ZMYND11 functions in bimodal regulation of latent genes and brain-like splicing to safeguard corticogenesis

Chang, Xuyao; Li, Wenqi; Matsui, Satoshi; Huynh, Cindy; Erickson, Craig; Guo, Feng; Cederquist, Gustav Y.; Studer, Lorenz; Iwafuchi, Makiko; Shillington, Amelle; Chronis, Constantinos; Tchieu, Jason

doi:10.1038/s41467-025-64063-0

Download PDF

Article
Open access
Published: 09 October 2025

ZMYND11 functions in bimodal regulation of latent genes and brain-like splicing to safeguard corticogenesis

Nature Communications volume 16, Article number: 9010 (2025) Cite this article

Subjects

Abstract

Despite the numerous pathogenic variants linked to neurodevelopmental disorders (NDDs) including autism (ASD) and intellectual disability, our understanding of the underlying mechanisms caused by risk genes remain unclear. Here, we show that mutations in ZMYND11, a newly implicated risk gene, impair human cortical progenitor and neuron production. ZMYND11, known for its tumor suppressor function, encodes a histone-reader that recognizes sites of transcriptional elongation and acts as a co-repressor. ZMYND11-deficient cortical neural stem cells upregulate inappropriate developmental pathways, leading to disrupted neurogenesis. In addition to its role on chromatin, ZMYND11 regulates a brain-specific RNA isoform switch involving the splicing regulator RBFOX2. Similar defects are observed in other chromatin-related ASD risk genes, some of which are partially rescued by enhancing ZMYND11 function. These findings uncover convergent pathways linking chromatin regulation and splicing to human brain development and advance our understanding of how genetic risk contributes to NDD.

Introduction

Chromatin and epigenetic-related genes play a crucial role in shaping the developmental landscape and influencing future cell fate decisions^1,2,3,4. Despite their broad importance, genomic studies have found that several chromatin-related genes are mutated in autism spectrum disorders (ASD), suggesting their potential contribution to neurodevelopmental vulnerability^5,6. De novo pathogenic variants in ZMYND11, a newly identified neurodevelopmental disorder (NDD) risk gene, are associated with intellectual disability, epilepsy, and autistic traits^7,8. ZMYND11 is a chromatin reader that targets transcriptional elongation sites by recognizing and binding to modified histones marked by H3K36me3^9,10. Once bound, ZMYND11 acts as a transcriptional repressor by inhibiting transcription, which has been observed in cancer models. However, its role in cortical brain development remains unexplored.

Elevated expression of ZMYND11 in breast cancer has been linked to improved survival, underscoring its crucial role as a tumor suppressor^10,11. However, it remains unknown whether similar repressive mechanisms, when disrupted, confer risk for NDD. In this study, we investigate the transcriptional repressor role of ZMYND11 during human cortical brain development, utilizing a pluripotent stem cell-derived model of cortical neurogenesis^12,13,14,15. We identify ZMYND11 as a key regulator at the intersection of epigenetic control and alternative splicing, mediating a critical switch between brain-specific and non-brain mRNA isoforms in cortical neural stem cells (NSCs). Notably, this dysregulation in tissue-specific mRNA isoforms is observed across multiple high-confidence chromatin-related ASD risk factors, suggesting broader implications for the etiology of NDD, including ASD.

Results

ZMYND11 is critical for the differentiation of radial glial neural stem cells (NSCs) into intermediate progenitor cells (IPCs), a process impaired in a subset of autism risk genes

Pathogenic variants within the coding region of ZMYND11 predominantly cause frameshift mutations, leading to haploinsufficiency^7,8 (Supplementary Fig. 1a). ZMYND11 expression is present throughout fetal brain development in cortical cell types such as NSCs and excitatory neurons¹⁶ (Supplementary Fig. 1b–d), suggesting de novo mutations may affect early cortical brain development. To determine its role in this process, we engineered loss-of-function and heterozygous/patient-like mutations in human embryonic stem cells (hESCs) to investigate its function in disease-relevant cell types within an isogenic background (Supplementary Fig. 1e). We validated the mutation status and confirmed that no off-target editing occurred (Supplementary Fig. 1f, g and Supplementary Data 1), along with verifying altered protein expression (Supplementary Fig. 2a–c). These novel lines were karyotypically normal (Supplementary Fig. 2d), did not reveal noteworthy changes in the expression of pluripotency markers such as NANOG and SOX2 (Supplementary Fig. 2e), nor were there significant differences in cell proliferation (Supplementary Fig. 2f).

We next asked whether there were any cellular phenotypes associated with ZMYND11 deficiency during the differentiation into cortical NSCs. Many ASD risk genes fall into classes where cortical NSCs delay (Class I) or accelerate (Class II) cortical development¹³. To determine which class ZMYND11 aligns with, we generated cortical organoids using a guided differentiation approach (Fig. 1a). We generated cortical organoids representing the anterior forebrain marked by FOXG1 expression, and containing TBR2 positive intermediate progenitors (IPCs), indicative of corticogenesis (Fig. 1b), with no noticeable differences in organoid size across genotypes (Supplementary Fig. 2g). We found a dramatic decrease in FOXG1 and TBR2 expression in the loss of function line suggesting that ZMYND11 could play a role in neural patterning (Fig. 1c).

**Fig. 1: Engineered human stem cell model of ZMYND11 NDD shows decreased production of intermediate progenitors (IPCs).**

To adopt a more quantitative approach and reduce the potential variability found within cortical organoid culture, we performed monolayer differentiations in conjunction with intracellular flow cytometry to measure changes in IPC generation. In contrast to cortical organoids, the directed differentiation of hESCs into prefrontal-like cortical NSCs (Supplementary Fig. 3a, b) was highly efficient in all lines (Fig. 1d and Supplementary Fig. 3c–e) with no differences observed in cell cycle profiles (Supplementary Fig. 3f). From these populations, we observed a ~ 4–7-fold decrease in IPC generation in ZMYND11 mutants (Fig. 1e), a pattern similarly found in Class I risk genes (Fig. 1f and Supplementary Data 5). The reduction in IPCs was not due to apoptosis (Supplementary Fig. 4a) but correlated with an abnormal retention of cortical NSC identity, marked by expression of CD133 (Supplementary Fig. 4b), indicating a resistance to progress normally through cortical differentiation rather than undergoing spontaneous differentiation into neurons. To confirm this impairment, we found that the generation of TBR1-positive deep-layer neurons was also impaired to a similar magnitude ( ~ 4.5–7-fold decrease, Supplementary Fig. 4c–e). To determine whether corticogenesis was delayed rather than permanently blocked, we analyzed a later time point (day 30) and observed a ~ 4–5-fold increase in IPCs and a ~ 2.5–3.5-fold increase in neurons in the mutants compared with the mutants at an early time point (day 16), indicating a delayed differentiation (Supplementary Fig. 4f, g). No differences in IPC or neuron differentiation were observed between heterozygous and the knockout lines, even though ZMYND11 protein expression was reduced by ~60% in heterozygous cells (Supplementary Fig. 4h), suggesting that once ZMYND11 levels fall below a critical threshold, further reduction does not exacerbate the impact on cortical differentiation.

To test whether re-expression of ZMYND11 in the knockout line could rescue the IPC phenotype, we cloned ZMYND11 with an N-terminal 3X FLAG tag into a doxycycline-inducible vector and activated it during differentiation (Fig. 1g and Supplementary Fig. 5a). We found that expression of FLAG-ZMYND11 throughout differentiation led to the highest yield of IPCs and subsequent neuron differentiation ( ~ 1.8-fold increase, Fig. 1h) compared to expression during only half of differentiation ( ~ 1.3-fold increase). However, we observed only a modest overall rescue, which we attributed to transgene silencing during differentiation where FLAG was expressed in >85% of ESCs but dropped to ~30% in cortical NSCs (Supplementary Fig. 5b). To account for this, we specifically analyzed FLAG positive cells and found that neuron differentiation exhibited a ~ 4-fold increase, similar to the magnitude of decrease in ZMYND11 knockout (red dotted line, Fig. 1i). This suggests that the IPC deficient and differentiation phenotypes in ZMYND11 mutants can be reversed by restoring expression. To better model the patient context, we overexpressed three ZMYND11 missense variants identified in patients (C574R, C575Y, and C598S) in the knockout background. While each variant partially rescued neuronal differentiation, the effect was consistently lower than that observed with wild-type ZMYND11 (Supplementary Fig. 5c).

Elevated expression of developmental signaling transcripts impairs corticogenesis in ZMYND11 deficient cortical NSCs

To determine the underlying molecular changes associated with ZMYND11 deficiency, we performed bulk RNA sequencing due to our highly homogeneous population of cortical NSCs, as well as the starting hESCs. Principal component analysis revealed that cell types clustered together based on their cell identity, regardless of genotype differences, but the magnitude of differential gene expression was more pronounced within the cortical NSC population (mutants vs controls) (Fig. 2a and Supplementary Data 2). Notably, ZMYND11 mRNA levels were not significantly different between wild-type and mutant lines, suggesting that protein reduction occurs through a post-transcriptional mechanism (Supplementary Fig. 6a). As expected, differential gene regulatory networks between hESCs and cortical NSCs converged on gene ontologies related to neural differentiation (Supplementary Fig. 6b). We next examined gene expression changes between controls and ZMYND11-deficient cortical NSCs, which showed a significant overlap (Fig. 2b). Since ZMYND11 is a known transcriptional repressor, we focused on genes differentially upregulated. ZMYND11-deficient NSCs displayed significant upregulation in gene networks associated with the BMP and WNT signaling pathways (Fig. 2c and Supplementary Fig. 6c), which may explain the enrichment for additional ontologies associated with non-nervous system developmental processes (Supplementary Fig. 6d). We validated the increase in BMP signaling in mutant lines by measuring levels of phosphorylated SMAD1/5/9 (Fig. 2d) and found that correlated with impaired neuron differentiation when we treated the cells with a γ-secretase inhibitor, DAPT (Fig. 2e, f).

**Fig. 2: Transcriptomic analysis reveals elevated levels of BMP and WNT signaling pathways in ZMYND11 deficient NSCs, with BMP inhibition partially rescuing IPC production.**

Due to these potent signaling molecules ability to alter cell fate, we modulated WNT and BMP signaling independently to retain cortical identity (Supplementary Fig. 7a, b). For the Hi-dose BMP inhibition, ZMYND11 deficient cortical NSCs were able to differentiate into deep-layer excitatory neurons identified by TBR1- and VGLUT2-expression after DAPT treatment ( > 90%, Fig. 2g and Supplementary Fig. 7c, d). We next wondered whether ectopic BMP signaling could also impair IPC generation. To test this, we performed differentiations with the Hi-dose level of BMP inhibitor and found an increase in IPC generation from ZMYND11 deficient NSCs with no change in IPC numbers in control cells (+/−: ~ 3.5-fold increase, −/−: ~ 1.8-fold increase, Fig. 2h). Conversely, when we extended the duration of the WNT signaling inhibitor, XAV939, we found no significant change in IPC generation (Supplementary Fig. 7e, f). Finally, to determine which lineages were enriched during spontaneous differentiation of ZMYND11 mutants and isogenic controls, we performed an embryoid body formation assay. Within 10 days of differentiation, we observed a profound upregulation of mesenchymal rather than epithelial gene expression suggesting that gene regulatory networks associated with mesendodermal-like fate dominate which could factor in disrupting the generation of cortical organoids from knockout lines (Supplementary Fig. 6e). These data demonstrate that ZMYND11 deficiency elevates the expression of developmental signaling pathways that impair cortical NSCs from differentiating into IPCs and neurons. This impairment can be partially rescued by further inhibition of BMP, but not WNT, signaling pathways.

ZMYND11 binding represses latent gene regulatory networks associated with BMP and WNT signaling pathways

ZMYND11 is known to bind sites of transcriptional elongation identified by histones modified with trimethylation of histone 3 (H3K36me3) to prevent transcription¹⁰. To determine whether these transcriptomic changes are directly caused by ZMYND11 deficiency, we performed CUT&RUN in cortical NSCs for ZMYND11 and H3K36me3. We found that ZMYND11 largely occupied promoter sites and positively correlated with H3K36me3 (Fig. 3a, b). To reduce uncertainty associated with the ZMYND11 and H3K36me3 antibodies in this assay, we performed CUT&RUN for ZMYND11 in haploinsufficient and loss-of-function lines and found a reduction in overall binding and peak height, respectively, confirming specificity and used the loss-of-function line as our negative control (Supplementary Fig. 8a, b). In parallel, we screened a panel of published H3K36me3 antibodies to identify those effective in CUT&RUN and chose the one with the highest enrichment at genes with active transcription (Supplementary Fig. 8c). While our H3K36me3 genomic binding profiles showed enrichment along the gene body, the ZMYND11 binding profiles differed somewhat from previous studies, potentially reflecting cell type-specific variations in recruitment. To confirm this, we generated a doxycycline-inducible FLAG-tagged ZMYND11 and assessed its binding pattern in cortical NSCs (Supplementary Fig. 8d). After 3 days of induction, we found that ~80% of peaks identified by the FLAG antibody overlapped with the ZMYND11 antibody. Notably, the FLAG antibody yielded a greater number of peaks, likely due to its higher specificity and lower background signal compared to the ZMYND11 antibody. Importantly, both datasets consistently showed enriched promoter binding, which may differ from its binding patterns in other cell types studied. To explore this, we reanalyzed all publicly available ZMYND11 binding data and found that four out of six studies^{9,10,17,18,19,20} also reported promoter enrichment (Supplementary Fig. 8e). While this might be due to the use of different ZMYND11 antibodies, these findings support the concept that ZMYND11 binding could be influenced by cell context.

**Fig. 3: ZMYND11 directly binds and transcriptionally represses BMP and WNT signaling-related genes.**

ZMYND11 and H3K36me3 bind to 7140 and 7706 genes in cortical NSCs (Fig. 3c, Supplementary Fig. 9a and Supplementary Data 3), respectively, with an overlap of 4478 genes (Fig. 3d). Combining our transcriptomic and binding data identified three major clusters that encompass ZMYND11 binding patterns. The “High” cluster binds across the gene body, the “Mid” cluster is strongly enriched at the promoter, and the “Low” cluster displays weak promoter binding (Fig. 3e). Several critical genes associated with cortical NSC identity such as PAX6 are in High cluster (Fig. 3f). By correlating binding with expression we found that ZMYND11 deficiency has little effect on global gene expression (Supplementary Fig. 9b). Interestingly, of the three clusters analyzed, we found that genes with weak ZMYND11 binding at the promoter (Low cluster) had significant changes in gene expression and are associated at sites with less H3K36me3 co-binding (Supplementary Fig. 9c). These ZMYND11 bound direct targets were enriched in non-nervous system developmental programs and signaling pathways associated with WNT and BMP signaling (Fig. 3g, h). Genes related to extracellular matrix and mesenchyme development, which were also differentially regulated, were not bound, suggesting that these changes are caused by an indirect mode of dysregulation (Supplementary Fig. 9d). These genomic data analyses suggest that ZMYND11 directly regulates genes associated with latent, non-nervous system development and related pathways at promoters in cortical NSCs.

ZMYND11 deficiency correlates with an alternative splicing switch from brain to non-brain isoform expression

In addition to the role ZMYND11 plays on chromatin, we next focused on the significant alternative splicing changes observed within cortical NSCs. Knockdown studies in cancer lines have shown that dysregulation of ZMYND11 increases differential alternative splicing (DAS) events that may contribute to tumorigenesis^9,11. To determine whether ZMYND11 deficient hESCs and cortical NSCs exhibit DAS, we performed splicing analysis using our transcriptomic datasets and found significant differential exon usage in ZMYND11 deficient cortical NSCs compared to controls and lower DAS events in hESCs (Fig. 4a and Supplementary Fig. 10a). We focused on cortical NSCs, and the DAS events predominantly involved spliced exon events. We next plotted principal components of exon usage, revealing a distinct separation of exon usage compared to controls (Fig. 4b). To minimize false positives from sequence alignment and statistical prediction algorithms, we performed additional analysis using two additional pipelines and considered only the intersecting events (Supplementary Fig. 10b). This approach identified 991 and 590 DAS events in heterozygous and homozygous mutations, respectively (Fig. 4c, Supplementary Fig. 10c, d and Supplementary Data 4), exhibiting a significant overlap (Supplementary Fig. 10e). The greater number of DAS events in heterozygous mutants likely reflects the increased number of biological replicates used in this group, which enhanced the statistical power and sensitivity for splicing detection. The same approach found only fewer overlapped DAS events in mutant hESCs (Supplementary Fig. 10f). Upon examining potential shared properties of the spliced isoforms in cortical NSCs, we found a significant correlation with the inclusion or exclusion of small exons (Supplementary Fig. 10g), which has been described in prior work on autism^21,22.

**Fig. 4: ZMYND11 promotes a brain-specific mRNA isoform switch.**

To investigate the potential link between expression changes and DAS events, we examined tissue-specific percent spliced in (PSI) values from VastDB²³ (Fig. 4d). We successfully mapped tissue-specific PSI values for 507 out of 590 DAS events identified in ZMYND11 −/− NSCs. Our analysis revealed a strong inverse relationship between PSI values in brain and non-brain tissues (Fig. 4e). Interestingly, for splicing events characterized by exon inclusion in non-brain tissues, the majority also showed exon inclusion in ZMYND11 −/− NSCs (205 out of 273 events). Conversely, for events typically exhibiting exon exclusion in non-brain tissues, a similar pattern was observed in ZMYND11 −/− NSCs (140 out of 233 events). Overall, approximately 70% of the DAS events in ZMYND11 −/− NSCs demonstrated a bias toward non-brain tissue splicing patterns. To further validate this observation, we correlated PSI values from ZMYND11 −/− cortical NSCs with non-brain tissue PSI values, finding a moderate correlation (R² = 0.23, Fig. 4f). In contrast, this correlation was nearly absent in ZMYND11 +/− NSCs (R² = 0.002, Supplementary Fig. 10h), suggesting that the loss of ZMYND11 disrupts brain-specific splicing, shifting it toward a non-brain splicing profile.

We further examined exon inclusion and exon exclusion DAS events in ZMYND11 −/− NSCs separately and found a dose-dependent PSI change (Fig. 4g). We validated high-confidence exon inclusion and exclusion events using semiquantitative PCR (Fig. 4h, i and Supplementary Fig. 11a, b). We then compared our data with previously reported DAS events from ZMYND11 knockdown studies^9,10, finding no significant overlap between these datasets (Supplementary Fig. 10i), which highlights the cell- and tissue-specific nature of alternative splicing regulation by ZMYND11.

To explore if ZMYND11 chromatin binding is associated with these alternative splicing changes, we identified direct binding to 68 out of 590 ( ~ 11%) high confidence DAS events in ZMYND11 −/− NSCs (Supplementary Fig. 10j). This number increased to 141 events upon profiling FLAG-ZMYND11 binding (Supplementary Fig. 10k). Interestingly, while not all DAS events exhibited direct ZMYND11 binding, the majority ( ~ 70%) were located near the transcription start site (TSS + /− 1 kb, Supplementary Fig. 10l), suggesting a potential role for ZMYND11 binding in regulating alternative splicing. Additionally, ZMYND11 binding was observed in both exon inclusion events (e.g., APP) and exon exclusion events (e.g., L1CAM, Supplementary Fig. 10m). This analysis suggests that ZMYND11 enables a switch in brain-specific alternative splicing.

ZMYND11 deficiency causes a tissue-specific isoform switch that affects migration and proliferation of cortical NSCs

Mapping the differentially spliced isoforms to biological processes revealed enrichment for genes important for synapse organization and microtubule-based movement, suggesting regulation of cytoskeletal machinery (Supplementary Fig. 12a). Specific exon usage in some of the genes with DAS events has been previously linked to cancer cell migration^24,25,26. To assess how these splicing changes might affect the migration of cortical NSCs, we designed shRNA constructs targeting either the brain or non-brain isoforms of L1CAM, CTTN and MEAF6 (Fig. 5a and Supplementary Fig. 12b) as these genes exhibited high differential PSI values, ZMYND11 dose dependency, and displayed significant expression (Supplementary Fig. 12c). For this experiment, we opted to use a stable NSC cell line (LTNSCs), which retain their progenitor state for over 100 passages while maintaining their neurogenic potential^27,28. By using these isoform-specific constructs, we selectively targeted the brain or non-brain isoforms (Supplementary Fig. 12d) and performed scratch assays on a confluent monolayer of LTNSCs to determine the effects on cell migration. While we did not observe migration defects associated with knockdown of CTTN isoforms, targeting the brain isoform of L1CAM (exon 3 or exon 28) impaired LTNSC migration compared to control (Fig. 5b). Contrary to our predictions, the knockdown of the non-brain isoform of MEAF6 not only blocked migration but also showed signs of low proliferation. To explore this further, we examined the cell cycle profiles of MEAF6 brain versus non-brain isoforms in LTNSCs. BrdU labeling demonstrated that the knockdown of the non-brain isoform of MEAF6 greatly decreased S phase (Fig. 5c). These results suggest that isoform switches impact both migration and cell proliferation in cortical NSCs.

**Fig. 5: Non-brain-like isoforms alter migration and proliferation of cortical NSCs.**

While ZMYND11 expression is high in the brain, it is not exclusive. To further understand how ZMYND11 regulates brain-specific splicing, we tested the roles of several RNA-binding proteins (SRSF1, SRSF10, SRSF4, EFTUD2, ELAVL1) that have been reported as ZMYND11 interactors⁹ and selected brain-specific splicing regulators (PTBP1, PTBP2, NOVA2, ELAVL4, RBFOX2)^29,30,31 using shRNA knockdown (Supplementary Fig. 12e). We also used ZMYND11 knockdown as the positive control. Expectedly, semi-quantitative PCR analysis of 14 high-confidence DAS events revealed that ZMYND11 depletion resulted in a gain of non-brain isoform pattern, similar to heterozygous mutants (Fig. 5d, e). However, upon analysis of its interactors and brain splicing regulators, only PTBP1 and RBFOX2 depletion resulted in a clear loss of non-brain isoforms. RBFOX2 expression was slightly increased in ZMYND11 mutants (Supplementary Fig. 12f). In contrast, knocking down other previously identified ZMYND11-interacting RNA-binding proteins did not significantly impact splicing, indicating that the splicing changes observed may involve multiple regulatory mechanisms. Finally, we explored whether BMP signaling might contribute to the observed splicing changes. We found that only high doses of BMP4 induced significant splicing changes, while low-dose BMP4 or BMP inhibitor treatments did not (Supplementary Fig. 12g, h). This suggests that despite ZMYND11 deficiency causing elevated expression of BMP signaling, it plays a limited role in regulating splicing in ZMYND11 knockout cortical NSCs.

Tissue-specific isoform switches found in additional risk factor lines for ASD can be improved by ZMYND11

Our data thus far have implicated that ZMYND11 acts as both a regulator of latent gene activity and alternative splicing to promote faithful cortical development by acting as a repressive blanket over the epigenome. Due to this shared phenotype with other autism related risk factors (Class I ASD), we performed a similar set of experiments where we generated transcriptomic data derived from cortical NSCs (Fig. 6a). Expectedly, Class I ASD lines showed downregulation of genes linked to neurogenesis driving the separation from isogenic controls within the PCA plot (Fig. 6b and Supplementary Fig. 13a). Despite little connection with ZMYND11, Class I ASD lines appeared to upregulate the expression of developmental pathways including BMP and WNT compared to controls (Fig. 6c). Although lower in magnitude, we observed 20 significant high confidence DAS events that switched isoform tissue-specificity, seven of which are shared with ZMYND11 mutants (Fig. 6d–f and Supplementary Fig. 13b). Interestingly, this non-brain isoform trend was not observed in Class II ASD lines (Fig. 6g and Supplementary Fig. 13c). Since we found this unexpected similarity in the Class I ASD lines, we thought that ZMYND11, through its transcriptional repression function, could rectify the cortical NSC deficits by targeting the silencing of developmental pathways as well as promoting a brain-biased switch in isoform expression. We infected our dox-inducible FLAG-ZMYND11 virus and differentiated individual Class I ASD lines as monolayer to cortical NSCs and assessed IPC generation (Fig. 6h). We found that while the percentage of IPC generation varied due to specific ASD mutations, there was indeed a positive correlation in the dox treated groups (Fig. 6i, j), which was not due to downregulation of ZMYND11 expression (Supplementary Fig. 13d–f). These data suggest that both repression of latent developmental pathways and restoration of brain-biased isoforms aid in faithful corticogenesis and potentially implicate the activity of ZMYND11 or genes with similar function as critical for maintaining a stable cortical progenitor identity. To explore this idea, we overexpressed ZMYND11 in WT cortical NSCs for 72 h to determine which gene regulatory networks are susceptible from being silenced (Supplementary Fig. 13g). Unexpectedly, transcriptome analysis of ZMYND11 overexpression in control lines revealed only ZMYND11 was significantly upregulated, suggesting that the homeostatic level of this chromatin reader was already at optimal levels, resulting in no significant impact on gene expression (Supplementary Fig. 13h). Splicing analysis also identified a few events that were not overlapped with high confidence DAS events identified in ZMYND11 KO (Supplementary Fig. 13i).

**Fig. 6: Several ASD risk factors share impaired IPC production and non-brain-like alternative isoform switches, which can be improved by ZMYND11.**

From these data, we conclude that a subthreshold of ZMYND11 levels exhibits defective generation of IPCs from cortical NSCs, a critical cell type for mammalian neurogenesis. Upregulation of latent developmental-related pathways and a switch from brain to non-brain alternative splicing impair the efficiency in generating these cells from cortical NSCs. Intriguingly, elevated levels of BMP and WNT signaling as well as the non-brain isoform switch are a shared phenotype found with mutants in other high-confidence ASD-risk genes whose corticogenesis can be improved by ZMYND11 overexpression. These findings highlight the complexity of chromatin-associated pathogenic variants, where dysregulation of gene expression is not the sole cause of NDD/ASD.

Discussion

Our study uncovers a pivotal role for ZMYND11 as a dual regulator of chromatin dynamics and splicing fidelity in cortical neural stem cells (NSCs), providing a novel lens through which to understand the molecular underpinnings of ASD and related NDD. We were able to correct isoform-specific splicing defects and restore neurogenesis to levels reflective of controls by modulating splicing factor expression or overactivation of ZMYND11. Leveraging this knowledge, we envision precision medicine strategies that may promote splicing modulation as a potential therapeutic route³².

A key finding was the abnormal activation of BMP and WNT signaling pathways in ZMYND11-deficient cortical NSCs. Prior studies have shown that these pathways play essential roles in regulating corticogenesis^33,34,35,36, and abnormal WNT signaling has been proposed as a major dysregulated pathway conferring risk for NDD/ASD pathogenesis^37,38. In agreement with this, ZMYND11-deficient NSCs displayed elevated levels of BMP and WNT signaling associated with increased expression of genes related to non-neural development. We found that ZMYND11 directly binds to many of these genes to repress their expression; however, this binding was not always correlated with H3K36me3 enrichment, suggesting additional regulatory mechanisms may influence ZMYND11 recruitment. One possibility is H3.3 S31 phosphorylation, which has been shown to eject ZMYND11¹⁸. Moreover, gene expression changes in mutants could not be fully explained by WT ZMYND11 binding. This is possibly due to other transcriptional machinery overriding its repressive function, with certain gene sets (e.g., latent developmental and signaling pathway genes) being particularly sensitive to ZMYND11 regulation. While BMP inhibition rescued IPC production, WNT inhibition did not, raising the question of why WNT pathway modulation was ineffective given its importance in NDD/ASD pathogenesis. We believe that this may be due to the specific role WNT signaling plays in regulating anterior-posterior patterning, which was not explored in our monolayer culture system³⁹. This may also explain the reduced forebrain lineage identity observed in our organoid cultures, where spatial patterning is more relevant. Despite this, patients with ZMYND11 mutations did not display major brain malformations^7,8, supporting our use of monolayer models for studying disease-relevant molecular phenotypes. Future studies could explore the contribution of non-canonical WNT signaling, particularly WNT5A, which was also highly elevated in ZMYND11-deficient lines⁴⁰.

One of the most striking findings from our study identified splicing alterations as a critical mechanism in ZMYND11-deficient cortical NSCs. Alterations in splicing, specifically, in ‘microexons’ have been previously reported as enriched in patients with ASD^21,29,41. ZMYND11 caused splicing changes that biased toward non-brain tissue identity, reflected a broader gene expression shift toward activation of non-neural-like development. Importantly, this switch in alternative splicing isoforms are not merely a byproduct but actively contribute to cellular deficits, affecting processes like migration and proliferation in NSCs. Surprisingly, we found that ZMYND11 did not largely regulate splicing, as few DAS events overlapped with chromatin binding. This suggests that the dysregulation of splicing is likely indirect, potentially mediated by splicing genes such as RBFOX2^42,43, since its expression was increased in ZMYND11 mutants. Overall, our findings point to a multi-faceted regulation of splicing by ZMYND11, with both direct and indirect mechanisms at play.

Although multiple studies have identified different convergent mechanisms in NDD/ASD pathogenesis^{44,45,46,47,48}, most studies are limited in phenotypic convergence or face problems such as low-depth sequencing. Moreover, as the idea of non-neural fate shift has also been previously reported in ASD cortex, where there was an enrichment of non-neural microexon²¹ and in an additional NDD/ASD risk gene, MYT1L, whose mutants elevate gene expression of non-neural fate and impair neurogenesis^49,50, it remains to be determined whether the same concept applies to a broader group of NDD/ASD mutations. To this end, we differentiated cortical NSCs from 6 separate lines harboring risk factors for ASH1L, DEAF1, RELN, CUL3, KDM5B, and ASXL3 along with multiple isogenic controls and revealed similar transcriptional changes as those found in ZMYND11 that included the impairment of IPC generation and upregulation of latent developmental pathways. Although splicing alterations in these lines were less pronounced, they exhibited a similar trend toward non-brain tissue bias, potentially due to the greater sequencing depth we used. Interestingly, this non-brain splicing trend did not apply to all NDD/ASD risk genes, as Class II mutants (ones that enhance neurogenesis) showed an opposite splicing trend. This suggests that NDD/ASD encompasses not just a spectrum of phenotypes but a range of underlying molecular mechanisms that correlate closely with the biological outcomes. Nevertheless, strategies to modulate RNA splicing that have already been successful in spinal muscular atrophy treatment⁵¹ serve as a novel therapeutic avenue for Class I mutants.

The strive for discovering common molecular mechanisms in polygenic disorders is to develop broad therapeutic strategies, reducing the need for personalized interventions. As a proof of principle, we demonstrated that overexpressing ZMYND11 in Class I mutant lines partially rescued cortical NSCs deficits by suppressing non-neural gene expression and restoring brain-biased splicing. Importantly, ZMYND11 overexpression in WT NSCs did not induce adverse effects, suggesting that ZMYND11 levels are not dose sensitive. This feature, coupled with ZMYND11’s relatively small size compared to other Class I genes (e.g., ASH1L, KDM5B, ASXL3), makes it a promising candidate for gene therapy approaches for which multiple studies have tried overexpression as a gene therapy^52,53. Future work will examine how overexpressing ZMYND11 partially restores the epigenetic landscape in Class I mutants by potentially compensating for other epigenetic gene dysregulation.

Methods

Maintenance and propagation of human embryonic stem cells (hESCs)

Experiments using the human embryonic stem cell (hESC) line H9 (WA09) and MEL1 were performed under protocol EIP190101, approved by the Cincinnati Children’s Hospital Medical Center (CCHMC) and Memorial Sloan Kettering Cancer Center (MSKCC) Embryonic Stem Cell Research Oversight (ESCRO) Committee. The H9 line was obtained from the WiCell Research Institute and the MEL1 line through Stem Cells Ltd under an appropriate license. All research was conducted in accordance with institutional guidelines, relevant regulations, and the principles outlined in the 2021 ISSCR Guidelines for Stem Cell Research and Clinical Translation. hESCs were maintained in Essential 8 or Essential 8 flex (E8) medium on Vitronectin-coated dishes following established protocols. Briefly, hESCs (H1, XY; H9, XX; MEL1, XY) were cultured at 37 °C in a 5% CO₂ environment and passaged as clumps using an EDTA dissociation solution. Cells were used for differentiation between passage 50–70. All hESC lines are karyotypically normal and tested for mycoplasma every 2–3 months.

Human embryonic kidney 293 T (HEK293T) cell culture and lentivirus production

HEK293T cells were cultured at 37 °C in a 5% CO₂ environment using DMEM medium supplemented with 10% fetal bovine serum (FBS), 1X Antibiotic-Antimycotic and 1X GlutaMAX. The medium was changed every 2 days, and cells were passaged using Trypsin-0.05% EDTA.

Lentiviruses were produced as previously described. The day before transfection, 5 million HEK293T cells were plated in a 10 cm dish coated with Poly-L-lysine & 0.1% gelatin. The lentivirus, packaging (psPAX2) and envelope (pMD2.G) vectors were transfected using X-tremeGene 9 in a 1:2:1 molar ratio, respectively. For the M2-rtTA and FLAG-ZMYND11 vectors, the medium was changed into E8 18 h post-transfection. For the pGreenPuro shRNA vectors, medium was changed into N2 medium supplemented with B27 (1:1000, without Vitamin A; N2-B27), 20 ng/ml EGF and 20 ng/ml FGF2 18 h post-transfection. Viruses were harvested at 48 and 72 h post-transfection, concentrated using AMICON Ultra-15 Centrifugal Filter Units, and snapped frozen in liquid nitrogen.

Three-dimensional guided brain organoid differentiation

Guided cortical organoid differentiation involved seeding 5000 hESCs into one well of ultra-low attachment 96-well plate in E8 supplemented with 10 μM ROCKi on day −1. EBs were treated with 100 nM LDN193189 (BMPi), 10 μM SB431542 (TGFβi) and 5 μM XAV939 (WNTi) in Essential 6 (E6) medium from day 0 to 6. The medium was then replaced with N2 medium supplemented with 20 ng/ml EGF, 20 ng/ml FGF2, and 1:1000 of B27 supplement without RA from day 7 to day 19. Cortical Spheroids were further matured in 20 μg/ml BDNF, 20 μg/ml GDNF and 200 μM Ascorbic acid in Neurobasal medium supplemented with B27 (1:50, without Vitamin A; NB-B27) and 1X GlutaMAX for up to 30 days. Organoid size was quantified by measuring the diameter using ImagJ.

Monolayer prefrontal (PFC) cortical differentiation and excitatory neuron induction

Dishes were coated with Matrigel (diluted 1:100 in DMEM/F12) and stored overnight at 4 °C. The next day, hESCs were dissociated to single cells using Accutase, resuspended with E8 and plated on the Matrigel-coated dish at a density of 300,000 cells /cm² in the presence of 10 μM ROCKi on day −1. From day 0 to 2, the cells were cultured in E6 medium supplemented with 100 nM LDN193189, 10 μM SB431542 and 5 μM XAV939 (i.e., LSBX) to promote anterior telencephalic patterning. From days 3 to 7, the cells were maintained in E6 with LSB with XAV939 removed. From days 8 to 18, the cells were treated with 50 ng/ml FGF8 in N2-B27 to promote a PFC-like cortical regionalization. After d18, the cells were treated with N2-B27 only to promote corticogenesis.

To induce cortical NSCs to neurons, 10 μM DAPT (NOTCHi) was added to day 18 of differentiation in NB-B27 for 5 days. The neurons were then replated onto dishes coated with Poly-ornithine (PO), Laminin (Lam) and Fibronectin (FN) and maintained in NB-B27 supplemented with 20 μg/ml BDNF, 20 μg/ml GDNF and 200 μM Ascorbic acid. To eliminate proliferative cells from the neuron culture, 1 μM Ara-C was added for 48 h. Fifty percent of the neuron medium was changed every 3 days.

Maintenance of long-term neural stem cells (LTNSCs)

Derivation of LTNSCs was as previously described^27,28. LTNSCs were maintained on PO-Lam-FN coated dishes with 20 ng/ml EGF and 20 ng/ml FGF2 in N2-B27. LTNSCs were maintained at high density, used between passages 15–25, and passaged every week using 0.05% Trypsin.

Spontaneous differentiation of hESCs

5000 hESCs were seeded into one well of ultra-low attachment 96-well plate in E8 supplemented with 10 μM ROCKi on day -1. From day 0 to 10, EBs were treated 20%KSR, 3%FBS, 1X GlutaMAX and 1X MEM-NEAA in DMEM/F12 medium before harvest at day 10.

Plasmid constructs, site-directed mutagenesis and molecular cloning

The lentiviral vector used to overexpress human ZMYND11 in our study, pLV[Exp]-Puro-TRE>3xFLAG-hZMYND11[NM_001370102.2], was constructed by VectorBuilder. The vector ID is VB211020-1233tqt, which can be used to retrieve detailed information about the vector on vectorbuilder.com. NEB Q5 Site-Directed Mutagenesis Kit was used to generate overexpression vectors for ZMYND11 missense variants (C574R: c.1720 T > C, C575Y: c.1724 G > A, C598S: c.1793 G > C).

Short hairpin RNA (shRNA) sequences were manually designed and cloned into the restriction sites EcoRI & BamHI in the pGreenPuro vector (System Biosciences, SI505A-1). shRNA duplexes were annealed and inserted using traditional ligation cloning. All shRNA sequences were listed in Supplementary Data 5.

ZMYND11 expression analysis in neurodevelopment

BrainSpan human fetal brain transcriptional atlas (https://brainspan.org) was used to extract ZMYND11 expression data in the developing prefrontal cortex (averaging: OFC, DFC, VFC and MFC), in comparison with GAPDH, the housekeeping gene, and UMOD, that is not expressed in the human brain.

Cell-type specific ZMYND11 expression in the developmental human brain was analyzed using publicly available single cell RNA-seq dataset (GSE120046). The Seurat R package was used to re-analyze the developmental human brain data, by inputting normalized count data and metadata. Clusters that represent early corticogenesis were extracted and plotted with “DimPlot” and “VlnPlot”.

Generation of ZMYND11 mutant hESCs

CRISPR/Cas9 was used to introduce frameshift mutations into exon 3 of ZMYND11 in H1 hESCs. Guide RNA (gRNA, IDT): 5’-AGCTAAGCTCAGCTGACGGG was chosen using predictive scores from a website tool (CRISPOR, http://crispor.tefor.net/). The gRNA-Cas9 complex was transfected using Lipofectamine Stem Transfection Reagent (Invitrogen) according to the manufacturer’s protocol. Individual clones were isolated by replating transfected cells at low density. Indels were screened using Sanger sequencing, and heterozygous and compound heterozygous clones were inferred bioinformatically using DECODR. To identify mutations on different alleles and to exclude mixed clones, TOPO cloning of the targeted region was applied. Selected heterozygous clones (e.g., Het2, Het4) were clonally expanded a second time to ensure there was no mixing of wildtype and homozygous clones. Off-target prediction was performed using CRISPOR. Top 5 off-target sites with appropriate primers in intergenic, intron and exon regions were amplified and Sanger Sequenced for analysis. All sequenced off-target sites were listed in Supplementary Data 1.

Generation of doxycycline inducible ZMYND11 overexpression lines

In experiments where we overexpressed the wildtype ZMYND11 construct, PX458-AAVS1 (Addgene #113194) and AAVS1-Neo-M2rtTA (Addgene #60843) were nucleofected into 3 million cells by LONZA 4D-Nucleofector using program CB-150. After 3 days of recovery with CloneR, the cells were selected using 200 μg/ml Geneticin and replated at low density for clonal expansion. Single colonies were isolated and verified for biallelic targeting by genomic PCR.

To overexpress ZMYND11 in ZMYND11 knockout or wildtype hESCs, validated and karyotypically normal M2-rtTA hESC clones were infected with 3X-FLAG-ZMYND11 lentivirus, followed by the addition of 2 μg/ml puromycin after 72 h to select for resistant clones. For overexpression of ZMYND11 in Class I gene mutant lines, FUW-M2rtTA (Addgene #20342) and 3X-FLAG-ZMYND11 lentiviruses were both transduced, with puromycin selection applied to isolate resistant clones.

Intracellular FACS analysis

Cells were dissociated using Accutase and washed with DPBS. Approximately 5 × 10⁵ cells were treated with Zombie Violet Live/Dead Dye (1:1000) in DPBS for 30 min. After washing with DPBS, the cells were stained with surface antibodies for 30 min (if needed). Next, the cells were fixed using 1X Fix/Perm buffer from the BD Pharmingen Transcription Factor Buffer Set and incubated at 4 °C for 1 h. Cells were then washed twice with 1X Perm/Wash buffer and incubated with primary antibodies for 1 h. After 2 more washes, secondary antibodies were added, and cells were incubated at 4 °C for 1 h in the dark. Finally, the cells were washed 2 times and prepared for analysis.

For staining of cytoplasmic marker cleaved caspase-3, the cells were instead fixed using IC fixation buffer from the eBioscience Intracellular Fixation & Permeabilization Buffer Set and washed using 1X Permeabilization buffer.

Cell cycle exit analysis

Cell cycle analysis was performed according to the manufacturer’s protocol. Briefly, cells were treated with 10 μM BrdU for 30 min and harvested using Accutase. The cells were then fixed with 70% ethanol on ice for 30 min, followed by treatment with 2 N HCl/Triton X-100 at room temperature for 30 min, and neutralized with 0.1 M borate buffer. After washing, the cells were stained with FITC anti-BrdU or AF647 anti-BrdU for 30 min at room temperature. Cells were washed again and resuspended in 5 μg/ml propidium iodide and prepared for analysis.

Histology and immunohistochemistry

For 3D culture, organoids were fixed in 4% PFA overnight at 4 °C, followed by three washes with DPBS the next day. The organoids were then dehydrated in 30% sucrose solution overnight at 4 °C and sectioned at a 20-micron thickness. Sections were mounted onto glass slides, blocked with 5% Normal Donkey Serum (NDS) for 30 min and incubated with primary antibodies overnight at room temperature. Sections were then washed 3 times with 0.2% Tween-20 and incubated with secondary antibodies for 2 h at room temperature. The sections were then washed again, stained with DAPI, embedded in Fluoromount, covered with coverslips, and stored at 4 °C until imaging.

For 2D culture, cells were replated onto a glass coverslip 24 h before fixation. The cells were fixed in 4% PFA for 10 min at 4 °C, permeabilized with 0.5% Triton X-100 for 5 min and blocked with 5% NDS in 0.2% Tween-20 for 1 h. The cells were then incubated with primary antibodies overnight at 4 °C. The cells were then washed 3 times with 0.2% Tween-20 and incubated with secondary antibodies for 30 min at room temperature, washed and stained with DAPI. The cells were then embedded in Fluoromount and stored at 4 °C until imaging.

Western blot

Cells were dissociated and counted. An equal number of cells were harvested per experiment, lysed in 1X LDS Sample Buffer supplemented with 1X Sample Reducing Agent, then denatured at 95 °C for 5 min. Samples were then loaded and run on a 4–12% pre-cast Bis-Tris gel and transferred onto a Nitrocellulose membrane for 2 h. Membranes were blocked in 5% milk, and primary antibodies were incubated overnight at 4 °C. The next day, membranes were washed with 1X TBS buffer with 0.1% Tween-20 and incubated with HRP-conjugated secondary antibodies for 1 h. Blots were visualized using SuperSignal Chemiluminescent Substrate.

RNA extraction and qPCR

RNA was extracted using TRIzol, followed by chloroform extraction. Purification of RNA was performed using RNeasy Kits and eluted in nuclease-free water. cDNA synthesis was performed with 1 μg of RNA using the High-Capacity cDNA Reverse Transcription Kit. qPCR was performed with SYBR Green using the QuantStudio 3 Real-Time PCR system. For semi-qPCR, isoform-specific primer pairs were designed and used to amplify both the inclusion and exclusion products. Products were resolved on a high percentage (3%) agarose gel electrophoresis. Calculation of percent spliced in was done by measuring the intensity of the inclusion isoform band divided by the sum of the intensity of the inclusion and exclusion isoform bands using ImageJ. Calculation of non-brain isoform percent was done similarly by measuring the intensity of the non-brain isoform band divided by the sum intensity. Heatmap was generated using “pheatmap” in R (row-wise scaled by Z-score transformation). All primer sequences were listed in Supplementary Data 5.

Scratch/Migration assay

LTNSCs were seeded onto 6-well plates at a density of 3 × 10⁵ cells/cm² and infected with shRNA viruses the next day before puromycin selection. After the cells became 100% confluent, a scratch was performed on the monolayer using a sterile 200 μl micropipette tip. The cells were carefully washed with the medium to remove non-adherent cells, and cultured for an additional 2 days, monitoring cell migration every 24 h. The rate of migration was measured by quantifying the distance that the cells moved from the edge of the scratch to the center as previously described⁵⁴.

RNA-seq and gene expression analysis

Total RNA for hESCs was poly-A enriched using oligodT Dynabeads (Invitrogen), and strand-specific RNA-seq libraries were constructed using TruSeq Stranded mRNA library kit, following the manufacturers’ instructions. Libraries were submitted to the Roy J. Carver Biotechnology Center at the University of Illinois at Urbana-Champaign (UIUC) for paired-end sequencing (150 bp). Total RNA for cortical neural stem cells was submitted to Novogene for paired-end sequencing (poly-A enriched, 150 bp). Raw FASTQ files were quality checked using FastQC. Adapter sequences were trimmed using fastp with parameters “-q 20 -n 7”. Trimmed reads were then aligned to ENSEMBL GRCh38 using HISAT2 with default parameters. A raw count file was generated using featureCounts in Subread with default parameters, and genes with no raw counts in >= 75% of samples were filtered out. Downstream analysis was done using R packages. Normalized counts per million (cpm) values were generated using “cpm” in “edgeR”. Principal component analysis was performed using “plotPCA” in “DESeq2”. Differential gene expression analysis was performed using “DESeq2” with P.adjust <0.05 and fold change > 1.5. Gene ontology analysis was performed using “clusterProfiler” with P.adjust calculated using the Benjamini-Hochberg method. Plots were generated using “ggplot2”.

Alternative splicing analysis using rMATS

Input alignment files for rMATS splicing analysis were generated either using HISAT2 with trimmed reads or STAR with raw reads with default parameters. Results from rMATS were further filtered by removing low coverage events (avg_read >5) using “maser” in R. Significant splicing events were identified with FDR < 0.05 and percent spliced-in change (ΔPSI) > 0.1. Sashimi plots were generated using rmats2sashimiplot. For exon usage principal component analysis, the PSI matrix was generated using rMATS with parameters “--statoff”. PSI matrix was then imputed by KNN method for missing values using “impute” in R. PCA was conducted using the “prcomp” function in R.

For RNA-seq datasets reanalysis in Guo et al.⁹ and Wen et al.¹⁰, we only used HISAT2-rMATS for the alternative splicing analysis.

Alternative splicing analysis using AltAnalyze

For AltAnalyze, we only input BAM files generated by STAR as indicated by the user manual. Significant splicing events were identified with raw P value < 0.05 and ΔPSI > 0.1. Events from AltAnalyze were more than those from rMATS as AltAnalyze summarized events at each individual exon-exon or exon-intron junction while rMATS summarized events of individual exon. Therefore, a significant changed spliced exon or intron might have more than one event reported in AltAnalyze.

Overlapping significant spliced exon events by STAR-AltAnalyze with those from HISAT2-rMATS or STAR-rMATS were performed using genomic coordinates of the target exon. To expand the potential overlap, events that have matched upstream & downstream exon coordinates and either matched 5’ target exon coordinate or 3’ target exon coordinate were also included. Plus, events predicted from different methods would have to be in the same direction (both predicted to be spliced-in or spliced-out) to be considered as overlapped. These overlapped events were subsequently called high-confidence differential alternative splicing (DAS) events for further analysis.

Tissue-specific PSI values for high-confidence DAS events were extracted from VastDB database (https://vastdb.crg.eu) according to target exon coordinates in ENSEMBL GRCh38 and categorized into either brain tissue or non-brain tissue for comparison. Heatmap was generated using “pheatmap” in R (row-wise scaled by Z-score transformation) after missing values were imputed by the KNN method using “impute” in R. 506 events with available tissue-specific PSI values are shown (1 event with the same values across tissues is omitted). Correlation plots were generated using Prism.

CUT&RUN, library preparation and analysis

CUT&RUN method was as previously described⁵⁵. Briefly, 5 × 10⁵ cells were harvested and bound to Concanavalin A beads. These cell-bound beads were incubated with antibodies overnight at 4 °C and treated with pAG-MNase for 1 h at 4 °C. Calcium was added to start pAG-MNase digestion for 30 min, and CUT&RUN fragments were released and purified by phenol-chloroform extraction. DNA concentration was measured by the Quantus fluorometer. For library preparation, we followed NEBNext Ultra II DNA Library Prep Kit for Illumina and performed size selection after PCR enrichment of adapter-ligated library using MagBind Magnetic Beads. Libraries were sent to Novogene for sequencing, aiming for 10–20 million reads.

CUT&RUN sequencing reads were quality checked using FastQC, and adapters were trimmed using fastp with parameters “-q 20 -n 7”. Reads were aligned to ENSEMBL GRCh38 using bowtie2 with parameters “--very-sensitive --no-unal --no-mixed --no-discordant -I 10 -X 700”. Low-quality (MAPQ < 10) reads were discarded using Samtools view with parameters “-q 10”. Duplicated reads were discarded using Picard MarkDuplicates with parameters “REMOVE_DUPLICATES=true”. Only uniquely mapped reads were used for subsequent analysis after these filters. Bigwig files were generated using bamCompare in deepTools with parameters “--binSize 20 --normalizeUsing RPKM”. For samples with replicates, BAM files were merged before generating bigwig files. Visualization was performed using computeMatrix with parameters “scale-regions --beforeRegionStartLength 5000 --regionBodyLength 5000 --afterRegionStartLength 5000 --skipZeros --missingDataAsZero” and plotHeatmap in deepTools. Spearman correlation was performed using multiBigwigSummary and plotCorrelation with default parameters in deepTools. Peak calling was performed using getDifferentialPeaksReplicates.pl in HOMER to call ZMYND11-bound regions versus the ZMYND11 KO sample and FLAG-ZMYND11-bound regions versus no doxycycline sample with poisson P value < 0.05 and fold change > 1. The KO sample was used as the background control for ZMYND11 peak calling because we observed a small but consistent enrichment at the TSS, which we interpreted as nonspecific pAG-MNase binding at open chromatin regions. For ZMYND11 binding calling in the FLAG-ZMYND11 overexpression experiment, we used rabbit IgG as the control and peaks were called using findPeaks.pl. H3K36me3 enriched peaks were called also against rabbit IgG control. Peaks overlapping with ENCODE blacklist regions were discarded. The blacklist file was downloaded from GitHub (https://github.com/Boyle-Lab/Blacklist/blob/master/lists/hg38-blacklist.v2.bed.gz). Overlapping of peaks was performed using bedtools intersect with default parameters. Annotation of peaks was performed using UROPA with parameters “feature: transcript, distance: 10000, attribute. value: protein_coding”. Clustering of ZMYND11-bound genes was done using plotHeatmap with parameters “--kmeans 3” in deepTools. Gene ontology analysis was similarly performed using “clusterProfiler” in R.

Re-analysis of previous ZMYND11 binding datasets was similarly performed using the same pipeline (mouse datasets mapped onto ENSMBL GRCm39, human datasets mapped onto ENSEMBL GRCh38) except during alignment without “-I 10 -X 700” parameters in bowtie2.

Overlapping of peaks with DAS events was similarly performed using bedtools intersect with default parameters. Distance of DAS events to TSS was calculated and plotted using “ChIPseeker” in R.

Quantification, statistical analysis and reproducibility

All data represent mean ± SEM. Statistical significance was determined by parametric tests including unpaired two-tailed t-test for two groups and analysis of variance (ANOVA) for more than two groups with post hoc tests as indicated. A P value < 0.05 was considered statistically significant. P.adjust values in gene ontology analysis were calculated using the Benjamini-Hochberg method. Prism 8 software was used for statistical analysis. For all cell culture experiments, n represents the total number of independent differentiations. Each experiment was repeated at least 3 times.

Materials availability

All materials used in our analysis is available to any researcher for purposes of reproducing or extending the analysis. Cell lines can be provided through materials transfer agreements (MTAs).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

All data used in our analysis is available to any researcher for purposes of reproducing or extending the analysis. The RNA and CUT&RUN sequencing data generated in this paper are uploaded to GEO with accession numbers: GSE279072 and GSE279068, respectively. This dataset also includes raw count tables of RNA-seq and bigWig files of CUT&RUN-seq. Source data are provided with this paper.

Code availability

All code used in our analysis is available to any researcher for purposes of reproducing or extending the analysis. This paper does not report original code.

References

Deevy, O. & Bracken, A. P. PRC2 functions in development and congenital disorders. Development 146, dev181354 (2019).
Article CAS PubMed PubMed Central Google Scholar
Ho, P. J., Lloyd, S. M. & Bao, X. Unwinding chromatin at the right places: how BAF is targeted to specific genomic locations during development. Development 146, dev178780 (2019).
Article CAS PubMed PubMed Central Google Scholar
G”kbuget, D. & Blelloch, R. Epigenetic control of transcriptional regulation in pluripotency and early differentiation. Development 146, dev164772 (2019).
Article Google Scholar
Ninova, M., Fejes T¢th, K. & Aravin, A. A. The control of gene expression and cell identity by H3K9 trimethylation. Development 146, dev181180 (2019).
Article CAS PubMed PubMed Central Google Scholar
De Rubeis, S. et al. Synaptic, transcriptional and chromatin genes disrupted in autism. Nature 515, 209–215 (2014).
Article PubMed PubMed Central Google Scholar
Satterstrom, F. K. et al. Large-scale exome sequencing study implicates both developmental and functional changes in the neurobiology of autism. Cell 180, 568–584.e523 (2020).
Article CAS PubMed PubMed Central Google Scholar
Yates, T. M. et al. ZMYND11-related syndromic intellectual disability: 16 patients delineating and expanding the phenotypic spectrum. Hum. Mutat. 41, 1042–1050 (2020).
Article CAS PubMed Google Scholar
Oates, S. et al. ZMYND11 variants are a novel cause of centrotemporal and generalised epilepsies with neurodevelopmental disorder. Clin. Genet 100, 412–429 (2021).
Article CAS PubMed Google Scholar
Guo, R. et al. BS69/ZMYND11 reads and connects histone H3.3 lysine 36 trimethylation-decorated chromatin to regulated pre-mRNA processing. Mol. Cell 56, 298–310 (2014).
Article CAS PubMed PubMed Central Google Scholar
Wen, H. et al. ZMYND11 links histone H3.3K36me3 to transcription elongation and tumour suppression. Nature 508, 263–268 (2014).
Article ADS CAS PubMed PubMed Central Google Scholar
Lian, C. et al. Epigenetic reader ZMYND11 noncanonical function restricts HNRNPA1-mediated stress granule formation and oncogenic activity. Signal Transduct. Target. Ther. 9, 258 (2024).
Article CAS PubMed PubMed Central Google Scholar
Tchieu, J. et al. A modular platform for differentiation of human pscs into all major ectodermal lineages. Cell Stem Cell 21, 399–410.e397 (2017).
Article CAS PubMed PubMed Central Google Scholar
Cederquist, G. Y. et al. A multiplex human pluripotent stem cell platform defines molecular and functional subclasses of autism-related genes. Cell Stem Cell 27, 35–49.e36 (2020).
Article CAS PubMed PubMed Central Google Scholar
Paşca, A. M. et al. Functional cortical neurons and astrocytes from human pluripotent stem cells in 3D culture. Nat. Methods 12, 671–678 (2015).
Article PubMed PubMed Central Google Scholar
Qian, X. et al. Brain-region-specific organoids using mini-bioreactors for modeling ZIKV exposure. Cell 165, 1238–1254 (2016).
Article CAS PubMed PubMed Central Google Scholar
Fan, X. et al. Single-cell transcriptome analysis reveals cell lineage specification in temporal-spatial patterns in human cortical development. Sci. Adv. 6, eaaz2978 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Plotnik, J. P. & Hollenhorst, P. C. Interaction with ZMYND11 mediates opposing roles of Ras-responsive transcription factors ETS1 and ETS2. Nucleic Acids Res 45, 4452–4462 (2017).
CAS PubMed PubMed Central Google Scholar
Armache, A. et al. Histone H3.3 phosphorylation amplifies stimulation-induced transcription. Nature 583, 852–857 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Bernasocchi, T. et al. Dual functions of SPOP and ERG dictate androgen therapy responses in prostate cancer. Nat. Commun. 12, 734 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Devoucoux, M. et al. Oncogenic ZMYND11-MBTD1 fusion protein anchors the NuA4/TIP60 histone acetyltransferase complex to the coding region of active genes. Cell Rep. 39, 110947 (2022).
Article CAS PubMed Google Scholar
Irimia, M. et al. A highly conserved program of neuronal microexons is misregulated in autistic brains. Cell 159, 1511–1523 (2014).
Article CAS PubMed PubMed Central Google Scholar
Gonatopoulos-Pournatzis, T. & Blencowe, B. J. Microexons: at the nexus of nervous system development, behaviour and autism spectrum disorder. Curr. Opin. Genet. Dev. 65, 22–33 (2020).
Article CAS PubMed Google Scholar
Tapial, J. et al. An atlas of alternative splicing profiles and functional associations reveals new regulatory programs and genes that simultaneously express multiple major isoforms. Genome Res. 27, 1759–1768 (2017).
Article CAS PubMed PubMed Central Google Scholar
Hauser, S. et al. Full-length L1CAM and not its Δ2Δ27 splice variant promotes metastasis through induction of gelatinase expression. PLoS One 6, e18989 (2011).
Article ADS CAS PubMed PubMed Central Google Scholar
Wang, Z. N. et al. High expression of PTBP1 promote invasion of colorectal cancer by alternative splicing of cortactin. Oncotarget 8, 36185–36202 (2017).
Article PubMed PubMed Central Google Scholar
Lee, A. R. et al. Alternative RNA splicing of the MEAF6 gene facilitates neuroendocrine prostate cancer progression. Oncotarget 8, 27966–27975 (2017).
Article PubMed PubMed Central Google Scholar
Tchieu, J. et al. NFIA is a gliogenic switch enabling rapid derivation of functional human astrocytes from pluripotent stem cells. Nat. Biotechnol. 37, 267–275 (2019).
Article CAS PubMed PubMed Central Google Scholar
Koch, P., Opitz, T., Steinbeck, J. A., Ladewig, J. & Brüstle, O. A rosette-type, self-renewing human ES cell-derived neural stem cell with potential for in vitro instruction and synaptic integration. Proc. Natl Acad. Sci. USA 106, 3225–3230 (2009).
Article ADS CAS PubMed PubMed Central Google Scholar
Zhang, X. et al. Cell-type-specific alternative splicing governs cell fate in the developing cerebral cortex. Cell 166, 1147–1162.e1115 (2016).
Article CAS PubMed PubMed Central Google Scholar
Su, C. H. D., D. & Tarn, W. Y. Alternative splicing in neurogenesis and brain development. Front. Mol. Biosci. 5, 12 (2018).
Article ADS PubMed PubMed Central Google Scholar
Mulligan, M. R. & Bicknell, L. S. The molecular genetics of nELAVL in brain development and disease. Eur. J. Hum. Genet. 31, 1209–1217 (2023).
Article CAS PubMed PubMed Central Google Scholar
Hong, D. & Iakoucheva, L. M. Therapeutic strategies for autism: targeting three levels of the central dogma of molecular biology. Transl. Psychiatry 13, 58 (2023).
Article CAS PubMed PubMed Central Google Scholar
Harrison-Uy, S. J. & Pleasure, S. J. Wnt signaling and forebrain development. Cold Spring Harb. Perspect. Biol. 4, a008094 (2012).
Article PubMed PubMed Central Google Scholar
Le Dréau, G. BuMPing into neurogenesis: how the canonical BMP pathway regulates neural stem cell divisions throughout space and time. Frontiers in Neuroscience 15,https://doi.org/10.3389/fnins.2021.819990 (2022).
Steinhart, Z. & Angers, S. Wnt signaling in development and tissue homeostasis. Development 145, dev146589 (2018).
Article PubMed Google Scholar
Wang, R. N. et al. Bone Morphogenetic Protein (BMP) signaling in development and human diseases. Genes Dis. 1, 87–105 (2014).
Article PubMed PubMed Central Google Scholar
Kwan, V., Unda, B. K. & Singh, K. K. Wnt signaling networks in autism spectrum disorder and intellectual disability. J. Neurodev. Disord. 8, 45 (2016).
Article PubMed PubMed Central Google Scholar
Caracci, M. O. et al. Wnt/β-catenin-dependent transcription in autism spectrum disorders. Front Mol. Neurosci. 14, 764756 (2021).
Article CAS PubMed PubMed Central Google Scholar
Karwacki-Neisius, V. et al. WNT signalling control by KDM5C during development affects cognition. Nature 627, 594–603 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
Subashini, C. et al. Wnt5a is a crucial regulator of neurogenesis during cerebellum development. Sci. Rep. 7, 42523 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Parikshak, N. N. et al. Genome-wide changes in lncRNA, splicing, and regional gene expression patterns in autism. Nature 540, 423–427 (2016).
Article ADS CAS PubMed PubMed Central Google Scholar
Weyn-Vanhentenryck, S. M. et al. HITS-CLIP and integrative modeling define the rbfox splicing-regulatory network linked to brain development and autism. Cell Rep. 6, 1139–1152 (2014).
Article CAS PubMed PubMed Central Google Scholar
Conboy, J. G. Developmental regulation of RNA processing by Rbfox proteins. Wiley Interdiscip Rev RNA 8, https://doi.org/10.1002/wrna.1398 (2017).
Paulsen, B. et al. Autism genes converge on asynchronous development of shared neuron classes. Nature 602, 268–273 (2022).
Article ADS CAS PubMed PubMed Central Google Scholar
Meng, X. et al. Assembloid CRISPR screens reveal impact of disease genes in human neurodevelopment. Nature 622, 359–366 (2023).
Article ADS CAS PubMed PubMed Central Google Scholar
Li, C. et al. Single-cell brain organoid screening identifies developmental defects in autism. Nature 621, 373–380 (2023).
Article ADS CAS PubMed PubMed Central Google Scholar
Jin, X. et al. In vivo Perturb-Seq reveals neuronal and glial abnormalities associated with autism risk genes. Science 370, eaaz6063 (2020).
Article CAS PubMed PubMed Central Google Scholar
Willsey, H. R. et al. Parallel in vivo analysis of large-effect autism genes implicates cortical neurogenesis and estrogen in risk and resilience. Neuron 109, 788–804.e788 (2021).
Article CAS PubMed PubMed Central Google Scholar
Weigel, B. et al. MYT1L haploinsufficiency in human neurons and mice causes autism-associated phenotypes that can be reversed by genetic and pharmacologic intervention. Mol. Psychiatry 28, 2122–2135 (2023).
Article CAS PubMed PubMed Central Google Scholar
Mall, M. et al. Myt1l safeguards neuronal identity by actively repressing many non-neuronal fates. Nature 544, 245–249 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Mercuri, E., Sumner, C. J., Muntoni, F., Darras, B. T. & Finkel, R. S. Spinal muscular atrophy. Nat. Rev. Dis. Prim. 8, 52 (2022).
Article PubMed Google Scholar
Luoni, M. et al. Whole-brain delivery of an instability-prone Mecp2 transgene improves behavioral and molecular pathological defects in mouse models of Rett syndrome. Elife 9, e52629 (2020).
Article CAS PubMed PubMed Central Google Scholar
Gholizadeh, S., Arsenault, J., Xuan, I. C. Y., Pacey, L. K. & Hampson, D. R. Reduced phenotypic severity following adeno-associated virus-mediated Fmr1 gene delivery in fragile X mice. Neuropsychopharmacology 39, 3100–3111 (2014).
Article CAS PubMed PubMed Central Google Scholar
Liang, C. C., Park, A. Y. & Guan, J. L. In vitro scratch assay: a convenient and inexpensive method for analysis of cell migration in vitro. Nat. Protoc. 2, 329–333 (2007).
Article CAS PubMed Google Scholar
Matsui, S. et al. Pioneer and PRDM transcription factors coordinate bivalent epigenetic states to safeguard cell fate. Mol. Cell 84, 476–489.e410 (2024).
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgments

We thank Drs. Rupa Sridharan (University of Wisconsin, Madison), Kenneth Campbell (CCHMC) and members of the Tchieu lab for their critical reading, helpful critiques, and advice on this study. Additionally, we would like to thank Dr. Qing (Richard) Lu (CCHMC) for providing the pGreenPuro vector. Funding: Cincinnati Children’s Research Foundation (JT, XC). U.S. Department of Defense grant W81XWH-22-1-0533 (JT). National Institutes of Health grant R01GM143161 (MI). Starr Foundation Tri-SCI-2023-013 (LS). Memorial Sloan Kettering Cancer Center grant P30CA008748 (LS).

Author information

Authors and Affiliations

Division of Developmental Biology and the Center for Stem Cell and Organoid Medicine (CuSTOM), Cincinnati Children’s Hospital Medical Center, Cincinnati, USA
Xuyao Chang, Satoshi Matsui, Cindy Huynh, Makiko Iwafuchi & Jason Tchieu
Graduate Program in Molecular and Developmental Biology, Cincinnati Children’s Hospital Medical Center, Cincinnati, USA
Xuyao Chang
Department of Pediatrics, University of Texas Southwestern Medical Center and Children’s Medical Center Dallas, Dallas, USA
Wenqi Li
Graduate Program in Immunology, University of Texas Southwestern Medical Center, Dallas, USA
Wenqi Li
WPI Premium Research Institute for Human Metaverse Medicine (WPI-PRIMe), Osaka University, Osaka, Japan
Satoshi Matsui
Division of Child and Adolescent Psychiatry, Cincinnati Children’s Hospital Medical Center, Cincinnati, USA
Craig Erickson
Department of Intelligent Systems Engineering, Indiana University Bloomington, Bloomington, USA
Feng Guo
Developmental Biology and Center for Stem Cell Biology, Memorial Sloan Kettering Cancer Center, New York, USA
Gustav Y. Cederquist & Lorenz Studer
Department of Pediatrics, University of Cincinnati College of Medicine, Cincinnati, USA
Makiko Iwafuchi, Amelle Shillington & Jason Tchieu
Division of Human Genetics, Cincinnati Children’s Hospital Medical Center, Cincinnati, USA
Amelle Shillington
Department of Biochemistry and Molecular Genetics, University of Illinois College of Medicine, Chicago, USA
Constantinos Chronis

Authors

Xuyao Chang
View author publications
Search author on:PubMed Google Scholar
Wenqi Li
View author publications
Search author on:PubMed Google Scholar
Satoshi Matsui
View author publications
Search author on:PubMed Google Scholar
Cindy Huynh
View author publications
Search author on:PubMed Google Scholar
Craig Erickson
View author publications
Search author on:PubMed Google Scholar
Feng Guo
View author publications
Search author on:PubMed Google Scholar
Gustav Y. Cederquist
View author publications
Search author on:PubMed Google Scholar
Lorenz Studer
View author publications
Search author on:PubMed Google Scholar
Makiko Iwafuchi
View author publications
Search author on:PubMed Google Scholar
Amelle Shillington
View author publications
Search author on:PubMed Google Scholar
Constantinos Chronis
View author publications
Search author on:PubMed Google Scholar
Jason Tchieu
View author publications
Search author on:PubMed Google Scholar

Contributions

Conceptualization: X.C. and J.T. Methodology: X.C., W.L., S.M., C.H., C.E., F.G., G.Y.C., L.S., M.I., A.S., C.C., and J.T. Investigation: X.C., C.C., and J.T. Visualization: X.C. and J.T. Funding acquisition: J.T. Project administration: J.T. Supervision: J.T. Writing – original draft: X.C. and J.T. Writing – review & editing: X.C., W.L., S.M., C.H., C.E., F.G., G.Y.C., L.S., M.I., A.S., C.C., and J.T.

Corresponding author

Correspondence to Jason Tchieu.

Ethics declarations

Competing interests

LS is a co-founder and consultant of BlueRock Therapeutics and DaCapo BrainSciences, which represent financial competing interests. The remaining authors declare that they have no competing financial or non-financial interests.

Peer review

Peer review information

Nature Communications thanks Yin Shen, Xingjie Ren and the other anonymous reviewer(s) for their contribution to the peer review of this work. [A peer review file is available.]

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information File

Description of Additional Supplementary Information

Supplementary Data 1

Supplementary Data 2

Supplementary Data 3

Supplementary Data 4

Supplementary Data 5

Supplementary Data 6

Reporting summary

Transparent Peer Review file

Source data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Chang, X., Li, W., Matsui, S. et al. ZMYND11 functions in bimodal regulation of latent genes and brain-like splicing to safeguard corticogenesis. Nat Commun 16, 9010 (2025). https://doi.org/10.1038/s41467-025-64063-0

Download citation

Received: 22 November 2024
Accepted: 08 September 2025
Published: 09 October 2025
DOI: https://doi.org/10.1038/s41467-025-64063-0

Subjects

Abstract

Introduction

Results

ZMYND11 is critical for the differentiation of radial glial neural stem cells (NSCs) into intermediate progenitor cells (IPCs), a process impaired in a subset of autism risk genes

Elevated expression of developmental signaling transcripts impairs corticogenesis in ZMYND11 deficient cortical NSCs

ZMYND11 binding represses latent gene regulatory networks associated with BMP and WNT signaling pathways

ZMYND11 deficiency correlates with an alternative splicing switch from brain to non-brain isoform expression

ZMYND11 deficiency causes a tissue-specific isoform switch that affects migration and proliferation of cortical NSCs

Tissue-specific isoform switches found in additional risk factor lines for ASD can be improved by ZMYND11

Discussion

Methods

Maintenance and propagation of human embryonic stem cells (hESCs)

Human embryonic kidney 293 T (HEK293T) cell culture and lentivirus production

Three-dimensional guided brain organoid differentiation

Monolayer prefrontal (PFC) cortical differentiation and excitatory neuron induction

Maintenance of long-term neural stem cells (LTNSCs)

Spontaneous differentiation of hESCs

Plasmid constructs, site-directed mutagenesis and molecular cloning

ZMYND11 expression analysis in neurodevelopment

Generation of ZMYND11 mutant hESCs

Generation of doxycycline inducible ZMYND11 overexpression lines

Intracellular FACS analysis

Cell cycle exit analysis

Histology and immunohistochemistry

Western blot

RNA extraction and qPCR

Scratch/Migration assay

RNA-seq and gene expression analysis

Alternative splicing analysis using rMATS

Alternative splicing analysis using AltAnalyze

CUT&RUN, library preparation and analysis

Quantification, statistical analysis and reproducibility

Materials availability

Reporting summary

Data availability

Code availability

References

Acknowledgments

Author information

Authors and Affiliations

Contributions

Corresponding author

Ethics declarations

Competing interests

Peer review

Peer review information

Additional information

Supplementary information

Source data

Rights and permissions

About this article

Cite this article

Share this article

Search

Quick links