A role for CHD8 in mRNA processing

Gompers et al. [12] provided evidence for abnormal mRNA splicing in the developing brain of Chd8^+/− mice, and recently showed that CHD8 is recruited to genes encoding mRNA splicing and processing factors in the mid-embryonic mouse brain [34]. CHD8 knockdown in human iPSC-derived neural progenitor cells also affected alternative splicing and levels of the histone modification, H3K36me3, were reduced [35]. CHD8 was found to interact with hnRNP (heterogeneous nuclear ribonucleoproteins) proteins, which have critical roles in pre-mRNA processing events such as RNA splicing [35]. The interaction of CHD8 with hnRNPL, a protein that regulates splicing and interacts with the H3K36me3 methyltransferase SETD2 was validated, providing a potential mechanism for splicing alterations and reduced H3K36me3 levels in Chd8-deficient cells [35].

CHD8-interacting proteins

In addition to hRNPs, CHD8 functions in complexes with other chromatin-associated proteins. Thompson et al. [36] reported the purification of 900 kDa complexes containing CHD8. Proteins associated with CHD8 included WDR, Co-REST, and other components of the SWI/SNF, and NuRD complexes, although the exact proteins identified were not reported in the paper. We have purified a similar size (~1 MDa) CHD8 complex from HeLA cell line nuclei and identified multiple components of the repressive NURD complex (CHD4, GATAD2A, GATAD2B, MTA2, MTA3 and HDAC1), as well as BAF complex components ACTL6A (BAF53A), SMARCC2 (BAF170), confirming these findings (Campos and Basson, unpublished). These interactions could endow CHD8 with either activating or repressive activities by increasing or reducing chromatin accessibility (Figure 1).

Yates et al. [37] further explored CHD8 interactions with the WDR-COMPASS (WDR5–ASH2L–rbBP5) complex in SF2 and NT2/D2 embryonal carcinoma cells. ASH2L recruitment to the HOXA2 promoter upon gene induction was dependent on CHD8, corresponding to reduced H3K4me3 levels at the HOXA2 promoter in CHD8-deficient cells. Unexpectedly, this was associated with increased HOXA2 expression, suggesting that CHD8 repressed HOXA2 expression, despite promoting the recruitment of WDR-COMPASS. Studies assessing total and genome-wide levels of H3K4me2/3 have not shown clear changes in H3K4me2/3 levels in Chd8-deficient cells (see e.g. [35]). However, global levels, or even genome-wide ChIP analysis of H3K4me3 at TSSs do not have the resolution to detect cell type-specific and gene-specific effects, so it remains possible that CHD8-WDR complexes are only recruited to a small subset of gene promoters where CHD8 deficiency might impact H3K4me3 levels.

Ceballos-Chávez et al. [38] established a functional relationship between CHD8 and the BAF complex by showing that CHD8 recruitment to progesterone receptor-regulated enhancers was regulated by the BAF complex. CHD8 can be recruited to neuronal enhancers by the H3K27me3 demethylase KDM6B (JMJD3), together with the TGFβ effector, SMAD3 [39]. KDM6B mutations are linked to a neurodevelopmental disorder that includes developmental delay, autistic features and intellectual disability, consistent with a possible mechanistic relationship between CHD8 and KDM6B [40]. Whether the H3K27me3 demethylase activity of KDM6B has any influence on the expression of CHD8-regulated genes remains unclear. However, it is intriguing to note that Chd8 deficiency is associated with reduced expression of Polycomb-regulated genes [16]. As the BAF complex can potentiate the H3K27me3-specific demethylase activity of KDM6B in the developing brain [41], the recruitment of CHD8, KDM6B and the BAF complex is predicted to function as a potent activating complex, antagonistic to Polycomb repressive mechanisms.

Other CHD8-interacting proteins of note include CTCF [42], an interaction likely to have neurodevelopmental consequences (Table 1). Indeed, we found up-regulation of a large number of clustered Protocadherin genes, known to be regulated by CTCF [43], in the P5 neocortex of Chd8^+/− mice [15].

Derafshi et al. [29] recently reported CHD8 recruitment to gene promoters in human iPSC-derived excitatory neurons. CHD8-associated regions were significantly enriched for ETS and YY1 motifs and active chromatin marks (H3K4me3, H3K36me3). Surprisingly, heterozygous CHD8 gene deletion was associated with the up-regulation of many immediate early genes, whereas homozygous CHD8 deletion was associated with the down-regulation of these genes. CHD8 loss was associated with reduced chromatin accessibility, especially at promoters containing ETS motifs. CHD8 directly interacted with the ETS transcription factor ELK1 in co-immunoprecipitation experiments and the recruitment of CHD8 was lost upon ELK1 knockdown. Although gene expression changes require further explanation, these findings provide compelling evidence that CHD8 also has an important role in post-mitotic neurons and that an important part of CHD8 function may be to modulate chromatin accessibility at ELK1-regulated gene promoters [29]. Indeed, Kawamura et al. recently reported that the conditional deletion of Chd8 in the adult (using a tamoxifen-inducible pan-Cre driver, CAGG-CreER) resulted in diminished induction of activity-dependent immediate early genes in response to the potent glutamate receptor agonist, Kainic acid. However, despite these findings, no learning and memory deficits could be detected in these mice [44] and the relevance of these observations to the intellectual disability associated with CHD8 deficiency remains unclear.

A recent pre-print by Subtil-Rodriguez et al. provides further functional confirmation of the ELK1-CHD8 interaction. They show that the serum/MEK/ERK-dependent recruitment of CHD8 to promoters is mediated by ELK1/ELK4 (https://www.biorxiv.org/content/10.1101/2022.09.09.507301v1). An analysis of available ChIP-seq datasets from a several different mouse and human models by the Nord laboratory also found an enrichment of ELF, ELK1, E2F, CTCF and YY1 motifs, consistent with these functional studies [45].

Thus, a similar picture is emerging than for the related factor CHD7. CHD7 is recruited to regulatory elements via tissue-specific transcription factors, including SOX2 [46], Runx1 [47] and Isl1 [48]. Intriguingly, in the case of CHD8, recruitment by different transcription factors appears to engage different mechanisms. For example, recruitment of CHD8 by p53 or β-catenin involves the co-recruitment of histone H1, chromatin compaction and gene repression. In contrast, CHD8 recruitment by ELK1 or KDM6B is associated with gene activation, presumably via the ATP-dependent chromatin remodelling activity of CHD8 ([29,39] and https://www.biorxiv.org/content/10.1101/2022.09.09.507301v1). It seems likely that this cell type-specific complexity also applies to other chromatin regulatory factors implicated in neurodevelopmental disorders.

Finally, Lasser et al. [49] recently reported that several ASD-associated chromatin remodelling factors associate with the mitotic spindle and by example, provided functional evidence for a role for CHD2 in cell cycle regulation, genome stability and apoptosis. Although the experimental evidence for a role for CHD8 in mitotic spindles is less convincing, the possibility that CHD8 is involved should be explored further.

Different mutations and sexual dimorphism

Given that ASD is diagnosed based on social and communication difficulties and the presence of restrictive and repetitive behaviours, a lot of research efforts have focused on the behavioural characterisation of Chd8^+/− mouse models. All groups reported behavioural phenotypes in Chd8^+/− mice, but the reported phenotypes were not always consistent [11–16,19,22]. Intriguingly, the Kim group showed that two different ASD-associated nonsense (frameshift) mutations in Chd8, Chd8^+/N2373K and Chd8^+/S62X [14,22], resulted in different behavioural phenotypes. Synaptic and molecular changes also differed [22]. Furthermore, sexual dimorphism was evident with males and females presenting with different phenotypes [14,22]. These findings suggest that different ASD-associated mutations in the same gene could manifest differently, even on the same genetic background and when characterised in the same laboratory. The reason for these differences is not known but seems to be unrelated to the reduction in wildtype CHD8 protein levels as these were reduced to an equivalent extent in both models [14,22]. One possibility is that truncated CHD8 protein fragments produced from these different alleles retain some activity and disrupt neuronal development and function in different ways. Further biochemical and molecular experiments will be necessary to test this hypothesis. It may be relevant to note that the C-terminal truncation CHD8^+/S62X, is not associated with megalencephaly in humans, whereas the Chd8^+/N2373K mutation is [53].

Tabbaa et al. [56] recently provided direct experimental proof that genetic background modifies neurodevelopmental phenotypes associated with Chd8 haploinsufficiency. They generated mice with a heterozygous frameshift mutation on 33 different genetic backgrounds. Whereas some phenotypes (increased brain size, social dominance and hypoactivity) were significant at a population level: others, (body weight, social behaviours and anxiety) were significantly modified by genetic background [56]. For instance, whereas Chd8^+/− mice on a C57BL/6J background spent more time sniffing a novel individual of the same sex and strain, recapitulating our findings in Chd8^+/− mice with a different mutation [15], this effect was only present in two of the other backgrounds. In tests for aggression, some genetic backgrounds elicited increased aggression, whilst others resulted in reduced aggression. Sex was a strong modifier of all the phenotypes [56]. For instance, whereas males in 18% of strains and females in 9% of strains showed increased anxiety, C57BL/6J females showed reduced anxiety.

It is intriguing to note that CHD8 variants are more prevalent in males than females, but no clear differences in phenotypic presentation have been noted in males and females presenting with pathogenic CHD8 variants [54]. As CHD8 regulates the expression of many other ASD-associated genes, genetic and epigenomic polymorphisms that affect the function or expression of these genes are likely responsible for modifying the expressivity or penetrance of CHD8-associated phenotypes.

Effects on brain growth

Increased brain size, or megalencephaly is a common feature associated with CHD8 heterozygosity in humans and has been recapitulated in different mouse models. Experiments in Chd8^+/− mouse models have reported relatively subtle effects on the proliferation and differentiation dynamics of neural progenitors arising from the ventricular zone [11,12,16]. To my knowledge, the effect of Chd8 haploinsufficiency on interneuron progenitors has not been assessed in the mouse.

Over the last couple of years, several groups have attempted to model neuronal differentiation in CHD8-deficient human cerebral organoids derived from pluripotent stem cell lines. Villa et al. showed that CHD8-deficient organoids were 50% larger than isogenic, wildtype controls after 4 months of culture, consistent with the megalencephalic phenotype in humans. Early neuronal stem cells remained proliferative for longer, thus expanding the progenitor cell pool leading to the production of increased neurons. Interestingly, organoids from cells with a human ASD-associated mutation not linked to macrocephaly, CHD8^S62X, did not show this phenotype, nicely validating this in vitro system [18]. Differentiation dynamics were differentially altered with accelerated production of interneurons and delayed differentiation of excitatory neurons. Transcriptomic analyses further suggest long-lasting effects on gene expression in excitatory neurons. In a separate study, Paulsen et al. [17] also reported evidence for accelerated differentiation of interneurons in CHD8^+/− organoids, and these effects were modified substantially by genetic background.

Finally, a report showed that a CRISPR/Cas9-induced heterozygous mutation in CHD8 in non-human primates was associated with megalencephaly, increased numbers of GFAP⁺ glial cells and OLIG2⁺ oligodendrocytes and more white matter than controls [57]. This surprising finding suggests that the increased brain size is primarily driven by the abnormal expansion of non-neuronal cells in the developing non-human primate brain. These findings contrast with the demonstration that Chd8 deficiency in mice is associated with reduced oligodendrocyte progenitor expansion, differentiation and myelination [28,58,59]. Further investigation is required to determine if these inconsistent findings represent species-specific differences in CHD8 function.

A role for CHD8 in the gut-brain axis

Chd8 is expressed in several developing organs during development, including the heart and gut [60]. It is well known that loss of function variants in many of the chromatin-associated factors linked to ASD, including CHD8, are also associated with congenital heart defects [61–63]. It has been postulated that developmental delay due to heart defects could contribute to neurodevelopmental phenotypes [64]. Clearly, congenital heart defects per se are not sufficient to cause neurodevelopmental phenotypes, but they may contribute to these phenotypes in already compromised individuals (Figure 2).

Open in a separate window

Figure 2.

Cellular/systemic dysregulation linked to CHD8 deficiency.

CHD8 deficiency affects multiple cell types and physiological systems in the body that could predispose to neurodevelopmental disorders. (1) CHD8 deficiency directly impacts synaptic mechanisms to dysregulate synaptic transmission during brain development and in the adult [13,14,21,22]. (2) CHD8 deficiency might cause congenital heart disease [54], leading to developmental delay and hypoxia. (3) Alterations in the gut epithelium might impact the microbiome, that could signal to the brain via the vagus nerve, or release neuroactive metabolites that can enter the circulation via a ‘leaky gut’, cross the blood–brain barrier and disrupt normal excitatory/inhibitory neurotransmitter levels in the brain and alter synaptic transmission [70].

Gut problems such as constipation are a common comorbidity of ASD [65,66], including in individuals with CHD8 variants [54]. Alterations in the gut microbiome have been suggested to contribute to ASD [67,68] via a (1) ‘leaky gut’, where increased permeability of the gut barrier results in bacteria and/or bacterial products entering the bloodstream and affecting the brain and (2) the vagus nerve, where bacteria in the gut somehow signals to the brain via the vagus nerve [69] (Figure 2).

Yu et al. [70] showed that the small intestines of Chd8^+/− mice had increased expression of the glutamine transporters Slc6a19 and Slc7a8. The microbiome in the gut was altered and intestinal inflammation was present in Chd8^+/− mice. They found a corresponding increase in glutamine levels in the serum, consistent with a ‘leaky gut’ phenotype. The levels of glutamine, as well as its metabolite in the brain, glutamate, were both elevated in the brain. The increased levels of the excitatory neurotransmitter glutamate in the brain were associated with an altered E/I balance in the medial prefrontal cortex of adult mice, characterised by an increased frequency of postsynaptic excitatory currents and reduced frequency of postsynaptic inhibitory currents on layer V neurons [70]. Intriguingly, supplementation with Bacteroides uniformis, the bacterial species with the largest fold change in the Chd8^+/− gut microbiome, rescued behavioural phenotypes and reduced the increased excitatory postsynaptic frequencies, serum glutamine levels and amino acid transporters in the gut of these mice [70].

As mentioned above, the effects of Chd8 mutation on synapses and brain circuits are dependent on genetic background, sex and even the type of genetic mutation. Only male mice were tested in the behavioural assays in this paper, and the sex of mice used in the other experiments are not reported.

These findings beg several related questions: (1) Are amino acid transporters, the gut microbiome, and glutamine and glutamate levels altered in other Chd8^+/− models too? (2) In models with an opposite synaptic phenotype in the prefrontal cortex [21], are these changes also in the opposite direction? (3) Would this mean that bacterial supplementation and treatments to restore a normal microbiome work for some individuals and not others with different variants in the same gene?

As mentioned above, a change in the gut microbiome can also signal to the brain via the vagus nerve (Figure 2). Sgritta et al. [71] have shown that gut permeability is not significantly affected in another ASD mouse model, Shank3B^−/− mice. Instead, treatment with Lactobacillus reuteri, sufficient to restore normal social behaviours in these mice, failed to rescue these deficits in mice after vagotomy (transection of the vagus nerve). The contribution of vagus nerve signals to the neurodevelopmental phenotypes of Chd8-deficient mice has not been evaluated.

I thank Laura Andreae, Nemanja Saric, Leticia Perez-Sisques and Shail Bhatt for comments on the manuscript. I apologise to those whose work on the topic could not be cited due to space constraints.