The Tousled-like kinases regulate genome and epigenome stability: implications in development and disease

The Tousled-like kinases (TLKs) are an evolutionarily conserved family of serine–threonine kinases that have been implicated in DNA replication, DNA repair, transcription, chromatin structure, viral latency, cell cycle checkpoint control and chromosomal stability in various organisms. The functions of the TLKs appear to depend largely on their ability to regulate the H3/H4 histone chaperone ASF1, although numerous TLK substrates have been proposed. Over the last few years, a clearer picture of TLK function has emerged through the identification of new partners, the definition of specific roles in development and the elucidation of their structural and biochemical properties. In addition, the TLKs have been clearly linked to human disease; both TLK1 and TLK2 are frequently amplified in human cancers and TLK2 mutations have been identified in patients with neurodevelopmental disorders characterized by intellectual disability (ID), autism spectrum disorder (ASD) and microcephaly. A better understanding of the substrates, regulation and diverse roles of the TLKs is needed to understand their functions in neurodevelopment and determine if they are viable targets for cancer therapy. In this review, we will summarize current knowledge of TLK biology and its potential implications in development and disease.


The Tousled-like kinases regulate genome and epigenome stability: implications in development and disease Summary
The Tousled-like kinases (TLKs) are an evolutionarily conserved family of serine-threonine kinases that have been implicated in DNA replication, DNA repair, transcription, chromatin structure, viral latency, cell cycle checkpoint control and chromosomal stability in various organisms. The functions of the TLKs appear to depend largely on their ability to regulate the H3/H4 histone chaperone ASF1, although numerous TLK substrates have been proposed.
Over the last few years, a clearer picture of TLK function has emerged through the identification of new partners, the definition of specific roles in development and the elucidation of their structural and biochemical properties. In addition, the TLKs have been clearly linked to human disease; both TLK1 and TLK2 are frequently amplified in human cancers and TLK2 mutations have been identified in patients with neurodevelopmental disorders characterized by intellectual disability (ID), autism spectrum disorder (ASD) and microcephaly. A better understanding of the substrates, regulation and diverse roles of the TLKs is needed to understand their functions in neurodevelopment and determine if they are viable targets for cancer therapy. In this review, we will summarize current knowledge of TLK biology and its potential implications in development and disease.

Identification of the Tousled kinase and Tousled-like kinases
The Tousled (TSL) kinase and Tousled-like kinases (TLKs) belong to a distinct branch of nuclear Ser-Thr kinases that are absent in yeast but appear to be constitutively expressed in most cells and tissues from plants and animals. TSL was first identified in Arabidopsis thaliana where mutations in the single TSL gene led to pleiotropic defects in morphogenesis, including delays in flowering time and leaf development [1]. Subsequent analysis of TSL showed that its deficiency led to cell cycle abnormalities but its mRNA and protein expression levels were stable throughout the cell cycle [2]. Apart from defects in plant development that could result from proliferation defects [1], A. thaliana TSL was reported to directly affect transcriptional gene silencing. Loss of TSL resulted in reduced H3K9me2, associated with heterochromatin, at reactivated gene loci, while no changes were observed in the mitosis-associated phosphorylation of histone H3 on Serine-10 (H3S10) [3]. Additionally, TSL mutants were hypersensitive to UV-B radiation and methyl methanesulfonate (MMS) and exhibited compromised siRNA-mediated silencing, indicating that TSL loss may lead to transcriptional deregulation and impaired DNA damage repair [4,3].
Following the identification of TSL in A. thaliana, Tousled-like kinases (TLKs) were identified in numerous organisms ( Figure 1A). This included Trypanosoma brucei, which encodes 2 distinct TLKs, as well as, Drosophila melanogaster and Caenorhabditis elegans that, like A. thaliana, encode a single TLK gene. The consequences of TLK depletion have been analyzed during development in each case, further implicating TLK activity in DNA repair, DNA replication, transcription and mitosis [5][6][7][8]. At the organismal level, the TLK-1 gene in C. elegans and Tlk gene in D. melanogaster are essential for viability, as their loss in rapidly dividing cells during early development led to severe chromatin abnormalities, proliferation defects and lethality [5,7]. In C. elegans, the major defects identified were transcriptional, reflected by reduced phosphorylation of RNA polymerase II and histone H3 Ser10, a marker of mitosis [5,6]. In D. melanogaster, Tlk mutation caused early arrest during embryonic development. This is likely due in part to defective chromatin maintenance, as the deleterious effects in eye development observed following expression of a kinase-dead TLK mutant could be rescued by the overexpression of the histone H3-H4 chaperone ASF1, now the most clearly defined substrate of the TLKs [7,9,10]. Apart from being required for proliferation during development, Tlk was also identified in a D. melanogaster RNAi screen for cell migration, identifying a requirement for TLK in JAK/STAT activation and the motility of polar cells [11].
Like trypanosomes, mammals encode 2 TLK genes, TLK1 and TLK2, located in different chromosomes (in humans 2q31.1 and 17q23.2, respectively). TLK1 and TLK2 share 84% identity at the amino acid (aa) level and 96% identity in the kinase domain ( Figure 1B) [12][13][14][15][16]. Each gene is reported to encode several isoforms of unknown relevance and an additional translationally regulated form of TLK1, termed TLK1B, has been characterized [17]. Consistent with data from other organisms, existing evidence suggests important roles for mammalian TLK1 and TLK2 in DNA replication, DNA repair, transcription and organismal development (proposed roles of TLKs in DNA repair were recently reviewed in [18] and will be further summarized here).

Domain organization and structural features of the TLKs
Both TLK1 and TLK2 exhibit the highest levels of activity during S-phase and are regulated by cell cycle checkpoint signaling in response to DNA damage [2,12,9,19,20]. Both TSL and TLKs have a C-terminal protein kinase catalytic domain and a large N-terminal regulatory domain defined by putative coiled coil (CC) domains ( Figure 1B) [12,1]. Analysis of the A.
thaliana TSL protein sequence first revealed the predicted CC regions, as well as three consensus nuclear localization signal (NLS) sequences in the N-terminus and in vitro assays showed that the CC regions of the TSL protein were required for oligomerization and full kinase activity [1].
The first X-ray crystal structure of a TLK family kinase domain was recently solved in complex with ATPγS, providing insight into the mode of TLK2 activation and a tool for modeling small molecule inhibitors [21]. In vitro analysis of TLK2 activity, as well as the identification of key autophosphorylation sites critical for its activity, indicated that TLK2 is activated through cis-autophosphorylation events in the kinase domain ( Figure 1C). These autophosphorylation events trigger a conformational change allowing the trans-and cisphosphorylation of sites in the N-terminal CC domains and C-tail, similar to what has been described for members of the closely related AGC kinase family, and suggesting that TLKs do not require an activating phosphorylation like members of the CDK family [21,22].
Biochemical studies indicated that monomeric TLK2 cannot achieve full activation and its dimerization and subsequent oligomerization are crucial for maximal activity ( Figure 1C). Activated TLK2 dimers led to the appearance of higher order oligomers, which are dependent on autophosphorylation in the loops joining the CC domains. Thus, oligomerization may not only trigger activation but also enzymatic activity by means of recruiting additional TLK2 molecules. It is worth noting that oligomeric constructs are capable of phosphorylating the substrate ASF1a while the kinase domain alone, lacking the N-terminal CC containing segment, cannot [21]. Thus, it is anticipated that either the CC region or its role in dimer/oligomerization are required for substrate recognition.
Most of the autophosphorylation sites identified in TLK2 are found in the loops joining the CC domains, suggesting that they are potentially important regulatory domains in vivo [21].
Numerous phosphorylation sites have been identified in the extreme N-terminus of TLK1 and TLK2, which was removed to promote solubility in the structural and biochemical study of the human TLK2 protein [21,23]. This N-terminal region also contains the NLS according to sequence analysis and consistent with the N-terminal mutants lacking the first 160 aa failing to localize to the nucleus [21,24]. Whether the N-terminal phosphorylation sites represent autophosphorylation or sites of regulation by other kinases remains to be determined. Many additional autophosphorylation sites were identified in the extreme C-terminus of TLK2, downstream of the kinase domain. These included sites analogous to those in TLK1 that were found to be negative regulatory sites targeted for phosphorylation by CHK1 in response to DNA damage [19,20,23]. TLK1 and TLK2 can homo and heterodimerize and this is critically dependent on the first coiled-coil (CC1) domain [21]. Therefore, apart from homodimerization within TLK2 molecules, heterodimerization with TLK1 appears to represent an additional layer of regulation ( Figure 1C). Whether the substrate selectivity, activity or regulatory inputs of TLK homodimers and heterodimers differ will be an important question to resolve in future studies.

Interactors and substrates of the Tousled-like kinases
The histone H3/H4 chaperone ASF1 has been identified as an interactor of TLKs in all organisms where it has been examined. In yeast and C. elegans, the C-terminus of ASF1 is highly acidic, possibly favoring its interaction with histones ( Figure 2). In D. melanogaster and mammalian homologs, the C-terminus is instead rich in Ser and Thr residues, which are TLK-dependent phosphorylation sites in mammalian ASF1a and ASF1b have been mapped and functionally investigated ( Figure 2). ASF1a is phosphorylated by TLKs during DNA replication on its C-terminal tail residues S166, S175, S192 and S199, while ASF1b is modified on residues S169 and S198 [10]. Although the precise mechanisms by which ASF1 is regulated remain unclear, in D. melanogaster, TLK phosphorylation of ASF1 controls its stability, while in vertebrates, TLK1-mediated phosphorylation of several sites on the Cterminal tail of ASF1 promote its binding affinity for the histone H3/H4 heterodimer [10,26].
These data suggested that TLK-mediated phosphorylation of ASF1a and ASF1b may promote histone delivery to downstream histone chaperones, such as CAF1 and HIRA, for replication-coupled and replication-independent chromatin assembly, respectively ( Figure 3) [27]. This is further supported by the observation that the de novo deposition of both H3.1 and H3.3 was impaired by TLK depletion [28].
Beyond ASF1, few well-validated substrates of TLKs have been described, although the proposed substrate spectrum of TLK1 consists of more than 150 proteins [29]. TLK1 has been shown to phosphorylate RAD9, a component of the RAD9-RAD1-HUS1 (9-1-1) alternative clamp loader that regulates DNA damage-induced CHK1 activation ( Figure 3) [30 -35]. Consistent with the TLKs being transiently inhibited by the checkpoint, a DNA damage-induced loss of phosphorylation was observed in RAD9 at S328 [33] and a mild reduction in RAD9 pS328 was reported following Bleomycin treatment of TLK1/2-depleted cells [25]. The interaction between TLK1 and RAD9 and its phosphorylation on T355 was enhanced by DNA damage and implicated in the checkpoint response, and S328 phosphorylation of RAD9 has been shown to regulate its subcellular localization [30,35]. Our own analysis of TLK2 using quantitative IP-mass spectrometry and BioID, as well as other approaches, failed to detect RAD9 as a TLK2 interactor, potentially due to differences in experimental conditions or specificity with TLK1 [36,29]. Future proteomic experiments following acute stress will be needed to fully understand the influence of TLKs on RAD9 and the details of the interaction in vivo.
The phosphorylation of H3S10 is required for chromosome condensation and is a widely used marker of mitosis [37]. TLKs have been proposed to mediate H3S10 phosphorylation in various organisms. In human cells, TLK1B phosphorylated H3S10 in vitro and TLK1B was capable of complementing a yeast mutant strain lacking the major yeast H3 kinase, Aurora B/IPL1 [13]. In C. elegans, TLK-1 promoted the Aurora B-mediated phosphorylation of H3S10 in a kinase independent manner, indicating that the influence on H3S10 is likely an indirect effect ( Figure 3) [6]. Tlk mutants in D. melanogaster displayed reduced levels of phosphorylated histone H3 [7], potentially the result of fewer cells entering mitosis, an effect that has also been observed in human cell lines and trypanosomes [7,28,8]. Therefore, while TLK depletion clearly influences cell cycle progression, it remains unclear whether H3S10 is a direct target of TLK activity in vivo.
Overexpression of a NEK1 T141A mutant influenced cell cycle checkpoint regulation in response to damage [29]. Given that NEK1 activity has been linked to ATR activation, these results may represent further regulatory integration into the checkpoint response [40].
One of the most consistent interactors we and others have identified, aside from ASF1, is LC8-type 1 and 2 (DYNLL1 and DYNLL2) which were originally identified as components of the axonemal dynein motor protein complex [41,36,42,43]. LC8 associates with multiple interaction partners independently of its motor protein functions, including the kinase NEK9 and proteins involved in double-strand break (DSB) repair, such as the MRE11-RAD50-NBS1 complex (MRN), ATMIN/ASCIZ and 53BP1 [44,41,[45][46][47]. LC8 has been proposed to play a general role as a multimerization hub that organizes or stabilizes different protein complexes ( Figure 3) [41]. The TLK2 binding domain for LC8 either lies within the CC1 domain or it requires heterodimerization with TLK1, as its binding is lost in CC1 deletion mutants of TLK2 that have impaired TLK1 interactions [21]. However, LC8 does not appear to be a TLK substrate in vitro, as purified LC8 is not phosphorylated by active TLK2 [36].
Whether loss of the LC8 interaction impairs other TLK2 interactions or modifies its activity in vivo will be interesting to determine, given that phosphorylation of ASF1a by the TLK2-ΔCC1 mutant is reduced in vitro and LC8 is implicated in the regulation of the NEK9 kinase that is also regulated via dimerization [21,48].
Another interactor of TLKs identified through unbiased proteomic analysis is the heterochromatin-associated protein RIF1 ( Figure 3) [36]. RIF1 was first identified in yeast as a regulator of telomere length [49] while in mammals, RIF1 does not have a telomere specific role but has been implicated in the control of DNA repair and the regulation of replication timing, the latter being a conserved function across species [50][51][52]. Immunoprecipitationmass spectrometry (IP-MS) analysis of ASF1a in human MCF-7 cells revealed extensive coverage of both TLK1/2 and RIF1, suggesting the RIF1-TLK interaction may occur through ASF1a [53]. Interestingly, both TLK1 and TLK2, but not ASF1a, were identified in RIF1associated complexes from IP-MS experiments in mouse ES cells [54]. As RIF1 acts in part through the recruitment of the PP1 phosphatase, it is tempting to speculate that it may act as a direct TLK regulator but further experiments are needed to address this possibility.
Finally, TLKs have been identified as proximal interactors of several key DNA replication and repair factors. While this does not indicate that they are necessarily direct interactors, it provides some insight into the cellular environment of the TLKs and is consistent with the proposition that despite being mostly nuclear soluble proteins they also localize to the vicinity of replication factories [10]. TLK1 and TLK2 were identified in a proteomics screen performed using BioID-PCNA in synchronized S-phase cells, but not in asynchronous cells, regardless of DNA damaging treatments [55]. Moreover, TLK1, TLK2, RIF1 and ASF1a were identified as proximal interactors of 53BP1 using APEX2 labeling, whereas TLK2 was identified as a proximal interactor of the DNA damage response factor MDC1, which also interacts with ASF1a [56,57].

Roles of the TLKs in mammalian development
The characterization of TLK1 and TLK2 knockout cells and mice suggested that TLK1 and TLK2 play largely redundant roles in genome maintenance [36] consistent with the fact that they form heterocomplexes [9,58]. Despite the fact that both kinases appear to be largely redundant in homeostatic somatic tissues, Tlk2-deficient mice perished during late embryogenesis due to placental failure, while no placental defects were observed in the absence of TLK1 [36]. Reduced ASF1 phosphorylation and impaired expression of placental markers were observed in Tlk2-deficient placental tissue. The observation that the knockout of ASF1a in mice leads to lethality by E9.5 [59], notably earlier than that observed with TLK2 deletion, is consistent with the incomplete effect of TLK2 loss on ASF1 phosphorylation we observed [36]. As mRNA levels of Tlk1 and Tlk2 in the placenta were similar but relative TLK1 protein levels were strongly reduced in placenta compared with embryonic tissue, it is possible that a translational or post-translational mechanism regulates TLK1 protein levels in this tissue. While several E3 ubiquitin ligases have been identified as TLK1 and TLK2 interactors, additional work is needed to validate these interactions and define their potential roles in post-translational regulation of TLKs in different tissues [60][61][62].
Bypass of placental development allowed the generation of Tlk2 null animals that did not show any overt phenotypes in homeostatic conditions, similar to Tlk1 null mice [36]. While mice appeared anatomically normal, the conditional knockout of Tlk1 and Tlk2 in stromal fibroblasts caused increased mammary gland branching and epithelial hyperproliferation [63]. Available data does not rule out specific functions for TLK1 or TLK2 in the context of particular cell types or in response to stress and suggests that they can largely compensate for each other. Moreover, neither gene acts as a strong tumor suppressor, despite their implication in DNA repair and genome instability. The conditional mouse models will no doubt play an important role in interrogating potential tissue and cell type specific roles of TLK1 and TLK2 in future work.

TLK activity is required for genome and epigenome stability
Human TLK expression is constitutive at both the mRNA and protein levels throughout the cell cycle, similar to what had been observed for TSL. However, TLK1/2 protein kinase activity oscillates during the cell cycle, peaking in S-phase [12]. Inhibition of DNA replication with numerous agents inhibited TLK kinase activity in a DNA damage response (DDR)-dependent manner, indicating that TLK activity is linked to ongoing DNA replication and regulated by the checkpoint [12,19,20]. TLK activity has been consistently implicated in the maintenance of genome stability across species but exactly how and why TLK activity is integrated into the DDR and how it promotes genome integrity remains to be fully elucidated.
In yeast, ASF1 interacts directly with and is a substrate of the checkpoint kinase Rad53 and this interaction has been implicated in genome stability and cell cycle checkpoint recovery [64,65]. In human cells, ASF1 does not appear to interact directly with the checkpoint kinases, but is instead regulated by them indirectly through the TLKs. CHK1 phosphorylates TLK1 at the C-terminal S695 residue, reversibly inhibiting its activity [19,20]. This CHK1dependent modulation of TLK1 potentially coordinates global ASF1 histone-binding capacity with the checkpoint response and allows chromatin restructuring during DNA repair. The attenuation of TLK1 activity upon checkpoint activation is transient and TLK1 was identified again as a direct CHK1 target using an analog sensitive CHK1 allele, although the physical interaction of CHK1 with TLK1B or TLK2 has not been observed in proteomics studies [36,29,66]. This may reflect the fact that most studies have been performed in asynchronous cells in the absence of DNA damage or that the interaction is too transient to be detected by the methods used. Although similar phosphorylated sites exist in the C-terminus of TLK2, it remains unclear if TLK2 is directly regulated by CHK1 or through heterodimerization with C-terminally phosphorylated TLK1.
Multiple lines of evidence have implicated TLK activity in the control of cell cycle progression, both in asynchronous cells and cells with DNA damage. The previously described interaction of TLK1 with RAD9, that has multiple roles in the response to DNA damage, has been linked to G2/M checkpoint recovery [30, 32,33], although other reports implicated TLK2, but not TLK1, in G2/M checkpoint recovery through ASF1a-mediated transcriptional regulation [67] (Figure 3). Conversely, TLK2 overexpression was also shown to impair the DNA damage-induced G2/M checkpoint in human cancer cells and Tlk overexpression prolonged G2 in D. melanogaster independently of its activity [68,69]. These may explain the observation that overexpression of TLK1B in mouse cells confers enhanced resistance to ionizing radiation, that is more toxic in highly proliferative cells [13]. Thus, the regulation of TLK levels and activity is required for normal cell cycle progression, likely involving numerous interactions and kinase-dependent and independent functions.
In addition to cell cycle progression, TLK activity has been implicated in chromosome segregation. The overexpression of a dominant negative form of TLK1B caused chromosome missegregation in mouse cells [70] and TLK1 was proposed to regulate myosin II regulatory light chain (MRLC) during mitosis to maintain correct chromosome segregation [71].
Further, mitotic defects have been observed in worms, flies and trypanosomes, although whether these are the result of under-replicated DNA, cell cycle progression defects or bona fide mitotic roles of TLK activity remains to be determined [8,7,6].
In cancer cells, depletion of TLK activity impaired nucleosome assembly and led to replication-coupled ssDNA accumulation and fork stalling [28], a state known as replication stress [72]. DSBs accumulated over time, inducing the DDR and provoking p53 activation and G1 arrest. TLK-depleted cells were sensitized to treatment with checkpoint kinase or Poly(ADP-ribose) Polymerase (PARP) inhibitors, indicating that ATR/CHK1 and PARP activity were crucial to prevent the rapid collapse of forks arrested due to chromatin assembly defects [28]. New histones incorporated during DNA replication can be identified by the lack of H4K20 methylation (H4K20me0) that acts to signal the recruitment of the TONSL-MMS22L homologous recombination complex [73]. Long term (72 hours) depletion of CAF1 and ASF1 impaired the recruitment of TONSL-MMS22L to DNA double-strand breaks [25]. Impaired nucleosome assembly in TLK-deficient cells would be predicted to have a similar effect on TONSL-MMS22L recruitment that could sensitize replication forks to collapse due to ATR/CHK1-dependent suppression of new origin firing and RPA exhaustion [74].
In addition to its well-established roles in transcription and replication, ASF1a was recently shown to regulate DSB repair. ASF1a is phosphorylated in a DNA damage-dependent manner by DNA-PKcs at S192, a residue previously identified to be a TLK target, indicating that multiple signaling pathways converge on the C-terminal tail of ASF1 ( Figure 2) [10,25].
This phosphorylation event promotes MMS22L-TONSL chromatin loading and subsequent recruitment of the RAD51 recombinase to promote homologous recombination (HR)mediated DNA repair. Paradoxically, ASF1a has also been proposed to suppress HR and promote non-homologous end-joining (NHEJ), a competing DSB repair pathway, through its ability to interact with MDC1 and promote the recruitment of several key factors, including the ubiquitin ligases RNF8 and RNF168 [57]. Considering that DNA damage-induced checkpoint activation transiently inhibits TLK activity [19] and ASF1a pS192 occurs after DNA damage, even in TLK1/2-depleted cells [25], it seems unlikely that TLK activity plays a major regulatory role. Nevertheless, as TLK activity plays an important role in genome stability and interacts with RIF1, a key regulator of DNA repair pathway choice, this possibility warrants further investigation [75].
Despite available evidence indicating that ASF1 is the primary TLK target in metazoans [9], it is notable that TLK depletion does not simply phenocopy ASF1 loss. Depletion of total ASF1 reduced replication fork speed and caused a strong S-phase arrest without causing RPA accumulation or DDR checkpoint activation [28,76]. Treatment with the deoxyribonucleotide reductase (RNR) inhibitor hydroxyurea (HU), that generates a robust DDR in cycling cells, including ssDNA/RPA accumulation and DNA breaks, failed to do so in ASF1-depleted cells [76]. These data, as well as the fact that the inhibition of DNA replication in TLK-depleted cells ameliorated the levels or replication stress and DNA damage, indicated that ongoing DNA replication underlies much of the genomic instability that accumulates in TLK-depleted cells, consistent with its peak activity in S-phase [28]. This may reflect that ASF1 has additional functions that are independent of TLK regulation, such as its interactions with the MCM2-7 helicase and role in histone recycling, and/or that additional TLK substrates influence the phenotypes [76,77].
In addition to promoting genome stability, several lines of evidence indicate that TLKs play an important role in epigenome maintenance [5,3,67]. TLK2 was identified in an siRNA screen for proteins required to maintain the silencing of Kaposi's sarcoma-associated herpesvirus (KSHV) and TLK1 depletion also resulted in reactivation of Epstein-Barr virus (EBV) [78]. In addition to exogenous viruses, the impaired de novo nucleosome deposition we observed in TLK-depleted cells [28] would be predicted to have potential consequences for epigenome maintenance that is required for cell identity programs, as well as the silencing of non-coding regions, such as endogenous viruses and telomeres, where ASF1 has been previously implicated [79,80]. Notably, we observed a strong decrease in H3.3 deposition in TLK-depleted cells and this replacement variant of H3 plays a key role in heterochromatin formation at telomeres and other transcriptionally silent genomic regions, as well as in promoters of developmentally regulated genes [81,82,28,83,84].

Roles of TLK activity in human disease
Despite their implication in replication stress, genome and epigenome instability and hyperproliferation, all of which play key roles in cancer etiology, both TLK1 and TLK2 are often maintained or amplified in human cancers and few recurrent mutations or copy number losses have been identified [28,85,86,63]. This pattern is reminiscent to that of ATR, and to a lesser extent CHK1, that is required by many cancers to tolerate increased replication stress [87]. These and other data have suggested that TLK activity may be a promising target to explore in cancer treatment. In addition, recent genetic studies have now implicated TLK2 mutations in several neurodevelopmental disorders, including intellectual disability (ID) and autism spectrum disorder (ASD), often associated with microcephaly [88][89][90], raising new interest in understanding the precise developmental roles of TLK activity.

TLK activity as a therapeutic target in cancer
In breast cancer, amplification of the 17q23 region, that contains several candidate oncogenes, including TLK2, occurs in more than 40% of tumors. In addition, several TLK2 single nucleotide polymorphisms (SNP) of unknown function, including rs733025 and rs2245092, were significantly associated with breast cancer risk and hormone receptorpositive breast tumors [91,92]. TLK2 was amplified in luminal ER+ breast cancer and was found to be hyper-phosphorylated in proteogenomics studies, potentially indicating increased activity [93,21,28,94]. TLK2 overexpression also correlated with increased chromosomal instability (CIN) in breast cancer [68] and promoted cell invasion and migration, both characteristics associated with metastasis in luminal breast cancer cells [93]. Moreover, a therapeutic effect of TLK2 inhibition or depletion was observed in xenograft models of breast cancer and glioblastoma [93,63].
Analysis of TLK1/2 copy number alterations across pan-cancer genomes showed that TLKs are more frequently maintained or amplified than lost [28]. In addition, high TLK1 and TLK2 expression levels correlated with poor prognosis in several cancer cohorts, including cervical squamous cell carcinoma and endocervical adenocarcinoma (TCGA-cesc) and uveal melanoma (TCGA-uvm). Targeting TLK activity was proposed as a potential therapeutic intervention in prostate cancer and shown to enhance the effects of some chemotherapeutic agents, including ATR/CHK1 and PARP inhibitors, as well as cisplatin, in different cancer types [28, [95][96][97]. How TLK1 or TLK2 expression is misregulated in cancer remains largely unclear. Aside from copy number alterations, miR-16 was shown to regulate TLK1 levels in oral squamous cell carcinoma and the circadian E3 ligase complex was demonstrated to regulate TLK2 stability, suggesting that alterations in these mechanisms could influence TLK levels in some types of cancer [98,62].
In addition to playing direct roles in DNA replication and chromatin maintenance, TLK1 and TLK2 were identified as non-cell autonomous modifiers of RAS pathway signaling in worms and mice [63]. Conditional depletion of TLK1 or TLK2 in mouse mammary fibroblasts caused hyperproliferation of surrounding mammary epithelial cells, indicating that loss of TLK activity leads to cellular crosstalk that may be relevant to their role in cancer and as therapeutic targets. Further supporting an important role for TLK depletion provoking a secretory response, depletion of Tlk in flies influenced cytokine-dependent signaling during cell migration and our recent work has demonstrated that TLK depletion leads to the loss of heterochromatin maintenance, desilencing of repetitive elements, including ERVs, and the activation of an innate immune response that included the secretion of inflammatory cytokines [84,11].
The emerging picture suggests that like ATR-CHK1, some degree of TLK activity is required for cancer cell proliferation and preventing the accumulation of toxic levels of replication stress. This requirement may be elevated in highly proliferative cancer cells and demand the amplification of the TLK-ASF1 pathway to avoid replication stress. Therefore, targeting TLK activity could have a therapeutic benefit in cancer and additionally, it could potentially augment novel selective therapies, including cell cycle checkpoint inhibitors, PARP inhibitors and DNA damaging agents [28]. Several small molecule TLK inhibitors have been reported, although most of these are highly promiscuous or target many non-kinase proteins, thus potent specific inhibitors have yet to be identified [21, 93,99]. Future work exploiting the TLK2 crystal structure as a tool for rational inhibitor design could conceivably identify clinically effective agents for use in cancer treatment [21].

TLK2 mutations in neurodevelopmental disorders
Neurodevelopmental disorders, including ASD and ID, are commonly caused by de novo spontaneous or inherited genetic mutations that affect brain development. A meta-analysis of data from over two thousand patients identified TLK2 as one of ten new candidate genes for ID and other neurodevelopmental disorders, such as ASD and schizophrenia [88]. These patients have de novo loss of function mutations (DNM) and exhibit severe clinical features, such as facial dysmorphisms and microcephaly. Previous studies had identified DNMs in TLK2 in sporadic ASD [100] and schizophrenia [101]. The fact that TLK2 DNM are significantly enriched in ASD was also confirmed by an independent study of a Japanese cohort [90].
A subsequent study involving patients from up to 7 countries characterized 38 unrelated individuals with two affected mothers with heterozygous variants in the TLK2 gene with a distinct neurodevelopmental disorder with a consistent pathological spectrum, including mild developmental delay, behavioral disorders, gastro-intestinal problems and facial dysmorphisms [89]. Mutations in the TLK2 gene include loss-of-function (LOF) variants (4 frameshift, 10 nonsense variants and one balanced translocation resulting in a disruption between TLK2 exon 2 and 3) and missense variants (9 missense and 12 canonical splice-site variants) ( Figure 4A). While TLK1 mutations have not been statistically linked to any distinct neurodevelopmental disorder, DNMs have been reported in isolated cases of neurodevelopmental disorder patients, suggesting that its role in brain development warrants further investigation [102,103,88,104]. It is worth noting that both TLK1 and TLK2 (MIM number 608438 and 608439 respectively) are significantly intolerant for both missense and truncating mutations in healthy individuals, similar to what has been observed in cancer genomes [89,28].
TLK2 mutations were found in heterozygosity, indicating that the neurodevelopmental defects presumably arise due to haploinsufficiency. We have previously speculated that the effects of the ID/ASD related TLK2 mutations could reflect placental defects, although we have not observed any clear placental phenotypes in Tlk2 heterozygous mice and the possibility that TLK1 mutations may underlie similar disorders would argue against this as being the sole cause [36,89]. Four previously described TLK2 ID mutations [88] strongly reduced kinase activity in vitro [21], suggesting that at least some of the TLK2 missense mutations could have a mild dominant negative effect. However, in other cases that involve larger truncations that include the CC1 domain involved in dimerization, haploinsufficiency more likely accounts for the related pathologies.
The neural progenitor population is particularly sensitive to cell cycle delays and DNA damage. Attrition of these cells is one of the major underlying causes of several neurodevelopmental disorders, including Seckel Syndrome, which is characterized by short stature and microcephaly [105], both of which are observed in a number of patients with heterozygous TLK2 mutations [89]. Hypomorphic mutations in both ATR and several MCM components of the replicative helicase have been linked to replication stress and placental defects, in the latter case associated with inflammatory responses [106,107]. However, analysis of Tlk2-deficient murine placentas did not uncover increased DNA damage signaling or proliferative defects, although this cannot be ruled out in humans due to major differences in placental development and gestation time [36].
H3.3 is required for H3K9me3 establishment in telomeres and endogenous retroviral elements (ERVs), as well as for H3K27me3 establishment at promoters of developmentally regulated genes [81][82][83]. Consistent with this, depletion of TLK1 delayed downregulation of pluripotency genes and impaired embryonic stem cell differentiation, suggesting impaired histone-mediated regulation of differentiation programs [108]. Numerous mutations in genes involved in epigenetic maintenance have been identified in ASD, including ATRX, that plays a prominent role in H3.3 deposition in heterochromatin at retrotransposons and telomeres [109,81], as well as KDM5B, KDM5C, SETD5, and DNMT3A [110]. Treatment with the histone deacetylase inhibitor valproic acid (VPA) led to increased ERV expression, and prenatal exposure to VPA has been linked to autism, identifying links between chromatin silencing, placental formation and autism [110][111][112].
An additional, non-exclusive possibility is that TLK activity regulates microexon splicing.
Brain-specific microexon splicing defects have been identified as a possible molecular mechanism underlying idiopathic ASD, given that a significant fraction of autistic brains analyzed by transcriptomic profiling showed misregulation of microexons and reduced levels of regulators of neuronal alternative splicing [113,114].  [116,117].
In future work, it will therefore be interesting to determine whether TLK2 mutations in human patients compromise ERV silencing, impair DNA replication, affect microexon splicing or elicit inflammatory responses that have been associated with ASD and other neurodevelopmental disorders, as well as placental defects in animal models ( Figure 4B) [118][119][120]107,121,122,84].

Conclusions and open questions
Despite the recent advances in our understanding of TLK structure, function and roles in human disease, many open questions remain about its regulation, targets and cell type specific roles. While TLK activity is clearly important for genomic stability and regulated by the cell cycle checkpoint machinery, why is it important to rapidly inhibit TLK following DNA damage? How does reduced TLK activity cause replication stress and how is ssDNA generated at forks stalled by reduced TLK activity? Are other TLK substrates aside from ASF1 relevant to fork progression? Is TLK activity important in post-mitotic cell populations? Structurally, we now have a clearer picture of the kinase domain but many questions remain. What is the structure of the CC domains and their polarity in the context of TLK dimers or multimers? How do they influence subcellular localization, activity or substrate selection? How does the phosphorylation of the C-terminus of TLK1 by CHK1 influence activity and does this apply to TLK2? Finally, as it is clear that TLKs are important during development and even appear to be selected for in cancer, a detailed analysis of their cell type and tissue specific roles will be needed. Is the influence of TLK mutations on neurodevelopment cell autonomous and if so, what cell types are affected and how? Would targeting TLK activity in cancer represent a viable strategy? Addressing these and other questions will further our insight into the important roles of these poorly-understood kinases in genome and epigenome maintenance.

Acknowledgements
We are grateful to members of the Stracker lab and A. Groth for discussions of unpublished data and suggestions. We apologize to those colleagues whose relevant work could not be specifically mentioned due to space constraints.

Competing interests
The authors declare no competing interests.    [10,25,26]. The total number of residues of each protein is shown. In the right panel, sequence identity and similarity between the 1-156 aa of yeast (y), C. elegans (ce) and human (h) ASF1 are displayed as a percentage and were assessed by NCBI Blast (blastp suite-2sequences).