MSL2 disruption leads to autism-like behaviors, impairs neurogenesis, and alters histone H4K16 acetylation

Abstract

Neurodevelopmental disorders (NDDs) are frequently associated with dysregulation of epigenetic mechanisms. Recent studies have implicated variants in components of the Male-Specific Lethal (MSL) complex, including MSL2, in NDD pathogenesis. However, the pathogenic mechanisms underlying MSL2-related neurodevelopmental disorders remain poorly defined. Here, we elucidate the role of MSL2 in neurodevelopment and NDDs through comprehensive analysis of conditional knockout mice and multi-omics profiling. Aggregation of published and unpublished genomic data and case series revealed that pathogenic variants in MSL complex components, including MSL2, presented with shared phenotypes. Using a nervous system-specific Msl2 conditional knockout mouse model, we demonstrate that Msl2 deficiency leads to impaired social novelty recognition, learning deficits, and spatial memory impairments. These behavioral abnormalities are characterized by disrupted neocortical lamination, compromised proliferation capacity and differentiative ability of neural progenitor cells, and reduced neuronal migration. Mechanistically, Msl2-deficient brains exhibit significantly reduced levels of histone H4 lysine 16 acetylation (H4K16ac), particularly at promoter regions. Integrated RNA-seq, ChIP-seq, and ATAC-seq analyses revealed widespread transcriptional downregulation and loss of chromatin accessibility at NDD-related genes. Variants in the MSL2 DNA-binding domain disrupted its ability to target key regulatory genes including FMR1, confirming direct epigenetic regulation. Together, these findings demonstrate that MSL2 promotes neurodevelopment primarily by maintaining chromatin accessibility through H4K16ac deposition, and its disruption underlies the molecular and behavioral hallmarks of NDDs.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00018-026-06237-7.

Introduction

Neurodevelopmental disorders (NDDs) constitute a spectrum of conditions marked by early-onset impairments in brain function and development, including autism spectrum disorder (ASD), attention-deficit/hyperactivity disorder (ADHD), intellectual disability (ID), specific learning disorders, and communication disorders as defined in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition [1]. These disorders affect diverse neurological functions such as cognition, communication, social interaction, and motor skills [2]. While each disorder presents unique symptoms and diagnostic criteria, they often exhibit overlapping features and frequently co-occur in individuals with NDDs. The etiology of NDDs is multifaceted, with evidence suggesting the involvement of both genetic and environmental factors [3]. Dysregulation of epigenetic mechanisms is increasingly implicated in the etiology of NDDs, particularly in ASD and ID [4]. Genome-wide genetic studies have identified numerous risk-associated genetic variants, implicating genes that are significantly enriched for functions in epigenetic regulation, such as chromosome remodeling [5]. However, the precise mechanisms by which these genes influence epigenetic processes remain largely unknown.

Msl2 is a component of the MSL histone acetyltransferase complex responsible for the majority of histone H4 lysine 16 acetylation (H4K16ac), facilitating transcriptional activation [6–8]. In Drosophila, Msl2 plays a crucial role in regulating dosage compensation of X-linked genes to balance gene dosage between sexes [9, 10]. In mammals, Msl2 not only regulates H4K16ac but also acts as an E3 ubiquitin ligase, promoting histone H2B monoubiquitination, leading to gene activation [11, 12]. Recent research has highlighted the involvement of MSL2 in regulating allelic expression in mammals [13]. De novo likely gene-disrupted (LGD) variants in MSL2 have been linked to developmental delay (DD) with genome-wide significance by integrating exome-sequence data from 31,058 parent–offspring trios [14]. Recent case series have further implicated that MSL2 variants lead to a new neurodevelopmental syndrome [15, 16]. Despite growing evidence linking MSL2 variants to NDDs, the precise pathogenic mechanism through which these variants disrupt brain development, cognition, and social function remain unclear.

In this study, we characterized nervous system specific Msl2 conditional knockout mice and identified deficits in social interaction, learning, and spatial memory. Conditional deletion of Msl2 in neural precursors disrupted cortical patterning and reduced H4K16 acetylation, underscoring its essential role in neurodevelopment and epigenetic regulation. Integrated RNA-seq and chromatin accessibility analysis further demonstrated that the MSL-H4K16ac pathway modulates genes critical for neurodevelopment, thereby providing mechanistic insight into MSL2-related NDDs.

Methods

Variants and clinical data

We collected de novo LGD/MIS variants and clinical phenotype data of MSL2, MSL3 and KAT8 from three case series studies [16–18] and large-scale four published and unpublished ASD/NDD studies [14, 17–20]. Whole genome/exome sequencing VCFs for the unpublished SPARK cohort (accession number: SPARK_iWES_v3 and SPARK_iWGS_v1.1) were retrieved from the SFARI Base repository (https://base.sfari.org/), along with the corresponding phenotypic files for both SPARK (accession number: SPARK_Collection_V15_2025—03—31) and SSC (accession number: SSC Version 15.3 Phenotype Dataset) participants used in this study. De novo LGD and MIS variants were filtered based on our previous report [21]. We removed potential duplicate variants and sample overlaps across studies. All variants were annotated using ANNOVAR [22]. The study protocol was approved by the Institutional Review Board of the State Key Laboratory of Medical Genetics, School of Life Sciences, Central South University.

Mice

Msl2 conditional knockout mice were generated by crossing Msl2-floxed mice (Biocytogen Pharmaceuticals) with Nestin-Cre mice (Beijing Weishang Lide Biotechnology). Mice on the C57BL/6 J genetic background were housed under a standard 12 h light/12 h dark regimen with free access to chow and water. Only littermates comparable in age and sex were allocated to experimental groups. All procedures involving animals were conducted in accordance with the Central South University Institutional Animal Care and Use Committee (protocol 2022—2—3) and followed the guidelines set by the Association for Assessment and Accreditation of Laboratory Animal Care International. Ethical compliance with all relevant regulations for animal research was additionally verified by the university IRB.

Mice behavior tests

Behavioral tests were performed on 6- to 8-week-old male mice. Mice were housed in standard filter-top cages (3–5 animals per cage) with ad libitum access to water and rodent chow, maintained on a 12:12-h light/dark cycle (09:00–21:00 lighting) at 22 °C with 50–60% relative humidity. All behavioral assays were conducted blind to genotype, with at least 24 h of rest between tests. Data were analyzed using the Animal Video Tracking Analysis System (AVTAS, version 5.0; AniLab Software and Instruments). Exact sample sizes for each behavioral assay are provided in the respective figure legends.

Three-chamber test for social interaction and novelty behavior

A transparent acrylic resin apparatus with three chambers (each 40 cm × 30 cm) was used. Mice were habituated in the center chamber for 10 min, followed by a 10-min social interaction test with an unfamiliar mouse (stranger I) in one chamber. For the novelty preference test, a second unfamiliar mouse (stranger II) replaced the cap in the opposite chamber, and the same mouse was video-tracked for a further 10-min period. Chambers were cleaned with 25% ethanol between trials. Behavioral data were automatically captured with a computerized video-tracking platform.

Marbles burying test

Mice were placed in standard cages with 5 cm of bedding. Glass marbles soaked in 25% ethanol and rinsed with deionized water were arranged in rows. The number of marbles buried (2/3 submerged) was recorded over 30 min.

Self-grooming, digging and standing

Mice were placed in an open-field apparatus (20 cm diameter × 30 cm height) with 1 cm of bedding. Movements were recorded for the last 6 min of a 10-min exploration period.

Morris water maze

Mice were tested in a circular tank (122 cm diameter × 51 cm height) filled with water colored with powdered milk. The experiment included a 4-day Hidden Platform Test and a Probe Trial. Mice were placed in the water once daily from each of four entry points, and the time to locate the hidden platform was recorded. After platform removal, spatial exploration was assessed by recording time spent in the target quadrant, distance traveled, and platform crossings.

Novel-object recognition test

Mice were placed in a plexiglass box (72 cm × 72 cm × 40 cm) with two identical objects. After a 10-min habituation, one object was replaced with a novel object, and exploration time was recorded for 10 min.

Y-maze test

Mice were placed in one arm of a Y-maze (40 cm × 8 cm × 20 cm) and allowed to explore freely for 5 min. Spontaneous alternation was calculated as the number of triadic arm entries divided by the maximum possible alternations (total arm entries minus 2).

Open-field test

Mice were placed in an open-field arena (72 cm × 72 cm × 40 cm) and allowed to explore for 10 min. Time spent in the central area (4 squares) and perimeter area (12 squares) was recorded.

Light–dark box test

Mice were placed in the dark chamber of a two-chamber apparatus (44 cm × 21 cm × 21 cm) and allowed to explore for 10 min. Time and distance spent in each chamber were measured.

Elevated plus maze test

Mice were placed in an elevated plus maze (two open and two closed arms, 5 cm × 30 cm) and recorded for 10 min. Time spent and entries into each arm were scored.

Western blot analysis

Western blot analysis was performed on cortical tissues harvested from embryonic and postnatal mice at the indicated stages (including E12.5, E14.5, E18.5, P0, P7, P14, P30, P45, and P60). Tissues were lysed in 1% Triton X-100 buffer. After SDS-PAGE separation and electro-transfer to PVDF membranes (Merck Millipore, IPVH00010), blots were blocked with 5% skim milk (Sigma Aldrich, A2058) at room temperature for 1 h and incubated overnight at 4 °C with primary antibodies, then for 1 h with HRP-conjugated secondaries (Jackson ImmunoResearch). Signals were developed with SuperSignal West Femto maximum-sensitivity reagent (Thermo Fisher). The following primary antibodies were used: MSL2 (Cell Signaling Technology, 44,006), H4K16ac (Cell Signaling Technology, 13,534), β-actin (Proteintech, 66,009—1).

Immunohistology

Embryonic mouse brains were fixed in 4% paraformaldehyde (PFA) at 4 °C for 24 h. After fixation, tissues were cryoprotected by sequential immersion in 15% and 30% sucrose solutions, embedded in optimal cutting temperature (OCT) compound, and snap-frozen at –80 °C. Cryo-preserved 20-µm coronal slices were first heat-fixed (60 °C, 120 min), rinsed in phosphate-buffered saline. For BrdU detection, sections underwent acid‑mediated antigen retrieval by treatment with 2 M HCl for 30 min at room temperature. To neutralize the acid, sections were then incubated in 0.1 M sodium borate buffer (pH 8.5) for 10 min. Subsequently, sections were washed extensively with PBS before proceeding to the blocking step. Non-specific binding was suppressed by 1 h incubation in 5% bovine serum albumin dissolved in 0.1% PBST. Slices were then probed with primary antibodies at 4 °C overnight and subsequently visualized with fluorophore-conjugated secondary reagents(Jackson ImmunoResearch) for 60 min at ambient temperature. Nuclear counterstaining was performed with 0.1 µg ml⁻¹ DAPI (Sigma-Aldrich, D9542). Finally, the sections were mounted with Fluoromount-G (SouthernBiotech, 0100—01). Confocal micrographs were collected on a Zeiss LSM 880 system and post-processed with ImageJ software. Analyses were performed on the primary somatosensory cortex (S1) at a coronal level corresponding to Bregma −1.5 mm. Cortical layers were identified based on DAPI nuclear staining and the characteristic distribution of layer-specific markers: Satb2 for layers II-IV, Ctip2 for layer V, and Tbr1 for layer VI. The boundaries between layers were defined by changes in cell density and marker expression patterns. Due to the well-characterized laminar organization of the S1 cortex, our quantitative analysis focused on this region.

Cell cycle kinetics analysis and nucleotide administration

Pregnant mice were injected intraperitoneally with 50 mg/kg of BrdU (Millipore-Sigma) or 5-Ethynyl-2’- deoxyuridine (EdU, Millipore-Sigma) thymidine analogs at E14.5 before being euthanized for embryo extraction and histological preparation. Ts was measured according to Houlihan and Feng (2014). EdU or BrdU-only pulses: timed-pregnant dams received i.p. EdU or BrdU (50 mg kg⁻¹) at E14.5 and were culled 0.5 h, 24 h or 4 d later. Embryonic brains were sectioned and EdU visualised with Click-iT kit C10338 (Thermo Fisher).

Neurosphere culture

E14.5 embryonic cortices were dissected, gently dissociated with Accutase (37 °C, 5 min; Thermo 00—4555—56), rinsed twice in DMEM, and brought to single-cell suspension in neurosphere medium. Cells were seeded at 3 × 10⁴–10⁵ cells/mL into untreated 6-well plates and grown in serum-free DMEM/F12-Gibco 11,320,033 plus B27 (Gibco 17,504,044), 10 ng/mL bFGF (Peprotech) and 20 ng/mL EGF (Invitrogen). For self-renewal assays, primary spheres were re-dissociated with Accutase, replated at 100 cells/mL in uncoated 24-well plates under identical conditions, and sphere diameters recorded on days 3, 5 and 7.

Cell cycle analysis

Neuronal cells were detached into single-cell suspension with Accutase, harvested, and pelleted (1,000 × g, 5 min). After discarding the supernatant, the cell pellet was rinsed with ice-cold PBS and then fixed by dropwise addition of 70% ethanol (pre-chilled) while gently vortexing, followed by overnight storage at 4 °C. The fixed samples were next incubated for 30 min at 37 °C under light protection in freshly prepared propidium iodide staining buffer (prepared from Beyotime Cell-Cycle Assay Kit, Cat# C1052) according to the supplier’s protocol. DNA content distributions were finally recorded on a Cytek DXP Athena flow cytometer and cell-cycle phases were modelled using ModFitLT 3.1 software.

Plasmids

The human MSL2 open reading frame(NM_018133.4), obtained from Youbio Biological Technology, was PCR-amplified and inserted into pcDNA3.1—3xFlag-C. To generate constructs for subsequent expression studies, the full-length MSL2 coding sequence was PCR-amplified using this plasmid as a template and subcloned into the mammalian expression vector pCAGGS. An HA tag sequence (5′-TAC CCA TAC GAC GTC CCA GAC TAC GCT-3′) was inserted in-frame immediately before the termination codon, resulting in the pCAGGS-MSL2-HA wild-type (WT) expression plasmid. All constructs were confirmed by Sanger sequencing.

Cell culture and transfection

HEK293T cells (catalog no. CBP60439, Cobioer) were maintained in DMEM supplemented with 10% fetal bovine serum (FBS; 10,100,147, Thermo Fisher Scientific) and 1% penicillin–streptomycin (15,140,122, Thermo Fisher Scientific) at 37 °C under 5% CO₂ in a humidified incubator. Transient transfections were carried out with Lipofectamine 3000 (L3000015, Thermo Fisher Scientific) according to the supplier’s instructions.

Antibodies

Primary antisera were sourced as follows: Satb2 (rabbit, 1:500, Abcam ab92446), Brn2 (mouse, 1:100, Santa Cruz sc—393,324), Ctip2 (rabbit, 1:100, Abcam ab18465), Tbr1 (rabbit, 1:400, CST 49661 s), Pax6 (rabbit, 1:100, BioLegend PRB-278P), Tbr2 (rabbit, 1:100, Abcam ab23345), BrdU (rat, 1:500, Abcam ab6326), Ki67 (rabbit, 1:400, CST 9129 s), MSL2 (rabbit, 1:1,000, CST 44006), H4K16ac (rabbit, 1:1,000, CST 13534) and HA-tag (rabbit, 1:1,000, CST 3724 s).

ChIP assays

For chromatin immunoprecipitation, confluent HEK293 cultures grown in 10 cm dishes were cross-linked for 10 min at room temperature with 1% formaldehyde (diluted from 16% stock, CST, 12,606), followed by 5 min quenching on ice with glycine. After two ice-cold PBS rinses, pellets were lysed in ChIP Sonication Cell Lysis Buffer; the crude suspension was gently transferred with a trimmed pipette tip into a Dounce homogeniser and stroked with a type-A pestle until translucent. Nuclei were then released by 10 min incubation on ice in ChIP Sonication Nuclear Lysis Buffer. Chromatin was sheared in Covaris microTUBES on a Covaris S2 ultrasonicator, and 10 µg aliquots were immunoprecipitated overnight at 4 °C with 2 µL anti-HA (CST, 3724S) or control IgG, plus Protein G magnetic beads (CST, 9006) added for the final 2 h. Bead–antibody complexes were resuspended in elution buffer and incubated at 65 °C, 1 200 rpm for 30 min (SCILOGEX SCI-100HCM-Pro). Cross-links were reversed by 2 h treatment with 5 M NaCl and proteinase K at 65 °C, and DNA was cleaned up on spin columns (CST, 10,010). Immunoprecipitated and input material were analysed by qPCR or high-throughput sequencing.

Multi-omics integrative analysis

This study integrated multi-omics data from two sources:

Publicly available datasets: RNA-seq (transcriptome), ChIP-seq (histone modification), and ATAC-seq (chromatin accessibility) data from Msl2-deficient neural progenitor cells were obtained from the Gene Expression Omnibus (GEO) under accession number GSE183764. These raw data were downloaded and reprocessed using standardized pipelines as described below.

In-house generated dataset: To assess the effect of patient-derived MSL2 mutations on DNA binding, we performed ChIP-seq using HA-tagged wild-type and mutant MSL2 constructs overexpressed in HEK293T cells. These data have been deposited in the Sequence Read Archive (SRA) under BioProject accession PRJNA1429780.

RNA-seq data analysis

Raw sequencing data were initially assessed for quality using FastQC (v0.12.1). Adapters were subsequently trimmed with Trim Galore (v0.6.10) to ensure high-quality sequences for further analysis. The processed reads were aligned to the GRCm38 mouse reference genome using HISAT2 (v2.2.1). The resulting BAM files were employed to quantify gene expression with HTSeq (v2.0.5), generating a count matrix suitable for differential expression analysis. Differentially expressed genes (DEGs) were identified using the DESeq2 package (v1.42.0), applying a significance threshold of |log2(Fold Change)|≥ 1 and p-value ≤ 0.05.

ChIP-seq/ATAC-seq data analysis

Raw sequencing data were evaluated for quality using FastQC. Adapters were trimmed with Trim Galore to ensure high-quality reads for subsequent analysis. The processed reads were aligned to the GRCm38 mouse reference genome or the human reference genome (GRCh38) using Bowtie2 (v2.5.4). Duplicated reads were then removed from the generated BAM files with Samtools to ensure accurate peak calling. Peaks were called from the deduplicated BAM files using MACS2 to identify enriched genomic regions. Differential peaks were determined with MACS2, applying a cutoff of likelihood ratio ≤ 0.001 and p-value ≤ 0.05 to define significant differences.

Statistics analysis

All experiments were performed at least three times. For normally distributed data, two-tailed unpaired or paired Student's t-test was used as appropriate. For non-normally distributed data, Mann–Whitney U test was used. For comparisons among more than two groups, one-way ANOVA was performed. For normally distributed data, Dunnett's multiple comparisons test was used when comparing all groups to a single control (WT), and Tukey's multiple comparisons test was used for comparisons among all groups. For non-normally distributed data, Kruskal–Wallis test with Dunn's post hoc test was used. For paired data, Wilcoxon matched-pairs signed rank test was used for comparisons between two conditions; Friedman test with Dunn's multiple comparisons test was used for comparisons among three or more paired groups. Two-way repeated measures ANOVA with Bonferroni's post hoc test was used for analysis of interaction effects between two factors. For behavioral data, a test for linear trend was incorporated within the one-way ANOVA framework to assess gene dosage effects, with genotype treated as an ordered variable (WT = 0, cHET = 1, cKO = 2). A two-tailed p-value < 0.05 was considered statistically significant.

Results

Genotype and phenotype patterns of MSL-related NDDs

Clinical genomic studies have established significant associations between functional variants in core MSL complex components-MOF(KAT8), MSL3, and MSL2 (Fig. 1A) and NDDs. To systematically delineate the variant pattern and compare genotype–phenotype correlations in MSL-related disorders, we analyzed published clinical data from three case series studies [16–18] and large-scale published and unpublished whole-exome or whole-genome sequencing cohorts [14, 19, 20] (see Methods), encompassing 35 pathogenic MSL2 variants, 20 MSL3 variants, and 12 KAT8 variants from 74 individuals (Supplementary Table 1, Fig. 1B). Our analysis revealed a shared phenotypic spectrum across all three genes, with near-complete penetrance of developmental delay/intellectual disability (DD/ID; 92.6%), speech delay (93.3%), and motor deficits (87.2%). Besides, we also noticed gene-specific phenotypic patterns. MSL2 variants carriers showed high prevalence of autism spectrum disorder (64.3%), while MSL3 variants were predominantly associated with feeding difficulties (81.3%) and hypotonia (76.5%). Patients with KAT8 variants frequently presented with seizures (70%) (Fig. 1C). This phenotypic convergence and divergence suggest both shared and subunit-specific pathogenic mechanisms. While no clinical cases and variants burden significance of MSL1 variants have been reported to date, our phenotypic clustering analysis suggests that the MSL complex regulates neurodevelopment through a coordinated epigenetic module.

Fig. 1.

Structural and mutational landscape of the mammalian MSL complex in neurodevelopmental disorders (A) Schematic representation of the core structure of the mammalian MSL complex, modified from Keller et al. [23]. The MSL complex comprises an MSL1 dimer, where the N-terminus interacts with MSL2 and the C-terminus binds to MOF(KAT8) and MSL3. Notably, the unstructured central region of MSL1 is contracted in mammals, resulting in a smaller complex size compared to Drosophila. (B) NDD-related variant patterns of three MSL complex components. (C) Phenotypic spectrum comparison of three MSL complex component related NDDs. Red boxes indicate clinical penetrance values. MRI, magnetic resonance imaging;/, not reported

Dysfunction of Msl2 in the nervous system leads to social and cognitive cognition impairments in mice

To investigate the role of MSL2 in brain development and MSL2-related disorder pathogenesis, we employed a conditionally loss-of-function approach by deleting Msl2 in the central nervous system. We generated mice harboring floxed Msl2 alleles and crossed them with Nestin-cre transgenic mice to achieve Cre-mediated recombination specifically in neuronal progenitors (Fig. S1A). Genotyping defined the experimental groups as follows: Msl2^f/f (wild-type controls, WT); Nestin-cre:Msl2^f/+ (conditional heterozygous, cHet), and Nestin-cre:Msl2^f/f (conditional homozygous, cKO). The cKO pups were born at expected Mendelian ratio and displayed no overt morphological abnormalities, with both cHet and cKO mice viable. Efficient deletion of Msl2 was confirmed by PCR analysis of genomic DNA and immunoblotting of protein extracts from E14.5 brains (Fig. S1 B-C). Notably, both cHet and cKO mice exhibited slower growth and decreased body weight compared to WT littermates (Fig. S1 D-E), mirroring the developmental delay observed in humans with MSL2 pathogenic variants.

To assess social behavior and social novelty recognition, we conducted the three-chamber test (Fig. 2A). All genotype groups demonstrated significantly greater interaction time with the stranger mouse compared to the inanimate objects chamber, with comparable social interaction preference index across groups. However, in the social novelty recognition phase, WT mice exhibited more time in the novel mouse chamber over the familiar one, while both cHet and cKO mice showed no such preference, indicating impaired social novelty recognition in Msl2-deficient mice. A significant linear trend was observed across genotypes (p for trend = 0.0036), demonstrating a gene dosage-dependent effect on social novelty behavior. The marble-burying test and a series of behavior observation was utilized to assess the repetitive behaviors. Quantitative analysis revealed no genotype dependent differences in marble bury frequency, standing, grooming, or digging (Fig. S2 A-B), suggesting that developmental Msl2 deficiency does not significant influence repetitive preference on the tested behaviors in mice.

Fig. 2.

Msl2 deficiency impairs cognitive and social behaviors in mice (A) Three-chamber test. Interaction time with object (O), stranger 1 (S1), and stranger 2 (S2) was compared, along with preference indexes. Social preference index calculated as (S1-O)/(S1 + O) and social novelty preference index calculated as (S2-S1)/(S2 + S1). n = 29 (Msl2^fl/fl), 17 (cHet), 22 (cKO). Statistical significance across genotypes was determined by one-way ANOVA followed by Tukey’s post-hoc test; comparisons between interaction times within each genotype were performed using two-tailed Student’s t test. Data are presented as mean ± SEM. (B) Morris water maze test. Escape latency during the learning phase, latency and number of exact crossings over the previously hidden platform during the probe phase, and swim speed during the probe phase were compared. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 indicate significant differences between WT and cKO groups on the indicated day. n = 32 (Msl2^fl/fl), 15 (cHet), 20 (cKO). Escape latency was analyzed using two-way repeated measures ANOVA followed by Sidak's multiple comparisons test. The latency to the platform, the number of platform crossings, and swim speed were compared using one-way ANOVA followed by Tukey's post-hoc test. Data are presented as mean ± SEM. (C) Novel-object recognition test. The time in compartments and discrimination index were compared. n = 25 (Msl2^fl/fl), 22 (cHet), 20 (cKO). Statistical significance across genotypes was determined by one-way ANOVA followed by Tukey’s post-hoc test; comparisons between interaction times within each genotype were performed using two-tailed Student’s t test. Data are presented as mean ± SEM. (D) Y maze test. The proportion of correct explorations of the three arms of the Y-maze during the experimental observation period was compared. n = 28 (Msl2.^fl/fl), 17 (cHet), 22 (cKO). Statistical analysis was performed using 1-way ANOVA. All data are represented as mean ± SEM. A test for linear trend was performed to assess gene dosage effects. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

Given that most patients with MSL2 variants exhibit intellectual disability, we evaluated cognitive function in our mouse model using established learning and memory paradigms. In the Morris water maze test, both cHet and cKO mice exhibited significantly prolonged escape latencies during spatial acquisition training and exhibited thigmotaxic behavior with disorganized search strategies during probe trials, indicating substantial deficits in spatial learning and long-term memory retention compared to WT controls (Fig. 2B). Linear trend analysis revealed significant gene dosage-dependent effects on both escape latency (p for trend = 0.0097) and platform crossings (p for trend = 0.0105), indicating that spatial memory is modulated by Msl2 gene dosage. Additionally, Msl2 cHET and cKO mice showed decreased exploration time and discrimination index in the novel object recognition test with significant linear trends (p for trend = 0.0014), as well as impaired spatial learning and short-term working memory in the Y-maze test with a significant linear trend across genotypes (p for trend = 0.0066) (Fig. 2 C-D). These consistent findings across multiple behavioral paradigms strongly support a critical role for Msl2 in cognitive function.

We employed anxiety-related behaviors using three standardized anxiety tests. While Msl2-cKO mice displayed increased anxiety-like behavior in the open field test, as evidenced by reduced center zone exploration (Fig. S2C), no significant genotype differences emerged in the elevated plus maze or light–dark box tests (Fig. S2 D-E). These findings indicate that Msl2 deficiency in the developing brain may selectively influence certain anxiety-related behaviors without causing generalized anxiety alterations.

Msl2 deficiency leads to altered neocortical lamination in mice

To explore the role of MSL2 in neurodevelopment, we characterized its spatiotemporal expression pattern in both human and mouse brains. Transcriptomic analysis of human developmental brain data (BrainSpan Atlas) demonstrates robust MSL2 mRNA expression during prenatal stage, with significant downregulation in postnatal periods (Fig. S3A). Similarly, examination of protein expression levels of Msl2 in the mouse cortex at different developmental stages via immunoblotting demonstrated elevated Msl2 expression from embryonic day 12.5 (E12.5) through birth (P0), with peak levels observed at E14.5 (Fig. S3B). This conserved expression pattern, marked by maximal expression during critical periods of neurogenesis, strongly implicates MSL2 in early neural development and cortical formation.

To assess Msl2's role in neurogenesis, we performed a comparative analysis of mouse brains at E18.5. Both cHET and cKO mice exhibited significant reductions in overall brain size, weight, and cortical area relative to WT littermates (Fig. 3A), demonstrating that Msl2 deletion disrupts normal brain development during embryogenesis. To substantiate these morphological findings, we conducted immunostaining to examine defects at the cellular level using layer-specific markers. Employing various antibodies specific to cortical lamination markers, we stained different layers of the neocortex. We observed a reduction in Satb2-positive and Brn2-positive cells, labelled the upper-layer cortex, in the cKO cerebral cortex at E18.5 compared to the WT cortex (Fig. 3 B-C). While no statistical difference was noted in cHET mice, a similar downward trend was evident. Furthermore, Tbr1-positive cells, labelled layer VI cortex, were decreased in the cKO cerebral cortex (Fig. 3 E). We did not observe significant differences in the numbers of Ctip2-positive cells, labelled layer V cortex, in cHET and cKO mouse cortex (Fig. 3D). These findings indicate altered cortical lamination in the Msl2-deficient brain at E18.5. To assess whether the embryonic cortical defects persist into adulthood, we examined whole brain morphology and cortical lamination in adult (P60) WT and Msl2 cHET mice. Similar to the observations at E18.5, adult cHET mice exhibited a modest but significant reduction in anterior–posterior length, while cortical length and cortical area remained unchanged (Fig. S4A). NeuN staining revealed that total cortical thickness was comparable between genotypes; however, consistent with the embryonic upper-layer neuron loss, regional analysis showed a selective reduction in neuronal density within the superficial layers (layers II–IV) of the primary somatosensory cortex in adult cHET mice (Fig. S4B). These findings indicate that partial loss of Msl2 leads to persistent structural abnormalities in the somatosensory cortex and has lasting effects on brain development.

Fig. 3.

Msl2 deficiency disrupts cortical development in a layer-specific manner at E18.5 (A) Quantitative comparison of brain weight, anterior–posterior (A-P) length, cortical length, and area in Msl2^fl/fl (n = 6), cHet (n = 6), and cKO (n = 10) mice at E18.5. Scale bar, 1 mm. (B) Littermate cortices stained with Satb2 from Msl2^fl/fl (n = 6), cHet (n = 5), and cKO (n = 6) mice at E18.5. The number of Satb2-positive cells per 10,000 μm² in layers 2–4 was compared. Scale bar, 20 μm. (C) Littermate cortices stained with Brn2 from Msl2^fl/fl (n = 6), cHet (n = 5), and cKO (n = 6) mice at E18.5. The number of Brn2-positive cells per 10,000 μm² in layers 2–4 was compared. Scale bar, 20 μm. (D) Littermate cortices stained with Ctip2 from Msl2^fl/fl (n = 6), cHet (n = 5), and cKO (n = 6) mice at E18.5. The number of Ctip2-positive cells per 10,000 μm² in layer 5 was compared. Scale bar, 20 μm. (E) Littermate cortices stained with Tbr1 from Msl2^fl/fl (n = 6), cHet (n = 5), and cKO (n = 6) mice at E18.5. The number of Tbr1-positive cells per 10,000 μm² in layer 6 was compared. Scale bar, 20 μm. (F) Littermate cortices stained with EdU from Msl2.^fl/fl (n = 6), cHet (n = 6), and cKO (n = 5) mice at E18.5. The distribution of EdU + cells in different cortical zones 4 days after EdU injection at E14.5 was calculated and compared. Scale bar, 20 μm. All data are represented as mean ± SEM. Statistical comparisons were performed using 1-way and 2-way ANOVA. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

To examine the effect of Msl2 deletion on the migration of neurons generated in developmental stages, we labeled newborn neurons by 5-ethynyl-2’-deoxyuridine (EdU) pulse at E14.5—a time point corresponding to the peak production of upper-layer neurons and analyzed the distribution of the labeled cells at E18.5, when migration is largely complete (Fig. 3F). We observed that the majority of the labeled control cells had entered the cortical plate (CP). In contrast, in the Msl2 deletion mice, a large number of the labeled cells remained in the intermediate zone (IZ). Quantification confirmed that neuronal migration was significantly affected at E18.5 by Msl2 deletion. These results suggest that Msl2 is required for the migration of neurons. Collectively, our data demonstrate that Msl2 plays a critical role in neocortical development, with its deficiency leading to altered cortical lamination and impaired neuronal migration. These findings highlight the importance of Msl2 in the proper formation of the cerebral cortex and provide insight into the developmental mechanisms that may contribute to MSL2-associated neurodevelopmental phenotypes.

Msl2 deficiency leads to decreased NPC proliferation and differentiation

The cortical abnormalities likely originate from impaired proliferation, differentiation, or migration of neural progenitor cells (NPCs). To examine these cellular processes, we performed immunostaining analysis of embryonic brain sections using radial glial cell (RGC) marker Pax6 and intermediate progenitor (IP) marker Tbr2. Quantitative assessment revealed significant reductions in both the thickness of the Pax6⁺ RGC layer and the density of Tbr2⁺ IP populations in cHET and cKO brains relative to WT controls at E14.5 (Fig. 4 A-B), demonstrating Msl2's essential role in maintaining the NPC pool during cortical development. To characterize NPC deficiencies, we systematically evaluated cell cycle processes. Cell cycle analysis via dual BrdU/EdU labeling showed prolonged S-phase duration (TS) in cKO NPCs (Fig. 4C), indicating proliferation defects. Complementary Ki67 staining revealed that a greater number of NPCs were retained within the cell cycle and failed to exit it, showing impaired NPC differentiation in Msl2-deficient brain. This disruption could have significant implications for brain development and function (Fig. 4D).

Fig. 4.

Msl2 deletion alters neural progenitor proliferation and cell-cycle kinetics in the E14.5 cortex (A) Littermate cortices stained with Pax6 and EdU from Msl2^fl/fl (n = 5), cHet (n = 6), and cKO (n = 5) mice at E14.5. The thickness and number of Pax6 + (VZ) cells, as well as the proportion of Pax6 + EdU + to EdU + cells in different cortical zones 30 min after EdU injection, were calculated and compared. Scale bar, 20 μm. (B) Littermate cortices stained with Tbr2 and EdU from Msl2^fl/fl (n = 5), cHet (n = 6), and cKO (n = 5) mice at E14.5. The thickness and number of Tbr2 + (SVZ) cells, as well as the proportion of Tbr2 + EdU + to EdU + cells in different cortical zones 30 min after EdU injection, were calculated and compared. Scale bar, 20 μm. (C) Schematic diagram of the experimental design and representative immunohistological (IH) images of S phase sequential labeling analysis of neural progenitor cells (NPCs) from Msl2^fl/fl (n = 5), cHet (n = 5), and cKO (n = 5) mice. Edu-Brdu double-stained cortical sections at E14.5 are shown. S phase durations were calculated (Ts = Ti/(Lcells/Scells)) and compared. Scale bar, 20 μm. (D) Littermate cortices stained with Ki67 and Brdu from Msl2^fl/fl (n = 6), cHet (n = 6), and cKO (n = 6) mice at E14.5. The number of Brdu-positive cells per 40,000 μm² 24 h after Brdu injection, as well as the proportion of Brdu +/Ki67– cells to total Brdu + cells, were calculated and compared. Statistical Analysis: All data are represented as mean ± SEM. Statistical comparisons were performed using 1-way ANOVA. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

To further confirm the impaired NPC proliferation in Msl2 deficiency mice, we conducted neurosphere formation assays and flow cytometric cell cycle analysis. Neurosphere formation capacity was markedly reduced in cKO-derived cultures, yielding smaller neurospheres than WT controls (Fig. 5A). Flow cytometric cell cycle analysis showed abnormal distribution with decreased G1-phase and increased S-phase populations in mutant NPCs (Fig. 5B). Together, these data suggest that Msl2 regulates cortical neurogenesis by coordinating NPC proliferation dynamics and differentiation, with its deficiency leading to fundamental disruptions in cortical development.

Fig. 5.

Msl2 deficiency impairs neurosphere formation and alters cell cycle distribution in mNPCs (A) Representative images of neurosphere formation from mNPCs cultured from Msl2^fl/fl and cKO embryos. Scale bar, 10 µm. The diameters of neurospheres were calculated and compared. Experiments were performed in three trials, with statistics based on the average of each condition across trials. (B) Flow cytometric analysis of cell cycle parameters in neurospheres. Upper panel, Msl2^fl/fl; lower panel, cKO. Bar graphs show the distribution of cell populations in G0/G1, S and G2/M phases. Statistical comparisons were performed using two-tailed t tests. All data are represented as mean ± SEM. ns, not significant; *P < 0.05, ***P < 0.001

Msl2 deletion causes H4K16 acetylation reduced in the cortex of mice

MSL2 serves as a pivotal component within the MSL complex, responsible for the bulk H4K16ac [23]. We hypothesized that molecular alterations regulated by H4K16ac contribute to the impaired neurogenesis and behaviors observed in Msl2-deficient mice. To delineate the H4K16ac expression pattern in Msl2-deficient mice, we conducted immunostaining in the embryonic cerebrocortical neuroepithelium and hippocampal primordium of Msl2-deficient and WT mice. Our observations revealed a significant reduction in H4K16ac levels in both conditional cHET and cKO mice compared to WT counterparts at E14.5 (Fig. 6A) and E18.5 (Fig. 6B). To corroborate these findings, immunoblotting was performed, demonstrating a consistent reduction in H4K16ac levels in cortical lysates of cHET and cKO mice at E14.5 (Fig. 6C). These results indicate that Msl2 deficiency leads to diminished H4K16ac levels in the cortex.

Fig. 6.

Msl2 loss reduces cortical H4K16ac (A) and (B) H4K16ac immunofluorescence of coronal brain sections from Msl2^fl/fl, cHet, and cKO mice at E14.5 (a, top) and E18.5 (b, top). Scale bars, 200 μm. The area within the white box is magnified on the right. Scale bars, 20 μm. Immunofluorescence intensity of H4K16ac was calculated and compared (A and B, bottom). The x-axis "Distance" represents the radial distance from the lateral ventricular wall outward into the cortical parenchyma. The bar graphs show the average fluorescence intensity quantified across the entire field of view (FOV). (C) Immunoblots of H4K16ac and Msl2 in lysates from brain cortical tissue of Msl2.^fl/fl, cHet, and cKO mice at E14.5. (D) Immunoblots of H4K16ac and Msl2 in the cerebral cortex during the embryonic and postnatal periods (left). Correlation analysis between H4K16ac and Msl2 expression levels (right). Statistical Analysis: Data are represented as mean ± SEM. Statistical comparisons were performed using 1-way ANOVA followed by Tukey’s multiple-comparison test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

Furthermore, to ascertain Msl2's role in regulating H4K16ac, we investigated their temporal expression and co-expression patterns of Msl2 and H4K16ac in the cortex. Immunostaining revealed high expression levels of both Msl2 and H4K16ac during embryonic stages, which significantly declined postnatally (Fig. 6D). Co-expression analysis unveiled a robust expression correlation between Msl2 and H4K16ac (Fig. 6D). These findings suggest that Msl2 may contribute to early neurogenesis by modulating H4K16ac levels.

Msl2 modulates NPC development by enhancing chromatin accessibility through H4K16 acetylation at promoter regions

To elucidate the molecular mechanisms by which Msl2 regulates neural progenitor cell (NPC) development, we integrated multi-omics data from a recently published study [13] (see Methods). First, we examined gene expression changes in Msl2-ablated mouse NPCs using RNA-seq data. Our analysis identified 1,005 genes significantly upregulated as well as 1,011 genes downregulated (Fig. 7A). Furthermore, we identified differential expression in 50 NDD-related genes, with 27 upregulated and 23 downregulated (Fig. 7B), indicating that MSL2 regulates NDD-related genes through both direct and indirect regulatory mechanisms. Gene set enrichment analysis revealed that the differentially expressed genes were significantly enriched in the 'MYC_TARGETS_V1' and 'MYC_TARGETS_V2' categories, which are well-established regulators of cellular proliferation (Fig. 7C). Notably, most enriched genes within both categories were downregulated, consistent with Msl2’s role as a transcriptional activator, suggesting that these genes may be direct targets of Msl2. Gene ontology analysis showed that the differentially expressed genes were enriched in terms related to transcription and neurodevelopment, such as nucleolus, chromatin, RNA binding and cell proliferation (Fig. 7D). Collectively, these results underscore the importance of Msl2 in NPC development.

Fig. 7.

Transcriptomic profiling of Msl2-knockout versus wild-type neural progenitor cells reveals dysregulated neurodevelopmental gene networks (A) Volcano plot of RNA-seq data comparing wild-type (WT) and Msl2 knockout (Msl2KO) samples. The plot displays log2(fold change) versus -log10(p-value). A total of 2,016 differentially expressed genes (DEGs) were identified, including 1,005 up-regulated genes (red dots) and 1,011 down-regulated genes (blue dots), using thresholds of P < 0.05 and |log2(fold change)|> 1. (B) Heatmap of neurodevelopmental disorder (NDD)-related genes within the DEGs. Genes were clustered based on their expression patterns, with color intensity reflecting relative expression levels (red = high, blue = low). (C) Gene Set Enrichment Analysis (GSEA) plots for the "Myc Targets V1(V2)" gene set. The enrichment score (ES) is plotted against the ranked list of total expressed genes. Positive values indicate enrichment at the top of the list, while negative values indicate enrichment at the bottom. The red and blue lines highlight the distribution of genes in the functional gene set, with red indicating high expression and blue indicating low expression. The gray area represents the signal-to-noise ratio, reflecting the overall enrichment strength. The normalized enrichment score (NES) was −2.37 (−2.02), with a P-value of 0.01, indicating significant enrichment of the Myc Targets developmental set in the Msl2KO sample. (D) Gene Ontology (GO) terms enriched among the DEGs. Terms were identified using DAVID, with significance determined by adjusted P-values (P < 0.05)

To assess the effect of Msl2 deficiency on the histone modification H4K16ac, we analyzed H4K16ac ChIP-seq data in NPCs. In wild-type NPCs, we identified 15,834 reproducible peaks, of which 31.9% were in promoter regions — a substantial enrichment given that promoters (defined as ± 2 kb around transcription start sites of RefSeq genes) cover only ~ 1.8% of the mouse genome (UCSC Genome Browser, mm10 assembly). This finding suggests the preferential association of H4K16ac with gene promoters (Fig. S5A). In Msl2-deficient NPCs, we observed a marked decrease in H4K16ac levels (Fig. 8A), consistent with our earlier immunoblotting and immunofluorescence data in mouse cortex (Figs. 6A and 6B). These findings further support the essential role of Msl2 in mediating H4K16ac in NPCs.

Fig. 8.

Multi-omics atlas of H4K16ac-mediated chromatin regulation (A) Heatmap showing the normalized H4K16ac-ChIP enrichment around transcription start sites (TSS), sorted by signal intensity. The TSS center was used as a reference point, with signals plotted ± 3 kb. The mean enrichment profile is shown above the heatmap. ChIP data were normalized to reads per million (RPM), as described in Methods. (B) Heatmap of ATAC-seq peaks grouped by accessibility changes between wild-type (WT) and knockout (KO) samples. Peaks are categorized as WT only (n = 7,380), unchanged (n = 9,616), or KO only (n = 1,031). Signal intensity is normalized to RPM. (C) As in (A), heatmap showing H4K16ac-ChIP enrichment in regions with closed (open) chromatin accessibility. The mean enrichment profile is shown above the heatmap. Chromatin accessibility was determined using ATAC-seq data. (D) Known transcription factor (TF)-binding motifs identified by Hypergeometric Optimization of Motif EnRichment (HOMER) analysis. Gene ontology (GO) and pathway enrichment analyses of the associated genes are shown. (E) Track plots of FMR1 genomic loci showing RNA-seq, ATAC-seq, and H4K16ac ChIP-seq signals, as well as human MSL2 ChIP-seq data. ChIP-seq analysis was performed in HEK293T cells expressing human MSL2 variants. Signals are normalized to RPM and displayed as wiggle tracks

Given the established relationship between histone acetylation and chromatin accessibility [24–27], we analyzed ATAC-seq data. In Msl2-deficient NPCs, our results demonstrated a pronounced reduction in chromatin accessibility, with approximately 43% of regions showing decreased accessibility compared to only 6% exhibiting increased accessibility (Fig. 8B). Notably, 13.8% of these reduced accessibility regions were localized to gene promoters (Fig. S5A), suggesting that Msl2 promotes gene expression by maintaining promoter accessibility in NPCs. We further observed a strong correlation between H4K16ac levels and chromatin accessibility, with regions of reduced accessibility displaying decreased H4K16ac, while regions with enhanced accessibility showed increased H4K16ac (Fig. 8C). This finding supports a mechanism whereby Msl2 modulates chromatin accessibility through H4K16ac deposition. Functional annotation of regions with diminished accessibility revealed significant enrichment for pathways involved in neural development and cell cycle regulation (Fig. 8D). Motif analysis identified binding sites for transcription factors critical for neural development, including members of the RFX family [28], as well as factors associated with stem cell maintenance and proliferation (TEAD4 [29], SOX2 [30], LHX2 [31], and JUND [32]). These results collectively indicate that Msl2 contributes to NPC developmental processes through the regulation of chromatin accessibility.

To further delineate the mechanisms by which Msl2 regulates NDD-associated and cell cycle-related genes, we investigated specific targets including Fmr1, Mcm6, Sox13, and Ncl. In Msl2-deficient NPCs, Fmr1—a gene critical for neuronal development and proliferation control [33–35]—showed significant transcriptional downregulation (Fig. 7B, 8E). This was accompanied by reduced H4K16ac levels and diminished chromatin accessibility at its locus (Fig. 8E), supporting Msl2's role in modulating neural development through epigenetic regulation. ChIP-seq analysis in MSL2-overexpressing 293 T cells revealed direct binding of MSL2 to the FMR1 promoter region (Fig. 8E). Notably, MSL2 DNA-binding domain variants (G465V, R480C, and C495R) identified in individuals with ASD exhibited impaired FMR1 promoter binding, whereas the K399R variant (located outside this domain) maintained wild-type binding capacity (Fig. 8E). These results demonstrate that MSL2's DNA-binding domain is essential for its recruitment to the FMR1 promoter. This recruitment is associated with elevated H4K16 acetylation and increased chromatin accessibility at this locus. Parallel mechanisms were observed for other Msl2 targets, such as Mcm6 (DNA replication) [36], Sox13 (cell proliferation and development) [37], and Ncl (ribosome biogenesis and cell cycle regulation) [38] (Fig. S5). These findings collectively establish MSL2 as a master regulator of neurogenesis, operating through coordinated epigenetic control of functionally diverse gene networks.

Discussion

Our multi-level analysis revealed that Msl2 ablation in mice recapitulates key behavioral phenotypes of individuals with MSL2 pathogenic variants. Specifically, we observed significant deficits in social novelty preference, as shown by reduced interaction time with a novel mouse versus a familiar one in the three-chamber test. Spatial working memory impairments were also evident, with decreased Y-maze spontaneous alternation rates. In the Morris water maze, Msl2-deficient mice exhibited longer escape latencies during acquisition trials and fewer platform crossings during probe trials, underscoring their spatial memory deficits. Moreover, novel object recognition was impaired, marked by a notable drop in the discrimination index. Collectively, these behavioral deficits mirror the social communication challenges and intellectual disabilities observed in patients with MSL2 variants. The behavioral evidence firmly establishes MSL2 as a critical regulator of development and complex behaviors. Notably, since both Msl2 cHET and cKO mice in our study carry the same Nestin-Cre transgene, the observed gene dosage-dependent phenotypes can be primarily attributed to Msl2 loss rather than confounding effects of Cre expression, a logic consistent with established literature [39]. Nevertheless, we candidly acknowledge the inherent limitations of the Nestin-Cre driver. As previously reported, the Nestin-Cre transgene itself may exert independent effects on metabolic parameters and specific cognitive domains, such as contextual fear memory [39]. While our experimental design—comparing cHET and cKO cohorts with identical Cre alleles—allows for the relative isolation of Msl2 dosage effects, we recognize that the absence of a dedicated Cre-only control group prevents us from fully disentangling more subtle Cre-related background noise. We agree that future studies incorporating Nestin-Cre-only controls would further strengthen these findings and provide a more granular delineation of the driver's potential contributions.

The precise laminar organization of the neocortex is essential for normal cortical circuit formation and function [40]. In Msl2-deficient mice, although cortical layers remained broadly identifiable based on layer-specific markers such as SATB2, CTIP2, and TBR1, quantitative morphometric analysis revealed significant alterations in cortical architecture, including reduced upper-layer neuronal density and decreased cortical thickness at E18.5. In addition, adult cHET mice showed a persistent reduction in neuronal density within the superficial layers of the primary somatosensory cortex. Together, these findings support an important role for Msl2 in cortical lamination and maturation. Although these structural abnormalities may be relevant to the behavioral phenotypes observed in the same model, the present data do not establish a direct causal relationship between lamination defects in the primary somatosensory cortex and specific behavioral abnormalities such as deficits in social novelty recognition or novel object recognition.

During neurodevelopment, the expansion and differentiation of neural progenitor cells (NPCs) must be tightly coordinated [41]. Our lineage tracing experiments demonstrate that Msl2 ablation causes a reduction in BrdU incorporation, coupled with the extended S-phase duration revealed by BrdU/EdU dual-pulse experiments, points to a profound dysregulation of cell cycle dynamics in Msl2-deficient neural progenitor cells. The flow cytometry data, showing decreased G1-phase cells with a concomitant increase in S-phase population while G2/M remains unchanged, suggests a specific bottleneck at the S-phase transition rather than a global cell cycle arrest [42]. This S-phase elongation, likely caused by replication stress or impaired DNA synthesis, creates a "traffic jam" effect, where cells accumulate in S-phase. This explains both the reduced BrdU labeling efficiency and the increased S-phase fraction in flow analysis. Importantly, the decreased cell cycle exit rate suggests that cells are not merely delayed but actively prevented from completing their differentiation program, potentially through sustained activation of cell cycle checkpoints. These findings collectively suggest that MSL2 plays a crucial role in maintaining efficient DNA replication kinetics, and its loss disrupts the balance between proliferation and differentiation during cortical development. The specific S-phase defect, rather than general proliferation failure, highlights a previously unrecognized aspect of MSL2 function in coordinating replication with neural progenitor cell fate decisions.

MSL2 is a key regulator of neurodevelopment, primarily through its role in mediating histone H4K16 acetylation (H4K16ac) as part of the MSL chromatin-modifying complex. We found that Msl2-deficient models show significant downregulation of cortical development genes. Complementary ChIP-seq analysis revealed widespread reduction of H4K16ac at promoter regions following MSL2 loss. It should be noted that a previous study indicated no effect of MSL2 knockdown on H4K16ac modification in HeLa cells[11]. However, our study along with recent research demonstrates that Msl2 deficiency does lead to decreased levels of H4K16ac in the brain[13]. This discrepancy may be attributed to the distinct roles and mechanisms of MSL2 in different cell types and tissues. Such tissue-specific epigenetic regulation is not unprecedented; for example, the histone methyltransferase SETD2 exhibits context-dependent functions in neural versus non-neural tissues [43, 44]. The brain-specific requirement for MSL2 may reflect unique chromatin landscapes, differential expression of MSL complex subunits, or cell-type-specific recruitment mechanisms that render neuronal loci particularly sensitive to MSL2 dosage. In contrast to HeLa cells, the brain is a highly complex and specialized organ where MSL2 plays a more critical and specific role in neurodevelopment, and the MSL complex may have tissue-specific regulatory subunits or interactors that contribute to its indispensable function in maintaining H4K16ac levels in the brain, thereby influencing cortical development genes. While HEK293T cells were utilized to efficiently evaluate the DNA-binding affinity of MSL2 variants, we acknowledge that this heterologous system limits our ability to fully recapitulate the neuronal chromatin landscape. Future studies using human iPSC-derived neurons would be valuable to further elucidate the cell-type-specific consequences of these variants. Further investigation is required to elucidate these tissue-specific differences in MSL2's function and regulation of H4K16ac. ATAC-seq further confirmed reduced accessibility at 43% of critical loci, particularly those involved in neural development. This coordinated decline in H4K16ac levels, chromatin openness, and gene expression supports a mechanistic link: MSL2-mediated histone acetylation maintains chromatin architecture to support transcriptional activation of neurogenic programs. Importantly, simulations of patient variants showed disrupted MSL2-DNA binding, directly linking genetic variants to epigenetic dysregulation. These findings identify MSL2 as a crucial molecular link between chromatin remodeling and transcriptional output during corticogenesis.

In summary, our integrated approaches demonstrated Msl2 as a master regulator of three fundamental neurodevelopmental processes. First, by depositing H4K16ac, it maintains chromatin flexibility to activate genes specifying cortical identity. Second, it modulates replication stress responses to ensure robust proliferation of neural progenitors during rapid neurogenesis. Third, disruptions in these processes can impair neurodevelopment, leading to functional abnormalities in social behavior and cognition. Together, these findings establish Msl2 as a critical nexus linking epigenetic regulation, neurodevelopmental integrity, and behaviors, offering new insights into the molecular basis of chromatin-related NDDs.

Supplementary Information

Below is the link to the electronic supplementary material.

Authors’ contributions

H.G., K.X., G.C., and S.L. designed the study. S.L. and S.Q. performed the research. G.C., S.T., Z.S., and Y.L. analyzed the data. H.G., G.C., S.L., S.T., X.J., J.T., and K.X. wrote and revised the manuscript. All authors provided feedback on the manuscript and approved the final version.

Funding

This study was supported by the National Natural Science Foundation of China (Nos. 82130043, 82330035, and 82361138573 to K.X.; Nos. 82222025 and 32271141 to H.G.; No. 82401388 to X.J.), the STI 2030-Major Project (No. 2021ZD0201704 to H.G.), the National Key Research and Development Program of China (No. 2021YFA0805200 to J.T.), and Hunan Provincial grants (Nos. 2023RC1020 and 2023SK2084 to H.G.; Nos. 2021SK1010 and 2023SK2114 to K.X.; No. 2024JJ6545 to X.J.).

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files. Additional data are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

All procedures involving animals were conducted in accordance with the Central South University Institutional Animal Care and Use Committee (protocol No. 2022—2—3) and followed the guidelines set by the Association for Assessment and Accreditation of Laboratory Animal Care International. Ethical compliance with all relevant regulations for animal research was additionally verified by the university IRB.

Consent for publication

Not applicable: This manuscript does not contain any individual person’s data in any form (including individual details, images, or videos).

Competing interests

The authors declare that they have no competing interests.