Cellular and molecular functions of SETD2 in the central nervous system

In eukaryotes, genomic DNA is compacted into chromatin by a group of proteins known as histones. These proteins assemble into octamers consisting of pairs of four core canonical histones: H2A, H2B, H3 and H4. There are several variants of each of these histones. For example, mammals have three main H3 variants, the canonical H3.1 and H3.2, and the variant H3.3, which is strongly associated with active genes (for a comprehensive review see Talbert and Henikoff, 2021). DNA wraps ∼1.65 turns around core histone octamers to form nucleosomes, the basic repeating units of chromatin (Luger et al., 1997). Enzymes known as ‘writers’ and ‘erasers’ catalyse the addition and removal, respectively, of posttranslational modifications (PTMs; e.g. methylation, acetylation and phosphorylation) to specific residues on histones, particularly those located in the N-terminal histone tail (Berger, 2002; Biswas and Rao, 2018). Histone modification plays a key role in the regulation of gene expression in eukaryotes. Modifications can influence physical interactions between histones, modifying chromatin structure and therefore accessibility of DNA to the transcriptional machinery. They can also indirectly regulate gene expression through the recruitment of gene-regulatory ‘reader’ proteins. These proteins contain specialised domains that can bind to histone modifications to facilitate a wide range of downstream cellular effects. For a more in-depth overview of histone modifications, we refer readers to a recent review by Ramazi et al. (2020).

Histone modifications are critical for proper central nervous system (CNS) function and development (Park et al., 2022). These include trimethylation of lysine 27 on histone H3 (H3K27me3), which is associated with transcriptional repression, and the acetylation of lysine 14 on histone H3 (H3K14ac), which is associated with active transcription (for an in-depth review see Park et al., 2022). Here, we focus on the trimethylation of lysine 36 on histone H3 (H3K36me3), as recent reports have revealed that this mark plays a key role in regulating diverse cellular processes and is dysregulated in some developmental disorders and a range of cancers. H3K36me3 influences gene expression via several mechanisms, including transcriptional regulation and mRNA processing; for an in-depth review see Sharda and Humphrey (2022). The writer protein responsible for this mark was first identified in Saccharomyces cerevisiae as the Su(var)3-9, Enhancer of Zeste and Trithorax (SET) domain-containing protein Set2 (Strahl et al., 2002). Set2 catalyses the mono-, di- and tri-methylation of lysine 36 on histone H3 (H3K36) (Venkatesh and Workman, 2013). The mammalian paralogue of Set2, SET domain-containing 2 (SETD2), also has methyltransferase activity (Sun et al., 2005). However, SETD2 is primarily responsible for only the trimethylation of H3K36 in vivo, as deletion of the protein in mammalian cells leads to a dramatic reduction in H3K36me3, but no changes in H3K36 monomethylation (H3K36me1) or dimethylation (H3K36me2) levels (Edmunds et al., 2008).

The function of SETD2 extends beyond H3K36me3, however, with studies identifying methylation of additional histone residues and a range of non-histone substrates for the protein (Fig. 1). Through H3K36me3 and these non-histone interactions, SETD2 is implicated in several different cellular processes. Many of these processes are critical for the development and function of the CNS. This is particularly evident in humans, where pathogenic SETD2 variants lead to neurodevelopmental disorders such as Luscan–Lumish syndrome (OMIM 616831). Here, we provide an overview of the cellular functions of SETD2, highlighting recently identified substrates, with a particular focus on its role in CNS. We will also discuss the involvement of SETD2 dysregulation in neurodevelopmental disorders.

Human SETD2 is a 2564-amino-acid protein that contains several functional domains (Fig. 2). These include a SET domain flanked by an ‘associated with SET’ (AWS) domain and a post-SET domain, which together provide methyltransferase activity (Sun et al., 2005). SETD2 also contains a Set2–Rpb1 interaction (SRI) domain, which interacts with RNA polymerase II (RNAPII) (Li et al., 2005), and a SETD2–hnRNP interaction (SHI) domain and a WW domain, which mediate interactions with heterogeneous nuclear ribonucleoprotein L (hnRNP L) and huntingtin (HTT), respectively (Bhattacharya et al., 2021; Faber et al., 1998). Within the SHI domain is a coiled-coil (CC) domain, which likely participates in interactions with hnRNP L (Bhattacharya et al., 2021). All these domains reside in the C-terminal half of the protein. The N-terminal half of the protein is largely unstructured but might regulate protein stability (Bhattacharya and Workman, 2020).

The most well-characterised function of SETD2 is trimethylation of H3K36, and this modification is chiefly added during transcription. The largest subunit of RNAPII contains a C-terminal repeat domain (CTD) that is hyperphosphorylated during active transcription. SETD2 interacts specifically with this phosphorylated form of the RNAPII CTD through its SRI domain (Li et al., 2005; Sun et al., 2005). This allows SETD2 to selectively associate with actively transcribed regions of the genome. As a result, H3K36me3 is generally deposited at the 3′ end of actively transcribed gene bodies and is associated with euchromatin (Bannister et al., 2005). However, one study on mice has shown that H3K36me3 is also present in heterochromatin, although the mechanisms behind this association are largely unknown (Chantalat et al., 2011). H3K36me3 levels are regulated in several ways (highlighted in Box 1).

During transcription, DNA partially unwraps itself from nucleosomes, exposing SETD2 target residues on histone H3. Subsequently, the SETD2 AWS domain binds to the N-terminal region of histone H3 (Liu et al., 2021). A regulatory L_IN loop between the SET and post-SET domains positions H3K36 in the active site of the SET domain, where methyl groups from the enzyme cofactor S-adenosylmethionine (SAM) are attached to H3K36 (Liu et al., 2021; Yang et al., 2016).

H3K36me3 deposited by SETD2 interacts with a large suite of H3K36me3-binding proteins to participate in numerous cellular processes (Table 1). One such process is de novo DNA methylation at genomic sites enriched with H3K36me3. This is performed by the reader protein DNA methyltransferase 3 beta (DNMT3B) (Baubec et al., 2015; Okano et al., 1999). DNA methylation plays an important role in regulating gene expression, predominantly through transcriptional repression. It also aids in repressing cryptic transcription, which is the inappropriate initiation of transcription from intragenic sites of protein-coding genes (Neri et al., 2017). Cryptic transcription can have deleterious effects on a range of cellular processes (for a review see McCauley and Dang, 2021).

The alternative splicing of pre-mRNA is also regulated by SETD2-mediated H3K36me3. The H3K36me3 reader protein MORF-related gene on chromosome 15 (MRG15, also known as MORF4L1) recruits the splice factor polypyrimidine tract-binding protein (PTB) to regulate the alternative splicing of various genes in mammalian cells (Luco et al., 2010). Another splicing protein, zinc finger MYND domain-containing 11 (ZMYND11), facilitates intron retention in human cells by interacting with U5 small nuclear ribonucleoprotein (snRNP) splice factors, and loss of SETD2 disrupts this process (Guo et al., 2014). ZMYND11 specifically interacts with trimethylated lysine 36 on the histone variant H3.3 (H3.3K36me3), a modification also catalysed by SETD2 (Guo et al., 2014). Notably, SETD2-mediated H3K36me3 might also regulate alternative splicing indirectly by regulating the expression of a splice factor. Indeed, H3K36me3 has been demonstrated in mice to directly regulate alternative splicing of the splice factor gene Srsf11, thus influencing downstream splicing (Xu et al., 2021b). Furthermore, there also appears to be a reciprocal effect of splicing on H3K36me3 deposition. The inhibition of pre-mRNA splicing in mammalian cells causes a reduction in SETD2 recruitment and subsequent H3K36me3 deposition (de Almeida et al., 2011; Kim et al., 2011). Indeed, SETD2 itself directly interacts with splice factors, such as hnRNP L, through its SHI domain, facilitating co-transcriptional splicing and influencing several alternative splicing events (Bhattacharya et al., 2021). The interaction with hnRNP L also regulates methyltransferase activity of SETD2. Deleting the SETD2 SHI domain or hnRNP L leads to a reduction in H3K36me3 (Bhattacharya et al., 2021; Yuan et al., 2009). Further mechanistic studies are required to elucidate how exactly hnRNP L and SETD2 influence each other's activity. Several biological functions are regulated by SETD2-dependent alternative splicing, including autophagosome–lysosome fusion (González-Rodríguez et al., 2022) and the inhibition of intestinal tumorigenesis (Yuan et al., 2017).

Other cellular processes regulated by SETD2-catalysed H3K36me3 include N⁶-methyladenosine (m6A) mRNA modification (Huang et al., 2019), DNA damage signalling and repair (Carvalho et al., 2014; Chu et al., 2020; Li et al., 2013; Pfister et al., 2014), and prevention of replication stress (Bleuyard et al., 2017; Pfister et al., 2015). These processes are mediated by several H3K36me3-binding proteins (Table 1).

SETD2 is additionally involved in the regulation of cell size. MicroRNA-mediated SETD2 knockdown in human cells causes an increase in cell size and total protein content accompanied by an increased protein synthesis rate in vitro (Molenaar et al., 2022). The mechanisms behind the observed cell size increases are unclear. SETD2 might indirectly regulate cell size by influencing cell cycle dynamics or by directly controlling protein synthesis rates (Molenaar et al., 2022). Further mechanistic studies are necessary to investigate this. It is also unclear whether this process is dependent on the ability of SETD2 to trimethylate H3K36. Overexpression of the H3K36 demethylase lysine demethylase 4A (KDM4A) or of the oncohistones H3.3K36M and H3.3K36A, which deplete H3K36me3, causes an increase in cell volume; however, despite H3.3K36M overexpression almost completely abolishing H3K36me3 expression, it results in a much smaller increase in cell size than that observed following SETD2 knockdown (Molenaar et al., 2022). This suggests that SETD2-mediated regulation of cell size is not entirely dependent on H3K36me3 and highlights that the function of SETD2 is not limited to H3K36me3 deposition.

Intriguingly, SETD2 has recently been identified to catalyse the trimethylation of lysine 14 on histone H3 (H3K14me3). Unlike the well-characterised H3K36me3 modification, which is highly enriched at sites of active transcription, the cellular functions of H3K14me3 and its association with transcription are largely unclear (Zhao et al., 2018). SETD2-deficient HeLa cells have reduced overall H3K14me3 levels, and in vitro methylation assays have demonstrated that SETD2, as well as the histone methyltransferase SUV39H1, is capable of trimethylating lysine 14 on histone H3 (Zhu et al., 2021). SETD2-mediated H3K14me3 levels have been observed to increase in HeLa cells under replication stress. In this context, the H3K14me3 modification has been shown to recruit a group of replication proteins known as the replication protein A (RPA) complex, which in turn activates ataxia telangiectasia related (ATR) protein, facilitating the restart of stalled replication forks and increasing cell survival compared to that of SETD2-deficient cells (Zhu et al., 2021). It remains to be seen whether SETD2 deposits H3K14me3 in vivo. The discovery that SETD2 is capable of methylating more than one histone residue raises key questions. Are there even more histone residues that SETD2 is capable of methylating? Indeed, a recent study has suggested that SETD2 might also monomethylate lysine 37 on histone H3 (H3K37me1). Deletion of SETD2 in human cells causes a reduction in H3K37me1 levels (Santos-Rosa et al., 2021). The histone modification is associated with the prevention of inappropriate replication in yeast, but it is currently unclear if this function is conserved in mammals (Santos-Rosa et al., 2021). What mechanisms govern the choice of histone substrate by SETD2? It is possible that cell signalling pathways activated by replication stress could alter SETD2 activity, potentially through the addition of posttranslational modifications (Zhu et al., 2021). Further mechanistic studies on SETD2-mediated H3K14 trimethylation might address these questions.

The methyltransferase activity of SETD2 is not limited to histones, as demonstrated by recent studies that have identified several non-histone substrates that are methylated by SETD2. One such substrate is α-tubulin. Dimers of α- and β-tubulin polymerise to produce microtubules, a critical component of the cellular cytoskeleton. SETD2 trimethylates α-tubulin at lysine 40 (α-TubK40me3), both in vitro (Park et al., 2016) and in vivo (Koenning et al., 2021). Currently, SETD2 is the only protein known to catalyse this modification. Distinct residues within the SRI domain of SETD2 are proposed to bind with the negatively charged C-terminal tail of α-tubulin. SETD2 then catalyses the methylation of α-TubK40 with its SET domain (Kearns et al., 2021). This mark is present on mitotic spindle microtubules during mitosis and cytokinesis, implicating SETD2 in regulation of these processes (Park et al., 2016). Loss of SETD2-mediated α-TubK40me3 in mouse embryonic fibroblasts results in mitotic defects, including failure of chromosome congression during prometaphase, formation of multipolar mitotic spindles, lagging of chromosomes at anaphase, and formation of chromosomal bridges and micronuclei at cytokinesis (Park et al., 2016).

SETD2 can methylate another cytoskeletal component: actin. Specifically, SETD2 performs trimethylation of lysine 68 on actin (ActK68me3) in mammalian cells (Seervai et al., 2020). The interaction between SETD2 and actin is mediated by the WW domain of SETD2, which binds the scaffolding protein HTT. HTT binds the adapter protein huntingtin-interacting protein 1 related (HIP1R), which binds to actin. This axis allows SETD2 to associate with actin and methylate it with its catalytic SET domain (Seervai et al., 2020). ActK68me3 has been observed to localise primarily to insoluble polymerised filamentous (F)-actin, and deletion of Setd2 in mouse embryonic fibroblasts results in a reduction in the ratio of F-actin to monomeric globular (G)-actin. These results suggest that SETD2-mediated ActK68me3 plays a role in actin polymerisation (Seervai et al., 2020). SETD2-deficient cells are observed to have cell migration defects, which could be attributed to the loss of ActK68me3 disrupting actin polymerisation (Seervai et al., 2020). However, the mechanisms underlying how ActK68me3 regulates actin dynamics remain unclear.

SETD2 also directly monomethylates lysine 525 on the transcription factor signal transducer and activator of transcription 1 (STAT1), via its catalytic SET domain (Chen et al., 2017). This is in contrast to other SETD2 substrates, which are primarily trimethylated, indicating that there is some level of flexibility in SETD2 methyltransferase activity. SETD2-mediated STAT1 methylation enhances the interferon α (IFNα)-mediated antiviral immune response in mammalian cells by promoting STAT1 phosphorylation by Janus kinase 1 (JAK1), a protein activated by IFNα signalling (Chen et al., 2017). Phosphorylated STAT1 then translocates to the nucleus to induce the transcription of genes involved in the IFNα-mediated antiviral immune response (Chen et al., 2017).

Another non-histone substrate of SETD2 is enhancer of zeste 2 (EZH2), a subunit of Polycomb repressive complex 2 (PRC2) (Yuan et al., 2020). PRC2 catalyses the deposition of H3K27me3, a modification primarily associated with transcriptional silencing. SETD2 monomethylates lysine 735 on EZH2. Methylated EZH2 is targeted for degradation through E3 ubiquitin ligase signalling, which is crucial for suppressing prostate cancer progression in mice (Yuan et al., 2020). SETD2 can thereby indirectly influence levels of further histone modifications.

Additionally, SETD2 appears to play a role in DNA damage signalling by directly interacting with the tumour suppressor protein p53 (also known as TP53). p53 inhibits the replication of cells containing damaged DNA and directly interacts with the C-terminal region of SETD2 in human cells (Xie et al., 2008). This interaction appears to enhance the transcriptional activity of p53, with overexpression of SETD2 causing a subsequent increase in the expression of various p53 target genes. SETD2 also indirectly enhances p53 stability by downregulating the transcription of HDM2 (also known as MDM2), which is involved in p53 degradation (Xie et al., 2008). Human cells lacking SETD2 have decreased p53 activation and therefore a lower proportion of cells entering cell cycle arrest at the G1-S DNA damage checkpoint (Carvalho et al., 2014). The mechanisms behind this interaction remain unclear. SETD2 might directly methylate p53 to enhance its activity. Indeed, another lysine methyltransferase, SET8 (also known as KMT5A), has been shown to methylate p53 and modulate its activity (Shi et al., 2007). Alternatively, SETD2 might direct the transcription factor to target genes marked with H3K36me3 (Xie et al., 2008).

The capacity of SETD2 to methylate other non-histone substrates remains largely unexplored. As such, it is possible that SETD2 has significant effects on a range of different cellular processes via non-histone substrates yet to be discovered. Indeed, substrate sequence specificity analysis has identified nine potential non-histone peptide substrates that could be methylated by SETD2; in vitro testing has confirmed that one of these substrates, lysine 666 of the extracellular matrix protein fibrillin-1, is methylated by SETD2 (Schuhmacher et al., 2020). It remains to be seen whether this process occurs in vivo and how it might influence biological function.

Given that SETD2 has a diverse range of cellular functions, it is perhaps unsurprising that it plays a critical role in the development and functioning of the mammalian CNS. Data from the Human Protein Atlas indicates that SETD2 RNA is ubiquitously expressed throughout the body of adult humans, including the CNS (Karlsson et al., 2021). SETD2 protein expression is predominantly localised to the nucleus and has been detected in multiple different cell and tissue types (Uhlén et al., 2015). SETD2 is expressed in all adult human brain regions, although it is particularly enriched in the cerebellum (Sjöstedt et al., 2020). Human Developmental Biology Resource expression data has also revealed SETD2 mRNA expression in foetal human brains (Lindsay et al., 2016). Setd2 expression patterns in adult mice largely reflect patterns observed in humans, although no enrichment in the cerebellum is observed in mice (Sjöstedt et al., 2020). Setd2 is also expressed widely across the developing mouse brain at embryonic day (E)13.5 and E15.5 (Xu et al., 2021a). Notably, western blot experiments show much higher SETD2 expression in the prenatal and neonatal mouse brain when compared to the adult brain (Koenning et al., 2021). This result has been corroborated with qPCR data from the developing mouse cortex, which shows that Setd2 mRNA expression is at its highest at around E14–E18 and then decreases postnatally (Xie et al., 2021). Despite this, immunohistochemistry has confirmed that SETD2 protein is still expressed across several brain regions in the adult mouse (Koenning et al., 2021).

Several recent studies have explored the role of SETD2 in the mammalian CNS. One such study reported the presence of the SETD2-mediated tubulin methylation mark, α-TubK40me3, in the developing and adult mouse brain using western blotting (Koenning et al., 2021). Further immunohistochemistry on adult mouse brains revealed that the mark is present in postmitotic cortical neurons and Purkinje cells, as well as mossy fibres in the hippocampus. In addition, immunostaining of primary neurons cultured from the cortex and hippocampus of postnatal day (P)0 mice revealed that the mark is particularly prevalent in the dendrites. The same study also probed the biological relevance of α-TubK40me3 by generating a mouse line containing a pathogenic variant of SETD2 with a mutant SRI domain. This variant was found to lack the ability to deposit α-TubK40me3, whilst critically maintaining the ability to deposit H3K36me3 in vivo, enabling the divergent roles of SETD2 in tubulin and histone methylation to be parsed. Homozygous mice showed embryonic lethality, but heterozygous mice were viable and displayed an anxiety-like phenotype. Primary neurons cultured from the cortex and hippocampus of the heterozygous mice at P0 showed reduced dendritic arborisation and axon length. Notably, RNA-seq and ChIP-seq using brains of heterozygous mice revealed minimal changes in gene expression and H3K36me3 deposition, further emphasising that the phenotypes observed likely resulted from reduced α-TubK40me3 deposition and not modification of chromatin (Koenning et al., 2021). Loss of α-TubK40me3 might disrupt microtubule formation in neurons, a process critical for dendritic and axonal organisation (Koenning et al., 2021). These abnormal neurons might contribute to the observed anxiety-like phenotype through the disruption of important neural circuits. Indeed, disruptions in dendritic arborisation have been observed in several other mouse models that display anxiety-like behaviours (Novais et al., 2013; Pla et al., 2013). Potential links between structural and behavioural abnormalities in the mice could be investigated further by performing electrophysiological experiments to probe neuronal function.

The role of α-TubK40me3 in microtubule formation has been explored further in another study (Xie et al., 2021). Setd2 knockdown in the developing cortex of mice led to neuronal migration deficits that were rescued with the expression of a truncated SETD2 construct containing only the AWS, SET and post-SET functional domains, and nuclear export signals for cytoplasmic localisation. This construct was found to retain the ability to trimethylate α-TubK40 but not H3K36. The deficits could also be rescued with the expression of an α-tubulin^K40F mutant that mimics K40 trimethylation. Furthermore, cultured neurons lacking SETD2 showed polarisation deficits and reduced microtubule numbers. This implicates the tubulin methyltransferase activity of SETD2 in the regulation of neuronal migration and polarity. α-TubK40me3 is primarily found on tubulin polymerised into microtubules, as opposed to soluble tubulin, suggesting that the mark plays a role in the formation of stable microtubules. The deficit in microtubule formation is proposed to impede the dendritic outgrowth required for neuronal polarisation. These unpolarised neurons lack the ability to efficiently migrate through the cortex, leading to lamination deficits and potentially abnormal cortical function (Xie et al., 2021).

SETD2 has also been implicated in the patterning of the cerebral cortex during development. Using a conditional Setd2 allele crossed to Emx1-Cre, Setd2 was conditionally ablated from the developing dorsal telencephalon of mice (Xu et al., 2021a). These mutants presented with a smaller cortex at P7. Whole-mount in situ hybridisation for several cortical area markers, such as Lmo4, Cad8 (also known as Cdh8) and Rorβ (also known as Rorb), revealed disrupted organisation of cortical areas, independent of any dysregulation of major patterning transcription factors such as Emx2, Pax6, Coup-TF1 (also known as Nr2f1) and Sp8. Furthermore, anterograde tracing experiments on adult Setd2 conditional knockout mice revealed disorganised corticothalamic and thalamocortical circuits. Behavioural tests revealed potential deficits in motor learning, spatial memory and social interaction in the conditional knockout mice. At a cellular level, disruptions to cortical arealisation were attributed to an H3K36me3-mediated disruption of DNMT3B activity, resulting in the downregulation of clustered protocadherin (cPcdh) gene expression during development (Xu et al., 2021a). It is possible that the expression of several other genes critical for brain function is also regulated by SETD2-dependent DNMT3B activity, although this has not yet been explored. Notably, the reason underlying the observed microcephaly in the Setd2 conditional knockout mice was not identified in this study. A reduction in brain size could be caused by reduced neurogenesis or increased apoptosis during development. Intriguingly, 5′-bromo-2′-deoxyuridine (BrdU) pulse chasing and cleaved caspase-3 (CC3) staining of the Setd2 conditional knockout mice revealed no observable changes in neurogenesis or apoptosis, respectively (Xu et al., 2021a). Further investigation is required to identify the cause of the brain size reduction in these mice. Other studies have identified roles for SETD2 in both neurogenesis and neuronal apoptosis, however. One such study revealed that SETD2 directly interacts with H2A.Z, a variant of histone H2A, to regulate neurogenesis in the developing mouse cortex (Shen et al., 2017). This interaction was found to promote the deposition of SETD2-mediated H3K36me3 at the promoter region of the transcription factor Nkx2.4. This enhances Nkx2.4 transcription, leading to the downstream transcription of genes involved in regulating neurogenesis. Disruption of the interaction by shRNA-mediated H2A.z knockdown in embryonic mouse brains was observed to increase the proliferation of neural progenitors at the expense of differentiation (Shen et al., 2017). It is worth noting, however, that although SETD2 dysfunction contributes to the neurogenesis deficits in the H2A.z knockdown mice, it might not be the sole cause. H2A.z knockdown might produce additional effects independently of SETD2, which could explain why neurogenesis deficits were not observed in the Setd2 conditional knockout mice from the aforementioned study (Xu et al., 2021a).

SETD2 has been implicated in the apoptosis of postmitotic neurons by using Nestin-Cre to selectively delete Setd2 from the developing nervous system of mice (Chu et al., 2020). CC3 staining and terminal deoxynucleotidyl transferase dUTP nick end labelling (TUNEL) assays revealed that these mice have increased levels of apoptosis throughout the brain at E18.5 and P0 (Chu et al., 2020). This directly contrasts with the lack of changes in apoptosis observed in P0 Emx1-Cre Setd2 conditional knockout mice (Xu et al., 2021a). The discrepancy might be due to differences in the mouse model. Emx1-Cre produces a dorsal-telencephalon-specific knockout, whereas Nestin-Cre produces a nervous-system-wide knockout. It is possible that Setd2 knockout cells derived from the dorsal telencephalon do not have apoptotic deficits, but cells derived from elsewhere in the CNS do. The increased apoptosis in the Nestin-Cre Setd2 conditional knockout mice has been attributed to SETD2 loss disrupting non-homologous end joining (NHEJ), which is the primary means of DNA double-stranded break repair in postmitotic neurons (Frappart and McKinnon, 2008). Loss of NHEJ machinery has been demonstrated to increase neuronal apoptosis (Barnes et al., 1998; Gao et al., 1998). It is unclear whether the observed apoptosis in the Nestin-Cre Setd2 conditional knockout mice was sufficient to significantly reduce brain size, however, as the study did not quantify brain size (Chu et al., 2020).

The role of SETD2 in alternative splicing of pre-mRNA might also influence brain function and development. By using ChIP-seq and RNA-seq, one study in mice has observed that H3K36me3 is strongly associated with different alternatively spliced exons in the nucleus accumbens, a brain region involved in the reward system (Hu et al., 2017). A related study has indicated that H3K36me3-mediated alternative splicing might play a role in cocaine-reward behaviour. Here, administration of cocaine to mice was found to cause alternative splicing in the nucleus accumbens, along with a concomitant enrichment of H3K36me3 at alternatively spliced exons (Xu et al., 2021b). One H3K36me3-enriched and alternatively spliced gene identified in the study was Srsf11, which encodes a splice factor. Targeted enrichment of H3K36me3 at the Srsf11 gene site in the nucleus accumbens was found to modulate alternative splicing of the gene, resulting in increased cocaine-reward behaviour in mice. These results suggest that H3K36me3 mediates the alternative splicing of the splice factor SRSF11, which then regulates downstream targets to control reward behaviour (Xu et al., 2021b).

Srsf11 is likely just one of many genes regulated by H3K36me3-mediated alternative splicing in the brain. Indeed, SETD2-mediated H3K36me3 has been implicated in the alternative splicing of several genes expressed in the hippocampus that are involved with neural function and linked with autism spectrum disorder (ASD) (Shah et al., 2020). Setd2 transcription has been observed to be regulated by fragile X messenger ribonucleoprotein 1 (FMRP). Loss of FMRP, which causes fragile X syndrome, leads to decreased stalling of ribosomes during transcription, increased SETD2 expression and, therefore, abnormal H3K36me3 deposition. In mice lacking FMRP, multiple genes implicated in neural function and development show changes in H3K36me3 deposition, as do genes linked to ASD. Several of these genes also display aberrant splicing, suggesting that H3K36me3 might play a role in regulating the alternative splicing of genes involved in neural development and function (Shah et al., 2020). Further study into how the alternative splicing of these genes affects neural function is necessary to better delineate the role of SETD2 in the brain. Nonetheless, this link between ASD-related genes and SETD2-dependent alternative splicing strongly implicates SETD2 in human pathology.

The important role of SETD2 in the CNS is exemplified by the effects of pathogenic SETD2 variants on humans (Table 2). Variants have been identified in individuals diagnosed with neurodevelopmental disorders such as ASD, intellectual disability and Luscan–Lumish syndrome (Fig. 2) (Chen et al., 2021). Pathogenic heterozygous SETD2 variants are the primary cause of the autosomal dominant Luscan–Lumish syndrome, an overgrowth disorder characterised by speech and motor delay, macrocephaly, autism-related symptoms, and intellectual disability (Chen et al., 2021). Magnetic resonance imaging (MRI) has revealed enlarged cerebral ventricles and structural abnormalities in the corpus callosum, hippocampus and cerebellum of some people with suspected Luscan–Lumish syndrome (Marzin et al., 2019; Zhang et al., 2023). The Genome Aggregation Database (gnoMAD) classifies SETD2 as intolerant to loss of function in humans, indicating haploinsufficiency (Karczewski et al., 2020). Many SETD2 variants, including missense and truncating variants, occur across the gene in people with Luscan–Lumish syndrome, although they are slightly more common around the SET domain (Fig. 2). All SETD2 variants associated with Luscan–Lumish syndrome are suspected to result in loss of function, inhibiting H3K36me3 deposition or other SETD2 functions (Marzin et al., 2019). However, comprehensive functional analysis of many of these variants is lacking, so definitive conclusions on their effects cannot be drawn.

Pathogenic variants of other genes result in overgrowth disorders similar to Luscan–Lumish syndrome, providing some insight into to its pathophysiology. Sotos syndrome (OMIM 117550), which is caused by pathogenic NSD1 variants, and Tatton-Brown–Rahman syndrome (OMIM 615879), which is caused by pathogenic DNMT3A variants, both have considerable overlap with Luscan–Lumish syndrome, suggesting that they might have mutual aetiologies (Baujat and Cormier-Daire, 2007; Tatton-Brown et al., 2014; Tlemsani et al., 2016). NSD1 has H3K36 methyltransferase activity responsible for depositing H3K36me1 and H3K36me2 (Li et al., 2009). The DNA methyltransferase DNMT3A also interacts with H3K36me2 and H3K36me3 to methylate DNA (Dhayalan et al., 2010). Both Sotos syndrome and Tatton-Brown–Rahman syndrome are associated with aberrant DNA methylation (Weinberg et al., 2019). Against this backdrop, it is possible that symptoms of Luscan–Lumish syndrome are caused by a loss of H3K36 methylation, resulting in abnormal DNA methylation and downstream dysregulation of gene expression. Recently, genome-wide profiling of several individuals with Luscan–Lumish syndrome has revealed widespread DNA hypomethylation, providing further evidence for this theory (Lee et al., 2023). Some of these hypomethylated genes, such as NRP2 and IGF2BP1, are linked to growth regulation, providing a possible mechanism for the observed overgrowth in Luscan–Lumish syndrome (Lee et al., 2023). Further investigation into the effects of these genes on growth is warranted.

The exact mechanisms by which SETD2 variants produce Luscan–Lumish syndrome remain largely unclear. Deficits in neuronal migration, cortical arealisation, neural circuit formation and alternative splicing like those observed in mouse models might also be present in humans and contribute to disorders such as ASD and intellectual disability. Several deficits observed in mouse models with irregular SETD2 function are associated with ASD, including abnormal cortical patterning (Lombardo et al., 2021), cPcdh expression (Anitha et al., 2013), neural circuit formation (Benkarim et al., 2021), neuronal migration (Reiner et al., 2016), microtubule structure (Gąssowska-Dobrowolska et al., 2022), as well as alternative splicing (Stamova et al., 2013).

Many questions also surround the development of macrocephaly in people with Luscan–Lumish syndrome. Limited data suggest that the symptom appears to mostly arise postnatally, with head circumference at birth often being normal (Marzin et al., 2019). It remains to be seen, however, whether prenatal brain abnormalities are present that later give rise to this postnatal macrocephaly. It is also unclear whether the macrocephaly is a result of enlarged brain size (megalencephaly), dysregulation of the ventricular system (hydrocephaly) or overgrowth of the skull (cranial hyperostosis). Observed ventriculomegaly in some individuals with SETD2 variants suggests that hydrocephaly might be a contributing factor (Luscan et al., 2014; Marzin et al., 2019), but more comprehensive MRI analysis would be necessary to discern the exact cause of the macrocephaly. Intriguingly, mice with cortex-specific homozygous Setd2 knockout show a reduction in mature cortex size, a seemingly opposing phenotype to humans with SETD2 loss of function (Xu et al., 2021a). Hence, it is possible that mouse and human SETD2 have different functions, or that macrocephaly arises from SETD2 dysregulation in areas outside the cortex, such as the ventricular system or the skull. However, there are two important differences between the mouse and human studies. First, the conditional deletion of Setd2 by Emx1-Cre in the mouse model will leave most, if not all, cortical interneurons and many oligodendrocytes with Setd2 intact. Second, the mice with reduced brain size in the conditional knockout study were homozygous for the Setd2 deletion, whereas human SETD2 disorders are all heterozygous. Heterozygosity might produce a completely different effect. Unlike in homozygous mice, no apparent changes in cortical organisation were identified in conditional heterozygote mice compared to controls in the aforementioned study (Xu et al., 2021a). Further study of heterozygous mouse models will be necessary to better understand the role of SETD2 in human disease.

Interestingly, a separate neurodevelopmental disorder associated with SETD2 variants has been recently identified, known as Rabin–Pappas syndrome (OMIM 620155) (Rabin et al., 2020). Individuals with this syndrome typically present with intellectual disability, microcephaly, and hypoplasia of the corpus callosum, pons and cerebellum (Rabin et al., 2020). Many of these features do not overlap with Luscan–Lumish syndrome. Of particular interest is the observed microcephaly, which is in direct contrast to the macrocephaly observed in Luscan–Lumish syndrome, but which might correlate with mouse Setd2 mutants. Rabin–Pappas syndrome was initially thought to be caused by a single de novo heterozygous SETD2 missense variant, p.R1740W (Rabin et al., 2020). However, a recent study has identified two other missense variants, p.E1718K and p.D2251E, that lead to a similar phenotype in humans (Parra et al., 2023). These variants are suspected to result in gain of function, although functional studies are lacking (Parra et al., 2023; Rabin et al., 2020). Recent genome-wide profiling has revealed that p.R1740W is associated with hypermethylation of DNA, which contrasts with the hypomethylation observed in Luscan–Lumish syndrome. Indeed, all the overlapping differentially methylated genes between the two syndromes show opposing methylation profiles, which might explain the opposing phenotypes (Lee et al., 2023). Furthermore, gain-of-function variants of DNMT3A are associated with hypermethylation and microcephalic dwarfism, further implicating Rabin–Pappas syndrome as a gain-of-function disorder (Heyn et al., 2019). Another reported missense variant, p.R1740Q, is associated with a less severe phenotype characterised by intellectual disability but no microcephaly (Lee et al., 2023). This disorder has been named intellectual development disorder, autosomal dominant 70 (OMIM 620157). Notably, codon 1740 does not lie in a known functional domain of SETD2. No individuals with Luscan–Lumish syndrome have been observed with a missense variant at or near this codon. It is possible that it resides in a functional domain that is yet to be described, and that dysregulation of functions mediated by this domain produces disease (Lee et al., 2023).

SETD2 has a wide range of cellular functions, which are mediated by its ability to deposit H3K36me3 and to interact with a variety of non-histone substrates. Many of these functions are critical to the development and function of the CNS, highlighted by the fact that pathogenic SETD2 variants in humans lead to neurodevelopmental disorders. Despite recent advances, it is still largely unclear what specific aspects of SETD2 function, when abrogated, contribute to disorders observed in humans. As mentioned previously, mice heterozygous for mutations that impair SETD2 tubulin methyltransferase activity, but not histone methyltransferase activity, present with neuronal and behavioural defects (Koenning et al., 2021). These results indicate that the tubulin methyltransferase activity of SETD2 is haploinsufficient in heterozygous mice, but evidence of haploinsufficiency for histone methyltransferase activity is lacking. In mouse embryonic fibroblast models, loss of one Setd2 allele does not significantly abrogate H3K36me3 deposition (Chiang et al., 2018). This suggests that human disorders like Luscan–Lumish syndrome might primarily be a result of defects in the ability of SETD2 to methylate non-histone substrates. However, this does not explain the functional overlap of Luscan–Lumish syndrome with Sotos syndrome and Tatton-Brown–Rahman syndrome, or the observed alterations in DNA methylation in individuals with pathogenic SETD2 variants, which could be attributed to H3K36me3-mediated DNA methylation deficits. Further study on mouse models and humans will be necessary to precisely distinguish the role of SETD2 in human disease.

Recent advances in mouse models and experimental techniques might provide insight into the role of SETD2 in the CNS in the future. Originally, one of the major roadblocks for studying SETD2 in the CNS of mice was the fact that total knockouts of Setd2 cause embryonic lethality at E10.5–E11.5 (Hu et al., 2010). Now, Setd2-floxed mice are readily available. These mice can be crossed with a range of Cre deleter strains to produce conditional knockouts that can survive into adulthood, providing more suitable models for interrogating the role of SETD2 in CNS function (Xu et al., 2021a). The identification of mutant forms of Setd2 that specifically abrogate certain aspects of SETD2 activity, such as its ability to methylate α-tubulin or H3K36, has allowed for the further delineation of the mechanisms underlying the role of this protein in the CNS (Koenning et al., 2021; Park et al., 2016; Xie et al., 2021). Understanding which specific SETD2 substrates contribute to CNS dysfunction might provide insight into possible treatments for disorders such as Luscan–Lumish syndrome. Emerging techniques such as single-cell RNA sequencing, ChIP sequencing and ATAC sequencing might provide further insight into the function of SETD2 in the brain by identifying how it regulates gene expression, epigenetic interactions with genes and chromatin architecture, respectively. These approaches could provide more answers on how observed deficits arise in current mouse models and might provide a better understanding of how SETD2 dysfunction causes disease, potentially leading to the development of treatments for Luscan–Lumish syndrome and other SETD2-associated pathologies.