Introduction

All molecular processes in the eukaryotic cell that involve transactions with genomic DNA have to deal with chromatin—the complex of DNA and histone proteins. In general, wrap** DNA around histone octamers to form nucleosomes acts as a barrier for processes that require DNA as a template such as transcription, replication and DNA repair. Indeed, histones need to be removed at least temporarily to allow access to the DNA for these processes (for reviews see [72, 131, 153]). At the same time, nucleosomes also act as an important docking platform for a myriad of factors that regulate DNA transactions. A critical layer of control for these processes is the post-translational modification of histone proteins. Histone post-translational modifications (PTMs) can directly influence the structure of chromatin, for example by neutralizing the positive charge of histone proteins, or act as a docking site for so called chromatin ‘reader’ proteins (for reviews see [24, 141, 176, 223]). Already early on it has been hypothesized that histone PTMs directly control chromatin processes and that a specific combination of histone PTMs can be viewed as a ‘code’ that specifies the function of a DNA region [177]. In the past decade, genome-wide maps for many histone PTMs in different cell types have been generated, which has indeed confirmed that specific histone PTMs often correlate with a DNA element that is in a particular state (i.e. active promoter, active enhancer, site of DNA damage) [10, 174]. However, correlation does not necessarily mean causation. One of the objectives in understanding chromatin is therefore to determine the functional roles of histone PTMs.

A large part of our understanding of chromatin modifying (‘writer’) enzymes comes from research on histone methyltransferases in budding yeast. In Saccharomyces cerevisiae, the SET domain-containing protein Set2 methylates histone H3 on lysine 36 (H3K36). This site is located at the base of the N-terminal tail of H3 and can be either mono-, di-, or trimethylated by Set2 [178]. H3K36 methylation was one of the first PTMs for which a clear function was found in transcription. Initial reports found Set2 to be enriched on the coding sequences of active genes suggesting a role in transcription elongation [164, 130]. In the absence of SETD2/H3K36me3, MORF4L1 might still be able to target PTB to exons by binding to H3K36me2, given its in vitro binding properties. However, because H3K36me2 is only enriched at the 5’ end of active gene bodies in humans [57], H3K36me2 is most likely not sufficient to promote PTB localization to the same target exons as H3K36me3.

Another factor that influences splicing through SETD2 activity is Zinc Finger MYND-Type Containing 11 (ZMYND11), which is a chromatin reader protein that has been reported to associate with spliceosome components and regulate intron retention [70]. ZMYND11 binds specifically to the replication-independent ‘gap-filler’ histone variant H3.3 when trimethylated at K36 (H3.3K36me3) through its tandem PHD-, bromo-, and PWWP (PBP) domain, which specifically recognizes both H3K36me3 as well as the S31 residue that is unique to H3.3 compared to canonical H3 [70, 209, 210, 221]. H3.3K36M also inhibits the H3K36 dimethyltransferase NSD2 (also known as MMSET) in a manner analogous to SETD2 and therefore also affects global H3K36me2 levels [57].

In chondroblastoma, H3.3K36M has been reported to contribute to tumor development by promoting colony formation, and inhibiting apoptosis and chondrocyte differentiation [57, 128]. Interestingly, H3.3K36M also leads to a redistribution of H3K27me3 away from developmentally silenced genes to regions normally enriched in H3K36me3, which may contribute to the derepression of PRC2 target genes that prevent differentiation [128]. In vitro, H3K36me3 nucleosomes are a poor substrate for PRC2 [216], providing a possible explanation for this redistribution of H3K27me3 in cells lacking H3K36me3. PRC2-mediated silencing is not strongly impaired in Drosophila H3K36R cells [140] so future studies are required to determine if defective PRC2-mediated gene repression is a general feature of cells lacking H3K36me3.

Other mutations that have been found in H3.3 (predominantly in H3F3A) are H3.3G34R/V in osteosarcoma [8] and glioblastoma [165, 202], [179], and H3.3G34W/L in giant cell tumor of the bone (GCTB; [8]. Unlike H3.3K36M, these H3.3G34 mutations do not affect global H3K36me3 levels and only inhibit SETD2 in cis [111]. H3.3G34 is involved in the binding of SETD2 to H3, fitting in a small pocket in the SET domain, and any substitution with a bulky amino acid residue blocks the interaction [209, 210, 221]. It is currently not clear how H3.3G34 mutations are mechanistically involved in tumor development. A similarity between H3.3K36M and H3.3G34 mutations is that they tend to occur in tumors found in children and young adults. In glioblastoma, H3.3G34R is associated with a developmental expression signature that includes genes that block differentiation (such as self-renewal genes; [18], which is reminiscent of H3.3K36M-mutant chondroblastoma [128]. A common theme might therefore be that H3.3 mutations that affect SETD2 function maintain precursor cells in a pluripotent state and block differentiation during development, although the exact mechanisms behind this process are likely different given that global H3K36me3 levels are unaltered in H3.3G34 mutant cells. In addition, H3.3G34W cells derived from GCTB patients were reported to have mRNA splicing defects which may contribute to tumorigenesis [121]. Expression of H3.3G34R (but not H3.3G34V), as well as H3.3K36M, also impairs DNA repair through HR [127, 154, 208] indicating that genomic instability might be a shared pathway contributing to tumorigenesis in H3.3 mutant and SETD2 deficient cells.

To summarize, mutations in H3.3 genes and SETD2 are found in distinct cancer types and both affect SETD2 activity but can do so in distinct manners (Fig. 6). Besides H3K36me3, SETD2 loss affects the methylation of non-histone substrates of SETD2, whereas H3.3 mutations can also affect other modification states of H3K36 (such as H3K36ac, H3K36me1, and -me2) either by directly preventing the modification or by in-trans inhibition of other writers such as NSD2.

Fig. 6
figure 6

Connection between mutations in SETD2 and histone H3.3 affecting H3K36 methylation. A SETD2 allele can be lost by one-copy deletion of the short arm of chromosome 3, which is a frequent event in ccRCC. Mono-allelic loss of SETD2 does not appear to affect global H3K36me3 levels in ccRCC indicating that SETD2 is haplo-sufficient for H3K36me3. However, α-tubulin methylation on K40 is lost upon mono-allelic SETD2 inactivation suggesting that SETD2-mediated maintenance of genomic stability through tubulin methylation might be frequently perturbed in ccRCC. Mutations in H3.3 (found in chondroblastoma, brain tumors and osteosarcoma among others) can either inhibit SETD2 in-cis (H3.3G34R/V) or in-trans (H3.3K36M). It is currently unknown to what extent H3.3K36M affects the methylation of non-histone substrates of SETD2 such as α-tubulin

Full size image

Overexpression of the H3K9me3/H3K36me3-demethylase KDM4A leads to genome instability

SETD2’s function can also be affected in cancer through misregulated expression of the demethylase KDM4A [69], which demethylates both H3K36me3 and H3K9me3 [100]. KDM4A is either deleted or overexpressed (predominantly through gene amplification) in several types of cancer including lung, breast, ovarian, and head and neck cancer [11, 20, 133]. KDM4A promotes S-phase progression and regulates replication timing [19, 20] and its function in cancer is best understood in the context of its overexpression (for a detailed review please see [107, 214]. Interestingly, KDM4A overexpression results in the (extrachromosomal) amplification of chromosome 1q12, through site-specific re-replication during a single cell cycle, and 1q12 amplification also correlates with KDM4A overexpression in tumor samples [20]. Chromosome 1q12 gain mediated by KDM4A overexpression depends on the catalytic activity of KDM4A, and can also be induced by expressing either H3.3K9M or H3.3K36M [20]. This suggests that both H3K36me3 and H3K9me3 prevent 1q12 re-replication during S-phase, although the exact mechanism remains to be determined. KDM4A also has a negative role in DNA repair, inhibiting the recruitment of 53BP1 to DNA damage sites, but this role is independent of its catalytic activity [132] suggesting it is not related to SETD2’s positive role in DNA repair. It is currently unknown if KDM4A further contributes to tumorigenesis by antagonizing SETD2’s role in promoting mRNA processing. Furthermore, it remains to be determined if KDM4A can demethylate non-histone substrates of SETD2 such as α-tubulin, which might be an additional pathway through which KDM4A negatively regulates genome stability. Thus, even though it is clear that both SETD2 loss and KDM4A overexpression perturb H3K36me3 levels, it is not yet entirely clear how much overlap there is in the mechanisms contributing to tumorigenesis in SETD2 deficient versus KDM4A overexpressing tumors.

Conclusion

The SETD2/Set2 enzymes have been a prime example of how different approaches in different model systems have led the way in unravelling the molecular role and regulation of a chromatin modifying system associated with RNA polymerase moving along genes. SETD2/Set2 is a key enzyme in the cell involved in a broad range of genome-associated processes. The studies on SETD2/Set2 and H3K36 methylation showcase that teasing apart the various functions requires perturbing not only the writer itself, but also the other domains of the enzymes, the opposing demethylation activity, and the substrate lysine. Moving beyond chromatin, the story of SETD2 emphasizes the importance of knowledge about non-histone substrates of so called ‘epigenetic writers’. An emerging theme in chromatin biology is that non-catalytic functions or activities towards non-histone substrates of epigenetic enzymes need to be considered to fully understand the physiological role of these enzymes. Looking forward, it will be important to develop tools to identify substrates of SETD2 in different cellular contexts in an unbiased way and to perturb SETD2 functions in a substrate-specific manner, e.g. by mutation of a substrate lysine, or by isolation of separation-of-function mutations. With the current advances in genome engineering and proteomics it can be expected that more SETD2 surprises will be discovered and that the function of SETD2 in normal cells and in disease will be further unraveled at a molecular level.