = > = > = > = > = > = < 6 > = > = > = < 6 > = > = > = > = > = < 6 > = > = > = > = > = > = > = > = < 6 > = > = > = > = > = > = > = > = > = > = > = > = > = < 6 > = > = > = < 6 < 6 > = > = > = > = < 6 > = < 6 < 6 > = > = > = < 6 > = > = > = > = > = > = < 6 > = < 6 > = > = > = < 6 < 6 < 6 < 6 < 6 < 6 < 6 > = > = > = > = > = < 6 < 6 > = > = > = > = < 6 > = > = > = > = < 6 < 6 < 6 < 6 < 6 < 6 > = < 6 < 6 > = < 6 > = < 6 < 6 < 6 > = < 6 < 6 < 6 < 6 < 6 < 6 < 6 < 6 < 6 < 6 > = < 6 > = > = > = < 6 < 6 < 6 < 6 < 6 < 6 > = > = < 6 < 6 < 6 < 6 < 6 < 6 > = < 6 > = < 6 < 6 < 6 > = > = < 6 < 6 < 6 < 6 > = > = < 6 > = > = < 6 > = > = > = > = > = > = > = < 6 < 6 > = > = 1222222222222111222222 2 11111111122222222222222222222222222222 122222222222222222222 122211111222222222211111111122222222222222 2121112111112111111111111222222 111111112222222222222222 1122222 22212122112222212 22 WWWWWWWWWWWWWWWWWWWWWWWWWWW M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M GWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWW M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M GWWWWWWWW M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M G M GWWWWW M G M G M G M G # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N111111111111111111331111111111111111111123 # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N # N232333331222223333333333333333333333 # N # N # N # N # N # N # N # N212222111111122222222222 # N # N # N # N # N2223 < 6 > = < 6 > = # N # N > = > = > = > = > = > = > = > = > = > = > = < 6 > = > = > = > = > = > = > = > = > = < 6 > = > = > = > = > = > = < 6 > = < 6 < 6 < 6 # N < 6 # N > = < 6 # N < 6 > = > = < 6 < 6 > = > = < 6 > = > = > = < 6 < 6 > = > = > = > = > = < 6 < 6 > = < 6 < 6 > = < 6 < 6 < 6 < 6 < 6 # N # N > = # N < 6 < 6 < 6 > = > = > = < 6 > = > = > = # N > = > = > = # N # N > = # N > = > = > = > = < 6 > = > = > = > = < 6 > = > = > = > = > = > = > = > = > = # N # N # N > = # N > = > = > = > = > = > = > = > = > = < 6 < 6 < 6 > = < 6 < 6 > = > = > = > = # N > = > = # N < 6 > = > = > = > = > = # N < 6 < 6 < 6 > = > = < 6 < 6 < 6 < 6 < 6 # N < 6 < 6 < 6 < 6 < 6 < 6 < 6 < 6 < 6 < 6 < 6 < 6 > = > = > = > = > = > = # N # N # N # N > = < 6 > = < 6 > = > = > = > = > = # N < 6 < 6 < 6 # N > = # N > = < 6 # N > = > = # N > = < 6 > = < 6 # N # N # N < 6 < 6 < 6 > = > = > = < 6 > = < 6 < 6 < 6 < 6 22220022211222212222222222222122222222202221122211121222221121222111221212001022222221220222002021122222222221212110002022222222222122221120220212221112122222121212222122222211211001122222221111112111122201111111122212221122 Mongolian MGL1 HBV HCV <60 2 Male Europe E Western MGL2 HBV/HDV Non-infected > 60 1 Female USA MGL3 HBV/HCV/HDV Cohort OriginViral etiologyMGL clusters Age Gender 1 NMF cluster 1 2 NMF cluster 2 1 Hierarchical cluster 1 3 Hierarchical cluster 2 4 Hierarchical cluster 3 2 Hierarchical cluster 4 5 Hierarchical cluster 5 6 Hierarchical cluster 6 NMF clusters Hierarchical clustering 1 NMF cluster 1 2 NMF cluster 2 1 Hierarchical cluster 1 3 Hierarchical cluster 2 4 Hierarchical cluster 3 2 Hierarchical cluster 4 5 Hierarchical cluster 5 6 Hierarchical cluster 6 NMF clusters Hierarchical clustering ANNEX B 204 Supplementary Figure 10. Transcriptomic profile of Mongolian and Western samples. a Nearest Template Prediction (NTP) and single sample gene set enrichment analysis (ssGSEA) of Mongolian (n = 106) and Western (n = 118) HCC samples delineating the main molecular differences between cohorts. b Single sample gene set enrichment analysis of Mongolian and Western HCC samples indicating distinct immune cell populations and the HCC immune class assessed by Nearest Template Prediction analysis. Mongolian Western Chiang 5 class Hoshida 3 class HCC Immune class GO glycosyl biosynthetic process GO pyrimidine metabolic process KEGG bile acid biosynthesis GO bile acid biosynthetic process GO polyamine metabolic process GO polyamine metabolic process GO positive regulation by host of viral process GO positive regulation of viral process GO modulation by virus of cellular process GO positive regulation of viral genome replication GO positive regulation of viral life cycle Hallmark inflammatory response KEGG antigen processing and presentation GO antigen processing and presentation KEGG T cell receptor signaling KEGG T cell receptor complex GO growth factor receptor binding GO growth factor binding GO regulation of cellular response to growth factor GO response to growth factor GO response to hepatocyte growth factor Gene expression Low High Country of origin Immune profiles 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 Immune subtypes Etiology MMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMM B.cells CD8.T.cells Cytotox Treg.cells T.helper.cells Th1.cells Th2.cells TFH.cells Tem.cells Tcm.cells iDC Macrophages NK.cells Eosinophils Neutrophils Mast.cells IFN_signature MDSC a b Mongolian Western Immune class Immune class Exhausted Active Exhaus Active 42% 29% HCC Immune class Im une subtypes 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 MEEEEEEEEEUUUEEEEEEEEEEEEUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEUUUUUUUUEEEEEEEEEEEEEEEEEEEEEEEE B cells CD8 T cells totoxic T cells Treg cells T helper cells Th1 cells Th2 cells TFH cells Tem cells Tcm cells iDC cells Macrophages NK cells Eo inophils Neutrophils Mast cells IFN signature MDSC Origin Mongolian Europe MGL1 CTNNB1 Poly 7 S1 Immune E Western USA MGL2 Proliferation 5 Unan. S2 Rest MGL3 IFN S3 Cohort Origin MGL clusters Chiang 5 classes Hoshida classes Immune class Exh. Active ANNEX B 205 Supplementary Figure 11. Survival of Mongolian HCC patients according to the presence of MGL clusters Kaplan-Meier estimates of overall survival in the Mongolian cohort. + + ++ + ++++ ++++++ + + ++ ++ + +++ +++ + +++ + + ++ + + +++ + p = 0.81 0.00 0.25 0.50 0.75 1.00 0 10 20 30 40 50 Follow−up (months) Cu m su rv iva l Survival (Mongolia cohort) 46 38 36 32 23 0 26 23 20 15 9 0 29 27 24 20 8 0MGL3 MGL2 MGL1 0 10 20 30 40 50 Follow−up (months) M G L clu st er s Number at risk MGL clusters + + + MGL1 MGL2 MGL3 ANNEX B 206 Supplementary Figure 12. Molecular classification of Mongolian HCC in the NCI cohort. a Consensus- clustered classification of Mongolian HCC samples (n = 70) using Non-negative Matrix Factorization. Clinico-pathological characteristics, Nearest Template Prediction (NTP) and single-sample gene set enrichment analyses (ssGSEA) are shown. b Characterization of inflammatory profile in the MGL clusters assessed by ESTIMATE analysis and ssGSEA capturing distinct immune populations. ANNEX B 207 Supplementary Figure 13. Comparison between MGL and MO classifications of Mongolian HCC. a Subclass mapping analysis comparing the MGL clusters in the Mongolian cohort (n = 106) and the MO classification in the Mongolian NCI cohort (n = 70). FDR-adjusted p-values are indicated. b HCC samples from the in-house Mongolian cohort classified according to the MGL clusters by non-negative matrix factorization (NMFc) and to the MO classification by nearest template prediction (NTP). The number of samples in each category is indicated. a b ANNEX B 208 SUPPLEMENTARY TABLES Supplementary Table 1. Baseline characteristics of the Western cohort by sample origin. Western Europe (n=117) Western USA (n=70)* P value Age (years) 67 (40-83) 65 (29-91) ns < 60 years (n, %) 19 (16) 13 (24) ns Gender (male, %) 92 (79) 45 (82) ns Etiology: 0.001 HBV+ (n, %) 23 (19) 18 (26) HBV/HDV+ (n, %) 1 (1) 2 (3) HCV+ (n, %) 56 (48) 13 (19) Non-infected (n, %) 37 (32) 37 (53) Bilirubin (mg/dL) 1 (0.4-3.2) 0.7 (0.3-3.8) ns Albumin (g/L) 40 (25-54) 42 (22-49) ns Platelets (109/L) 161 (29-493) 154 (27-460) ns < 150x109/L (n, %) 55 (47) 26 (47) ns AFP > 400 IU/mL (n, %) 18 (16) 8 (17) ns Tumor size (cm) 4.2 (1.5-20) 4.4 (1-19) ns > 5 cm (n, %) 45 (41) 22 (42) ns BCLC stage (0-A, %) 87 (80) 35 (66) ns Multinodular disease (n, %) 29 (26) 13 (25) ns Advanced liver fibrosis (F3-4, %) 72 (88) 34 (64) 0.002 Cirrhosis (F4, %) 57 (70) 24 (45) 0.007 Microvascular invasion (yes, %) 44 (38) 36 (66) 0.001 Tumor grade (G3-4, %) 24 (27) 17 (31) ns *Baseline characteristics, except etiology, are missing for 15 (21%) patients in the Western USA subcohort ANNEX B 209 Supplementary Table 2. Viral genotypes in the Mongolian and Western cohorts. Mongolian Cohort (N=106) Western Cohort (N=44) p value HBV genotype Genotype A (n, %) 0 (0) 2 (4.5) <0.001 Genotype B (n, %) 0 (0) 2 (4.5) Genotype C (n, %) 0 (0) 12 (27.3) Genotype D (n, %)‡ 95 (89.6) 19 (43.2) Non-genotypable* (n, %) 11 (10.4) 9 (20.5) HDV genotype Genotype 1 (n, %) 85 (95.5) 1 (33.3) <0.001 Genotype 2 (n, %) 0 (0) 1 (33.3) Non-genotypable* (n, %) 4 (4.5) 1 (33.4) HBV mutations BCP A1762T (yes, %)† 10 (13.7) 21 (60) <0.001 BCP G1764A (yes, %)† 15 (20.5) 23 (65.7) <0.001 Precore G1896A (yes, %)† 21 (28.8) 16 (45.7) ns HBV, hepatitis B virus; HDV, hepatitis delta virus; BCP, basal pre-core ‡2 patients showed recombinant forms of the C and D genotypes * non-genotypable due to technical failure †HBV mutations were evaluated in 73 (69%) patients in the Mongolian cohort and 35 (80%) patients in the Western cohort ANNEX B 210 Supplementary Table 3. Baseline characteristics of HBV-infected Mongolian patients according to the 25th quartile of HBV-DNA load. Low HBV-DNA (<4 log copies/μg total DNA) (n=30) High HBV-DNA (≥4 log copies/μg total DNA) (n=75) p value Age (years) 57.7 (18-71.3) 56,2 (41.1- 75.8) ns Gender (male, %)‡ 15 (51.7) 37 (52.1) ns Etiology ns HBV (n, %) 8 (26.7) 7 (9.3) HBV/HDV (n, %) 20 (66.7) 56 (74.7) HBV/HCV/HDV (n, %) 1 (3.3) 11 (14.7) HBV/HCV (n, %) 1 (3.3) 1 (1.3) HDV+ (n, %) 21 (70) 67 (89.3) 0.021 Region ns Western (n, %) 3 (11.5) 17 (25.8) Central (n, %) 12 (46.2) 25 (37.9) Eastern (n, %) 4 (15.4) 4 (6.1) Ulaanbaatar (n, %) 7 (26.9) 20 (30.3) Liver fibrosis (F3-4, %)* 6 (20.7) 34 (47.2) 0.015 Tumor size (cm) 6 (3-14.9) 6.5 (1.6-20) ns Multinodular (yes, %) 0 (0) 12 (17.4) 0.032 BCLC stage (0-A, %) 24 (96) 50 (72.5) 0.02 BCP A1762T (yes, %)† 2 (11.1) 8 (14,5) ns BCP G1764A (yes, %)† 2 (11.1) 13 (23.6) ns Precore G1896A (yes, %)† 7 (38.9) 14 (25,5) ns HBV, hepatitis B virus; HCV, hepatitis C virus; HDV, hepatitis delta virus; BCLC, Barcelona Clinic Liver Cancer; BCP, basal pre-core ‡Gender information was missing in 9 individuals in the Mongolia cohort *Fibrosis stage was evaluated in 168 (88%) in the Mongolia cohort †HBV mutations were evaluated in 73 (69%) patients in the Mongolian cohort (18 HBV-DNA low and 55 HBV-DNA high) ANNEX B 211 Supplementary Table 4. Baseline characteristics of HDV-infected Mongolian patients according to the 25th quartile of HDV-RNA load. Low HDV-RNA (<2.4 log IU/ng total RNA)(n=23) High HDV-RNA (≥2.4 log IU/ng total RNA) (n=66) p value Age (years) 57 (41.1-72.6) 56 (44.2-75.6) ns Gender (male, %) 8 (36.4) 33 (53.2) ns Etiology ns HBV/HDV (n, %) 17 (73.9) 60 (90.9) HBV/HCV/HDV (n, %) 6 (26.1) 6 (9.1) Region ns Western (n, %) 7 (35) 7 (12.3) Central (n, %) 5 (25) 28 (49.1) Eastern (n, %) 3 (15) 5 (8.8) Ulaanbaatar (n, %) 5 (25) 17 (29.8) Liver fibrosis (F3-4, %)* 8 (34.8) 30 (45.5) ns Multinodular (yes, %) 1 (4.8) 12 (20.3) ns BCLC stage (0-A, %) 18 (85.7) 43 (72.9) ns AFP > 400 IU/mL (n, %) 7 (35) 9 (19.6) ns ALT (IU/L) 49.5 (14-426) 72.3 (10-452) 0.013 BCP A1762T (yes, %) 2 (14.3) 6 (13.3) ns BCP G1764A (yes, %) 3 (21.4) 8 (17.8) ns Precore G1896A (yes, %) 4 (28.6) 8 (17.8) ns HBV, hepatitis B virus; HCV, hepatitis C virus; HDV, hepatitis delta virus; BCLC, Barcelona Clinic Liver Cancer; AFP, alfa-fetoprotein; ALT, alanine aminotransferase;BCP, basal pre-core †‡Gender information was missing in 9 individuals in the Mongolia cohort *Fibrosis stage was evaluated in 168 (88%) in the Mongolia cohort †HBV mutations were evaluated in 68 (64%) patients in the Mongolia cohort ANNEX B 212 Supplementary Table 5. Baseline characteristics of HBV-infected Western patients classified according to the median HBV-DNA load. Low HBV-DNA (<5 log copies/μg total DNA) (n=20) High HBV-DNA (≥5 log copies/μg total DNA) (n=21) p value Age (years) 63 (29-87) 62 (41-78) ns Gender (male, %)‡ 12 (70.5) 17 (100%) 0.016 HDV+ (n, %) 1 (5) 2 (10) ns Liver fibrosis (F3-4, %)* 5 (41.7) 13 (76.5) ns Tumor size (cm) 5 (1.8-16) 5.5 (2-18) ns Multinodular (yes, %) 4 (23.5) 4 (21.1) ns BCLC stage (0-A, %) 11 (64.7) 16 (84.2) ns BCP A1762T (yes, %)† 8 (50) 13 (72.2) ns BCP G1764A (yes, %)† 9 (56.3) 14 (77.8) ns Precore G1896A (yes, %)† 7 (43.8) 9 (50) ns HBV, hepatitis B virus; HCV, hepatitis C virus; HDV, hepatitis delta virus; BCLC, Barcelona Clinic Liver Cancer; BCP, basal pre-core ‡Gender information was missing in 15 individuals in the Western cohort *Fibrosis stage was evaluated in 135 (72%) in the Western cohort †HBV mutations were evaluated in 35 (80%) patients in the Western cohort ANNEX B 213 Supplementary Table 6. Focal copy number alterations. P values correspond to Fisher test comparing Mongolian and Western cohorts. There was no difference in overall CNV burden between cohorts. Chromosome arm Mongolia (n=151) Western (n=112) Total (n=263) p value Mongolia (n=151) Western (n=112) Total (n=263) p value 8p 18 (12%) 11 (10%) 29 (11%) ns 59 (39%) 66 (59%) 125 (48%) 0.002 9q 11 (7%) 2 (2%) 13 (5%) 0.047 14 (9%) 24 (21%) 38 (14%) 0.007 1q 72 (48%) 55 (49%) 127 (48%) ns 2 (1%) 6 (5%) 8 (3%) 0.01 1p 29 (19%) 9 (8%) 38 (14%) 0.013 8 (5%) 17 (15%) 25 (10%) ns 2p 17 (11%) 10 (9%) 27 (10%) ns 7 (5%) 8 (7%) 15 (6%) ns 2q 15 (10%) 8 (7%) 23 (9%) ns 9 (6%) 8 (7%) 17 (6%) ns 3p 8 (5%) 5 (4%) 13 (5%) ns 7 (5%) 8 (7%) 15 (6%) ns 3q 13 (9%) 5 (4%) 18 (7%) ns 3 (2%) 8 (7%) 11 (4%) ns 4p 6 (4%) 5 (4%) 11 (4%) ns 26 (17%) 20 (18%) 46 (17%) ns 4q 2 (1%) 3 (3%) 5 (2%) ns 37 (25%) 29 (26%) 66 (25%) ns 5p 35 (23%) 31 (28%) 66 (25%) ns 5 (3%) 5 (4%) 10 (4%) ns 5q 24 (16%) 27 (24%) 51 (19%) ns 12 (8%) 5 (4%) 17 (6%) ns 6p 40 (26%) 28 (25%) 68 (26%) ns 4 (3%) 4 (4%) 8 (3%) ns 6q 19 (13%) 11 (10%) 30 (11%) ns 22 (15%) 17 (15%) 39 (15%) ns 7p 53 (35%) 30 (27%) 83 (32%) ns 1 (1%) 3 (3%) 4 (2%) ns 7q 55 (36%) 30 (27%) 85 (32%) ns 1 (1%) 3 (3%) 4 (2%) ns 8q 66 (44%) 61 (54%) 127(48%) ns 8 (5%) 5 (4%) 13 (5%) ns 9p 14(9%) 4 (4%) 18 (7%) ns 23 (15%) 25 (22%) 48 (18%) ns 10p 12 (8%) 10 (9%) 22 (8%) ns 11 (7%) 10 (9%) 21 (8%) ns 10q 6 (4%) 4 (4%) 10 (4%) ns 26 (17%) 19 (17%) 45 (17%) ns 11p 7 (5%) 4 (4%) 11 (4%) ns 11 (7%) 14 25 (10%) ns 11q 7 (5%) 4 (4%) 11 (4%) ns 12 (8%) 15 27 (10%) ns 12p 14 (9%) 9 (8%) 23 (9%) ns 13 (9%) 14 27 (10%) ns 12q 14 (9%) 9 (8%) 23 (9%) ns 9 (6%) 11 20 (8%) ns 13q 2 (1%) 5 (4%) 7 (3%) ns 26(17%) 26 52 (20%) ns 14q 7 (5%) 6 (5%) 13 (5%) ns 16 (11%) 13 29 (11%) ns 15q 6 (4%) 2 (2%) 8 (3%) ns 18 (12%) 10 28 (11%) ns 16p 5 3%) 6 (5%) 11 (4%) ns 35 (23%) 28 63 (24%) ns 16q 2 (1%) 6 (5%) 8 (3%) ns 47 (31%) 41 88 (33%) ns 17p 9 (6%) 6 (5%) 15 (6%) ns 36 (24%) 23 59 (22%) ns 17q 21(14%) 18 (16%) 39 (15%) ns 7 (5%) 9 16 (6%) ns 18p 11 (7%) 3 (3%) 14 (5%) ns 20 (13%) 13 33 (13%) ns 18q 8 (5%) 2 (2%) 10 (4%) ns 23 (15%) 16 39 (15%) ns 19p 19 (13%) 8 (7%) 27 (10%) ns 18 (12%) 19 37 (14%) ns 19q 20 (13%) 12 (11%) 32 (12%) ns 13 (9%) 16 29 (11%) ns 20p 33 (22%) 22 (20%) 55 (21%) ns 5 (3%) 6 11 (4%) ns 20q 37 (25%) 23 (21%) 60 (23%) ns 4 (3%) 2 6 (2%) ns 21q 5 (3%) 10 (9%) 15 (6%) ns 39 (26%) 24 63 (24%) ns 22q 12 (8%) 8 (7%) 20 (8%) ns 23 (15%) 21 44(17%) ns ANNEX B 214 Supplementary Table 7. Protein-coding mutations and tumor mutational burden. (TMB) in the in-house and external cohorts. TMB is shown as mutations/30 MB (Alexandrov, Nature 2013) and mutations/50 MB (Schulze, Nat Gen 2015) for comparison with previously published data. In-house cohorts External cohorts Mongolian Western Mongolian NCI TCGA European (Schulze) Korean (Ahn) Mutations 121 70 111 76 61 63 TMB (Mutations/30 Mb) 4.0 2.3 3.7 2.5 2.0 2.1 TMB (Mutations/50 Mb) 2.4 1.4 2.2 1.5 1.2 1.3 ANNEX B 215 Supplementary Table 8. Mutations in DNA damage repair (DDR) genes in the Mongolian and Western cohorts. Western cohort Mongolian cohort Gene Patients harboring mutations (n) Patients harboring mutations (%) Patients harboring mutations (n) Patients harboring mutations (%) TP53 36 32.14 70 46.36 ATM 7 6.25 13 8.61 BRCA2 4 3.57 3 1.99 ATR 3 2.68 4 2.65 HERC2 3 2.68 3 1.99 POLE 3 2.68 2 1.32 ATRX 2 1.79 9 5.96 POLD1 2 1.79 4 2.65 REV3L 2 1.79 4 2.65 TP53BP1 2 1.79 4 2.65 PTEN 2 1.79 3 1.99 CUL3 2 1.79 2 1.32 HELQ 2 1.79 1 0.66 PER1 2 1.79 1 0.66 SMARCA4 1 0.89 7 4.64 SHPRH 1 0.89 6 3.97 HFM1 1 0.89 5 3.31 RIF1 1 0.89 5 3.31 FANCA 1 0.89 3 1.99 POLA1 1 0.89 3 1.99 SLX4 1 0.89 3 1.99 SMARCAD1 1 0.89 2 1.32 MSH6 1 0.89 1 0.66 PARP4 1 0.89 1 0.66 SMC5 1 0.89 1 0.66 SMC6 1 0.89 1 0.66 WRN 1 0.89 1 0.66 FANCD2 0 0.00 6 3.97 ASCC3 0 0.00 5 3.31 FANCM 0 0.00 5 3.31 POLQ 0 0.00 5 3.31 BLM 0 0.00 4 2.65 MDC1 0 0.00 3 1.99 PALB2 0 0.00 2 1.32 RAD50 0 0.00 2 1.32 DDB1 0 0.00 1 0.66 MLH3 0 0.00 1 0.66 RFC1 0 0.00 1 0.66 TOPB1 0 0.00 0 0.00 LIGA4 0 0.00 0 0.00 ANNEX B 216 Supplementary Table 9. 100 genes with statistical differences between Mongolian and Western cohorts. Gene % mutations Mongolian % mutations Western Odds ratio Mongolian vs Western P value Mongolian vs Western % mutations Europe % mutations USA P value Europe vs USA TP53 46.4% 32.1% 0.548 0.022 26.1% 41.9% 0.098 TTN 36.4% 17.0% 0.357 0.001 20.3% 11.6% 0.305 RYR2 15.9% 7.1% 0.407 0.036 8.7% 4.7% 0.708 APOB 15.2% 4.5% 0.26 0.005 1.4% 9.3% 0.071 HMCN1 15.2% 6.3% 0.371 0.03 1.4% 14.0% 0.013 SYNE1 13.2% 5.4% 0.371 0.038 4.3% 7.0% 0.674 LAMA1 11.9% 2.7% 0.203 0.006 2.9% 2.3% 1 FLG 10.6% 0.9% 0.076 0.001 1.4% 0.0% 1 ABCA13 10.6% 2.7% 0.232 0.015 4.3% 0.0% 0.284 KMT2A 10.6% 2.7% 0.232 0.015 2.9% 2.3% 1 DNAH7 10.6% 3.6% 0.313 0.036 5.8% 0.0% 0.296 NOTCH3 6.6% 0.9% 7.828 0.027 1.4% 0.0% 1 KRT7 9.3% 0.0% 0 0 0.0% 0.0% NA DNAH8 9.3% 0.9% 0.088 0.003 0.0% 2.3% 0.384 TSC2 9.3% 0.9% 0.088 0.003 0.0% 2.3% 0.384 AHNAK2 9.3% 0.9% 0.088 0.003 0.0% 2.3% 0.384 PTPN13 9.3% 1.8% 0.178 0.016 2.9% 0.0% 0.523 DNAH9 9.3% 2.7% 0.269 0.041 4.3% 0.0% 0.284 COL6A3 8.6% 1.8% 0.193 0.028 1.4% 2.3% 1 LRBA 7.9% 0.9% 0.104 0.009 0.0% 2.3% 0.384 ANK3 7.9% 1.8% 0.211 0.029 1.4% 2.3% 1 KIAA1109 7.9% 1.8% 0.211 0.029 1.4% 2.3% 1 CMYA5 7.3% 0.0% 0 0.003 0.0% 0.0% NA RNF213 7.3% 0.9% 0.115 0.015 0.0% 2.3% 0.384 PLXNA4 7.3% 0.9% 0.115 0.015 0.0% 2.3% 0.384 ALMS1 7.3% 1.8% 0.231 0.047 1.4% 2.3% 1 MYO15A 7.3% 1.8% 0.231 0.047 1.4% 2.3% 1 SLC7A8 6.6% 0.0% 0 0.006 0.0% 0.0% NA MYH13 6.6% 0.9% 0.127 0.027 1.4% 0.0% 1 MGAM 6.6% 0.9% 0.127 0.027 0.0% 2.3% 0.384 GPR112 6.6% 0.9% 0.127 0.027 1.4% 0.0% 1 SDK2 6.0% 0.0% 0 0.011 0.0% 0.0% NA NAV3 6.0% 0.9% 0.142 0.047 1.4% 0.0% 1 MKI67 6.0% 0.9% 0.142 0.047 1.4% 0.0% 1 NFE2L2 6.0% 0.9% 0.142 0.047 1.4% 0.0% 1 OTOG 6.0% 0.9% 0.142 0.047 1.4% 0.0% 1 LAMC3 6.0% 0.9% 0.142 0.047 1.4% 0.0% 1 PXDNL 6.0% 0.9% 0.142 0.047 0.0% 2.3% 0.384 GTF3C1 5.3% 0.0% 0 0.023 0.0% 0.0% NA NOS1 5.3% 0.0% 0 0.023 0.0% 0.0% NA KRT6A 5.3% 0.0% 0 0.023 0.0% 0.0% NA JMY 5.3% 0.0% 0 0.023 0.0% 0.0% NA PTPRS 5.3% 0.0% 0 0.023 0.0% 0.0% NA HCN1 5.3% 0.0% 0 0.023 0.0% 0.0% NA MAP2 5.3% 0.0% 0 0.023 0.0% 0.0% NA DLEC1 5.3% 0.0% 0 0.023 0.0% 0.0% NA CHD6 5.3% 0.0% 0 0.023 0.0% 0.0% NA PDZD2 5.3% 0.0% 0 0.023 0.0% 0.0% NA PLXNA3 5.3% 0.0% 0 0.023 0.0% 0.0% NA ANNEX B 217 TELO2 5.3% 0.0% 0 0.023 0.0% 0.0% NA TMEM132D 4.6% 0.0% 0 0.022 0.0% 0.0% NA PCDHA6 4.6% 0.0% 0 0.022 0.0% 0.0% NA DMXL1 4.6% 0.0% 0 0.022 0.0% 0.0% NA PRRC2A 4.6% 0.0% 0 0.022 0.0% 0.0% NA ZFP36L1 4.6% 0.0% 0 0.022 0.0% 0.0% NA SLIT1 4.6% 0.0% 0 0.022 0.0% 0.0% NA MAP1B 4.6% 0.0% 0 0.022 0.0% 0.0% NA SYCP2 4.6% 0.0% 0 0.022 0.0% 0.0% NA CCDC30 4.6% 0.0% 0 0.022 0.0% 0.0% NA IPO9 4.6% 0.0% 0 0.022 0.0% 0.0% NA ABCA1 4.6% 0.0% 0 0.022 0.0% 0.0% NA TNN 4.6% 0.0% 0 0.022 0.0% 0.0% NA LMTK3 4.0% 0.0% 0 0.04 0.0% 0.0% NA PDGFRA 4.0% 0.0% 0 0.04 0.0% 0.0% NA CD1C 4.0% 0.0% 0 0.04 0.0% 0.0% NA FAM184B 4.0% 0.0% 0 0.04 0.0% 0.0% NA SLC23A1 4.0% 0.0% 0 0.04 0.0% 0.0% NA TTLL5 4.0% 0.0% 0 0.04 0.0% 0.0% NA LILRA2 4.0% 0.0% 0 0.04 0.0% 0.0% NA PCSK5 4.0% 0.0% 0 0.04 0.0% 0.0% NA ASTN1 4.0% 0.0% 0 0.04 0.0% 0.0% NA ZRSR2 4.0% 0.0% 0 0.04 0.0% 0.0% NA DLGAP3 4.0% 0.0% 0 0.04 0.0% 0.0% NA ABHD17A 4.0% 0.0% 0 0.04 0.0% 0.0% NA BAZ2B 4.0% 0.0% 0 0.04 0.0% 0.0% NA FAR2 4.0% 0.0% 0 0.04 0.0% 0.0% NA KCNT1 4.0% 0.0% 0 0.04 0.0% 0.0% NA GABRB2 4.0% 0.0% 0 0.04 0.0% 0.0% NA HEPACAM2 4.0% 0.0% 0 0.04 0.0% 0.0% NA SLC44A5 4.0% 0.0% 0 0.04 0.0% 0.0% NA ANKRD31 4.0% 0.0% 0 0.04 0.0% 0.0% NA PHACTR4 4.0% 0.0% 0 0.04 0.0% 0.0% NA FANCD2 4.0% 0.0% 0 0.04 0.0% 0.0% NA COL16A1 4.0% 0.0% 0 0.04 0.0% 0.0% NA CYP2A13 4.0% 0.0% 0 0.04 0.0% 0.0% NA SPEN 4.0% 0.0% 0 0.04 0.0% 0.0% NA PABPC5 4.0% 0.0% 0 0.04 0.0% 0.0% NA STK31 4.0% 0.0% 0 0.04 0.0% 0.0% NA CCDC146 4.0% 0.0% 0 0.04 0.0% 0.0% NA IGSF9B 0.7% 8.0% 13.107 0.002 0.0% 0.0% NA FLNB 0.7% 6.3% 10 0.012 0.0% 0.0% NA PTPN21 0.7% 5.4% 8.491 0.044 0.0% 0.0% NA FRG1B 0.0% 5.4% inf 0.006 0.0% 0.0% NA KAT6A 0.0% 4.5% inf 0.013 0.0% 0.0% NA TDO2 0.0% 3.6% inf 0.032 0.0% 0.0% NA TAF1A 0.0% 3.6% inf 0.032 0.0% 0.0% NA ASB14 0.0% 3.6% inf 0.032 0.0% 0.0% NA KIF20A 0.0% 3.6% inf 0.032 0.0% 0.0% NA SFSWAP 0.0% 3.6% inf 0.032 0.0% 0.0% NA IGSF3 0.0% 3.6% inf 0.032 0.0% 0.0% NA ANNEX B 218 Supplementary Table 10. Genes more frequently mutated in Mongolia versus other cohorts. Genes significantly mutated in 1 or more external non-Mongolian cohorts are shown. Western and asian non-Mongolian cohorts Mongolian NCI cohort Gene % Mongoli an in- house cohort Number of cohorts with Significa nt diff % Weste rn in- house cohort P VALUE Mongoli an vs Western % Korea n P VALUE Mongoli an vs Korean % Europe an P VALUE Mongoli an vs Europea n % TCG A P VALUE Mongoli an vs TCGA % Mongoli an NCI P VALUE in-house Mongoli an vs Mongoli an NCI TP53 46 4 32 0.022 31 0.00 22 0.000 28 0.000 30 0.020 LAMA1 12 4 3 0.006 6 0.03 3 0.001 4 0.003 8 0.496 KRT7 9 4 0 0.000 0 0.00 1 0.000 1 0.000 0 0.006 PTPN13 9 4 2 0.016 1 0.00 1 0.000 4 0.015 3 0.099 COL6A3 9 4 2 0.028 3 0.04 3 0.036 4 0.026 8 1.000 GPR112 7 4 1 0.027 0 0.00 1 0.007 0 0.000 0 0.033 ALMS1 7 4 2 0.047 3 0.04 2 0.006 3 0.026 3 0.233 MYO15A 7 3 2 0.047 3 0.04 2 0.006 1 0.001 4 0.556 PRRC2A 5 3 0 0.022 0 0.00 1 0.0497 0 0.001 0 0.100 DLEC1 5 4 0 0.023 1 0.02 1 0.016 1 0.025 4 1.000 JMY 5 4 0 0.023 0 0.00 0 0.003 0 0.000 0 0.057 KRT6A 5 4 0 0.023 1 0.03 0 0.000 1 0.008 0 0.057 TELO2 5 4 0 0.023 0 0.00 1 0.016 1 0.001 3 0.508 PLXNA3 5 4 0 0.023 0 0.00 1 0.016 1 0.025 3 0.508 ARID1A 17 3 10 0.107 3 0.00 10 0.043 7 0.001 4 0.002 KMT2C 9 3 3 0.062 3 0.01 0 0.000 3 0.008 6 0.591 CSMD3 15 3 13 0.718 8 0.04 6 0.008 8 0.034 13 0.674 AGRN 4 3 1 0.244 0 0.02 0 0.015 0 0.001 6 0.730 CUBN 13 3 7 0.217 6 0.04 3 0.000 5 0.005 10 0.496 AKAP9 8 3 4 0.193 3 0.02 2 0.023 1 0.000 0 0.011 PRDM5 5 3 3 0.524 0 0.01 1 0.031 1 0.018 1 0.441 PIK3R4 5 3 1 0.143 1 0.03 1 0.031 1 0.009 3 0.722 CELSR1 7 3 3 0.163 3 0.04 1 0.001 2 0.009 1 0.109 MUC6 7 3 5 0.797 2 0.02 1 0.002 2 0.012 1 0.181 ARAP3 5 3 2 0.197 0 0.00 0 0.003 2 0.047 3 0.508 BTAF1 5 3 1 0.083 1 0.02 1 0.016 2 0.033 6 1.000 GABRA1 5 3 1 0.083 0 0.00 0 0.003 2 0.033 0 0.057 KIAA1731 5 3 1 0.083 1 0.03 1 0.026 0 0.000 0 0.057 BAI1 5 3 1 0.143 0 0.01 1 0.031 0 0.000 0 0.100 BEND5 3 3 0 0.074 0 0.01 0 0.033 1 0.025 0 0.180 C10orf118 3 3 0 0.074 0 0.04 0 0.033 0 0.002 0 0.180 CCDC147 3 3 0 0.074 0 0.04 0 0.033 0 0.002 0 0.180 CCT8 3 3 0 0.074 0 0.04 0 0.033 0 0.002 1 0.667 CDC42EP4 3 3 0 0.074 0 0.04 0 0.008 0 0.002 1 0.667 FAM83E 3 3 1 0.244 0 0.04 0 0.033 0 0.010 1 0.667 FERD3L 3 3 1 0.244 0 0.04 0 0.033 0 0.010 0 0.180 GART 3 3 1 0.244 0 0.04 0 0.008 1 0.025 0 0.180 HOMER3 3 3 0 0.074 0 0.04 0 0.033 0 0.002 1 0.667 LMNA 3 3 1 0.244 0 0.04 0 0.008 1 0.025 1 0.667 LRRC36 3 3 0 0.074 0 0.01 0 0.033 0 0.010 3 1.000 MAP6 3 3 0 0.074 0 0.01 0 0.033 0 0.002 1 0.667 AADAT 3 3 1 0.398 0 0.02 0 0.021 0 0.028 1 1.000 ARFIP1 3 3 0 0.139 0 0.02 0 0.021 0 0.028 1 1.000 ATP6AP1 3 3 0 0.139 0 0.02 0 0.021 0 0.028 0 0.309 PRKAR1B 3 3 1 0.398 0 0.02 0 0.021 0 0.007 0 0.309 PNPLA2 3 3 0 0.139 0 0.02 0 0.021 0 0.007 0 0.309 OSMR 3 3 0 0.139 0 0.02 0 0.021 0 0.007 1 1.000 CSMD1 0 3 0 1.000 4 0.01 8 0.000 6 0.002 10 0.001 DMD 0 3 0 1.000 6 0.00 3 0.047 4 0.008 8 0.003 DNAH17 0 3 0 1.000 5 0.00 3 0.026 4 0.013 4 0.032 ANNEX B 219 DST 0 3 0 1.000 10 0.00 5 0.004 4 0.013 3 0.101 FRAS1 0 3 0 1.000 4 0.01 5 0.002 5 0.003 6 0.010 MUC16 0 3 0 1.000 23 0.00 12 0.000 16 0.000 30 0.000 MUC2 0 3 0 1.000 3 0.045 3 0.047 5 0.002 0 1.000 PCLO 0 3 0 1.000 15 0.000 9 0.000 11 0.000 15 0.000 NLRP8 5 3 3 0.363 1 0.03 1 0.016 1 0.001 0 0.057 PHACTR4 4 3 0 0.040 1 0.06 0 0.015 1 0.022 0 0.180 HMCN1 15 3 6 0.030 13 0.55 5 0.000 7 0.008 11 0.302 SYNE1 13 3 5 0.038 10 0.32 2 0.000 4 0.000 13 1.000 LRBA 8 3 1 0.009 5 0.29 1 0.002 1 0.000 4 0.397 KMT2A 11 3 3 0.015 5 0.07 2 0.001 3 0.001 6 0.316 CMYA5 7 3 0 0.003 3 0.08 1 0.003 2 0.009 1 0.109 RNF213 7 3 1 0.015 4 0.25 2 0.006 2 0.009 1 0.109 MGAM 7 3 1 0.027 6 0.83 2 0.030 2 0.018 4 0.558 LAMC3 6 3 1 0.047 2 0.09 2 0.038 2 0.017 0 0.061 CHD6 5 3 0 0.023 3 0.18 1 0.026 2 0.047 0 0.057 CCDC30 5 3 0 0.022 1 0.06 0 0.001 1 0.009 1 0.441 PCDHA6 5 3 0 0.022 1 0.06 0 0.006 1 0.047 0 0.100 ABHD17A 4 3 0 0.040 1 0.06 0 0.003 0 0.001 0 0.180 CD1C 4 3 0 0.040 1 0.06 0 0.015 0 0.003 0 0.180 CYP2A13 4 3 0 0.040 1 0.06 0 0.003 1 0.010 1 0.435 FAM184B 4 3 0 0.040 1 0.16 0 0.015 1 0.022 1 0.435 FANCD2 4 3 0 0.040 2 0.20 0 0.003 1 0.022 1 0.435 ZRSR2 4 3 0 0.040 1 0.16 0 0.015 0 0.003 1 0.435 CCDC146 4 3 0 0.040 0 0.00 0 0.003 1 0.071 4 1.000 NOTCH3 7 3 1 0.027 0 0.00 2 0.064 3 0.047 1 0.181 SLC23A1 4 3 0 0.040 0 0.00 1 0.059 0 0.003 0 0.180 OTOG 6 3 1 0.047 2 0.04 2 0.105 1 0.006 6 1.000 HCN1 5 3 0 0.023 0 0.00 2 0.067 2 0.033 0 0.057 MYH13 7 3 1 0.027 2 0.02 2 0.064 1 0.003 4 0.558 TSC2 9 3 1 0.003 3 0.01 5 0.088 3 0.008 7 0.798 FAR2 4 3 0 0.040 0 0.00 1 0.059 0 0.003 1 0.435 HEPACAM2 4 3 0 0.040 0 0.02 1 0.059 1 0.022 0 0.180 TTLL5 4 3 0 0.040 0 0.00 1 0.059 0 0.003 3 1.000 ANNEX B 220 Supplementary Table 11. Assignment of the four de novo extracted HCC signatures to all single and linear combinations of two COSMIC v3 signatures. Signature mapping highlighted in yellow indicated the assignment used in our analysis, selected based on cosine similarity (Cos sim). de novo signature 1 de novo signature 2 de novo signature 3 de novo signature 4 (SBSM) Reference signature Weight Cos sim Reference signature Weight Cos sim Reference signature Weight Cos sim Reference signature Weight Cos sim SBS22 SBS22=1 0.9735 SBS6 + SBS40 SBS6=0.12; SBS40=0.88 0.91406 SBS16 + SBS26 SBS16=0.45; SBS26=0.55 0.92379 SBS28 + SBS40 SBS28=0.09; SBS40=0.91 0.81766 SBS3 + SBS22 SBS3=0.12; SBS22=0.88 0.97496 SBS15 + SBS40 SBS15=0.09; SBS40=0.91 0.90401 SBS12 + SBS16 SBS12=0.55; SBS16=0.45 0.91218 SBS17b + SBS40 SBS17b=0.07; SBS40=0.93 0.81288 SBS22 + SBS25 SBS22=0.86; SBS25=0.14 0.97488 SBS30 + SBS40 SBS30=0.13; SBS40=0.87 0.90294 SBS16 + SBS54 SBS16=0.67; SBS54=0.33 0.89886 SBS9 + SBS40 SBS9=0.3; SBS40=0.7 0.80978 SBS8 + SBS22 SBS8=0.09; SBS22=0.91 0.97469 SBS23 + SBS40 SBS23=0.1; SBS40=0.9 0.9012 SBS16 + SBS46 SBS16=0.59; SBS46=0.41 0.88026 SBS40 + SBS55 SBS40=0.9; SBS55=0.1 0.8047 SBS4 + SBS22 SBS4=0.07; SBS22=0.93 0.97468 SBS19 + SBS40 SBS19=0.09; SBS40=0.91 0.8984 SBS16 + SBS33 SBS16=0.76; SBS33=0.24 0.87846 SBS40 + SBS43 SBS40=0.93; SBS43=0.07 0.78171 SBS22 + SBS40 SBS22=0.9; SBS40=0.1 0.97462 SBS40 + SBS42 SBS40=0.85; SBS42=0.15 0.89789 SBS16 + SBS21 SBS16=0.74; SBS21=0.26 0.86621 SBS40 + SBS60 SBS40=0.98; SBS60=0.02 0.77619 SBS22 + SBS46 SBS22=0.95; SBS46=0.05 0.97415 SBS40 + SBS84 SBS40=0.89; SBS84=0.11 0.89751 SBS5 + SBS16 SBS5=0.59; SBS16=0.41 0.86563 SBS40 + SBS54 SBS40=0.97; SBS54=0.03 0.77316 SBS18 + SBS22 SBS18=0.04; SBS22=0.96 0.97412 SBS32 + SBS40 SBS32=0.11; SBS40=0.89 0.89374 SBS5 + SBS26 SBS5=0.55; SBS26=0.45 0.85408 SBS40 + SBS41 SBS40=0.93; SBS41=0.07 0.77314 SBS22 + SBS45 SBS22=0.97; SBS45=0.03 0.97409 SBS1 + SBS40 SBS1=0.05; SBS40=0.95 0.89348 SBS16 + SBS25 SBS16=0.56; SBS25=0.44 0.85369 SBS37 + SBS40 SBS37=0.06; SBS40=0.94 0.77238 SBS22 + SBS39 SBS22=0.94; SBS39=0.06 0.97407 SBS40 + SBS44 SBS40=0.88; SBS44=0.12 0.89218 SBS16 + SBS37 SBS16=0.57; SBS37=0.43 0.85156 SBS40 + SBS51 SBS40=0.97; SBS51=0.03 0.77162 SBS22 + SBS36 SBS22=0.97; SBS36=0.03 0.97406 SBS24 + SBS40 SBS24=0.13; SBS40=0.87 0.89124 SBS3 + SBS16 SBS3=0.46; SBS16=0.54 0.84754 SBS40 + SBS57 SBS40=0.98; SBS57=0.02 0.77099 SBS22 + SBS38 SBS22=0.98; SBS38=0.02 0.97392 SBS11 + SBS40 SBS11=0.08; SBS40=0.92 0.89086 SBS5 + SBS12 SBS5=0.55; SBS12=0.45 0.84635 SBS1 + SBS40 SBS1=0; SBS40=1 0.7703 SBS17a + SBS22 SBS17a=0.02; SBS22=0.98 0.97391 SBS7b + SBS40 SBS7b=0.07; SBS40=0.93 0.88954 SBS16 + SBS44 SBS16=0.74; SBS44=0.26 0.84225 SBS2 + SBS40 SBS2=0; SBS40=1 0.7703 SBS22 + SBS35 SBS22=0.96; SBS35=0.04 0.97391 SBS7a + SBS40 SBS7a=0.05; SBS40=0.95 0.88584 SBS16 + SBS40 SBS16=0.62; SBS40=0.38 0.83552 SBS3 + SBS40 SBS3=0; SBS40=1 0.7703 SBS22 + SBS56 SBS22=0.98; SBS56=0.02 0.9739 SBS29 + SBS40 SBS29=0.1; SBS40=0.9 0.88549 SBS16 + SBS17a SBS16=0.89; SBS17a=0.11 0.83266 SBS4 + SBS40 SBS4=0; SBS40=1 0.7703 SBS5 + SBS22 SBS5=0.06; SBS22=0.94 0.97389 SBS5 + SBS40 SBS5=0.22; SBS40=0.78 0.88366 SBS26 + SBS40 SBS26=0.62; SBS40=0.38 0.83076 SBS5 + SBS40 SBS5=0; SBS40=1 0.7703 SBS22 + SBS54 SBS22=0.98; SBS54=0.02 0.97388 SBS31 + SBS40 SBS31=0.07; SBS40=0.93 0.88193 SBS9 + SBS16 SBS9=0.27; SBS16=0.73 0.83046 SBS6 + SBS40 SBS6=0; SBS40=1 0.7703 SBS22 + SBS29 SBS22=0.97; SBS29=0.03 0.97387 SBS18 + SBS40 SBS18=0.07; SBS40=0.93 0.88079 SBS8 + SBS26 SBS8=0.28; SBS26=0.72 0.8304 SBS7a + SBS40 SBS7a=0; SBS40=1 0.7703 SBS22 + SBS24 SBS22=0.97; SBS24=0.03 0.97383 SBS20 + SBS40 SBS20=0.05; SBS40=0.95 0.88025 SBS4 + SBS16 SBS4=0.23; SBS16=0.77 0.8302 SBS7b + SBS40 SBS7b=0; SBS40=1 0.7703 SBS22 + SBS53 SBS22=0.98; SBS53=0.02 0.97381 SBS2 + SBS40 SBS2=0.02; SBS40=0.98 0.88024 SBS26 + SBS58 SBS26=0.78; SBS58=0.22 0.82948 SBS7c + SBS40 SBS7c=0; SBS40=1 0.7703 SBS13 + SBS22 SBS13=0.01; SBS22=0.99 0.97381 SBS21 + SBS40 SBS21=0.04; SBS40=0.96 0.8797 SBS7d + SBS16 SBS7d=0.13; SBS16=0.87 0.82906 SBS7d + SBS40 SBS7d=0; SBS40=1 0.7703 SBS20 + SBS22 SBS20=0.02; SBS22=0.98 0.97378 SBS36 + SBS40 SBS36=0.04; SBS40=0.96 0.87903 SBS3 + SBS26 SBS3=0.38; SBS26=0.62 0.82839 SBS8 + SBS40 SBS8=0; SBS40=1 0.7703 SBS10a + SBS22 SBS10a=0.01; SBS22=0.99 0.97378 SBS4 + SBS40 SBS4=0.07; SBS40=0.93 0.87889 SBS8 + SBS16 SBS8=0.24; SBS16=0.76 0.82801 SBS10a + SBS40 SBS10a=0; SBS40=1 0.7703 SBS9 + SBS22 SBS9=0.04; SBS22=0.96 0.97375 SBS40 + SBS52 SBS40=0.98; SBS52=0.02 0.87874 SBS16 + SBS20 SBS16=0.86; SBS20=0.14 0.82791 SBS10b + SBS40 SBS10b=0; SBS40=1 0.7703 SBS22 + SBS57 SBS22=0.98; SBS57=0.02 0.97371 SBS7d + SBS40 SBS7d=0.03; SBS40=0.97 0.87816 SBS4 + SBS26 SBS4=0.24; SBS26=0.76 0.82751 SBS11 + SBS40 SBS11=0; SBS40=1 0.7703 SBS22 + SBS52 SBS22=0.99; SBS52=0.01 0.97369 SBS40 + SBS50 SBS40=0.96; SBS50=0.04 0.87811 SBS16 + SBS42 SBS16=0.82; SBS42=0.18 0.82731 SBS12 + SBS40 SBS12=0; SBS40=1 0.7703 SBS22 + SBS33 SBS22=0.99; SBS33=0.01 0.97369 SBS35 + SBS40 SBS35=0.05; SBS40=0.95 0.87799 SBS16 + SBS35 SBS16=0.81; SBS35=0.19 0.82729 SBS13 + SBS40 SBS13=0; SBS40=1 0.7703 SBS22 + SBS44 SBS22=0.98; SBS44=0.02 0.97368 SBS14 + SBS40 SBS14=0.03; SBS40=0.97 0.87768 SBS25 + SBS26 SBS25=0.33; SBS26=0.67 0.82666 SBS14 + SBS40 SBS14=0; SBS40=1 0.7703 SBS22 + SBS41 SBS22=0.97; SBS41=0.03 0.97368 SBS10b + SBS40 SBS10b=0.02 ; SBS40=0.98 0.87748 SBS16 + SBS18 SBS16=0.84; SBS18=0.16 0.82605 SBS15 + SBS40 SBS15=0; SBS40=1 0.7703 SBS22 + SBS55 SBS22=0.99; SBS55=0.01 0.97366 SBS40 + SBS46 SBS40=0.96; SBS46=0.04 0.87704 SBS14 + SBS16 SBS14=0.11; SBS16=0.89 0.82554 SBS16 + SBS40 SBS16=0; SBS40=1 0.7703 SBS22 + SBS49 SBS22=0.99; SBS49=0.01 0.97365 SBS40 + SBS45 SBS40=0.98; SBS45=0.02 0.87686 SBS16 + SBS24 SBS16=0.83; SBS24=0.17 0.82548 SBS17a + SBS40 SBS17a=0; SBS40=1 0.7703 SBS22 + SBS31 SBS22=0.98; SBS31=0.02 0.97365 SBS40 + SBS48 SBS40=0.99; SBS48=0.01 0.87654 SBS16 + SBS31 SBS16=0.85; SBS31=0.15 0.82525 SBS18 + SBS40 SBS18=0; SBS40=1 0.7703 SBS14 + SBS22 SBS14=0.01; SBS22=0.99 0.97364 SBS40 + SBS49 SBS40=0.99; SBS49=0.01 0.87642 SBS16 + SBS29 SBS16=0.85; SBS29=0.15 0.82461 SBS19 + SBS40 SBS19=0; SBS40=1 0.7703 SBS22 + SBS42 SBS22=0.98; SBS42=0.02 0.97363 SBS40 + SBS59 SBS40=0.99; SBS59=0.01 0.8761 SBS16 + SBS36 SBS16=0.88; SBS36=0.12 0.8245 SBS20 + SBS40 SBS20=0; SBS40=1 0.7703 ANNEX B 221 SBS22 + SBS26 SBS22=0.98; SBS26=0.02 0.97362 SBS40 + SBS56 SBS40=0.99; SBS56=0.01 0.87605 SBS16 + SBS43 SBS16=0.89; SBS43=0.11 0.82397 SBS21 + SBS40 SBS21=0; SBS40=1 0.7703 SBS7d + SBS22 SBS7d=0.01; SBS22=0.99 0.97362 SBS40 + SBS53 SBS40=0.99; SBS53=0.01 0.87588 SBS16 + SBS57 SBS16=0.87; SBS57=0.13 0.82355 SBS22 + SBS40 SBS22=0; SBS40=1 0.7703 SBS21 + SBS22 SBS21=0.01; SBS22=0.99 0.9736 SBS3 + SBS40 SBS3=0.05; SBS40=0.95 0.87583 SBS16 + SBS41 SBS16=0.84; SBS41=0.16 0.82332 SBS23 + SBS40 SBS23=0; SBS40=1 0.7703 SBS22 + SBS30 SBS22=0.99; SBS30=0.01 0.97359 SBS38 + SBS40 SBS38=0.01; SBS40=0.99 0.87583 SBS16 + SBS22 SBS16=0.89; SBS22=0.11 0.82315 SBS24 + SBS40 SBS24=0; SBS40=1 0.7703 SBS22 + SBS28 SBS22=0.99; SBS28=0.01 0.97359 SBS25 + SBS40 SBS25=0.01; SBS40=0.99 0.87559 SBS16 + SBS58 SBS16=0.87; SBS58=0.13 0.82201 SBS25 + SBS40 SBS25=0; SBS40=1 0.7703 SBS7b + SBS22 SBS7b=0.01; SBS22=0.99 0.97358 SBS33 + SBS40 SBS33=0.01; SBS40=0.99 0.87558 SBS7c + SBS16 SBS7c=0.09; SBS16=0.91 0.82177 SBS26 + SBS40 SBS26=0; SBS40=1 0.7703 SBS15 + SBS22 SBS15=0.01; SBS22=0.99 0.97358 SBS7c + SBS40 SBS7c=0; SBS40=1 0.87558 SBS16 + SBS39 SBS16=0.82; SBS39=0.18 0.82168 SBS27 + SBS40 SBS27=0; SBS40=1 0.7703 SBS22 + SBS37 SBS22=0.98; SBS37=0.02 0.97358 SBS8 + SBS40 SBS8=0; SBS40=1 0.87558 SBS16 + SBS45 SBS16=0.93; SBS45=0.07 0.82148 SBS29 + SBS40 SBS29=0; SBS40=1 0.7703 SBS22 + SBS50 SBS22=0.99; SBS50=0.01 0.97357 SBS9 + SBS40 SBS9=0; SBS40=1 0.87558 SBS16 + SBS50 SBS16=0.9; SBS50=0.1 0.82125 SBS30 + SBS40 SBS30=0; SBS40=1 0.7703 SBS22 + SBS23 SBS22=0.99; SBS23=0.01 0.97356 SBS10a + SBS40 SBS10a=0; SBS40=1 0.87558 SBS16 + SBS17b SBS16=0.95; SBS17b=0.05 0.82121 SBS31 + SBS40 SBS31=0; SBS40=1 0.7703 SBS12 + SBS22 SBS12=0.02; SBS22=0.98 0.97356 SBS12 + SBS40 SBS12=0; SBS40=1 0.87558 SBS12 + SBS58 SBS12=0.79; SBS58=0.21 0.82118 SBS32 + SBS40 SBS32=0; SBS40=1 0.7703 SBS22 + SBS32 SBS22=0.99; SBS32=0.01 0.97355 SBS13 + SBS40 SBS13=0; SBS40=1 0.87558 SBS26 + SBS35 SBS26=0.8; SBS35=0.2 0.82115 SBS33 + SBS40 SBS33=0; SBS40=1 0.7703 SBS6 + SBS22 SBS6=0.01; SBS22=0.99 0.97353 SBS16 + SBS40 SBS16=0; SBS40=1 0.87558 SBS16 + SBS56 SBS16=0.95; SBS56=0.05 0.82099 SBS34 + SBS40 SBS34=0; SBS40=1 0.7703 SBS22 + SBS84 SBS22=0.99; SBS84=0.01 0.97353 SBS17a + SBS40 SBS17a=0; SBS40=1 0.87558 SBS10a + SBS16 SBS10a=0.04; SBS16=0.96 0.82091 SBS35 + SBS40 SBS35=0; SBS40=1 0.7703 SBS19 + SBS22 SBS19=0.01; SBS22=0.99 0.97353 SBS17b + SBS40 SBS17b=0; SBS40=1 0.87558 SBS16 + SBS53 SBS16=0.93; SBS53=0.07 0.82086 SBS36 + SBS40 SBS36=0; SBS40=1 0.7703 SBS22 + SBS43 SBS22=0.99; SBS43=0.01 0.97353 SBS22 + SBS40 SBS22=0; SBS40=1 0.87558 SBS7b + SBS16 SBS7b=0.07; SBS16=0.93 0.82053 SBS38 + SBS40 SBS38=0; SBS40=1 0.7703 SBS11 + SBS22 SBS11=0.01; SBS22=0.99 0.9735 SBS26 + SBS40 SBS26=0; SBS40=1 0.87558 SBS16 + SBS38 SBS16=0.95; SBS38=0.05 0.8205 SBS39 + SBS40 SBS39=0; SBS40=1 0.7703 SBS1 + SBS22 SBS1=0; SBS22=1 0.9735 SBS27 + SBS40 SBS27=0; SBS40=1 0.87558 SBS16 + SBS30 SBS16=0.92; SBS30=0.08 0.82029 SBS40 + SBS42 SBS40=1; SBS42=0 0.7703 SBS2 + SBS22 SBS2=0; SBS22=1 0.9735 SBS28 + SBS40 SBS28=0; SBS40=1 0.87558 SBS16 + SBS51 SBS16=0.91; SBS51=0.09 0.82005 SBS40 + SBS44 SBS40=1; SBS44=0 0.7703 SBS7a + SBS22 SBS7a=0; SBS22=1 0.9735 SBS34 + SBS40 SBS34=0; SBS40=1 0.87558 SBS15 + SBS16 SBS15=0.06; SBS16=0.94 0.81994 SBS40 + SBS45 SBS40=1; SBS45=0 0.7703 SBS7c + SBS22 SBS7c=0; SBS22=1 0.9735 SBS37 + SBS40 SBS37=0; SBS40=1 0.87558 SBS16 + SBS23 SBS16=0.94; SBS23=0.06 0.81989 SBS40 + SBS46 SBS40=1; SBS46=0 0.7703 SBS10b + SBS22 SBS10b=0; SBS22=1 0.9735 SBS39 + SBS40 SBS39=0; SBS40=1 0.87558 SBS16 + SBS32 SBS16=0.92; SBS32=0.08 0.81977 SBS40 + SBS47 SBS40=1; SBS47=0 0.7703 SBS16 + SBS22 SBS16=0; SBS22=1 0.9735 SBS40 + SBS41 SBS40=1; SBS41=0 0.87558 SBS18 + SBS26 SBS18=0.17; SBS26=0.83 0.81968 SBS40 + SBS48 SBS40=1; SBS48=0 0.7703 SBS17b + SBS22 SBS17b=0; SBS22=1 0.9735 SBS40 + SBS43 SBS40=1; SBS43=0 0.87558 SBS11 + SBS16 SBS11=0.05; SBS16=0.95 0.81942 SBS40 + SBS49 SBS40=1; SBS49=0 0.7703 SBS22 + SBS27 SBS22=1; SBS27=0 0.9735 SBS40 + SBS47 SBS40=1; SBS47=0 0.87558 SBS12 + SBS40 SBS12=0.65; SBS40=0.35 0.81938 SBS40 + SBS50 SBS40=1; SBS50=0 0.7703 SBS22 + SBS34 SBS22=1; SBS34=0 0.9735 SBS40 + SBS51 SBS40=1; SBS51=0 0.87558 SBS6 + SBS16 SBS6=0.05; SBS16=0.95 0.81914 SBS40 + SBS52 SBS40=1; SBS52=0 0.7703 SBS22 + SBS47 SBS22=1; SBS47=0 0.9735 SBS40 + SBS54 SBS40=1; SBS54=0 0.87558 SBS16 + SBS84 SBS16=0.94; SBS84=0.06 0.81912 SBS40 + SBS53 SBS40=1; SBS53=0 0.7703 SBS22 + SBS48 SBS22=1; SBS48=0 0.9735 SBS40 + SBS55 SBS40=1; SBS55=0 0.87558 SBS12 + SBS25 SBS12=0.67; SBS25=0.33 0.8191 SBS40 + SBS56 SBS40=1; SBS56=0 0.7703 SBS22 + SBS51 SBS22=1; SBS51=0 0.9735 SBS40 + SBS57 SBS40=1; SBS57=0 0.87558 SBS16 + SBS55 SBS16=0.96; SBS55=0.04 0.81907 SBS40 + SBS58 SBS40=1; SBS58=0 0.7703 SBS22 + SBS58 SBS22=1; SBS58=0 0.9735 SBS40 + SBS58 SBS40=1; SBS58=0 0.87558 SBS16 + SBS85 SBS16=0.93; SBS85=0.07 0.81886 SBS40 + SBS59 SBS40=1; SBS59=0 0.7703 SBS22 + SBS59 SBS22=1; SBS59=0 0.9735 SBS40 + SBS60 SBS40=1; SBS60=0 0.87558 SBS16 + SBS19 SBS16=0.96; SBS19=0.04 0.81872 SBS40 + SBS84 SBS40=1; SBS84=0 0.7703 SBS22 + SBS60 SBS22=1; SBS60=0 0.9735 SBS40 + SBS85 SBS40=1; SBS85=0 0.87558 SBS16 + SBS59 SBS16=0.97; SBS59=0.03 0.81868 SBS40 + SBS85 SBS40=1; SBS85=0 0.7703 SBS22 + SBS85 SBS22=1; SBS85=0 0.9735 SBS40 SBS40=1 0.87558 SBS7a + SBS16 SBS7a=0.03; SBS16=0.97 0.81845 SBS40 SBS40=1 0.7703 SBS25 + SBS27 SBS25=0.9; SBS27=0.1 0.7411 SBS3 + SBS30 SBS3=0.8; SBS30=0.2 0.86242 SBS16 + SBS52 SBS16=0.98; SBS52=0.02 0.81845 SBS3 + SBS9 SBS3=0.51; SBS9=0.49 0.74308 SBS25 + SBS34 SBS25=0.96; SBS34=0.04 0.72535 SBS5 + SBS29 SBS5=0.77; SBS29=0.23 0.85806 SBS22 + SBS26 SBS22=0.14; SBS26=0.86 0.81825 SBS3 + SBS28 SBS3=0.86; SBS28=0.14 0.74092 SBS25 + SBS35 SBS25=0.91; SBS35=0.09 0.72431 SBS5 + SBS18 SBS5=0.77; SBS18=0.23 0.85756 SBS10b + SBS16 SBS10b=0.02; SBS16=0.98 0.81815 SBS4 + SBS9 SBS4=0.27; SBS9=0.73 0.72368 SBS8 + SBS25 SBS8=0.04; SBS25=0.96 0.72021 SBS4 + SBS5 SBS4=0.27; SBS5=0.73 0.85722 SBS16 + SBS49 SBS16=0.98; SBS49=0.02 0.81808 SBS9 + SBS55 SBS9=0.85; SBS55=0.15 0.72333 SBS25 + SBS47 SBS25=0.98; SBS47=0.02 0.72016 SBS5 + SBS36 SBS5=0.84; SBS36=0.16 0.85037 SBS16 + SBS60 SBS16=0.99; SBS60=0.01 0.81805 SBS9 + SBS39 SBS9=0.67; SBS39=0.33 0.71522 SBS1 + SBS25 SBS1=0; SBS25=1 0.71986 SBS5 + SBS24 SBS5=0.76; SBS24=0.24 0.84979 SBS4 + SBS12 SBS4=0.23; SBS12=0.77 0.8179 SBS9 + SBS18 SBS9=0.78; SBS18=0.22 0.71505 ANNEX B 222 SBS2 + SBS25 SBS2=0; SBS25=1 0.71986 SBS3 + SBS7a SBS3=0.89; SBS7a=0.11 0.84274 SBS16 + SBS47 SBS16=0.98; SBS47=0.02 0.81789 SBS8 + SBS9 SBS8=0.29; SBS9=0.71 0.71393 SBS3 + SBS25 SBS3=0; SBS25=1 0.71986 SBS2 + SBS3 SBS2=0.07; SBS3=0.93 0.83827 SBS1 + SBS16 SBS1=0; SBS16=1 0.81776 SBS9 + SBS29 SBS9=0.79; SBS29=0.21 0.71238 SBS4 + SBS25 SBS4=0; SBS25=1 0.71986 SBS3 + SBS32 SBS3=0.82; SBS32=0.18 0.83301 SBS2 + SBS16 SBS2=0; SBS16=1 0.81776 SBS5 + SBS9 SBS5=0.4; SBS9=0.6 0.71213 SBS5 + SBS25 SBS5=0; SBS25=1 0.71986 SBS3 + SBS5 SBS3=0.49; SBS5=0.51 0.83275 SBS13 + SBS16 SBS13=0; SBS16=1 0.81776 SBS9 + SBS50 SBS9=0.82; SBS50=0.18 0.71142 SBS6 + SBS25 SBS6=0; SBS25=1 0.71986 SBS5 + SBS52 SBS5=0.92; SBS52=0.08 0.83155 SBS16 + SBS27 SBS16=1; SBS27=0 0.81776 SBS9 + SBS51 SBS9=0.8; SBS51=0.2 0.71029 SBS7a + SBS25 SBS7a=0; SBS25=1 0.71986 SBS3 + SBS6 SBS3=0.87; SBS6=0.13 0.83149 SBS16 + SBS28 SBS16=1; SBS28=0 0.81776 SBS9 + SBS24 SBS9=0.79; SBS24=0.21 0.70762 SBS7b + SBS25 SBS7b=0; SBS25=1 0.71986 SBS5 + SBS8 SBS5=0.75; SBS8=0.25 0.83072 SBS16 + SBS34 SBS16=1; SBS34=0 0.81776 SBS3 + SBS17b SBS3=0.9; SBS17b=0.1 0.7051 SBS7c + SBS25 SBS7c=0; SBS25=1 0.71986 SBS5 + SBS30 SBS5=0.83; SBS30=0.17 0.82942 SBS16 + SBS48 SBS16=1; SBS48=0 0.81776 SBS9 + SBS36 SBS9=0.85; SBS36=0.15 0.70343 SBS7d + SBS25 SBS7d=0; SBS25=1 0.71986 SBS5 + SBS45 SBS5=0.9; SBS45=0.1 0.82939 SBS16 SBS16=1 0.81776 SBS5 + SBS28 SBS5=0.83; SBS28=0.17 0.70299 SBS9 + SBS25 SBS9=0; SBS25=1 0.71986 SBS3 + SBS19 SBS3=0.87; SBS19=0.13 0.82841 SBS26 + SBS36 SBS26=0.87; SBS36=0.13 0.81735 SBS2 + SBS9 SBS2=0.07; SBS9=0.93 0.70113 SBS10a + SBS25 SBS10a=0; SBS25=1 0.71986 SBS3 + SBS18 SBS3=0.81; SBS18=0.19 0.82689 SBS12 + SBS54 SBS12=0.82; SBS54=0.18 0.81686 SBS9 + SBS30 SBS9=0.86; SBS30=0.14 0.6986 SBS10b + SBS25 SBS10b=0; SBS25=1 0.71986 SBS5 + SBS35 SBS5=0.81; SBS35=0.19 0.82575 SBS8 + SBS12 SBS8=0.25; SBS12=0.75 0.81658 SBS9 + SBS25 SBS9=0.73; SBS25=0.27 0.69554 SBS11 + SBS25 SBS11=0; SBS25=1 0.71986 SBS5 + SBS42 SBS5=0.8; SBS42=0.2 0.82529 SBS26 + SBS29 SBS26=0.84; SBS29=0.16 0.81623 SBS9 + SBS42 SBS9=0.84; SBS42=0.16 0.69521 SBS12 + SBS25 SBS12=0; SBS25=1 0.71986 SBS5 + SBS50 SBS5=0.85; SBS50=0.15 0.82463 SBS3 + SBS12 SBS3=0.35; SBS12=0.65 0.81582 SBS9 + SBS35 SBS9=0.82; SBS35=0.18 0.69516 SBS13 + SBS25 SBS13=0; SBS25=1 0.71986 SBS5 + SBS19 SBS5=0.88; SBS19=0.12 0.82431 SBS24 + SBS26 SBS24=0.17; SBS26=0.83 0.81564 SBS9 + SBS32 SBS9=0.86; SBS32=0.14 0.69422 SBS14 + SBS25 SBS14=0; SBS25=1 0.71986 SBS3 + SBS11 SBS3=0.88; SBS11=0.12 0.82371 SBS26 + SBS42 SBS26=0.84; SBS42=0.16 0.81558 SBS9 + SBS45 SBS9=0.9; SBS45=0.1 0.69358 SBS15 + SBS25 SBS15=0; SBS25=1 0.71986 SBS3 + SBS36 SBS3=0.86; SBS36=0.14 0.82323 SBS26 + SBS31 SBS26=0.85; SBS31=0.15 0.81503 SBS7a + SBS9 SBS7a=0.08; SBS9=0.92 0.69292 SBS16 + SBS25 SBS16=0; SBS25=1 0.71986 SBS3 + SBS23 SBS3=0.88; SBS23=0.12 0.82169 SBS12 + SBS35 SBS12=0.8; SBS35=0.2 0.815 SBS9 + SBS43 SBS9=0.89; SBS43=0.11 0.69124 SBS17a + SBS25 SBS17a=0; SBS25=1 0.71986 SBS8 + SBS30 SBS8=0.61; SBS30=0.39 0.82138 SBS12 + SBS26 SBS12=0.45; SBS26=0.55 0.81435 SBS9 + SBS13 SBS9=0.93; SBS13=0.07 0.69049 SBS17b + SBS25 SBS17b=0; SBS25=1 0.71986 SBS4 + SBS30 SBS4=0.59; SBS30=0.41 0.82111 SBS26 + SBS56 SBS26=0.93; SBS56=0.07 0.81104 SBS9 + SBS28 SBS9=0.92; SBS28=0.08 0.68905 SBS18 + SBS25 SBS18=0; SBS25=1 0.71986 SBS3 + SBS15 SBS3=0.89; SBS15=0.11 0.82079 SBS12 + SBS18 SBS12=0.84; SBS18=0.16 0.8105 SBS9 + SBS38 SBS9=0.94; SBS38=0.06 0.68465 SBS19 + SBS25 SBS19=0; SBS25=1 0.71986 SBS3 + SBS7b SBS3=0.89; SBS7b=0.11 0.82037 SBS12 + SBS42 SBS12=0.84; SBS42=0.16 0.80975 SBS9 + SBS52 SBS9=0.95; SBS52=0.05 0.68463 SBS20 + SBS25 SBS20=0; SBS25=1 0.71986 SBS3 + SBS42 SBS3=0.8; SBS42=0.2 0.82002 SBS26 + SBS39 SBS26=0.82; SBS39=0.18 0.80972 SBS9 + SBS58 SBS9=0.89; SBS58=0.11 0.68384 SBS21 + SBS25 SBS21=0; SBS25=1 0.71986 SBS3 + SBS24 SBS3=0.8; SBS24=0.2 0.81946 SBS26 + SBS50 SBS26=0.89; SBS50=0.11 0.80971 SBS9 + SBS54 SBS9=0.93; SBS54=0.07 0.68267 SBS23 + SBS25 SBS23=0; SBS25=1 0.71986 SBS5 + SBS7a SBS5=0.92; SBS7a=0.08 0.81882 SBS26 + SBS45 SBS26=0.93; SBS45=0.07 0.80947 SBS9 + SBS53 SBS9=0.94; SBS53=0.06 0.68235 SBS24 + SBS25 SBS24=0; SBS25=1 0.71986 SBS29 + SBS30 SBS29=0.53; SBS30=0.47 0.81861 SBS14 + SBS26 SBS14=0.08; SBS26=0.92 0.80904 SBS9 + SBS60 SBS9=0.97; SBS60=0.03 0.6821 SBS25 + SBS26 SBS25=1; SBS26=0 0.71986 SBS5 + SBS56 SBS5=0.93; SBS56=0.07 0.81833 SBS12 + SBS22 SBS12=0.87; SBS22=0.13 0.80897 SBS9 + SBS11 SBS9=0.94; SBS11=0.06 0.68133 ANNEX B 223 Supplementary Table 12. The 96 trinucleotide frequency for SBS Mongolia. trinucleotide change Relative contribution (SBS Mongolia) A[C>A]A 0.035955461 A[C>A]C 0.013847095 A[C>A]G 0.001721218 A[C>A]T 0.004252817 C[C>A]A 0.014587099 C[C>A]C 0.012362044 C[C>A]G 0.00118355 C[C>A]T 0.003425482 G[C>A]A 0.014750146 G[C>A]C 0.004370825 G[C>A]G 0.001760215 G[C>A]T 0.008717436 T[C>A]A 0.01448695 T[C>A]C 0.035744186 T[C>A]G 0.005123383 T[C>A]T 0.018882905 A[C>G]A 0.006779861 A[C>G]C 0.002203666 A[C>G]G 0.001205332 A[C>G]T 0.017434403 C[C>G]A 0.006238753 C[C>G]C 0.004975743 C[C>G]G 0.004117009 C[C>G]T 0.007098419 G[C>G]A 0.001594505 G[C>G]C 4.75875E-20 G[C>G]G 0.000473898 G[C>G]T 0.003988173 T[C>G]A 0.003795465 T[C>G]C 0.004487134 T[C>G]G 0.001583128 T[C>G]T 0.023982606 A[C>T]A 0.014945584 A[C>T]C 0.008389485 A[C>T]G 0.004551594 A[C>T]T 0.011651419 C[C>T]A 0.009402314 C[C>T]C 0.006496308 C[C>T]G 0.002564394 C[C>T]T 0.009626158 G[C>T]A 0.010533685 G[C>T]C 0.009922544 G[C>T]G 0.002524279 G[C>T]T 0.001925169 T[C>T]A 0.028736617 T[C>T]C 0.011600356 T[C>T]G 0.004565619 T[C>T]T 0.011004112 A[T>A]A 0.006363384 ANNEX B 224 A[T>A]C 4.75875E-20 A[T>A]G 0.011583259 A[T>A]T 0.001046362 C[T>A]A 4.75875E-20 C[T>A]C 0.008078025 C[T>A]G 0.009598124 C[T>A]T 0.013844478 G[T>A]A 0.004196896 G[T>A]C 0.000210043 G[T>A]G 0.0048592 G[T>A]T 0.000784631 T[T>A]A 0.007414714 T[T>A]C 0.005257449 T[T>A]G 4.75875E-20 T[T>A]T 0.012157859 A[T>C]A 4.75875E-20 A[T>C]C 0.004967699 A[T>C]G 0.030530339 A[T>C]T 6.78691E-16 C[T>C]A 0.004469143 C[T>C]C 0.011343202 C[T>C]G 0.009519282 C[T>C]T 0.004018189 G[T>C]A 0.000606475 G[T>C]C 1.49375E-11 G[T>C]G 0.002786154 G[T>C]T 0.004888446 T[T>C]A 0.009366417 T[T>C]C 0.000592396 T[T>C]G 0.001751501 T[T>C]T 0.014951023 A[T>G]A 0.027779096 A[T>G]C 0.020987142 A[T>G]G 0.056400253 A[T>G]T 0.045994861 C[T>G]A 0.007120867 C[T>G]C 0.00877218 C[T>G]G 0.01775835 C[T>G]T 0.037815563 G[T>G]A 0.00780353 G[T>G]C 0.003159935 G[T>G]G 0.014371422 G[T>G]T 0.012311793 T[T>G]A 0.019668356 T[T>G]C 0.014263082 T[T>G]G 0.037929245 T[T>G]T 0.053111095 ANNEX B 225 Supplementary Table 13. Number of signature occurrences above bootstrap exposure cutoff of 0.1 (at P = 0.1, one sided) in the Mongolian and Western cohorts. Differences between Western samples from different origins (Europe and USA) are also indicated. Statistical differences between cohorts were assessed by Fisher test corrected by FDR. Mongolian % Mongolian Western (all) % Western (all) adj. p val Europe % Europe USA % USA adj. p val SBS1 0 0.0% 0 0.0% 1.000 0 0.0% 0 0.0% NA SBS4 2 1.3% 2 1.8% 1.000 2 2.9% 0 0.0% 0.834 SBS5 10 6.6% 13 11.6% 0.515 9 13.0% 4 9.3% 0.834 SBS6 1 0.7% 0 0.0% 1.000 0 0.0% 0 0.0% NA SBS12 2 1.3% 0 0.0% 0.800 0 0.0% 0 0.0% NA SBS16 0 0.0% 2 1.8% 0.515 2 2.9% 0 0.0% 0.834 SBS18 0 0.0% 1 0.9% 0.781 0 0.0% 1 2.3% 0.834 SBS22 21 13.9% 6 5.4% 0.137 0 0.0% 6 14.0% 0.018 SBS26 3 2.0% 0 0.0% 0.580 0 0.0% 0 0.0% NA SBS29 1 0.7% 0 0.0% 1.000 0 0.0% 0 0.0% NA SBS40 49 32.5% 33 29.5% 0.944 21 30.4% 12 27.9% 0.834 SBSM 38 25.2% 5 4.5% 0.00003 4 5.8% 1 2.3% 0.834 ANNEX B 226 Supplementary Table 14. Number of environmental signature occurrences above weight cutoff of 0.1 in the Mongolian and Western cohorts. Statistical differences between cohorts were assessed by Fisher test corrected by FDR. Mongolian % Mongolian Western % Western adj. p val MNU_350_uM 61 40.4% 41 36.6% 0.79922 DMS_0_078_mM 57 37.7% 21 18.8% 0.041921 AFB1_0_25_uM_plus_S9 46 30.5% 24 21.4% 0.461537 DES_0_938_mM 42 27.8% 33 29.5% 0.940622 DMH_11_6_mM_plus_S9 36 23.8% 32 28.6% 0.715437 ENU_400_uM 36 23.8% 23 20.5% 0.781518 Formaldehyde_120_uM 29 19.2% 32 28.6% 0.412997 S_6_Nitrochrysene_12_5_uM_plus_S9 29 19.2% 31 27.7% 0.46342 DBP_0_0039_uM 27 17.9% 12 10.7% 0.461537 AZD7762_1_625_uM 23 15.2% 13 11.6% 0.758845 Mechlorethamine_0_3_uM 22 14.6% 21 18.8% 0.715437 Methyleugenol_1_25_mM 21 13.9% 14 12.5% 0.99728 S_1_8_DNP_8_uM 20 13.2% 10 8.9% 0.691591 AAII_37_5_uM 18 11.9% 18 16.1% 0.715437 Temozolomide_200_uM 11 7.3% 16 14.3% 0.460788 S_6_Nitrochrysene_50_uM_plus_S9 10 6.6% 2 1.8% 0.412997 S_5_Methylchrysene_1_6_uM_plus_S9 9 6.0% 5 4.5% 0.940622 Furan_100_mM_plus_S9 7 4.6% 3 2.7% 0.781518 DBPDE_0_000625_uM 7 4.6% 1 0.9% 0.46342 MX_7_uM_plus_S9 5 3.3% 11 9.8% 0.372154 OTA_0_08_uM_plus_S9 5 3.3% 6 5.4% 0.781518 AAI_1_25_uM 4 2.6% 4 3.6% 0.924609 S_1_8_DNP_0_125_uM 2 1.3% 1 0.9% 1 SSR_1_25_J 2 1.3% 10 8.9% 0.108054 S_3_NBA_0_1_uM 2 1.3% 2 1.8% 1 DBPDE_0_000156_uM 1 0.7% 0 0.0% 1 DBP_0_0313_uM_plus_S9 1 0.7% 0 0.0% 1 Benzidine_200_uM 1 0.7% 2 1.8% 0.781518 Cyclophosphamide_18_75_uM_plus_S9 1 0.7% 3 2.7% 0.691591 Potassium_bromate_260_uM 1 0.7% 3 2.7% 0.691591 S_6_Nitrochrysene_0_78_uM 1 0.7% 0 0.0% 1 S_1_6_DNP_0_09_uM 1 0.7% 2 1.8% 0.781518 N_Nitrosopyrrolidine_50_mM 1 0.7% 0 0.0% 1 Cisplatin_3_125_uM 1 0.7% 3 2.7% 0.691591 Carboplatin_5_uM 1 0.7% 6 5.4% 0.372154 S_6_Nitrochrysene_50_uM 1 0.7% 4 3.6% 0.473602 DBADE_0_109_uM 0 0.0% 0 0.0% 1 Potassium_bromate_875_uM 0 0.0% 0 0.0% 1 S_4_ABP_300_uM_plus_S9 0 0.0% 5 4.5% 0.186098 Semustine_150_uM 0 0.0% 2 1.8% 0.473602 DBADE_0_0313_uM 0 0.0% 0 0.0% 1 BPDE_0_125_uM 0 0.0% 0 0.0% 1 BaP_0_39_uM_plus_S9 0 0.0% 0 0.0% 1 Ellipticine_0_375_uM_plus_S9 0 0.0% 3 2.7% 0.412997 DBA_75_uM_plus_S9 0 0.0% 0 0.0% 1 PhIP_3_uM_plus_S9 0 0.0% 0 0.0% 1 S_3_NBA_0_025_uM 0 0.0% 0 0.0% 1 Propylene_oxide_10_mM 0 0.0% 2 1.8% 0.473602 DBAC_5_uM_plus_S9 0 0.0% 0 0.0% 1 Cisplatin_12_5_uM 0 0.0% 1 0.9% 0.715437 PhIP_4_uM_plus_S9 0 0.0% 1 0.9% 0.715437 BaP_2_uM_plus_S9 0 0.0% 0 0.0% 1 ANNEX B 227 Supplementary Table 15. Clinico-pathological characteristics of the Mongolian gene expression-based clusters in the Mongolian cohort. MGL1 (n=47, 44%) MGL2 (n=27, 26%) MGL3 (n=32, 30%) p value Age (years) 62 (43-76) 56 (44-72) 56 (41-75) 0.028 <60 years (n, %) 18 (39.1) 20 (76.9) 19 (63.3) 0.005 Gender (male, %) 29 (63) 8 (30.8) 19 (63.3) 0.017 BMI > 25 kg/m2 (n, %) 26 (56.5) 12 (46.2) 11 (39.3) ns Etiology: <0.001 • HBV (n, %) 4 (8.5) 0 (0) 0 (0) • HBV/HDV (n, %) 27 (57.4) 19 (70.4) 26 (81.3) • HBV/HCV/HDV (n, %) 1 (2.1) 6 (22.2) 5 (15.6) • HCV (n, %) 15 (31.9) 2 (7.4) 1 (3.1) Region ns • Western (n, %) 9 (20.5) 6 (26.1) 4 (14.8) • Central (n, %) 18 (40.9) 9 (39.1) 13 (48.1) • Eastern (n, %) 1 (2.3) 3 (13) 3 (11.1) • Ulaanbaatar (n, %) 16 (36.4) 5 (21.7) 7 (25.9) BCLC stage (0-A, %) 35 (77.8) 17 (68) 24 (85.7) ns AFP >400 IU/mL (n, %) 5 (13.5) 9 (42.9) 4 (16.7) 0.026 Liver fibrosis (F3-4, %) 17 (36.2) 12 (44.4) 12 (37.5) ns Microvascular invasion (yes, %) 16 (41) 13 (65) 9 (34.6) 0.099 Tumor grade (G3-4, %) 4 (10) 4 (26.7) 0 (0) 0.045 BMI, body mass index; HBV, hepatitis B virus; HCV, hepatitis C virus; HDV, hepatitis delta virus; BCLC, Barcelona Clinic Liver Cancer, AFP, alfa-fetoprotein Some variables have missing values for MGL1, MGL2 and MGL3, respectively: Age, gender: 1, 1, and 2 patients. BMI: 1, 1, and 4 patients. Region: 3, 4, and 5 patients. BCLC stage: 2, 2, and 4 patients. AFP: 10, 6, and 8 patients. Microvascular invasion: 8, 7, and 8 patients. Tumor grade: 7, 12, and 16 patients. Supplementary Table 16. Clinico-pathological characteristics of the Mongolian gene expression-based clusters in the Mongolian NCI cohort. MGL1 MGL2 MGL3 p value (n=34, 49%) (n=18, 26%) (n=18, 26%) Age (years) 62 (41-76) 56.5 (45-77) 59 (23-68) 0.19 <60 years old (n, %) 13 (38.2) 10 (55.6) 9 (50.0) 0.546 Gender (male, %) 21 (61.7) 8 (44.4) 8 (44.4) 0.45 BMI > 25 kg/m2 (n, %) 13 (38.2) 6 (33.3) 7 (38.9) 0.83 Viral infection: HBV + (n, %) 22 (64.7) 14 (77.8) 13 (72.2) 0.804 HDV + (n, %) 14 (41.2) 5 (27.8) 8 (44.4) 0.626 HCV + (n, %) 23 (67.6) 10 (55.6) 8 (44.4) 0.473 AFP >20 IU/ml (n, %) 10 (29.4) 10 (55.6) 5 (27.8) 0.023 Liver cirrhosis (F4, %) 14 (41.2) 9 (50.0) 5 (27.8) 0.135 BMI, body mass index; HBV, hepatitis B virus; HCV, hepatitis C virus; HDV, hepatitis delta virus; AFP, alfa- fetoprotein ANNEX B 228 Supplementary Table 17. HCC specific COSMIC v3 signatures, along with its prevalence across a total of 493 HCC samples and its cosine similarity against COSMIC v2 version. Note that SBS16 changes greatly from v2 to v3. COSMIC v3 signature HCC Sample Number Cosine Similarity against COSMIC v2 SBS1 303 0.95 SBS3 35 0.96 SBS4 88 0.94 SBS5 491 0.96 SBS6 5 0.95 SBS9 4 0.98 SBS12 214 0.94 SBS16 83 0.79 SBS17a 4 NA SBS17b 6 NA SBS18 45 0.99 SBS19 5 0.89 SBS22 51 0.96 SBS23 3 1 SBS24 19 0.94 SBS26 2 0.92 SBS28 2 0.92 SBS29 139 0.97 SBS30 3 0.96 SBS31 1 NA SBS35 19 NA SBS37 18 NA SBS40 45 NA ANNEX B 229 Supplementary Table 18. Publicly available gene sets and signatures used in the study. Name Reference HCC classification Sia HCC immune class Sia D, et al. Gastroenterology 2017;153:812-826 Chiang classification Chiang D, et al. Cancer Res 2008;68:6779-6788 Hoshida classification Hoshida Y, et al. Cancer Res 2009;69:7385-7392 Cluster A signature Lee JS, et al. Hepatology 2004;40:667–76 Molecular pathways TGFB late signature Coulouarn C, et al. Hepatology 2008: 47 2059-2067 MET signature Kaposi-Novak P, et al. J Clin Invest 2006;116:1582-95 NOTCH signature Villanueva A, et al. Gastroenterology 2012;143: 1660-1669 RB1 signature Bollard J, et al. Gut 2017;66: 1286-1296 Antitumor immune response PD1 signaling signature Quigley M, et al. Nat Med 2010;16:1147-51 Exhaustion signature Quigley M, et al. Nat Med 2010;16:1147-51 Stromal enrichment score Yoshihara K, et al. Nat Commun 2013;4:2612 Immune enrichment score Yoshihara K, et al. Nat Commun 2013;4:2612 IFN signature Ayers M, et al. J Clin Invest 2017; 127:2930–40 Immune cell infiltrate gene sets Bindea G, et al. Immunity 2013;39:782-95 ANNEX B 230 Supplementary Data Study 2 Figure S1 Figure S1. Unsupervised clustering analysis. Tumor samples and background liver samples clustered together but in separate branches. Female tumors clustered with the background liver even though they had the Rian signature. (A) Hierarchical clustering and non-negative matrix factorization (NMF) from all the samples and the samples separated by tumor or adjacent tissue. (B) Cophenetic coefficient of the NFM clustering from all samples and tumor or background liver samples alone. Figure S2. Tomato positive tumor exhibit Rian signature. (A) Histology of tumor section from a mouse on HFD that received the AAV-tdTomato virus. Brown stains in the anti-RFP panel shows tdTomato positive hepatocytes. Scale bars = 200 µM. (B) Expression levels of mRNA transcripts near the vector insertion site. Chromosome position is shown the x-axis chr12:110,100,000-111,600,000 and tag density is on the y-axis. TdTomato tumor has a slightly different expression pattern near the integration site suggesting random integration in the Rian locus. ANNEX B 231 Figure S2 Figure S1. Unsupervised clustering analysis. Tumor samples and background liver samples clustered together but in separate branches. Female tumors clustered with the background liver even though they had the Rian signature. (A) Hierarchical clustering and non-negative matrix factorization (NMF) from all the samples and the samples separated by tumor or adjacent tissue. (B) Cophenetic coefficient of the NFM clustering from all samples and tumor or background liver samples alone. Figure S2. Tomato positive tumor exhibit Rian signature. (A) Histology of tumor section from a mouse on HFD that received the AAV-tdTomato virus. Brown stains in the anti-RFP panel shows tdTomato positive hepatocytes. Scale bars = 200 µM. (B) Expression levels of mRNA transcripts near the vector insertion site. Chromosome position is shown the x-axis chr12:110,100,000-111,600,000 and tag density is on the y-axis. TdTomato tumor has a slightly different expression pattern near the integration site suggesting random integration in the Rian locus. Table S1: Summary of different treatment groups used in the neonatal experimental paradigm. Neonatal mice were infected with rAAV, started on HFD or RD at 3 weeks of age and euthanized at 6 months of age. Vector Diet # of Males # of Females Total AAV-CAG-tdTomato HFD 4 5 9 AAV-CAG-tdTomato RD 3 3 6 AAV-Rian-CMV HFD 14 5 19 AAV-Rian-CMV RD 7 8 15 No AAV HFD 3 3 6 Table S2: Summary of different treatment groups used in the adult experimental paradigm. 3 weeks old mice were started on HFD or RD, injected with rAAV at 10 weeks of age, and euthanized at 9 months. Vector Diet # of mice (Male) AAV-CAG-tdTomato HFD 10 AAV-CAG-tdTomato RD 5 AAV-Rian-CMV HFD 10 AAV-Rian-CMV RD 20 AAV-Rian-CMV + 2/3 PH RD 5 No AAV HFD 5 ANNEX B 232 Table S3. Summary of the samples used for RNAseq. HCC Tumor samples Virus inoculated AAV-Rian AAV-Tomato (control) Gender Male Female Male Female Diet RD HFD RD HFD RD HFD RD HFD n of samples 3 3 1 1 - 1 - - Short name used in the figures RD -R ia n- T H FD -R ia n- T RD -R ia n- T- F H FD -R ia n- T- F H FD -T om at o- T Background liver samples Virus inoculated N/A AAV-Tomato (control) Treatment Estrogen Vehicle N/A N/A Gender Male Female Male Female Male Female Male Female Diet RD HF D RD HF D RD HF D RD HF D RD HF D RD HF D RD HF D RD HF D n of samples 1 1 - - 1 1 1 - 1 1 - - - - - - Short name used in the figures R D -E 2 H FD -E 2 RD -V eh H FD -V eh RD -V eh -F RD -T om at o H FD -T om at o ANNEX B 233 Supplemental materials and methods: Murine samples: Neonatal male and female mice were infected with an AAV targeting the Rian locus (AAV- Rian) or a control AAV (AAV-Tomato), fed with RD or HFD for six months, and three weeks old mice on fed HFD or RD and treated with estrogen or vehicle for one month. Table below summarizes the samples used for RNAseq. All background liver samples correspond to a pooled mix of 3 samples sequenced together. Tumor samples were sequenced individually. Unsupervised cluster analysis: RNA-seq data were filtered to remove all the genes that were not expressed in any of the samples (FPKM=0). Unsupervised clustering of the samples was performed using non-negative matrix factorization (NMF) and hierarchical clustering GenePattern modules.1, 2 For the NMF, the k with greater cophenetic coefficient was used for each analysis. Gene expression profile analysis and generation of an AAV-Rian gene signature: Murine gene expression data were humanized using the mouse-human orthologues extracted from the BioMart database through BiomaRt R package.3 Next, class prediction using previously published HCC molecular classifications was conducted by Nearest Template Prediction (NTP) analysis (Gene Pattern modules).4 The evaluation of the enrichment of distinct molecular pathways and gene expression signatures was performed using single-sample Gene Set Enrichment Analysis (ssGSEA). To this end, Molecular Signature Database gene sets (MSigDB, www.broadinstitute.org/msigdb) and previously reported gene-expression signatures representing different states of inflammation and liver function were tested (Table S3).5 To evaluate the similarity to the tumors developed in Wang et al. 2012, a gene signature was generated using the top differentially expressed genes in Wang et al. 2012 tumors compared to healthy liver (AAV-Rian).6 The generated gene signature (from now on called AAV-Rian ANNEX B 234 signature) was composed of 199 up-regulated genes (AAV-Rian-UP) and 100 down-regulated genes (AAV-Rian-DOWN). ssGSEA and NTP were used to assess similarity between tumors reported in this report and AAV-Rian signature. In addition, to further evaluate the expression of the genes located in the Rian Locus, the “Rian-Locus” geneset was generated including all the genes in the murine genome region chr12:108860000-1104180000, where the Rian Locus is located (genome of reference GRCm38) (Table S2). The Rian-Locus geneset and the Wang gene signature (AAV-Rian) were evaluated in non-humanized gene expression data. Differentially expressed genes between groups of samples were identified using the Limma R package11 (FDR<0.05 and Fold-change [FC]> 2 or <0.5).7 Supplemental references: 1. Brunet, JP, Tamayo, P, Golub, TR, and Mesirov, JP (2004). Metagenes and molecular pattern discovery using matrix factorization. Proc Natl Acad Sci U S A 101: 4164-4169. 2. Eisen, MB, Spellman, PT, Brown, PO, and Botstein, D (1998). Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A 95: 14863-14868. 3. Smedley, D, Haider, S, Ballester, B, Holland, R, London, D, Thorisson, G, et al. (2009). BioMart--biological queries made easy. BMC Genomics 10: 22. 4. Reich, M, Liefeld, T, Gould, J, Lerner, J, Tamayo, P, and Mesirov, JP (2006). GenePattern 2.0. Nat Genet 38: 500-501. 5. Llovet, JM, Montal, R, Sia, D, and Finn, RS (2018). Molecular therapies and precision medicine for hepatocellular carcinoma. Nat Rev Clin Oncol 15: 599-616. 6. Wang, PR, Xu, M, Toffanin, S, Li, Y, Llovet, JM, and Russell, DW (2012). Induction of hepatocellular carcinoma by in vivo gene targeting. Proc Natl Acad Sci U S A 109: 11264-11269. 7. Smyth, G, Gentleman, R, Carey, V, Dudoit, S, Irizarry, R, and Huber, W (2005). Limma: linear models for microarray data, Bioinformatics and Computational Biology Solutions Using R and Bioconductor. Springer: 397-420. ANNEX B 235 Supplementary Data Study 3 SUPPLEMENTARY MATERIALS AND METHODS Treatment strategy and sample collection Lenvatinib (Eisai, Ibaraki, Japan) and anti-PD1 (anti-murine PD-1 mab clone J43 BioXCell BE0033-2) treatments were administered in accordance to current bibliography and provider’s recommendations [1–4]. Lenvatinib was diluted in distilled water, while anti-PD1 and IgG were diluted in InVivoPure dilution buffer (BioXCell IP0065) in agreement with manufacturer’s recommendations. Lenvatinib (10 mg/kg) was administered daily by oral gavage [1,2], and anti-PD1 (10 mg/kg) was intraperitoneally administered every 3 days for a total of 5 doses [4]. Sample size calculation was based on power analysis and on our previous studies using similar models [5]. In order to analyze the mechanisms of action of the drugs, 5 mice per arm from the Hepa1-6 model were euthanized 1 day after the last anti-PD1 dose (day 13 post-randomization, early timepoint) and 2 hours after the last lenvatinib administration, in accordance with pharmacodynamic studies [6–9]. Tumors were weighed and samples were collected for molecular analyses and immune population characterization. Blood samples were collected by cardiac puncture from mice under deep terminal anesthesia. The remaining mice were monitored until a tumor volume of 1000 mm3 was reached or until study termination at day 125 post-randomization after sacrificing the last placebo animal (late timepoint). Tumor samples were collected, and tumor growth and survival, defined as time to reach 1000 mm3 tumor volume, were measured. Response to the treatment was measured according to the percentage change thresholds of the RECIST criteria adapted for murine models [10–12]. Time to objective response was defined as time to achieve a 30% decrease in tumor volume [11]. Animals developing tumor ulcers were censored from the survival and response analysis at the time of sacrifice (placebo n=1, combination n=1). Potential treatment-related toxicity was evaluated by monitoring body weight. The Hep53.4 models were used as validation and all animals were sacrificed at day 13 post-randomization. Tumor samples were harvested and response to treatment was evaluated as described above. Assessment of potential gender effect in the syngeneic model The syngeneic mouse models of HCC were used for this study in accordance with current bibliography supporting the use of this model to investigate the effect of immunotherapy and combination treatments [13]. Additional experiments in both male and female mice were conducted to determine whether there are gender differences in the subcutaneous HCC model. Specifically, the Hepa1-6 model was generated in male and female mice (n=30) by injecting 5x106 Hepa1-6 cells in 100ml PBS in the right flank of in 5-6- week old C57BL/6J mice (Charles River Laboratories). Animals were sacrificed once a tumor volume of 1000 mm3 was reached or at study termination (40 days post-injection). No differences were observed in terms of tumor penetrance, time to tumor onset, tumor growth or survival (Supplementary Figure 1A-D). Histological analysis in haematoxylin and eosin slides did not reveal differences in histopathological features or immune infiltrate (Supplementary Figure 1E). Considering this, we discard any effect of mice gender in the animal model presented in the study. Sample processing for flow cytometry analysis Tumor and blood samples from the Hepa1-6 model were collected to perform flow cytometry analysis. Tumor samples were minced and digested using Accumax cell detachment solution (Innovative Cell Technologies) for 45 min at 37°C and cells were filtered using a 70 µm nylon cell strainer. Blood samples were collected in EDTA-coated tubes and peripheral blood leukocytes were isolated after red blood cell lysis using RBC lysis buffer (eBioscience). Cells from tumor and blood samples were then washed, filtered and stained according to standard flow cytometer protocols (Supplementary Table 1,2). ANNEX B 236 Immunohistochemistry staining of tumor samples Only tumors with sufficient tumoral material were processed for histological analysis. From the Hepa1-6 model, 14 and 17 samples were analyzed at the early and late timepoints, respectively. Samples from the early timepoint included 4 placebo, 4 lenvatinib, 3 anti-PD1 and 3 combination. Samples from the late timepoint included 8 placebo, 4 lenvatinib, 2 anti-PD1 and 3 combination. 6 tumors from the Hep53.4 subcutaneous model were used for validation. The primary antibodies used were CD3 (Abcam ab16669), CD4 (Abcam ab183685), CD8 (Abcam ab203035), FOXP3 (Abcam ab215206), CTLA4 (Abcam ab237712), PD1 (Sino Biological INC 50124-RP02), PDL1 (Novus Biologicals MAB90782), CD31 (Abcam ab28364), CD163 (Abcam ab182422) and CD68 (Abcam ab125212) (Supplementary Table 2, Supplementary Table 3). All immunohistochemical (IHC) stainings were carried out on 3 μm-thick FFPE tissue sections after heat-induced antigen retrieval. CTLA4 and FOXP3 staining was quantified by the digital pathology imaging software QuPath (version 0.2.0) [14]. To ensure representative sampling of the entire tumor, 3 to 5 regions of interest (ROI) were analyzed. Each ROI had a magnification of 200X and each slide contained an average of 2700 cells in total. The number of positive cells per total cells was determined for each ROI. FOXP3 was used to assess the presence and distribution of Regulatory T-cells (Treg). Positivity for the immune checkpoints PD1, PDL1 and CTLA4 was also analyzed. To detect angiogenesis, CD31 positivity and presence of vessels encapsulating tumor clusters (VETC) were analyzed [15]. Interaction between Treg cells and tumor vasculature was assessed by double staining of FOXP3 and CD31. the interaction of FOXP3 and CD31 was evaluated following the ensuing criteria: a) when FOXP3 positive cells were localized close to a well-formed vessel the interaction was considered marked; b) when FOXP3 was localized close to vessel sprouts the interaction was considered moderate; c) when FOXP3 positive cells were neither close to a well-formed vessel nor to a vessel sprout, the interaction was considered minimal; and d) when there was no FOXP3 staining, the interaction was considered absent. Finally, the M2 per total macrophage ratio was assessed by measuring CD68 and CD163 markers in an automatized IHC stainer (Autostainer 48 AS48030, Agilent). RNA preparation and transcriptome analysis Tumor samples from each treatment arm in the Hepa1-6 model collected at the early timepoint or from mice reaching the survival endpoint at day 13 post-randomization were included (n=21). Samples were cut, collected in RNAlater solution and then stored at -80ºC. Total RNA was extracted from 30 mg of tissue using Trizol reagent (Invitrogen) and purified with RNeasy columns (Qiagen, Valencia, CA). RNA sample concentration and quality were assessed by NanoDrop ND-1000 spectrometer (NanoDrop, Wilmington, DE) and bioanalyzer (Agilent, Palo Alto, CA), respectively. Gene expression microarray studies were conducted using the Clariom S Mouse Array (Affymetrix, Santa Clara, CA). The raw CEL files were background corrected and normalized using the Robust Multiarray Averaging (RMA) procedure using the oligo R package [16]. To analyze the gene expression profile of the tumors, intensity values were log transformed and mice genes were humanized. Briefly, mouse-human orthologues were obtained from the BioMart database through the BiomaRt R package [17] and expression values for each human gene were calculated using the CollapseDataset module from GenePattern. The gene expression profile was analyzed using the NTP, GSEA and ssGSEA modules from GenePattern [18] (gene sets available in MSigDB or previously reported [19,20], Supplementary Table 10). For GSEA analysis, only gene sets with enrichment score >1.5 were analyzed. Differentially expressed genes between treatment arms were identified (Bonferroni p<0.05, fold change [FC] >1.5) and pathway analysis was performed using the DAVID functional annotation tool. The ESTIMATE score of stromal infiltration and relative tumor purity was assessed using the ESTIMATE R package [21]. ANNEX B 237 Identification of potential responders to combination treatment in independent human cohorts Gene expression data from a cohort of 228 surgically resected fresh-frozen HCC samples and 169 paired non-tumoral samples (Heptromic dataset, GSE63898) was analyzed. Full descriptions of the cohort and RNA profiling data are available in previous publications [22,23]. The upregulation and downregulation of the combination rescue signature and HCC immune signature [22] were assessed in tumor samples by NTP to classify patients as combination-only responder class, potential responders to ICI or rest. The gene expression profile of each group was assessed using the NTP, GSEA and ssGSEA modules from GenePattern. NTP predictions have been classified as significant using nominal p-value <0.05 and FDR <0.1 for the combination rescue signature and using FDR<0.05 for the remaining signatures. The relative fraction of immune cells in the tumor tissue was estimated using CIBERSORT tool [24]. Similarities between murine and human HCC tumors were assessed by principal component analysis (PCA) and sub-map analysis. SUPPLEMENTARY REFERENCES 1. Kato Y, Tabata K, Kimura T, et al. Lenvatinib plus anti-PD-1 antibody combination treatment activates CD8 + T cells through reduction of tumor-associated macrophage and activation of the interferon pathway. PLoS One. 2019;14:e0212513. 2. Kimura T, Kato Y, Ozawa Y, et al. Immunomodulatory activity of lenvatinib contributes to antitumor activity in the Hepa1-6 hepatocellular carcinoma model. Cancer Sci. 2018;109:3993–4002. 3. European Medicines Agency. Keytruda Assessment Report. Procedure No. EMEA/H/C/003820/0000. 2015. 4. Böttcher JP, Reis e Sousa C. The Role of Type 1 Conventional Dendritic Cells in Cancer Immunity. Trends in Cancer. 2018;4:784–92. 5. Martinez-Quetglas I, Pinyol R, Dauch D, et al. IGF2 Is Up-regulated by Epigenetic Mechanisms in Hepatocellular Carcinomas and Is an Actionable Oncogene Product in Experimental Models. Gastroenterology. 2016;151:1192–205. 6. Lindauer A, Valiathan CR, Mehta K, et al. Translational Pharmacokinetic/Pharmacodynamic Modeling of Tumor Growth Inhibition Supports Dose-Range Selection of the Anti-PD-1 Antibody Pembrolizumab. CPT pharmacometrics Syst Pharmacol. 2017;6:11–20. 7. Okamoto K, Kodama K, Takase K, et al. Antitumor activities of the targeted multi-tyrosine kinase inhibitor lenvatinib (E7080) against RET gene fusion-driven tumor models. Cancer Lett. 2013;340:97–103. 8. Kato Y, Bao X, Macgrath S, et al. Lenvatinib mesilate (LEN) enhanced antitumor activity of a PD-1 blockade agent by potentiating Th1 immune response. Ann Oncol. 2016;27:(suppl 6). 9. Kato Y, Tabata K, Hori Y, et al. Effects of lenvatinib on tumor-associated macrophages enhance antitumor activity of PD-1 signal Inhibitors. 2015;14:(Suppl 2). 10. Eisenhauer EA, Therasse P, Bogaerts J, et al. New response evaluation criteria in solid tumours: Revised RECIST guideline (version 1.1). Eur J Cancer. 2009;45:228–47. 11. Gao H, Korn JM, Ferretti S, et al. High-throughput screening using patient-derived tumor xenografts to predict clinical trial drug response. Nat Med. 2015;21:1318–25. 12. Byrne AT, Alférez DG, Amant F, et al. Interrogating open issues in cancer precision medicine with patient-derived xenografts. Nat Rev Cancer. 2017;17:254–68. 13. Brown ZJ, Heinrich B, Greten TF. Mouse models of hepatocellular carcinoma: an overview and highlights for immunotherapy research. Nat Rev Gastroenterol Hepatol. 2018;15:536–54. 14. Bankhead P, Loughrey MB, Fernández JA, et al. QuPath: Open source software for digital pathology image analysis. Sci Rep. 2017;7:16878. 15. Itoh S, Yoshizumi T, Yugawa K, et al. Impact of Immune Response on Outcomes in Hepatocellular Carcinoma: Association with Vascular Formation. Hepatology. 2020; 16. Irizarry RA, Hobbs B, Collin F, et al. Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics. 2003;4:249–64. ANNEX B 238 17. Smedley D, Haider S, Ballester B, et al. BioMart - Biological queries made easy. BMC Genomics. 2009;10:22. 18. Reich M, Liefeld T, Gould J, et al. GenePattern 2.0. Nat Genet. 2006;38:500–1. 19. Bindea G, Mlecnik B, Tosolini M, et al. Spatiotemporal dynamics of intratumoral immune cells reveal the immune landscape in human cancer. Immunity. 2013;39:782–95. 20. Jerby-Arnon L, Shah P, Cuoco MS, et al. A Cancer Cell Program Promotes T Cell Exclusion and Resistance to Checkpoint Blockade. Cell. 2018;4:984–97. 21. Yoshihara K, Shahmoradgoli M, Martínez E, et al. Inferring tumour purity and stromal and immune cell admixture from expression data. Nat Commun. 2013;4:2612. 22. Sia D, Jiao Y, Martinez-Quetglas I, et al. Identification of an Immune-specific Class of Hepatocellular Carcinoma, Based on Molecular Features. Gastroenterology. 2017;153:812–26. 23. Villanueva A, Portela A, Sayols S, et al. DNA Methylation-based prognosis and epidrivers in hepatocellular carcinoma. Hepatology. 2015;61:1945–56. 24. Newman AM, Liu CL, Green MR, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods. 2015;12:453–7. ANNEX B 239 Supplementary Fig. 1. Comparison of male and female subcutaneous syngeneic HCC models. (A) Time to randomization (time to reach 200 mm3) after injection of the cells. Plots represent mean values plus standard error. (B) Mean tumor growth after randomization in males vs females and (C) p-values as determined by Mann-Whitney, with Bonferroni corrected p-values also displayed. (D) Kaplan-Meier curve showing survival (time to reach 1000 mm3) between male and female mice. Mice sacrificed at study termination (40 days post tumor injection) are represented as censored. (E) Histopathological characteristics of male and female syngeneic murine tumors. Representative images of H&E stained tumors captured with 40X magnification. A B C D E n.s. ANNEX B 240 Supplementary Fig. 2. Gating strategy for the flow cytometric analysis in myeloid and T cell panels. Immune cell populations analyzed by flow cytometry in the lymphoid and myeloid antibody panels. ANNEX B 241 Supplementary Fig. 3. Antitumoral effect of lenvatinib plus anti-PD1. (A) Tumor volume at randomization in the subcutaneous models. (B) Body weight monitoring in treated mice from the Hepa1- 6 model. (C) Response to treatment at late timepoint (n=37). Upper part indicates differences in progressive disease rate and lower part, differences in objective response rate. Right table shows number of mice per group and response. # indicates animal with no measurable tumor but with tumor tissue found in the necropsy. (D) Tumor viability assessed in H&E slides at the early and late timepoints. (E) Representative images of tumor viability in each treatment arm. Images were captured with 20X magnification. (F) Percentage of positive staining for Ki67 in tumors from animals in each treatment arm. Boxplots indicate median and quartiles. *p<0.05, **p<0.01, ***p<0.001 vs placebo. PD, progressive disease, SD, stable disease, OR, objective response. ** * ** ** Tumor response rates PD SD OR Placebo 9 0 0 Lenvatinib 4 0 6 Anti-PD1 1 0 9 Combination 0 0 8 C A B D E Placebo CombinationAnti-PD1 Lenvatinib # F n.s. n.s. Tumor volume at randomization ANNEX B 242 Supplementary Fig. 4. Response to treatment in the Hep53.4 models. (A) Response to treatment in the Hep53.4 subcutaneous model (n = 40). Upper part indicates differences in progressive disease rate and lower part, differences in objective response rate. Right table shows the number of mice per group and response. (B) Change in bioluminescence in the orthotopic model at day 12 post-randomization. * p<0.05, **p<0.01, ***p < 0.001. PD, progressive disease, SD, stable disease, OR, objective response. ANNEX B 243 Supplementary Fig. 5. Immune cell populations in blood samples detected by flow cytometry analysis. (A) Lymphoid and (B) myeloid immune cell populations from blood samples collected at the early timepoint. Results for each treatment arm are shown (n=5 samples per arm). Boxplots indicate median and quartiles. *p<0.05. Blood lymphocyte panel Blood myeloid panel Figure S6 B A ANNEX B 244 Supplementary Fig. 6. Histological analysis of lymphocytic tumor infiltrate. (A) Mean percentage of CD4 and CD8 staining in tumors from animals in each treatment arm in the Hepa1-6 model. (B) Percentage of samples with intra-tumoral or peripheral CD4 staining location in the Hepa1-6 model and (C) representative images. Images were captured with 40X. (D) Percentage of positive cells for CD3 and (E) mean percentage of CD4 and CD8 staining in the Hep53.4 subcutaneous model. (F) Percentage of samples with intra-tumoral or peripheral CD8 staining in the Hep53.4 model. * p<0.05, ** p<0.01. A B ** * C Figure S4 n.s. D E F ANNEX B 245 Supplementary Fig. 7. Histological analysis of Treg cell tumor infiltrate and interaction with tumor vasculature. (A) Percentage of positive cells for FOXP3 staining in tumor samples from the Hepa1-6 model. Boxplots indicate median and quartiles. (B) Percentage of samples with intra-tumoral or peripheral FOXP3 staining and (C) representative images. (D) Percentage of samples with marked, moderate, minimal Treg and tumor vasculature co-localization, or absence of Treg, assessed by FOXP3 and CD31 co-staining. (E) Representative images of FOXP3 (pink) and CD31 (brown) co-staining. Images were captured with 40X. * Images were captured with 40X. * p<0.05, ** p<0.01. * Figure S5 A B C D E n.s. ANNEX B 246 Supplementary Fig. 8. Histological analysis of tumor microenvironment at the early and late timepoints. (A) Percentage of cells double-stained with CD68 and CD163 (M2 macrophages) out of total CD68-stained samples (total macrophages) in tumor samples from treated animals. (B) Percentage of positive cells for CTLA4 and positive area for (C) PD1, and (D) PDL1 staining in tumor samples from treated animals. (E) Percentage of samples with vessels encapsulating tumor clusters in tumor samples (VETC) from treated animals. (F) Representative images of CD31 staining and (G) percentage of positive area in each treatment arm. Images were captured with 40X. Boxplots indicate median and quartiles. * p<0.05, *** p<0.001. E * *** *** n.s. n.s. * * * n.s. n.s. A B C D * ** * Presence of VETC Absence of VETC Placebo Lenvatinib Anti-PD1 Combination G * * F ANNEX B 247 Supplementary Fig. 9. Transcriptomic differences between treatment groups. (A) Principal component analysis including samples from murine HCC, human HCC tumors and adjacent non-tumoral cirrhotic tissue (Heptromic cohort). (B) Number of differentially expressed genes among treatment groups (Bonferroni p<0.05, FC>1.5 or <0.67). (C) GSEA analysis of gene sets associated with pathways downstream of Lenvatinib targets in tumor samples from the lenvatinib group compared to placebo. (D- F) Representative differentially expressed gene sets between lenvatinib (D), anti-PD1 (E) or combination (F) and placebo assessed by GSEA. (G) Top differentially expressed gene sets between combination and anti-PD1 assessed by GSEA (Bonferroni test, all p<0.05). 1265 540 30 1621 560 30 Lenvatinib vs Placebo Combination vs anti-PD1Combination vs Placebo Anti-PD1 vs Placebo A B D E F G p < 0.0001 FDR < 0.0001 p = 0.002 FDR = 0.021 p = 0.008 FDR = 0.032 p = 0.008 FDR = 0.039 C ANNEX B 248 Supplementary Fig. 10. Enrichment of gene sets associated with lenvatinib target pathway activation in human HCC. Transcriptomic profile of HCC samples classified as HCC Immune class (potential responders to ICI) or combination-only responder class. The heatmap represents the enrichment score of VEGF, FGF, RET, or immune-related gene sets. ANNEX B 249 SUPPLEMENTARY TABLES Supplementary Table 1. Cell populations assessed by flow cytometry. * Population analyzed in blood only. # Population analyzed in tumor tissue only. Supplementary Table 2. Antibody panels for flow cytometry analysis. Antibody Fluorochrome Color Clone Test (μg in 100 ul) Company Lymphocyte panel (tumor) CD45 PE-Cy7 Yel 5 30F11 0.125 Biolegend CD3 PerCP/Cy5.5 Blue 4 145-2C11 0.25 Biolegend CD19 PE Yel 1 1D3/CD19 0.03125 Biolegend CD4 APC/Fire 750 Red 3 GK 1.5 0.125 Biolegend CD8 Brilliant Violet 510 Viol 2 53–6.7 0.5 Biolegend FoxP3 APC Red 1 FJK-16s 1 eBioscience PD1 FITC Blue 1 J43 1 eBioscience Ki67 eFluor 450 Viol 1 SolA15 0.1 eBioscience Live Dead Red - Yel 2 - 1 ul per 106 cells ThermoFisher Lymphocyte panel (blood) CD45 PE-Cy7 Yel 5 30F11 0.125 Biolegend CD3 PerCP/Cy5.5 Blue 4 145-2C11 0.25 Biolegend CD19 PE Yel 1 1D3/CD19 0.03125 Biolegend CD4 APC/Fire 750 Red 3 GK 1.5 0.125 Biolegend CD8 Brilliant Violet 510 Viol 2 53–6.7 0.5 Biolegend FoxP3 APC Red 1 FJK-16s 1 eBioscience CD44 FITC Blue 1 IM7 0.5 Biolegend Ki67 eFluor 450 Viol 1 SolA15 0.1 eBioscience Live Dead Red - Yel 2 - 1 ul per 106 cells ThermoFisher Myeloid panel CD45 PE-Cy7 Yel 5 30F11 0.125 Biolegend CD3 PerCP/Cy5.5 Blue 4 145-2C11 0.25 Biolegend CD11b FITC Blue 1 M1/70 0.25 Biolegend CD11c Alexa fluor 700 Red 2 N418 0.5 Biolegend Ly6C PE Yel 1 HK1.4 0.25 Biolegend Ly6G APC Red 1 1A8 0.06 Biolegend Zombie Aqua stain - Viol 2 - 1 ul per 106 cells Biolegend Lymphoid cell populations Markers Leukocytes CD45+ T cells CD45+ CD3+ Ag-experienced T cells* CD45+ CD3+ CD44+ CD8 T cells CD45+ CD3+ CD8+ PD1+ CD8 T cells# CD45+ CD3+ CD8+ PD1+ Proliferating CD8 T cells CD45+ CD3+ CD8+ Ki67+ CD4 T cells CD45+ CD3+ CD4+ T regulatory cells CD45+ CD3+ CD4+ Foxp3+ B cells CD45+ CD3- CD19+ Myeloid cell populations Markers Type 1 dendritic cells# CD45+ CD3- CD11b- CD11c+ Ly6c-Ly6g- Macrophages CD45+ CD3- CD11b+ Ly6c- Ly6gLow MDSC CD45+ CD3- CD11b+ Ly6c+ Ly6g+ ANNEX B 250 Supplementary Table 3. List of antibodies used for flow cytometry and immunohistochemical analyses. Antibody Supplier Cat no. Clone no. CD45 (PE/Cyanine7) Biolegend 103114 30F11 CD3ε (PerCP/Cyanine5.5) Biolegend 100327 145-2C11 CD19 (PE) Biolegend 152407 1D3/CD19 CD4 (APC/Fire™ 750) Biolegend 100459 GK 1.5 CD8a (Brilliant Violet 510™) Biolegend 100752 53–6.7 FOXP3 (APC) eBioscience 17-5773-82 FJK-16s PD1 (FITC) eBioscience 11-9985-81 J43 Ki67 (eFluor450) eBioscience 48-5698-80 SolA15 CD44 (FITC) Biolegend 103021 IM7 CD11b (FITC) Biolegend 101205 M1/70 CD11c (Alexa Fluor 700) Biolegend 117319 N418 Ly-6C (PE) Biolegend 128007 HK1.4 Ly-6G (APC) Biolegend 127613 1A8 CD3 Abcam ab16669 SP7 CD4 Abcam ab183685 EPR19514 CD8 Abcam ab203035 Polyclonal FOXP3 Abcam ab215206 EPR22102-37 CTLA4 Abcam ab237712 CAL49 PD1 Sino Biological 50124-RP02 Polyclonal PDL1 Novus Biologicals MAB90782 2096A CD31 Abcam ab28364 Polyclonal CD163 Abcam ab182422 EPR19518 CD68 Abcam ab125212 Polyclonal Anti-rabbit IgG (HRP) Dako P0448 Polyclonal Ki67 Abcam Ab16667 SP6 ANNEX B 251 Supplementary Table 4. Immune cell population in tumor samples from treated mice. Tumor immune infiltrate in tumor samples from treated mice in the Hepa1-6 model measured by flow cytometry analysis represented as percentage of CD45+ events. Values indicate mean ± standard deviation. Placebo (n=5) Lenvatinib (n=5) Anti-PD1 (n=5) Combination (n=5) Tumor T cells 50.74 ± 9.37 52.96 ± 19.67 79.01 ± 8.44 75.80 ± 14.84 CD4 T cells 14.24 ± 8.43 4.84 ± 1.74 11.45 ± 5.28 8.08 ± 2.81 Regulatory T cells 3.32 ± 0.98 1.67 ± 1.64 2.49 ± 1.04 0.48 ± 0.25 CD8 T cells 26.54 ± 5.12 19.22 ± 5.01 31.82 ± 11.72 27.28 ± 17.34 Proliferating CD8 T cells 5.92 ± 4.97 6.91 ± 3.12 11.07 ± 3.45 13.83 ± 9.86 PD1+ CD8 T cells 11.27 ± 8.10 10.45 ± 2.75 2.20 ± 1.09 1.85 ± 1.22 B cells 15.19 ± 12.39 10.05 ± 8.01 5.11 ± 6.27 5.93 ± 4.74 Type 1 dendritic cells 0.60 ± 0.52 0.71 ± 0.41 2.45 ± 1.13 2.89 ± 1.17 Macrophages 3.07 ± 3.44 2.78 ± 1.92 1.63 ± 1.06 2.95 ± 3.49 MDSC 2.23 ± 2.60 9.58 ± 7.57 2.58 ± 2.25 3.50 ± 3.95 Blood T cells 37.59 ± 14.07 40.14 ± 5.17 41.69 ± 11.47 35.92 ± 12.55 Antigen experienced T cells 4.27 ± 1.55 5.48 ± 1.98 5.20 ± 2.36 7.81 ± 3.52 CD4 T cells 11.82 ± 4.25 13.00 ± 3.35 14.75 ± 6.46 14.78 ± 6.25 Regulatory T cells 1.05 ± 3.91 0.63 ± 0.32 0.65 ± 0.27 0.44 ± 0.34 CD8 T cells 11.57 ± 3.91 13.03 ± 1.58 14.16 ± 2.69 12.57 ± 5.03 Proliferating CD8 T cells 2.42 ± 3.38 3.14 ± 2.94 3.28 ± 2.18 2.53 ± 2.59 B cells 40.80 ± 17.57 40.23 ± 12.19 39.31 ± 5.68 50.06 ± 11.85 Macrophages 12.29 ± 8.73 10.67 ± 8.72 12.46 ± 6.04 18.72 ± 12.77 MDSC 2.21 ± 1.70 4.08 ± 2.87 4.18 ± 5.28 1.59 ± 0.96 ANNEX B 252 Supplementary Table 5. Top 30 dfferentially expressed genes in tumors from each treatment arm (Bonferroni p <0.05, FC>1.5 or <0.67). Lenvatinib vs Placebo Anti-PD1 vs Placebo Combination vs Placebo Combination vs anti-PD1 gene FC p-val gene FC p-val gene FC p-val gene FC p-val TSC22D3 2.891 0.040 CLEC12A 10.070 0.004 JCHAIN 14.257 0.004 ABHD1 1.627 0.008 SLC16A3 2.143 0.005 JCHAIN 9.016 0.049 CLEC12A 12.422 0.001 ACVR1 0.652 0.050 LAP3 1.782 0.038 ATP6V0D2 7.975 0.004 CLEC4D 10.589 0.001 BIN1 1.829 0.028 AMPD3 1.779 0.018 SLAMF7 7.136 0.012 ATP6V0D2 9.702 0.003 BMP2 0.587 0.015 GAL3ST4 1.749 0.027 EMB 6.618 0.007 IL7R 8.912 0.002 CALD1 0.619 0.038 GADD45B 1.748 0.018 OGN 6.542 0.036 SLAMF7 8.560 0.005 CHST10 1.562 0.013 S1PR3 1.746 0.020 LUM 6.533 0.043 TREM2 7.679 0.001 CMC2 1.529 0.043 CRELD2 1.579 0.028 IL7R 6.367 0.009 ADAM8 6.776 0.001 COG4 0.628 0.050 ISG15 1.573 0.042 CLEC4D 5.732 0.024 SAMSN1 6.629 0.002 CORO6 1.621 0.024 SGK1 1.538 0.043 PTPN22 5.650 0.004 EMB 6.455 0.010 ENDOU 1.762 0.032 HERPUD1 1.514 0.028 NOS2 5.529 0.029 MZB1 6.425 0.003 FMN2 0.662 0.018 ITIH4 1.512 0.041 SAMSN1 5.493 0.008 RGS1 6.326 0.002 FST 0.444 0.038 MAB21L3 1.508 0.039 HGF 5.489 0.019 PLA2G7 6.181 0.002 GCNT4 0.617 0.021 CDON 0.640 0.018 ADAM8 5.488 0.008 PTPN22 6.110 0.002 HERC6 0.489 0.043 PCID2 0.637 0.015 CD226 5.408 0.046 CD69 6.055 0.001 INPP5J 0.583 0.011 GPN1 0.629 0.006 PLA2G7 5.056 0.009 CD226 5.948 0.013 KRT10 0.604 0.028 PCDH12 0.624 0.022 TREM2 4.998 0.020 CD84 5.868 0.001 MAGIX 1.651 0.009 PLSCR4 0.600 0.029 ARHGAP15 4.952 0.006 TLR7 5.732 0.001 MYOM1 1.957 0.043 HLCS 0.597 0.012 CD84 4.803 0.005 ITK 5.598 0.001 NDUFS7 1.533 0.032 WDFY3 0.595 0.042 RGS1 4.766 0.017 MEF2C 5.570 0.046 NR4A1 1.697 0.038 CROCC 0.581 0.046 ATP8B4 4.740 0.017 NCEH1 5.542 0.004 PAF1 1.677 0.001 ANKRD16 0.563 0.042 ITGAX 4.657 0.005 MMP3 5.409 0.047 RMDN2 0.658 0.013 TRIT1 0.561 0.023 ICOS 4.490 0.012 PLA2G2D 5.407 0.003 SLC22A2 1.590 0.011 EPHB3 0.546 0.030 MPEG1 4.428 0.012 ATP8B4 5.397 0.003 STAB2 1.627 0.008 FBXO36 0.542 0.034 IL2RG 4.421 0.002 LY9 5.372 0.001 STK38L 0.524 0.028 GPIHBP1 0.524 0.045 ITK 4.335 0.005 ARHGEF6 5.317 0.019 TDRP 1.504 0.038 ZMYM4 0.512 0.042 KCNN4 4.270 0.035 ARHGAP15 5.265 0.010 TGM2 0.586 0.028 GJA1 0.477 0.044 DPEP2 4.253 0.008 KCNN4 5.191 0.007 TRIM7 1.791 0.028 HMGCS1 0.467 0.042 CD3G 4.152 0.016 MERTK 5.190 0.002 TXLNB 2.230 0.050 MEST 0.399 0.008 KLRD1 4.119 0.008 TM6SF1 5.149 0.003 USP14 0.640 0.032 Supplementary Table 6. Functional annotation of overexpressed genes only in the combination group compared to placebo. Category Term Count Nominal p value Bonferroni p Fold change KEGG pathway hsa04662:B cell receptor signaling pathway 11 2.21E-05 0.0049 5.57 KEGG pathway hsa04750:Inflammatory mediator regulation of TRP channels 12 9.75E-05 0.0215 4.28 KEGG pathway hsa04660:T cell receptor signaling pathway 12 0.0001 0.0258 4.19 KEGG pathway hsa04142:Lysosome 13 0.0002 0.0345 3.75 KEGG pathway hsa04062:Chemokine signaling pathway 17 7.23E-05 0.0160 3.19 UP keywords Lysosome 18 5.74E-05 0.0177 3.19 GO term CC direct GO:0005887~integral component of plasma membrane 60 9.05E-06 0.0030 1.81 GO term CC direct GO:0005886~plasma membrane 132 4.74E-05 0.0157 1.37 UP keywords Cytoplasm 141 3.21E-05 0.0099 1.36 UP keywords Membrane 215 9.62E-08 2.99E-05 1.34 ANNEX B 253 Supplementary Table 7. GSEA results. Top pathways enriched in lenvatinib vs placebo (A), anti-PD1 vs placebo (B), anti-PD1 vs placebo (C) and combination vs anti-PD1 (D). A. Lenvatinib vs Placebo NAME SIZE ES NES p-val FDR ON NIVOLUMAB UP 131 0.657 2.964 0.000 0.000 HALLMARK ALLOGRAFT REJECTION 174 0.620 2.860 0.000 0.000 IMMUNE INFILTRATE 91 0.660 2.808 0.000 0.000 T CELL 95 0.619 2.626 0.000 0.000 KEGG PRIMARY IMMUNODEFICIENCY 31 0.722 2.478 0.000 0.000 GO ADAPTIVE IMMUNE RESPONSE 183 0.529 2.467 0.000 0.000 GO ACTIN MYOSIN FILAMENT SLIDING 38 0.685 2.466 0.000 0.000 GO CELLULAR DEFENSE RESPONSE 43 0.661 2.457 0.000 0.000 HALLMARK INTERFERON GAMMA RESPONSE 174 0.523 2.422 0.000 0.000 GO REGULATION OF ADAPTIVE IMMUNE RESPONSE 100 0.568 2.416 0.000 0.000 HALLMARK INFLAMMATORY RESPONSE 191 0.505 2.406 0.000 0.000 GO STRUCTURAL CONSTITUENT OF MUSCLE 40 0.672 2.395 0.000 0.000 GO POSITIVE REGULATION OF LEUKOCYTE MEDIATED IMMUNITY 69 0.593 2.376 0.000 0.000 T CD4 40 0.645 2.373 0.000 0.000 T CD8 43 0.638 2.370 0.000 0.000 KEGG RIBOSOME 74 0.569 2.346 0.000 0.000 GO POSITIVE REGULATION OF LYMPHOCYTE MEDIATED IMMUNITY 57 0.594 2.338 0.000 0.000 GO CHEMOKINE MEDIATED SIGNALING PATHWAY 45 0.628 2.322 0.000 0.000 BIOCARTA NO2IL12 PATHWAY 17 0.783 2.307 0.000 0.000 GO REGULATION OF LEUKOCYTE MEDIATED IMMUNITY 126 0.520 2.306 0.000 0.001 NIVOLUMAB (MOLECULAR) RESISTANT MELANOMA DN 353 0.459 2.292 0.000 0.000 GO REGULATION OF INTERFERON GAMMA PRODUCTION 80 0.548 2.282 0.000 0.001 GO REGULATION OF LYMPHOCYTE MEDIATED IMMUNITY 90 0.541 2.272 0.000 0.001 HALLMARK TNFA SIGNALING VIA NFKB 187 0.481 2.267 0.000 0.000 GO RESPONSE TO INTERFERON GAMMA 94 0.534 2.266 0.000 0.001 GO MYELOID LEUKOCYTE ACTIVATION 87 0.543 2.261 0.000 0.001 BIOCARTA CTLA4 PATHWAY 16 0.786 2.257 0.000 0.001 GO LEUKOCYTE ACTIVATION 356 0.442 2.255 0.000 0.001 HALLMARK INTERFERON ALPHA RESPONSE 83 0.543 2.255 0.000 0.000 T CD8 EXHAUSTED 68 0.560 2.243 0.000 0.000 GO MYOFILAMENT 24 0.702 2.224 0.000 0.001 GO CELLULAR RESPONSE TO INTERFERON GAMMA 74 0.545 2.221 0.000 0.001 GO POSITIVE REGULATION OF ADAPTIVE IMMUNE RESPONSE 62 0.570 2.211 0.000 0.001 GO IMMUNOLOGICAL SYNAPSE 31 0.645 2.210 0.000 0.001 GO B CELL RECEPTOR SIGNALING PATHWAY 31 0.652 2.203 0.000 0.001 BIOCARTA NKT PATHWAY 27 0.656 2.196 0.000 0.001 GO INNATE IMMUNE RESPONSE 390 0.427 2.189 0.000 0.001 ENDOTHELIAL 179 -0.525 -2.200 0.000 0.000 STROMA 89 -0.552 -2.111 0.000 0.000 CAF 189 -0.356 -1.514 0.001 0.029 KEGG VALINE LEUCINE AND ISOLEUCINE DEGRADATION 41 -0.697 -2.318 0.000 0.000 KEGG PROPANOATE METABOLISM 30 -0.675 -2.101 0.000 0.000 KEGG BUTANOATE METABOLISM 31 -0.622 -1.921 0.000 0.005 KEGG PEROXISOME 77 -0.476 -1.782 0.000 0.030 KEGG GLYCINE SERINE AND THREONINE METABOLISM 29 -0.577 -1.770 0.007 0.027 KEGG BETA ALANINE METABOLISM 22 -0.614 -1.743 0.002 0.032 KEGG HISTIDINE METABOLISM 26 -0.572 -1.730 0.003 0.032 KEGG CITRATE CYCLE TCA CYCLE 28 -0.561 -1.696 0.005 0.039 NIVOLUMAB (MOLECULAR) RESISTANT MELANOMA UP 178 -0.367 -1.541 0.000 0.012 ANTI-PD-1 RESISTANT MELANOMA UP 421 -0.319 -1.451 0.000 0.019 ANNEX B 254 B. Anti-PD1 vs Placebo NAME SIZE ES NES p-val FDR IMMUNE INFILTRATE 91 0.842 3.168 0.000 0.000 T CELL 95 0.816 3.087 0.000 0.000 ON NIVOLUMAB UP 131 0.749 2.942 0.000 0.000 GO ADAPTIVE IMMUNE RESPONSE 294 0.682 2.876 0.000 0.000 NIVOLUMAB (MOLECULAR) RESISTANT MELANOMA DN 353 0.658 2.823 0.000 0.000 GO B CELL RECEPTOR SIGNALING PATHWAY 49 0.809 2.731 0.000 0.000 GO T CELL ACTIVATION 389 0.626 2.714 0.000 0.000 HALLMARK INTERFERON GAMMA RESPONSE 175 0.674 2.704 0.000 0.000 GO REGULATION OF ANTIGEN RECEPTOR MEDIATED SIGNALING PATHWAY 53 0.788 2.668 0.000 0.000 GO POSITIVE REGULATION OF CELL ACTIVATION 270 0.635 2.659 0.000 0.000 GO ANTIGEN RECEPTOR MEDIATED SIGNALING PATHWAY 203 0.647 2.650 0.000 0.000 GO POSITIVE REGULATION OF LEUKOCYTE CELL CELL ADHESION 183 0.653 2.641 0.000 0.000 GO LYMPHOCYTE MEDIATED IMMUNITY 182 0.646 2.636 0.000 0.000 GO LEUKOCYTE CELL CELL ADHESION 286 0.621 2.621 0.000 0.000 GO IMMUNE RESPONSE REGULATING CELL SURFACE RECEPTOR SIGNALING PATHWAY 322 0.612 2.614 0.000 0.000 GO REGULATION OF LEUKOCYTE MEDIATED IMMUNITY 147 0.655 2.604 0.000 0.000 GO T CELL DIFFERENTIATION 213 0.626 2.601 0.000 0.000 GO LEUKOCYTE MEDIATED CYTOTOXICITY 71 0.737 2.601 0.000 0.000 GO IMMUNE RESPONSE REGULATING SIGNALING PATHWAY 468 0.598 2.600 0.000 0.000 HALLMARK INFLAMMATORY RESPONSE 192 0.633 2.595 0.000 0.000 GO INTERFERON GAMMA PRODUCTION 94 0.689 2.593 0.000 0.000 GO LYMPHOCYTE DIFFERENTIATION 306 0.611 2.592 0.000 0.000 GO ADAPTIVE IMMUNE RESPONSE BASED ON SOMATIC RECOMBINATION OF IMMUNE RECEPTORS BUILT FROM IMMUNOGLOBULIN SUPERFAMILY DOMAINS 195 0.629 2.579 0.000 0.000 GO LEUKOCYTE PROLIFERATION 235 0.618 2.578 0.000 0.000 GO B CELL ACTIVATION 203 0.627 2.567 0.000 0.000 GO CELL KILLING 92 0.685 2.561 0.000 0.000 GO REGULATION OF LYMPHOCYTE MEDIATED IMMUNITY 104 0.673 2.561 0.000 0.000 GO REGULATION OF T CELL ACTIVATION 267 0.612 2.555 0.000 0.000 GO CELL CHEMOTAXIS 235 0.611 2.552 0.000 0.000 GO POSITIVE REGULATION OF LYMPHOCYTE ACTIVATION 222 0.618 2.548 0.000 0.000 HALLMARK INTERFERON ALPHA RESPONSE 85 0.692 2.546 0.000 0.000 GO REGULATION OF IMMUNE EFFECTOR PROCESS 312 0.599 2.545 0.000 0.000 KEGG T CELL RECEPTOR SIGNALING PATHWAY 107 0.664 2.539 0.000 0.000 GO LEUKOCYTE MIGRATION 356 0.592 2.537 0.000 0.000 GO LEUKOCYTE CHEMOTAXIS 167 0.634 2.536 0.000 0.000 GO POSITIVE REGULATION OF CYTOKINE PRODUCTION 387 0.584 2.535 0.000 0.000 GO REGULATION OF B CELL RECEPTOR SIGNALING PATHWAY 24 0.852 2.529 0.000 0.000 GO T CELL ACTIVATION INVOLVED IN IMMUNE RESPONSE 76 0.700 2.526 0.000 0.000 GO T CELL DIFFERENTIATION INVOLVED IN IMMUNE RESPONSE 60 0.690 2.402 0.000 0.000 GO INTERLEUKIN 4 PRODUCTION 32 0.772 2.402 0.000 0.000 GO ALPHA BETA T CELL PROLIFERATION 28 0.696 2.112 0.000 0.000 KEGG APOPTOSIS 77 0.576 2.107 0.000 0.000 GO TRANSFORMING GROWTH FACTOR BETA PRODUCTION 35 0.515 1.627 0.014 0.037 HALLMARK P53 PATHWAY 191 0.376 1.529 0.002 0.014 HALLMARK APICAL JUNCTION 190 0.365 1.473 0.004 0.026 HALLMARK COAGULATION 127 0.362 1.406 0.034 0.049 HALLMARK MYC TARGETS V1 192 -0.307 -1.491 0.000 0.037 ANNEX B 255 HALLMARK MYOGENESIS 196 -0.332 -1.640 0.000 0.010 HALLMARK E2F TARGETS 191 -0.377 -1.824 0.000 0.002 CELL CYCLE G2 M 52 -0.542 -2.144 0.000 0.000 C. Combination vs Placebo NAME SIZE ES NES p-val FDR IMMUNE INFILTRATE 91 0.845 3.097 0.000 0.000 T CELL 95 0.804 2.980 0.000 0.000 MACROPHAGE 350 0.668 2.926 0.000 0.000 NIVOLUMAB (MOLECULAR) RESISTANT MELANOMA DN 353 0.672 2.926 0.000 0.000 ON NIVOLUMAB UP 131 0.726 2.796 0.000 0.000 GO ADAPTIVE IMMUNE RESPONSE 294 0.646 2.751 0.000 0.000 GO B CELL RECEPTOR SIGNALING PATHWAY 49 0.822 2.696 0.000 0.000 GO ANTIGEN RECEPTOR MEDIATED SIGNALING PATHWAY 203 0.653 2.679 0.000 0.000 GO POSITIVE REGULATION OF CELL ACTIVATION 270 0.634 2.672 0.000 0.000 HALLMARK ALLOGRAFT REJECTION 175 0.656 2.648 0.000 0.000 GO B CELL ACTIVATION 203 0.638 2.617 0.000 0.000 GO T CELL ACTIVATION 389 0.590 2.570 0.000 0.000 GO POSITIVE REGULATION OF LEUKOCYTE CELL CELL ADHESION 183 0.629 2.567 0.000 0.000 GO REGULATION OF CELL ACTIVATION 464 0.579 2.564 0.000 0.000 KEGG NATURAL KILLER CELL MEDIATED CYTOTOXICITY 88 0.696 2.562 0.000 0.000 GO LYMPHOCYTE DIFFERENTIATION 306 0.601 2.554 0.000 0.000 GO B CELL DIFFERENTIATION 106 0.677 2.535 0.000 0.000 GO REGULATION OF ANTIGEN RECEPTOR MEDIATED SIGNALING PATHWAY 53 0.745 2.529 0.000 0.000 GO POSITIVE REGULATION OF LYMPHOCYTE ACTIVATION 222 0.607 2.524 0.000 0.000 GO REGULATION OF LYMPHOCYTE ACTIVATION 354 0.577 2.521 0.000 0.000 GO IMMUNE RESPONSE REGULATING CELL SURFACE RECEPTOR SIGNALING PATHWAY 322 0.587 2.520 0.000 0.000 KEGG B CELL RECEPTOR SIGNALING PATHWAY 72 0.716 2.518 0.000 0.000 GO B CELL PROLIFERATION 75 0.703 2.504 0.000 0.000 GO SPECIFIC GRANULE MEMBRANE 85 0.682 2.488 0.000 0.000 GO T CELL DIFFERENTIATION 213 0.600 2.479 0.000 0.000 KEGG T CELL RECEPTOR SIGNALING PATHWAY 107 0.652 2.471 0.000 0.000 GO IMMUNE RESPONSE REGULATING SIGNALING PATHWAY 468 0.556 2.460 0.000 0.000 GO REGULATION OF B CELL DIFFERENTIATION 26 0.824 2.458 0.000 0.000 BIOCARTA DC PATHWAY 16 0.846 2.169 0.000 0.000 HALLMARK APOPTOSIS 153 0.381 1.522 0.004 0.014 ENDOTHELIAL 179 -0.329 -1.465 0.000 0.037 HALLMARK WNT BETA CATENIN SIGNALING 40 -0.441 -1.540 0.024 0.023 GO CELL CYCLE G2 M PHASE TRANSITION 249 -0.422 -1.962 0.000 0.008 KEGG CELL CYCLE 121 -0.519 -2.169 0.000 0.000 GO CHROMOSOMAL REGION 305 -0.488 -2.293 0.000 0.000 KEGG MATURITY ONSET DIABETES OF THE YOUNG 23 -0.763 -2.299 0.000 0.000 GO CHROMOSOME SEPARATION 83 -0.581 -2.306 0.000 0.000 GO REGULATION OF CHROMOSOME SEPARATION 58 -0.629 -2.327 0.000 0.000 GO NUCLEAR CHROMOSOME SEGREGATION 224 -0.515 -2.346 0.000 0.000 GO REGULATION OF UBIQUITIN PROTEIN LIGASE ACTIVITY 21 -0.812 -2.358 0.000 0.000 GO METAPHASE ANAPHASE TRANSITION OF CELL CYCLE 54 -0.643 -2.366 0.000 0.000 GO CHROMATIN REMODELING AT CENTROMERE 29 -0.746 -2.371 0.000 0.000 GO REGULATION OF CHROMOSOME SEGREGATION 97 -0.596 -2.395 0.000 0.000 GO NEGATIVE REGULATION OF CHROMOSOME SEGREGATION 42 -0.688 -2.401 0.000 0.000 GO SISTER CHROMATID SEGREGATION 165 -0.568 -2.480 0.000 0.000 GO MITOTIC NUCLEAR DIVISION 259 -0.533 -2.504 0.000 0.000 GO CHROMOSOME CENTROMERIC REGION 177 -0.566 -2.506 0.000 0.000 GO MITOTIC SISTER CHROMATID SEGREGATION 135 -0.614 -2.597 0.000 0.000 ANNEX B 256 GO CONDENSED CHROMOSOME CENTROMERIC REGION 102 -0.653 -2.692 0.000 0.000 HALLMARK G2M CHECKPOINT 184 -0.638 -2.850 0.000 0.000 CELL CYCLE G2 M 52 -0.852 -3.053 0.000 0.000 Combination vs Anti-PD1 NAME SIZE ES NES p-val FDR IMMUNE INFILTRATE 91 0.610 2.446 0.000 0.000 KEGG CARDIAC MUSCLE CONTRACTION 72 0.560 2.171 0.000 0.000 B CELL 64 0.570 2.152 0.000 0.000 HALLMARK OXIDATIVE PHOSPHORYLATION 174 0.480 2.108 0.000 0.000 B CELLS 23 0.666 2.009 0.000 0.003 NIVOLUMAB (MOLECULAR) RESISTANT MELANOMA DN 353 0.420 2.005 0.000 0.000 KEGG PRIMARY IMMUNODEFICIENCY 31 0.595 1.915 0.000 0.014 KEGG DILATED CARDIOMYOPATHY 88 0.482 1.903 0.000 0.011 KEGG HYPERTROPHIC CARDIOMYOPATHY HCM 82 0.481 1.897 0.000 0.010 KEGG B CELL RECEPTOR SIGNALING PATHWAY 71 0.468 1.773 0.002 0.034 KEGG VIRAL MYOCARDITIS 49 0.491 1.764 0.002 0.032 T CELL 95 0.377 1.514 0.009 0.058 T CD4 40 0.426 1.478 0.037 0.058 TFH CELLS 25 0.374 1.150 0.276 0.792 CD8 T CELLS 24 0.365 1.128 0.272 0.573 NK CELLS 20 0.331 0.950 0.534 0.794 T HELPER CELLS 19 0.290 0.843 0.673 0.710 TH1 CELLS 24 -0.300 -0.874 0.650 0.729 NEUTROPHILS 17 -0.351 -0.957 0.518 0.775 TCM CELLS 27 -0.436 -1.293 0.150 0.309 IDC 22 -0.458 -1.305 0.127 0.391 HALLMARK APOPTOSIS 152 -0.359 -1.484 0.002 0.021 HALLMARK TNFA SIGNALING VIA NFKB 187 -0.358 -1.526 0.002 0.016 ENDOTHELIAL 179 -0.372 -1.584 0.002 0.024 HALLMARK ESTROGEN RESPONSE EARLY 192 -0.373 -1.586 0.000 0.011 TREG CELLS 147 -0.386 -1.594 0.000 0.093 STROMA 89 -0.416 -1.599 0.005 0.028 HALLMARK WNT BETA CATENIN SIGNALING 40 -0.489 -1.599 0.015 0.010 HALLMARK KRAS SIGNALING UP 183 -0.387 -1.634 0.000 0.008 HALLMARK MYC TARGETS V2 58 -0.478 -1.699 0.002 0.004 KEGG ECM RECEPTOR INTERACTION 81 -0.453 -1.705 0.000 0.034 IFN SIGNATURE 26 -0.572 -1.719 0.004 0.059 RESPONDERS ON NIVOLUMAB DN 437 -0.370 -1.725 0.000 0.000 KEGG ADHERENS JUNCTION 68 -0.475 -1.739 0.000 0.028 HALLMARK APICAL SURFACE 42 -0.531 -1.759 0.007 0.002 HALLMARK ESTROGEN RESPONSE LATE 189 -0.433 -1.822 0.000 0.001 KEGG AXON GUIDANCE 125 -0.457 -1.840 0.000 0.008 KEGG PATHWAYS IN CANCER 309 -0.411 -1.844 0.000 0.008 HALLMARK TGF BETA SIGNALING 54 -0.557 -1.955 0.000 0.000 KEGG P53 SIGNALING PATHWAY 58 -0.552 -1.961 0.002 0.003 BIOCARTA CELLCYCLE PATHWAY 23 -0.681 -1.969 0.000 0.019 HALLMARK INTERFERON GAMMA RESPONSE 174 -0.480 -2.022 0.000 0.000 KEGG TGF BETA SIGNALING PATHWAY 83 -0.546 -2.069 0.000 0.001 KEGG CELL CYCLE 121 -0.573 -2.276 0.000 0.000 HALLMARK INTERFERON ALPHA RESPONSE 83 -0.606 -2.315 0.000 0.000 HALLMARK MITOTIC SPINDLE 193 -0.588 -2.494 0.000 0.000 CELL CYCLE G1 S 42 -0.745 -2.501 0.000 0.000 HALLMARK E2F TARGETS 190 -0.647 -2.752 0.000 0.000 CELL CYCLE G2 M 52 -0.820 -2.870 0.000 0.000 HALLMARK G2M CHECKPOINT 187 -0.681 -2.879 0.000 0.000 ANNEX B 257 Supplementary Table 8. Combination rescue gene signature. Gene Group p-value (Bonferroni) FC Score (Log2FC) MZB1 Upregulation by combination 0.003 6.425 2.684 MEF2C Upregulation by combination 0.046 5.570 2.478 NCEH1 Upregulation by combination 0.004 5.542 2.470 MMP3 Upregulation by combination 0.047 5.409 2.435 ARHGEF6 Upregulation by combination 0.019 5.317 2.411 TM6SF1 Upregulation by combination 0.003 5.149 2.364 TSC22D3 Upregulation by combination 0.000 5.005 2.323 GPNMB Upregulation by combination 0.007 4.718 2.238 DNASE1L1 Upregulation by combination 0.002 4.481 2.164 FOLR2 Upregulation by combination 0.011 4.372 2.128 CD79B Upregulation by combination 0.014 4.290 2.101 RGS2 Upregulation by combination 0.036 4.244 2.085 MYLIP Upregulation by combination 0.014 4.192 2.068 GPR65 Upregulation by combination 0.001 4.087 2.031 B3GALNT1 Upregulation by combination 0.019 4.002 2.001 HVCN1 Upregulation by combination 0.004 3.928 1.974 HCST Upregulation by combination 0.002 3.846 1.943 PRRG1 Upregulation by combination 0.044 3.775 1.916 SLPI Upregulation by combination 0.011 3.770 1.915 SCARB2 Upregulation by combination 0.016 3.747 1.906 GDE1 Upregulation by combination 0.026 3.671 1.876 TRPV2 Upregulation by combination 0.006 3.578 1.839 RAB29 Upregulation by combination 0.003 3.556 1.830 GNGT2 Upregulation by combination 0.012 3.520 1.816 LY6K Upregulation by combination 0.006 3.508 1.811 SLC9A9 Upregulation by combination 0.002 3.494 1.805 FGD4 Upregulation by combination 0.008 3.477 1.798 ADSSL1 Upregulation by combination 0.018 3.472 1.796 HMOX1 Upregulation by combination 0.003 3.449 1.786 FLI1 Upregulation by combination 0.009 3.432 1.779 LST1 Upregulation by combination 0.035 3.415 1.772 CD8A Upregulation by combination 0.030 3.403 1.767 PTPN6 Upregulation by combination 0.027 3.351 1.744 RASSF4 Upregulation by combination 0.002 3.347 1.743 TNFAIP2 Upregulation by combination 0.030 3.343 1.741 DOCK10 Upregulation by combination 0.020 3.321 1.731 SLC43A2 Upregulation by combination 0.010 3.298 1.722 MMP27 Upregulation by combination 0.011 3.283 1.715 KBTBD11 Upregulation by combination 0.010 3.245 1.698 GPM6B Upregulation by combination 0.049 3.196 1.676 PIP4K2A Upregulation by combination 0.007 3.179 1.669 CD38 Upregulation by combination 0.006 3.171 1.665 RINL Upregulation by combination 0.008 3.160 1.660 ANNEX B 258 CELF2 Upregulation by combination 0.012 3.106 1.635 IKZF3 Upregulation by combination 0.007 3.105 1.635 SNX29 Upregulation by combination 0.014 3.079 1.623 MAFB Upregulation by combination 0.010 3.078 1.622 SYT11 Upregulation by combination 0.003 3.075 1.620 CNR2 Upregulation by combination 0.002 3.071 1.619 LIPA Upregulation by combination 0.011 3.045 1.606 RSPO3 Upregulation by combination 0.004 3.042 1.605 HFE Upregulation by combination 0.029 3.027 1.598 GTPBP10 Upregulation by combination 0.038 3.015 1.592 EPHA7 Downregulation by combination 0.029 0.121 -3.048 LECT2 Downregulation by combination 0.043 0.121 -3.047 AMBP Downregulation by combination 0.011 0.126 -2.986 CPN1 Downregulation by combination 0.045 0.163 -2.614 SLC16A4 Downregulation by combination 0.021 0.185 -2.433 TM4SF20 Downregulation by combination 0.028 0.185 -2.432 TTR Downregulation by combination 0.041 0.186 -2.425 ABCC2 Downregulation by combination 0.033 0.194 -2.368 LMO7 Downregulation by combination 0.005 0.195 -2.360 THSD4 Downregulation by combination 0.005 0.206 -2.277 ELOVL7 Downregulation by combination 0.002 0.214 -2.223 PLCH1 Downregulation by combination 0.021 0.217 -2.206 EPB41L4B Downregulation by combination 0.023 0.221 -2.177 SERPIND1 Downregulation by combination 0.044 0.222 -2.173 ITGA3 Downregulation by combination 0.031 0.224 -2.156 ROBO1 Downregulation by combination 0.022 0.228 -2.134 PKHD1 Downregulation by combination 0.042 0.229 -2.124 ZIC5 Downregulation by combination 0.016 0.231 -2.116 TSPAN8 Downregulation by combination 0.041 0.236 -2.082 CATSPERD Downregulation by combination 0.009 0.239 -2.067 ALDOB Downregulation by combination 0.029 0.247 -2.020 ACSL3 Downregulation by combination 0.045 0.251 -1.997 RAMP3 Downregulation by combination 0.029 0.251 -1.993 TROAP Downregulation by combination 0.015 0.252 -1.988 TMED6 Downregulation by combination 0.025 0.257 -1.962 CCDC136 Downregulation by combination 0.007 0.258 -1.953 SQLE Downregulation by combination 0.044 0.262 -1.930 AXIN2 Downregulation by combination 0.037 0.268 -1.900 RALGAPA2 Downregulation by combination 0.033 0.269 -1.894 KHDRBS3 Downregulation by combination 0.025 0.270 -1.889 SEMA6A Downregulation by combination 0.011 0.272 -1.877 ALAD Downregulation by combination 0.008 0.273 -1.872 EFNA1 Downregulation by combination 0.011 0.274 -1.866 SOX9 Downregulation by combination 0.010 0.275 -1.861 ENPP2 Downregulation by combination 0.050 0.276 -1.858 IRS1 Downregulation by combination 0.011 0.278 -1.844 ANNEX B 259 SLC39A8 Downregulation by combination 0.042 0.281 -1.834 TSPAN12 Downregulation by combination 0.027 0.283 -1.819 LAMB3 Downregulation by combination 0.049 0.284 -1.819 TPX2 Downregulation by combination 0.037 0.286 -1.807 GRB7 Downregulation by combination 0.016 0.289 -1.789 STRA6 Downregulation by combination 0.035 0.291 -1.782 BMP4 Downregulation by combination 0.008 0.293 -1.773 PTPN3 Downregulation by combination 0.009 0.300 -1.735 CLDN4 Downregulation by combination 0.009 0.305 -1.712 PRKG2 Downregulation by combination 0.025 0.306 -1.711 FAM222A Downregulation by combination 0.040 0.307 -1.705 DIAPH3 Downregulation by combination 0.036 0.309 -1.694 BAIAP2L1 Downregulation by combination 0.018 0.310 -1.691 PSRC1 Downregulation by combination 0.022 0.310 -1.690 DHCR24 Downregulation by combination 0.015 0.311 -1.686 RCAN2 Downregulation by combination 0.018 0.311 -1.684 NR5A2 Downregulation by combination 0.006 0.314 -1.671 TNNT2 Downregulation by combination 0.027 0.315 -1.668 DSP Downregulation by combination 0.033 0.316 -1.663 PLCD3 Downregulation by combination 0.027 0.317 -1.659 PALMD Downregulation by combination 0.002 0.317 -1.656 MYBL2 Downregulation by combination 0.046 0.319 -1.648 ZC3H12C Downregulation by combination 0.047 0.320 -1.646 PTPRF Downregulation by combination 0.024 0.320 -1.645 BCAR1 Downregulation by combination 0.018 0.323 -1.630 UBE2C Downregulation by combination 0.042 0.325 -1.624 PAX2 Downregulation by combination 0.022 0.325 -1.620 DUSP4 Downregulation by combination 0.027 0.326 -1.618 GPR87 Downregulation by combination 0.029 0.326 -1.616 SULT1C2 Downregulation by combination 0.047 0.328 -1.606 MIA2 Downregulation by combination 0.014 0.329 -1.605 CHSY3 Downregulation by combination 0.029 0.330 -1.601 CHRNB1 Downregulation by combination 0.037 0.330 -1.599 RASSF6 Downregulation by combination 0.009 0.330 -1.598 PHGDH Downregulation by combination 0.032 0.333 -1.589 CCND1 Downregulation by combination 0.043 0.333 -1.586 ANNEX B 260 Supplementary Table 9. Correlation between classification according to the combination rescue signature and clinic-pathological parameters. Immune class (n=55) Combination responder class (n=52) Rest (n=121) p-value Median age (IQR) 67 (62-73) 67 (63-71) 65 (61-71) 0.25 Gender, male (%) 42 (76) 37 (71) 101 (84) 0.12 Etiology (%) Hepatitis C 25 (46) 26 (52) 52 (44) 0.65 Hepatitis B 14 (26) 7 (14) 27 (23) 0.32 Alcohol 7 (13) 4 (8) 22 (19) 0.21 Others 8 (15) 13 (26) 17 (14) 0.19 Tumor size, median (IQR) 3.5 (2.5-6.5) 3.5 (2.8-5) 3.5 (2.5-7) 0.97 Multiple nodules (%) Absent 43 (80) 37 (71) 89 (74) 0.61 Present 11 (20) 15 (29) 31 (26) Vascular invasion (%) Absent 40 (74) 33 (65) 73 (61) 0.27 Present (micro or macro) 14 (26) 18 (35) 46 (39) Satellites (%) Absent 42 (78) 35 (67) 87 (72) 0.48 Present 12 (22) 17 (33) 34 (28) Degree of tumor differentiation (%) Well 10 (24) 4 (9) 19 (19) 0.14 Moderate/poor 31 (76) 41 (91) 79 (81) Bilirubin, >1 mg/dL (%) 29 (54) 23 (45) 62 (52) 0.63 Albumin, <3.5 g/L (%) 5 (9) 6 (12) 14 (12 0.92 Platelet count, <100,000/mm3 (%) 10 (19) 9 (18) 23 (19) 1.00 AFP, >100 mg/dL (%) 19 (35) 9 (18) 26 (22) 0.09 Events (%) Recurrence 37 (70) 35 (67) 81 (70) 0.96 Death 28 (51) 34 (65) 71 (60) 0.30 ANNEX B 261 Supplementary Table 10. Publicly available gene sets and signatures used in the study. Name Reference HCC classification Sia HCC immune class Sia D, et al. Gastroenterology 2017;153: 812-826 Chiang classification Chiang D, et al. Cancer Res 2008;68:6779–6788 Hoshida classification Hoshida Y, et al. Cancer Res 2009;69:7385–7392 Boyault classification Boyault S, et al. Hepatology 2007;45:42–52 TGFß pathway Fibroblast response to TGFß Calon A, et al. Cancer Cell 2012;22:571-84 T cell response to TGFß Calon A, et al. Cancer Cell 2012;22:571-84 Tumor composition Immune cells Bindea Bindea G, et al. Immunity 2013;39:782-95 T cells Bindea Bindea G, et al. Immunity 2013;39:782-95 Th1 cells Bindea Bindea G, et al. Immunity 2013;39:782-95 Th2 cells Bindea Bindea G, et al. Immunity 2013;39:782-95 Regulatory T cells Bindea Bindea G, et al. Immunity 2013;39:782-95 iDC Bindea Bindea G, et al. Immunity 2013;39:782-95 B cells Bindea Bindea G, et al. Immunity 2013;39:782-95 Activated dendritic cells Charoentong Charoentong P, et al. Cell Rep 2017;18:248-262 Cytotoxic T cells Jerby Arnon Jerby-Arnon L, et al. Cell 2018;175:984-997 Endothelial cells Jerby Arnon Jerby-Arnon L, et al. Cell 2018;175:984-997 M1/M2 Coates Coates PJ, et al. Cancer Res 2008;68:450-6 Response to ICI IFN signature Ribas Ribas A, et al. J Clin Oncol 33, 2015 (suppl; abstr 3001) Anti-PD1 resistant melanoma Riaz N, et al. Cell 2017;171:934-949 Nivolumab (molecular) resistant melanoma Riaz N, et al. Cell 2017;171:934-949 On nivolumab Riaz N, et al. Cell 2017;171:934-949 Responders on nivolumab Riaz N, et al. Cell 2017;171:934-949