MeCP2 gene dosage-dependent neurodevelopmentally restricted defects arise by aberrant activation of cell fate-determining bivalent genes

Abstract

The overexpression of MECP2 leads to severe neurological deficits in MECP2 duplication syndrome, and its dosage is considered a risk factor in gene therapy for Rett syndrome. However, in MECP2 duplication syndrome patients, MECP2 dysregulation arises at the embryonic stage while in Rett syndrome gene therapy, MECP2 is delivered into the mature brain. Here, we show that MeCP2 overexpression induces transcriptional alterations in neural progenitor cells, but has minimal effects in neurons in both mouse and human contexts. Consequently, MeCP2 overexpression in neural progenitor cells, but not mature neurons, leads to functional changes. Mechanistically, we observe that both endogenous and overexpressed Mecp2 bind to the same CpG island repertoire. In neurons, where endogenous Mecp2 is highly expressed, ectopic protein expression leads to reduced CpG island binding and accelerated protein degradation. In contrast, in neural progenitor cells, lower endogenous Mecp2 levels facilitate stronger deposition of the ectopic protein onto CpG islands, driving the transcriptional activation of many developmental bivalent genes. We show that this activation is mediated by the interaction with the SWI/SNF chromatin remodeling complex. Our findings establish that increased gene dosage-dependent effects are highly influenced by cell type, levels of proteins and their mechanisms of action.

MECP2 has gene dosage-dependent differing neurological effects in MECP2 duplication syndrome and Rett syndrome. Here, the authors show that these effects are context dependent, with MeCP2 overexpression affecting gene expression and function differently in neural progenitors and mature neurons.

Introduction

The establishment and maintenance of neuronal identity and function rely on the precise, dynamic regulation of gene expression, safeguarded by a multitude of epigenetic regulators. These factors act in a cell type- and time-dependent manner, shaping the transcriptional landscape throughout neurodevelopment^1–3. Major efforts have been devoted to unveiling the mechanistic roles of these factors in developmental processes, however their functions in differentiated cells remain less investigated. Among them, methyl-CpG-binding protein 2 (MeCP2) is an epigenetic reader that plays a uniquely dosage-sensitive role^4,5. While ubiquitously expressed, MeCP2 levels rise sharply postnatally and become highly enriched in mature neurons, aligning with critical phases of synaptic maturation and circuit refinement^6,7. Loss-of-function mutations in MECP2 cause Rett syndrome (RTT), a severe neurodevelopmental disorder characterized by intellectual disability, motor dysfunction, and autistic features^8,9. In contrast, MECP2 gene duplication leads to MECP2 duplication syndrome (MDS), which manifests with hypotonia, neurodevelopmental delay, and epilepsy^10–13. The broad overlapping pathological spectrum of these disorders has led to the prevailing hypothesis that MeCP2 levels must be precisely regulated to ensure normal brain functions.

Historically, MeCP2 has been viewed as a pure transcriptional repressor that binds methylated CpG dinucleotides to recruit corepressor complexes^5,14,15. However, recent studies have expanded this model, showing that MeCP2 can also bind unmethylated DNA and promote transcriptional activation by facilitating RNA polymerase II recruitment^16,17. These findings support a more complex and context-dependent model, in which MeCP2 modulates gene expression programs in response to the cellular and epigenetic environment.

Restoring physiological levels of Mecp2 in mouse models of both RTT and MDS is sufficient to rescue key pathological features^18,19, establishing gene therapy as the most promising therapeutic approach. However, its safe clinical implementation is thought to rely on the strict avoidance of MeCP2 overexpression, especially in RTT patients, where about half of the cells already express the wild-type allele and could also be affected by excessive MeCP2 levels, potentially leading to neuronal dysfunction. AAV-mediated delivery of MECP2/Mecp2 in juvenile Mecp2-mutant mice has been shown to ameliorate behavioral and molecular phenotypes, highlighting its therapeutic potential^20–22. Interestingly, even under supraphysiological viral dosing, no neurological alterations directly linked to Mecp2 overexpression were detected in either Mecp2-mutant or wild-type mice, with reported adverse effects instead arising from unrelated systemic factors^20–22. These results stand in stark contrast to MDS mouse models, where chronic Mecp2 overexpression beginning in embryogenesis leads to profound transcriptional dysregulation and neurological symptoms^19,23–25. This discrepancy led us to speculate that the timing and cellular context of MeCP2 overexpression play a critical role in determining its pathological outcome. During embryonic development, when endogenous MeCP2 is expressed at very low levels^6,7, cells may be especially vulnerable to its premature activity. In contrast, overexpression in the mature brain, where MeCP2 is already abundant, could be better tolerated, possibly due to reduced plasticity and proteostatic mechanisms that limit accumulation of the exogenous protein. Altogether, these considerations raise the possibility that the postnatal brain may be more resilient to elevated MeCP2 levels than previously thought. This hypothesis implicates that the functional consequences and tolerability of altering epigenetic regulator dosage are likely to vary substantially across developmental stages and cellular contexts. As such, many other epigenetic factors implicated in neurodevelopmental disorders may exhibit similarly dynamic and context-dependent functions. Dissecting these temporal and cell-type-specific roles is thus essential, not only for understanding disease mechanisms, but also for guiding the rational design of gene therapies that are both effective and safe.

To address this fundamental issue, we integrated RNA sequencing, CUT&Tag²⁶, and 4f-SAMMY-seq²⁷ analyses to examine the chromatin landscape and transcriptional effects of Mecp2 overexpression. Moreover, we validated our findings in human neural cells. Our results demonstrate that the cellular response to Mecp2 overexpression is highly context-dependent, elucidating the mechanisms that dictate its chromatin binding and transcriptional regulation. This study not only provides unexpected insights into MeCP2 biology but also has direct implications for the future development of gene therapies for brain disorders caused by impaired epigenetic regulation.

Results

Differential gene deregulation and phenotypic changes induced by Mecp2 overexpression in primary neuronal progenitor cells and neurons

To study and compare the effect of Mecp2 overexpression in uncommitted and differentiated cells, we derived primary neuronal progenitor cells (NPCs) and neurons (Ns) from E13.5 and E18.5 mouse embryos, respectively (Fig. 1a). NPCs and Ns exhibited distinct levels of Mecp2 expression, with primary neurons showing significantly higher expression levels compared to NPCs (Fig. S1a)^6,7. We then transduced these two cellular models using lentiviral vectors expressing either Mecp2 or EGFP (control) through the Ef1α constitutive promoter. To prevent excessively high Mecp2 levels, we tested different multiplicities of infection (MOI) and selected the lowest one that ensured >90% transduction efficiency (Fig. S1b, c). We then assessed Mecp2 expression 5 days post-transduction by Western blot, using an MOI of 10 in NPCs and 5 in neurons, which resulted in a 3-4-fold overexpression compared to baseline levels (Figs. 1b and S1a). Once the conditions were optimized, we repeated the same experimental protocol and performed transcriptomics analysis by RNA-sequencing. In Ns, we identified approximately 500 deregulated genes (DEGs).However, the majority displayed a fold change (FC) < 1, suggesting only minimal alterations in gene expression (Fig. 1c). In contrast, NPCs exhibited approximately 5000 DEGs, with 1137 of them being upregulated (FC > 1), challenging the notion that Mecp2 primarily acts as a repressor factor (Fig. 1c, Supplementary datasets 1, 2). Gene Ontology (GO) analysis revealed a strong enrichment of these upregulated genes in developmental processes, particularly in neural differentiation (Fig. 1d). Conversely, the set of downregulated genes with a fold change above 1 was about one-third smaller and was predominantly enriched for broader functional categories such as ribogenesis and RNA processing. (Fig. 1c, d). Phenotypic analysis with Ki67 staining revealed a significant reduction in the growth rate of Mecp2-overexpressing NPCs in culture (Fig. S2a), a finding that aligns with the transcriptional dysregulation. Additionally, in the absence of proliferative factors (bFGF2 and EGF), a greater proportion of cells differentiated in post-mitotic neurons as assessed by βIII-tubulin immunostaining (Fig. S2b).

Fig. 1. Mecp2 overexpression differentially alters the global transcriptome in NPCs and Ns.

a Schematic overview of the experimental workflow: primary NPCs and Ns derived from wild-type (WT) embryos were transduced with lentiviral vectors to overexpress Mecp2 or EGFP (used as a negative control). b Western blot analysis of Mecp2 protein levels in NPCs (Left) and Ns (Right) transduced with the Mecp2 transgene or EGFP. Densitometric quantification of Mecp2 relative to Actin (Left for NPCs; Right for Ns) is shown as fold change compared to the EGFP control (n = 3 independent biological replicates per condition). Data are presented as mean values ± SD. *p < 0.05, two-sided t-test. (NPCs Mecp2 vs EGFP p = 0.044 (*); Ns Mecp2 vs EGFP p = 0.013 (*)). c Volcano plots showing log₂ fold changes (log₂FC) for all detected genes in NPCs and Ns, comparing Mecp2-overexpressing cells to EGFP controls. Red dots indicate differentially expressed genes (DEGs) with p < 0.05 (n = 3 independent biological replicates per condition). d Representative Gene Ontology (GO) categories enriched among upregulated and downregulated genes in NPCs and Ns upon Mecp2 overexpression relative to EGFP control. Adjusted p-values were calculated using Fisher’s exact test for overrepresentation, followed by Benjamini–Hochberg correction. Images were created in BioRender. Vania Broccoli.

Mecp2 localization into CpG islands in both NPCs and Ns

To investigate the molecular basis behind this cell type-specific effect of Mecp2 overexpression on gene transcription, we decided to explore Mecp2 binding to the genome using CUT&Tag²⁶. Specifically, we examined the binding of endogenous Mecp2 (referred to as Mecp2-endo) in both EGFP-transduced control NPCs and Ns, as well as the binding of the exogenous overexpressed Mecp2 (referred to as Mecp2-exo), isolated by V5 epitope tagging (Fig. 2a). Distribution of Mecp2 was initially evaluated for the various genomic regions where Mecp2 binding occurred. Both Mecp2-endo and -exo exhibited strikingly similar distribution patterns in both NPCs and Ns, with most of the binding localized in gene regions. Notably, around 50% of the binding was observed in promoters or in genomic regions near the transcriptional start sites (TSSs) of genes (Fig. 2b). To further examine the positioning of the peaks, we generated heatmaps spanning the TSS to transcriptional end site (TES) regions, which revealed a pronounced enrichment of both Mecp2-endo and -exo binding just upstream of the TSSs, indicating a strong preference for these regulatory regions (Fig. 2c). Importantly, we obtained highly comparable peak distributions using two independent antibodies, which supports the robustness and specificity of our CUT&Tag approach. We also observed a strong enrichment of both Mecp2-endo and -exo on CpG islands in the two cellular models, in line with the findings by Liu et al. (2024), who also reported significant Mecp2 enrichment into CpG islands in human pluripotent stem cell (PSC)-derived neurons¹⁶ (Fig. 2d, e). CpG islands are, in fact, predominantly located upstream of TSSs and generally hypomethylated²⁸. Cross-referencing our data with publicly available datasets describing DNA methylation in NPCs and Ns²⁹, we confirmed that CpG islands targeted by Mecp2 are overall hypomethylated in both cell types (Fig. S2a). However, we observed a moderately higher methylation level in Ns compared to NPCs (Fig. S2a). Because Mecp2 localization and function are also influenced by non-CG methylation, particularly mCA, as reported by Gabel et al.³⁰, we further assessed mCA levels by dot blot in our cells of interest (Fig. S3a). We detected the presence of mCA in our 7-day neurons (the same time point used for RNA-seq and CUT&Tag), while its levels were extremely low in NPCs, as expected³¹. Notably, mCA levels in 7-day neurons were lower compared with 14-day neurons and adult mouse cortex, which were used as positive controls, also as expected³¹. To verify the specificity of the antibody used for mCA detection, we included control probes representing unmethylated DNA, mCA-only, and mCG-only sequences (Fig. S3a). In summary, our findings align with recent studies confirming that Mecp2 is enriched at hypomethylated promoter regions^16,32.

Fig. 2. Endogenous and exogenous Mecp2 enriches at CpG islands near TSSs.

a Schematic overview of the CUT&Tag assay performed to map endogenous (endo) and exogenous (exo) Mecp2 binding in primary NPCs and Ns derived from wild-type (WT) embryos (n = 2–3 biological replicates per group). b Genomic distribution of Mecp2 binding peaks for both endogenous and exogenous protein. Color code: yellow = promoter; pink = distal intergenic regions; light orange = 3′ UTR; dark orange = 5′ UTR; light blue = first intron/exon; dark blue = gene body. c Heatmaps showing Mecp2 binding signal intensity from transcription start site (TSS) to transcription end site (TES) for endogenous and exogenous Mecp2 in NPCs and Ns. d Genome browser snapshots showing CUT&Tag read coverage of Mecp2-endo and -exo in correspondence of CpG islands in NPCs and Ns. e Metagene plots showing Mecp2 binding enrichment around CpG islands across the mouse genome. Curves represent EGFP-treated NPCs (endo NPCs, light blue) and Ns (endo Ns, light red), and Mecp2-overexpressing NPCs (exo NPCs, blue) and Ns (exo Ns, red). Images were created in BioRender. Vania Broccoli.

Differential occupancy of CpG islands by Mecp2-endo and -exo in NPCs and Ns

Despite the difference in Mecp2 expression levels between NPCs and Ns, the overall genomic distribution of Mecp2 binding appeared remarkably similar (Fig. 2b–e).

To further explore this, we analyzed Mecp2-endo and -exo binding profiles across promoters, gene bodies, and intergenic regions, with signal intensities normalized to region length to ensure comparability across genomic features (Fig. S3b). In both cell types, Mecp2-endo and -exo displayed predominant enrichment at promoter regions, with this preference being particularly strong in neurons (Fig. S3b). In NPCs, the difference was less marked, and Mecp2-exo showed a relatively broader distribution, with enrichment also observed within gene bodies and intergenic regions (Fig. S3b). However, when integrating these binding data with RNA-seq results, only promoter-associated Mecp2-exo enrichment correlated significantly with transcriptional changes, whereas enrichment within gene bodies did not display a consistent association with gene expression (Fig. S3c).

To gain deeper insight into Mecp2 promoter binding, we initially focused on the number of CpG islands bound by Mecp2, quantifying the overlap between the Mecp2-endo and -exo isoforms in both cellular models. Surprisingly, we observed that Mecp2 targeted approximately 15,000 CpG islands, with more than 95% of them being occupied by both Mecp2-endo and -exo in NPCs and Ns, demonstrating a high concordance in CpG island targeting (Fig. 3a).

Fig. 3. Differential enrichment of exogenous Mecp2 at CpG islands in NPCs and Ns.

a Schematic representation of CpG islands co-occupied by both Mecp2-endo and -exo in NPCs and Ns. Color legend: light blue = Mecp2-endo in NPCs; light red = Mecp2-endo in Ns; blue = Mecp2-exo in NPCs; red = Mecp2-exo in Ns. b Distribution plots of Mecp2 coverage across CpG islands. The x-axis represents the percentage of CpG island coverage (0–100%), and the y-axis indicates Mecp2 signal density. c Quantification of Mecp2-exo enrichment at CpG islands co-targeted relative to Mecp2-endo in both NPCs (light purple) and Ns (purple). In brief, for each condition, we calculated the signal intensity of both Mecp2-endo and -exo at shared CpG island targets and computed the enrichment score (ES: exo/endo). CpG islands with ES > 2 (dash red line) were considered significantly enriched by exogenous Mecp2. d Pie charts show the genomic annotation of Mecp2-exo-enriched CpG islands (ES < 2) in NPC (right) and Ns (left). e Bar plot showing the number of genes linked to promoter-associated CpG islands with an enrichment score (ES) < 2 in Mecp2-endo Ns/NPCs (light orange), Mecp2 exo/endo NPCs (light purple), and Mecp2 exo/endo Ns (purple). f Bar plots showing the overlap between genes from the Mecp2 exo/endo NPC-enriched and exo/endo Ns-enriched lists with the transcriptome data from RNA-seq. Bars indicate the number of genes that are not expressed (gray), expressed but not deregulated (light gray), upregulated (pink), or downregulated (green). g Representative table of DEGs up- (green) and down-regulated (yellow) with a FC > 0.5 and FC < 1. h Representative gene ontology (GO) categories significantly enriched among genes associated with Mecp2 exo/endo NPCs-enriched CpG islands. Adjusted p-values were calculated using Fisher’s exact test for overrepresentation, followed by Benjamini–Hochberg correction. I Genome browser snapshots showing CUT&Tag read coverage at selected Mecp2-exo NPC-enriched loci: Neurod1, Emx1, and Zic3. Color legend: light blue = Mecp2-endo; blue = Mecp2-exo; black = control IgG.

Since no substantial differences were found in the number of targeted CpG islands between the two cellular models, we next assessed the degree of CpG island physical coverage by Mecp2 (from 0 to 100%). As expected, the fraction of CpG islands fully covered by Mecp2-endo was higher in Ns than in NPCs, consistent with the higher levels of endogenous Mecp2 expression in this cellular type. Interestingly, an opposite trend was observed for Mecp2-exo, which showed greater CpG island coverage in NPCs compared to Ns (Fig. 3b).

To quantitatively investigate these differences and validate our approach, we focused on Mecp2 enrichment over targeted CpG islands, starting by comparing the Mecp2-endo signal between control EGFP-transduced Ns and NPCs, as an internal reference. We first calculated the Mecp2 signal density over CpG islands commonly targeted in both cell types (99% overlap) (Fig. 3a), as described in Bartsovic et al.³³ (Fig. S4a, b). Then, we computed an enrichment score (ES) for each CpG island as the ratio of signal intensity in Ns relative to NPCs (Ns/NPCs Mecp2-endo signal). CpG islands with an ES > 2 were considered enriched for Mecp2 in Ns (Fig. S4c). Focusing on these CpG islands, we found that the majority (91.6%) were localized within promoter regions (Fig. S4d). This observation guided the generation of a gene list derived specifically from these promoter-associated CpG islands enriched in Mecp2-endo (Supplementary dataset 3). This list comprised approximately 4500 genes, including well-known Mecp2-regulated genes such as Chd8, Mtor, and Bdnf, which are crucial for cellular pathways where Mecp2 is known to play a central role, such as mRNA processing, regulation of translation, chromatin remodeling, neuronal maturation, and activity^34–37 (Fig. S4e, f). The identification of these genes, which are strongly implicated in executing Mecp2 functions, served as a robust quantitative validation of our analysis, demonstrating that the detected Mecp2 binding sites are of biological relevance.

Thus, we repeated the same analysis for Mecp2-exo binding enrichment on common targeted CpG islands in NPCs and Ns (Ns and NPCs Mecp2-exo signal vs Mecp2-endo signal). In line with previous observations, we found more enriched CpG islands for Mecp2-exo in NPCs compared to Ns (Fig. 3b, c). In NPCs, we identified 2,287 CpG islands enriched (ES < 2; Mecp2-exo/endo) of which 89% localized within promoter regions (Fig. 3c, d). From this, we derived a list of 2007 genes, which we refer to as Mecp2-exo/endo NPC enriched (Fig. 3e, Supplementary dataset 3). In contrast, in Ns, we found far fewer CpG islands enriched for Mecp2-exo, of which 81.5% located in promoters and associated with 685 genes (Fig. 3c–e, Supplementary dataset 3). To better understand the transcriptional impact of this differential enrichment, we intersected the Mecp2-exo/-endo-enriched gene lists from NPCs and Ns with our RNA-seq dataset (Fig. 1). In both cases, around 75% of the genes were transcriptionally expressed (Fig. 3f). However, while nearly 50% of these expressed genes were deregulated in NPCs, fewer than 5% were transcriptionally affected in Ns (Fig. 3f). Strikingly, within the Mecp2-exo/endo NPC-enriched gene list, upregulated genes doubled the number of downregulated ones (Fig. 3f). Moreover, not only upregulated genes were more abundant, but many (203 out of 449) also exhibited a FC > 1, indicating a stronger transcriptional deregulation (Fig. 3g). In contrast, only a minority of the downregulated genes showed a FC > 1 (57 out of 288), suggesting a generally milder repressive effect, ultimately pointing to a potential direct role of Mecp2 in gene upregulation (Fig. 3g). To further characterize the biological relevance of these transcriptional changes, we performed Gene Ontology (GO) analysis on the Mecp2-exo/endo NPC-enriched gene list. This revealed again a strong association with neurogenesis, consistently reflecting the functional identity of the genes most deregulated in our RNA-seq analysis (Figs. 3h and 1d). Among these, we identified key regulators of neuronal differentiation such as Neurod1, Emx1, and Zic3 (Fig. 3i).

Exogenous Mecp2 upregulated bivalent genes in NPCs by interacting with the SWI/SNF complex

We observed that many of the genes presenting Mecp2 enrichment at CpG islands in NPCs were transcriptionally altered, with the predominant group being upregulated, whereas this effect was largely absent in Ns. To better understand the significant discrepancy between NPCs and Ns, we analyzed the promoter epigenetic features of these genes, focusing on the histone markers H3K4me3, associated with transcriptional activation, and H3K27me3, associated with transcriptional repression, using datasets generated from the same cellular models³⁸. This analysis highlighted that over one third of the Mecp2-exo/endo enriched genes in NPCs displayed a bivalent chromatin state (H3K4me3 and H3K27me3) and, by crossing with the RNA-seq dataset, many of these genes resulted to be up-regulated following Mecp2 overexpression (Fig. 4a–c). Specifically, when narrowing the focus to genes with a fold change FC > 1, we found that the subset of bivalent up-regulated genes was predominant compared to the other subsets (Fig. 4c). In contrast, in Ns, approximately 85% of the genes enriched for Mecp2-exo exhibited a defined chromatin state, either decorated with H3K4me3 (63%) or with H3K27me3 (22%) (Fig. 4a). Interestingly, although a small proportion of exogenous Mecp2-enriched genes in Ns were bivalent (16%) (Fig. 4a), RNA-seq analysis revealed that the majority of them (83%) were already transcriptionally active (Fig. 4b). Conversely, in NPCs, a significant fraction of bivalent Mecp2-enriched genes (about 40%) was not expressed (Fig. 4c). These findings point to intrinsic differences in the regulation of bivalent genes between NPCs and Ns. Chromatin bivalency is believed to confer a “poised” transcriptional state to lineage-specifying genes in undifferentiated cells, where these genes are typically expressed at low levels but are primed for activation^38–41. Additionally, the promoters of these genes are rich in CpG islands⁴², the primary target of Mecp2. The activation of bivalent genes in NPCs generally occurs through the action of the SWI/SNF complex, which counteracts Polycomb proteins to remove the repressive histone marker H3K27me3^43–45. Thus, we hypothesized that Mecp2 could interact with this complex, inducing the transcriptional activation of these genes. Although an interaction between Mecp2 and the SWI/SNF complex, specifically with one of its ATPase subunits, Brm, was initially reported in 2005⁴⁶, a subsequent study failed to confirm this finding, raising doubts about the existence of such an interaction⁴⁷. Nevertheless, we noted that SMARCB1 emerged as a putative MeCP2 interactor in a recently published MeCP2 interactome proteomics dataset (https://mecp2-neuroatlas.wi.mit.edu)¹⁶. Given the well-known involvement of the SWI/SNF complex in the recognition and controlled activation of bivalent genes⁴³, we decided to directly test this interaction via co-immunoprecipitation. To this end, we co-expressed EGFP-tagged MeCP2 and V5-tagged SMARCB1 in HEK293 cells, confirming their physical association (Fig. 4d). To further validate this observation, we performed CUT&Tag analysis of endogenous Smarcb1 in NPCs transduced with EGFP or Mecp2 lentiviral vectors (Fig. 4e). We then analyzed the peak distribution of Smarcb1 near the TSS, categorizing genes in Mecp2 targets and non-targets. In the EGFP control condition, Smarcb1 distribution showed few differences between the two groups (Fig. 4f), whereas in the Mecp2 overexpression condition, Smarcb1 was significantly enriched at Mecp2 target genes in correspondence of CpG islands (Fig. 4g, h). Specifically, we observed a strong enrichment of Smarcb1 across all genes, but particularly on bivalent genes, which was in line with our expectations (Fig. 4i). Indeed, we found Smarcb1 enriched at bivalent genes targeted by Mecp2-exo in NPCs, including transcription factors involved in neurogenesis such as Zic3, genes related to synaptic activity like Grin3a, and regulators of neuronal migration such as Reln (Fig. 4j). Notably, all these genes were found to be upregulated in our RNA-seq analysis. Thus, these observations support the hypothesis of a functional interaction between these two chromatin factors, and the enrichment of Smarcb1 at the TSS of bivalent genes could explain their stimulation upon Mecp2 overexpression.

Fig. 4. Mecp2 recruits the SWI/SNF complex to activate the expression of bivalent genes.

a Pie charts showing the distribution of histone modifications at the promoters of genes from the Mecp2-exo/-endo NPCs and Ns-enriched lists. Blue = active (H3K4me3); purple = repressed (H3K27me3); yellow = bivalent (both H3K4me3 and H3K27me3). b Bar plots showing the overlap between Mecp2-enriched genes (categorized as active, repressed, or bivalent) and RNA-seq expression data. c Representative table of DEGs up- (pink) and down-regulated (green) with a FC > 0.5 and FC < 1 in active, repressed, or bivalent genes group. d Schematic of the co-immunoprecipitation (co-IP) experiment designed to test the interaction between EGFP-tagged Mecp2 and V5-tagged Smarcb1 (right). Representative Western blots from EGFP IP (top) and V5 IP (bottom) showing input, unbound, and IP fractions in conditions where only Mecp2, only Smarcb1, or both proteins were expressed (left). e Schematic of the CUT&Tag experiment for Smarcb1 in NPCs overexpressing either EGFP or Mecp2. Heatmaps showing Smarcb1 binding enrichment in Mecp2 target and non-target gene sets in NPCs overexpressing EGFP (f) or Mecp2 (g). h Genome browser snapshots showing CUT&Tag read coverage of Smarcb1 in correspondence of CpG islands in EGFP and Mecp2-overexpressing NPCs. i Metagene plots showing Smarcb1 binding enrichment around TSS of bivalent (solid line) and non-bivalent genes (dash line) in EGFP-treated NPCs (yellow) and Mecp2-treated NPCs (orange). Shaded areas represent mean ± SEM across genes in each category. j Genome browser snapshots of CUT&Tag read coverage for Smarcb1 in representative Mecp2 target genes in NPCs treated with EGFP or Mecp2: Zic3, Grin3a, and Reln. Images were created in BioRender. Vania Broccoli.

Mecp2 overexpression accelerated cortical neurogenesis in vivo

To confirm that the activation of neurogenesis following Mecp2 overexpression in NPCs is not restricted to cellular models in vitro, we moved to test this effect in vivo. Hence, we overexpressed Mecp2 together with EGFP, to track Mecp2-overexpressing cells, or EGFP alone as a control, by in utero electroporation of E14.5 mouse embryos (Fig. S5a). To maintain consistency with the in vitro experiments, we co-electroporated two separate plasmid constructs corresponding to the same lentiviral vectors used in vitro, ensuring that both overexpression and control conditions were driven by the same constitutive Ef1α promoter (Fig. S5a). At E18.5, we analyzed the brains to determine the number of EGFP+ progenitors that migrated from the proliferative (SVZ/VZ) zone to the cortical plate (CP) and differentiated into neurons. We observed a significantly higher number of EGFP+ cells in the cortical plate (Tubb3 positive area) compared to Pax6+ NPCs in Mecp2-EGFP treated animals (Fig. S5b), thereby confirming that Mecp2 overexpression accelerates neuronal differentiation processes with premature neuronal migration in the cortical plate.

Exogenous Mecp2 exerted a more pronounced effect on the chromatin state in NPCs than in Ns

We then decided to investigate the overall chromatin organization of NPCs and Ns overexpressing Mecp2 to assess the occurrence of chromatin alterations. One of the proposed roles of Mecp2 is to act as a chromatin regulator, shaping the structural and functional properties of heterochromatin in neurons during development and maturation^14,48,49. However, recent studies have questioned its involvement in chromatin state regulation, suggesting that heterochromatin assembly might occur independently of Mecp2 function⁵⁰. We analyzed the chromatin landscape of our cellular models through 4f-SAMMY-seq, a recently developed method that enables the fractionation of DNA based on chromatin accessibility^27,51. The protocol separately analyzes the DNase-sensitive chromatin fraction, corresponding to euchromatin (S2S), and the DNase-resistant chromatin fraction, corresponding to heterochromatin (S3) (Fig. 5a). Initially, we assessed the concordance of the S2S and S3 fractions across biological replicates, observing a high degree of consistency between samples. At approximately 150 Kb resolution, the mean Spearman correlation between samples was 0.97 for the euchromatin fraction in NPCs, 0.86 for the same fraction in Ns, 0.89 for the heterochromatin fraction in NPCs, and 0.78 for Ns (Fig. S5c). This indicates that, across biological replicates, we can accurately separate euchromatin and heterochromatin.

Fig. 5. 4f-SAMMY-seq analysis reveals limited chromatin changes in NPCs and Ns overexpressing Mecp2.

a Schematic overview of the 4f-SAMMY-seq assay performed to analyze chromatin state in S2S (euchromatin) and S3 (heterochromatin) fractions in primary NPCs and Ns derived from wild-type (WT) embryos treated with EGFP or Mecp2 (n = 3 independent biological replicates per group). Plots showing the percentage of 150 kb genomic regions that shift from compartment A (euchromatin) to compartment B (heterochromatin) and vice versa in Ns (b; EGFP vs Mecp2), in NPCs (c; EGFP vs Mecp2), and in Ns vs NPCs (d; EGFP-treated). e Genome browser snapshots of 4f-SAMMY-seq signal in the S2S and S3 fractions of control (EGFP-treated) and Mecp2-overexpressing NPCs and Ns. f Representative gene ontology (GO) categories significantly enriched among genes associated with genomic regions shifting from compartment S3 to S2S in NPCs overexpressing Mecp2 vs EGFP control. Adjusted p-values were calculated using Fisher’s exact test for overrepresentation, followed by Benjamini–Hochberg correction. g Bar plot showing the overlap between genes located in genomic regions shifting from compartment S3 to S2S in Mecp2-overexpressing NPCs (compared to EGFP controls) and Mecp2 target genes. h Representative table of DEGs up (green) and down-regulated (yellow) with a FC > 0.5 and FC < 1 identified within the gene list shifting from compartment S3 to S2S. i Pie charts showing the distribution of histone modifications at the promoters of genes shifting from compartment S3 to S2S. Color legend: blue = active (H3K4me3); purple = repressed (H3K27me3); yellow = bivalent (both H3K4me3 and H3K27me3). Images were created in BioRender. Vania Broccoli.

To compare the EGFP and Mecp2 overexpression conditions, we calculated, for each biological replicate, the solubility score by normalizing reads associated with euchromatin against heterochromatin (S2S/S3). Genomic bins with solubility scores above +0.1 were classified as euchromatin (S2S), whereas those with scores below −0.1 were classified as heterochromatin (S3). Differentially enriched genomic regions between the two conditions were identified as bins in which all solubility scores fell outside the 0.99 reference consensus interval (Fig. 5b–d). We identified chromatin regions that shifted from S2S to S3 and vice versa between the EGFP and Mecp2 conditions. In Ns, we observed almost no changes (S2S to S3 = 0%; S3 to S2S = 0.04%) (Fig. 5b–e), whereas in NPCs, although minimal, we found more regions transitioning from heterochromatin to euchromatin (S2S to S3 = 0.01%; S3 to S2S = 0.42%) (Fig. 5c–e). As a positive control for the analysis, we included the comparison between NPCs and Ns (EGFP-treated), where—as expected—we observed a much more significant number of regions changing their chromatin status (S2S to S3 = 8.57%; S3 to S2S = 5.42%) (Fig. 5d, e). Notably, we observed more regions transitioning from euchromatin to heterochromatin rather than the opposite. This finding aligns with the general trend of chromatin compaction during cellular differentiation⁵². However, focusing on the approximately 2000 genes located in regions that underwent the opposite transition (from S3 to S2S), GO analysis confirmed their enrichment in pathways related to neuronal development and activity (Fig. S5D, Supplementary dataset 4). Next, we further investigated the 0.42% of regions shifting from heterochromatin to euchromatin in NPCs upon Mecp2 overexpression compared to the EGFP condition (Fig. 5c). Within these regions, we identified around 300 genes and, through GO analysis, we found several terms associated with cell-cycle transition and neuronal differentiation (Fig. 5f, Supplementary dataset 4). Among these, approximately two-thirds were direct targets of Mecp2 at their CpG islands, and, in line with the transition from heterochromatin to euchromatin, intersecting with the RNA-seq data, we found more upregulated genes with a higher fold change compared to downregulated genes (Fig. 5g, h). Notably, around 30% of these genes were also identified as bivalent (Fig. 5i). Overall, while Mecp2 overexpression does not appear to cause major chromatin rearrangements in these contexts, we observed subtle changes in NPCs that align with its proposed role in actively promoting transcription, whereas again no detectable alterations were found in Ns.

Exogenous Mecp2 showed reduced DNA affinity and accelerated protein degradation in Ns

The lack of both transcriptional and chromatin alterations in Mecp2-overexpressing Ns is consistent with the minimal enrichment of Mecp2 at its target CpG islands in these cells (Figs. 3d, e and 4a, c). However, these findings did not clarify the exact fate of the overexpressed Mecp2 protein, since its localization and target regions remain unchanged compared to the endogenous protein (Fig. 2b, c). For this reason, we decided to investigate Mecp2 affinity for DNA to assess potential differences between Mecp2-endo and -exo in NPCs and Ns. In fact, many of the CpG islands targeted by Mecp2 were already fully occupied by Mecp2-endo in Ns (Fig. 3b), leading us to speculate that the accessibility of its -exo counterpart to its primary target sites might be hindered and that its intrinsic instability could further reduce its overall DNA-binding affinity²¹. To address this hypothesis, we employed an incremental salt-solution DNA extraction method to measure the ionic strength required to disrupt the interaction between the protein-of-interest and DNA (Fig. 6a). In Ns, Mecp2-exo was almost entirely released in the 400 mM fraction, whereas Mecp2-endo showed a broader distribution, with a substantial portion also present in the subsequent 600 mM fraction, suggesting a stronger interaction (Fig. 6b). In contrast, in NPCs, although Mecp2 affinity for DNA was different compared to Ns, we did not observe any significant differences in the avidity between Mecp2-endo and -exo (Fig. 6a). These results imply that overexpressed Mecp2 in Ns established a weaker association with DNA compared to the endogenous Mecp2. On this line, it is well known that many pathological mutations in MECP2, particularly those affecting the methyl-binding domain (MBD), impact its ability to interact with the DNA, and this reduced affinity leads to a more rapid degradation of the protein^53,54. To further investigate this, we assessed protein stability by cycloheximide (CHX) treatment, comparing Mecp2-endo and -exo levels after 4 and 8 h of CHX exposure. To exclude a possible influence of the V5 tag on Mecp2-exo stability, we also tested the same construct bearing a FLAG tag instead. In NPCs, no differences were observed among the three forms, consistent with the results of DNA-binding assays (Fig. 6c). Conversely, in neurons, Mecp2-exo was degraded approximately 50% faster than its -endo counterpart with both tags, confirming a differential protein turnover (Fig. 6d).

Fig. 6. Affinity and stability of ectopically overexpressed Mecp2 differ between NPCs and Ns.

a, b Representative Western blots of Mecp2-endo (detected with anti-Mecp2 antibody in EGFP-treated cells) and Mecp2-exo (detected with anti-V5 antibody in Mecp2-treated cells) from chromatin-associated protein fractions extracted with increasing concentrations of NaCl (200–800 mM). Quantification is shown in the adjacent bar plots. Mecp2 intensity in each fraction was calculated by densitometric analysis and expressed as a percentage of the total Mecp2 signal. Right panels: NPCs; left panels: Ns. n = 3 independent biological replicates per group. Data are presented as mean values ± SD. **p < 0.01 ANOVA-one way with Tukey’s post hoc test. (Ns Mecp2-Endo vs Ns Mecp2-Exo [NaCl 200 mM] p = 0.007 (**); Ns Mecp2-Endo vs Ns Mecp2-Exo [NaCl 200 mM] p = 0.006 (**)). c, d Representative Western blots of Mecp2-endo (anti-Mecp2, EGFP-treated) and Mecp2-exo (anti-V5 and anti-FLAG, Mecp2-treated) under basal conditions and after 4 or 8 h of cycloheximide (CHX) treatment. Quantification is shown in the bar plots below. Mecp2 or V5 or FLAG signals were normalized to actin and expressed as fold change relative to the basal (time 0) condition. Right panels: NPCs; left panels: Ns. n = 3 independent biological replicates per group. Data are presented as mean values ± SD. *p < 0.05, **p < 0.01 ANOVA-one way with Tukey’s post hoc test. (Ns Mecp2-Endo vs Ns Mecp2-Flag CHX 4 h p = 0.027 (*); Ns Mecp2-Endo vs Ns Mecp2-V5 and Ns Mecp2-Flag CHX 8 h p = 0.005 (*)).

To explore potential mechanisms underlying this difference, we analyzed the two well-characterized Mecp2 phosphorylation sites in Serine 80 and 421 (pS80 and pS421)^55–58 by Western blot, normalizing phospho-specific signals to total Mecp2 levels in control and Mecp2-overexpressing NPCs and Ns. Interestingly, pS421, which is typically associated with neuronal activity, was reduced in both NPCs and Ns overexpressing Mecp2 (Fig. S6). Low pS421 levels in NPCs have been linked to decreased proliferation⁵⁹, whereas this modification does not appear to influence Mecp2 localization or chromatin binding⁴⁹. In contrast, pS80 showed an opposite trend, with a marked decrease in NPCs and an increase in Ns overexpressing Mecp2 (Fig. S6). Notably, pS80 has been recently associated with a faster chromatin turnover, consistent with our observation that Mecp2-exo displays reduced DNA-binding affinity in this cellular model⁴⁹. Although preliminary, these data are consistent with previous findings and support the notion that differential Mecp2 phosphorylation could contribute to the distinct biochemical behavior of endogenous and exogenous Mecp2 observed in our models.

Mecp2 overexpression did not alter excitatory post-synaptic currents in the adult mouse brain

To further corroborate our in vitro findings, we decided to assess the effects of Mecp2 overexpression in vivo at the functional level. Given the absence of major molecular alterations observed in Ns, we aimed to determine whether a similar resilience could be detected in the intact brain, where circuit-level dynamics and homeostatic mechanisms may further shape the cellular response to increased Mecp2 levels. Several studies have shown that in MDS mouse models, cerebral cortical neurons exhibit increased glutamatergic synapses and enhanced spontaneous excitatory transmission^60,61. To achieve broad Mecp2 overexpression in the adult mouse brain, we utilized the AAV-PHP.eB vector, which efficiently crosses the blood-brain barrier after intravascular administration, equipped with a strong CBA constitutive promoter^21,62 (Fig. 7a). Furthermore, using the same plasmids employed for AAV production, we performed in utero electroporation in E14.5 mouse embryos, as described in Fig. S3e. Thus, these two models allowed us to discriminate the Mecp2 overexpression dependent effects on neuronal functions elicited in either embryonic NPCs or directly in adult neurons (Fig. 7a). In 8-week-old treated mice, we assessed Mecp2 distribution using immunostaining with a V5 tag and EGFP (Fig. S7a, b). Given the higher number of targeted cells observed in the somatosensory cortex following in utero electroporation and AAV transduction, we focused our patch-clamp analysis on this region. Consistent with previous studies in MDS mouse models, Mecp2 overexpression in NPCs led to a significant increase in spontaneous excitatory post-synaptic currents (sEPSCs) both in frequency and amplitude in the recorded adult brain tissue^23,61 (Fig. 7b). In contrast, adult mice treated with either AAV:Mecp2 or AAV:EGFP did not exhibit differences in these currents, suggesting no functional alterations caused by Mecp2 overexpression in mature neurons (Fig. 7c).

Fig. 7. Mecp2 overexpression does not induce functional alterations in neurons in vivo.

a Schematic overview of the experimental design to assess the functional effects of Mecp2 overexpression in vivo. To overexpress Mecp2 in NPCs, wild-type embryos were electroporated in utero at E14.5 with a plasmid expressing EGFP alone (control) or in combination with Mecp2. To overexpress Mecp2 in mature neurons, 4-week-old mice were systemically injected with an AAV-PHP.eB vector carrying either EGFP (control) or the same Mecp2-expressing construct used in NPCs. At 6 weeks of age, mice were sacrificed, and spontaneous excitatory postsynaptic currents (sEPSCs) were recorded in the somatosensory cortex. Right: representative traces of sEPSCs recorded from cortical neurons of mice overexpressing Mecp2 in NPCs (b) or in mature neurons (c). Bar graphs on the right show the mean sEPSC frequency and median amplitude. For the E14.5 cohort: 3 EGFP-injected mice and 4 Mecp2-injected mice. For the adult cohort: 3 EGFP-injected mice and 3 Mecp2-injected mice. Data are presented as box plots displaying mean (+), median (internal horizontal line), first and third quartiles (upper and lower box edges), and minimal and max values (whiskers) of the data distribution. In (B) n = 15 cells/3animals in the EGFP group and 18cells/4animals for the Mecp2 group; for sEPSC mean frequency p = 0.008 (**) and for sEPSC median amplitude p = 0.001 (**) by two-sided t-test with Welch’s correction; in (C) n = 15 cells/3animals in the EGFP group and 17cells/3animals for the Mecp2 group; for sEPSC mean frequency p = 0.708 (ns) by two-sided t-test with Welch’s correction and for sEPSC median amplitude p = 0.7627 (ns) by Mann–Whitney U test). d Volcano plots showing log₂FC for all detected genes in human NPCs and Ns, comparing MeCP2-overexpressing cells to EGFP controls. Red dots indicate differentially expressed genes (DEGs) with p < 0.05 (n = 3 independent biological replicates per condition). e Representative Gene Ontology (GO) categories enriched among upregulated and downregulated genes in human NPCs and Ns upon MeCP2 overexpression relative to EGFP control. Adjusted p-values were calculated using Fisher’s exact test for overrepresentation, followed by Benjamini–Hochberg correction. Images were created in BioRender. Vania Broccoli.

Mecp2 overexpression alters transcription in human NPCs but not in neurons, mirroring the mouse model

Finally, to validate our findings in the human context, human iPSCs were differentiated into both NPCs (hNPCs) and telencephalic cortical neurons (hNs) using a dual-SMAD inhibition protocol. Both cell types were transduced with lentiviral vectors expressing either human MeCP2 or EGFP as a control, using the same vector backbone employed for murine cell cultures. As for the mouse cellular models, we first tested different multiplicities of infection (MOIs) to achieve >90% transduction efficiency and then validated the level of overexpression obtained with the selected MOI (10 for hNPCs and 20 for hNs) by Western blot (Fig. S8a–c).

After 5 days post-transduction, both hNPCs and hNs were lysed to isolate total RNA. Global transcriptomic analysis in hNs revealed minimal gene deregulation, with only 19 genes upregulated and 35 downregulated, all showing a fold change below 1 except for MeCP2 itself (Fig. 7d; Supplementary dataset 5). In contrast, hNPCs exhibited approximately 2500 DEGs, with a higher number of upregulated genes and larger fold changes, closely resembling the pattern observed in mouse NPCs (Fig. 7d; Supplementary data 6). In line with this observation, GO analysis revealed a strong enrichment of these upregulated genes in neuronal differentiation pathways (Fig. 7e).

Intersecting the up- and downregulated genes from both experimental groups (mNPCs and hNPCs) revealed a strong correlation among the upregulated genes, confirming the conserved transcriptional impact of Mecp2/MeCP2 overexpression in progenitor cells (Fig. S8d). These results confirm that MeCP2 overexpression exerts only minor transcriptional effects in neurons, while profoundly altering gene expression in NPCs.

Discussion

In this study, we report that Mecp2 overexpression elicits profoundly divergent outcomes depending on the cellular context, with NPCs displaying significantly greater sensitivity than fully differentiated Ns. In progenitor cells, Mecp2 overexpression triggers widespread gene deregulation, converging in activating genes controlling neuronal differentiation. This results in accelerated neurogenesis both in vitro and in vivo, ultimately producing neurons with altered synaptic activity, a deficit reminiscent of the electrophysiological defects observed in MDS mouse models and consistent with the high incidence of epilepsy in MDS patients^63,64. In stark contrast, Mecp2 overexpression in neurons induces only modest transcriptional and functional alterations both in vitro and in vivo, indicating a higher tolerance of mature cells to Mecp2 dosage. Mechanistically, we show that Mecp2 preferentially binds CpG islands, regions typically hypomethylated and located near TSSs. Its genomic localization appears largely unaffected by global DNA methylation levels or by the differential mCA content between NPCs and neurons. This observation is consistent with recent findings from CUT&Tag and CUT&Run analyses in vitro and in vivo, respectively, which revealed Mecp2 affinity for hypomethylated DNA^16,32, a feature that previous ChIP-seq approaches failed to fully capture⁶⁵. Interestingly, Mecp2 genomic distribution appears largely conserved across developmental stages (NPCs vs Ns) and largely comparable between endogenous and ectopic proteins. Indeed, the main difference we observe lies in the relative occupancy state of these CpG islands. In Ns, the most target CpG islands are already saturated with endogenous Mecp2, limiting the binding of the ectopic protein. In contrast, the lower endogenous Mecp2 levels in NPCs make these regions more accessible, enabling broader recruitment of Mecp2-exo. Intriguingly, despite Mecp2 binding to around 15,000 CpG islands, the enrichment does not occur uniformly across them. Both endogenous and overexpressed Mecp2 preferentially accumulate on genes associated with neurogenesis and neuronal maturation. The reasons behind this selective binding need to be investigated. Intriguingly, among the most enriched genes bound by Mecp2 in NPCs, nearly half are bivalent, characterized by the simultaneous presence of activating and repressive histone marks in their promoters, and are strongly implicated in the regulation of cell fate decisions^38–41. These genes also show marked upregulation in our transcriptomic analysis following Mecp2 overexpression, supporting the hypothesis that Mecp2 binding at these loci contributes directly to their transcriptional activation. Further confirming this, we identify Mecp2 interaction with the SWI/SNF chromatin remodeling complex, which is a crucial molecular player in promoting the bivalent genes activation^43–45. These results support the dual role of Mecp2 as both a repressor and an activator, emphasizing the dynamic nature of its function across development. However, it remains to be experimentally proven that Mecp2-dependent bivalent gene activation can also occur during native neuronal differentiation. It is important to note that this early activation of bivalent genes could offer some insights into the early neurodevelopmental signs seen in MDS patients. Indeed, patients often present with craniofacial and brain dysmorphisms, which are typically associated with disrupted developmental processes, further supporting the assumption that Mecp2 overexpression may interfere with early developmental processes^10,11. Additionally, the evidence that Mecp2 overexpression accelerates neurogenesis has already been described in the MDS mouse model, but the underlying molecular mechanism has not been elucidated⁶⁶. However, these results do not necessarily imply that restoring MeCP2 levels in mature MDS neurons would be ineffective. Indeed, correcting MeCP2 expression at later stages may still lead to substantial functional improvements, as already indicated by previous studies^19,67, although it is unlikely to fully rescue all deficits, as many of them originate during development. Regarding Mecp2-dependent chromatin solubility, changes observed in our in vitro models were modest but primarily involved transitions from heterochromatin to euchromatin in NPCs, corroborating a role for Mecp2 in promoting chromatin accessibility and transcriptional activation. It is possible that the extent of this effect in vitro is mitigated by culture conditions, which may dampen the neuronal differentiation potential driven by Mecp2 overexpression.

On the contrary, in mature Ns, Mecp2 overexpression results in minimal CpG islands binding enrichment, due to the saturation of these sites by the endogenous protein. As a consequence, the ectopic Mecp2 is also more rapidly degraded, consistent with its lower chromatin affinity, and fails to induce significant transcriptional or chromatin changes. Notably, differential phosphorylation patterns observe at Mecp2 Ser80 and Ser421 suggest that post-translational modifications may further modulate its DNA-binding dynamics and stability, contributing to the distinct biochemical behavior of the exogenous protein in NPCs and neurons. In line with these observations, Mecp2 overexpression in human iPSC-derived neurons also causes minimal transcriptional alterations, suggesting that this molecular framework is evolutionarily conserved and has direct relevance for medical applications in humans.

Our work demonstrates that Mecp2 overexpression can be well tolerated, highlighting the critical role of cellular context in shaping its functional outcome. Such context-dependent effects are likely not unique to Mecp2 and may extend to other chromatin regulators. Indeed, gene duplications involving factors such as RAI1, CHD4, CHD8, and EHMT1 have been implicated in rare neurodevelopmental disorders, each associated with a broad spectrum of neurological deficits^68–71. As for MeCP2, these factors are expressed during neural development and their alteration heavily impact on the transcriptional program of progenitor cells, altering their developmental differentiation program^72,73. In adulthood, several chromatin factors act as tumor suppressor genes, since their inactivation promotes tumorigenesis in patients and experimental models^74–76. However, much less is known about the consequences of their overexpression in terms of tolerability and underlying mechanisms in adult mature cell types. Thus, our work is uncovering an intriguing layer of chromatin factor regulation, which can complement our understanding on the mechanisms controlling key aspects of their functions related to gene network flexibility, compensatory mechanisms and cell type specificity.

Despite these significant insights, some limitations should be acknowledged in our study. The acute nature of our model may not fully capture long-term consequences of Mecp2 overexpression. However, in previous work, we found no major transcriptional abnormalities months after Mecp2 overexpression in wild-type mouse brains, supporting the hypothesis of neuronal tolerance also at prolonged time points²¹. Moreover, while this study focuses on neurons, Mecp2 is also expressed in other cell types, including astrocytes, oligodendrocytes, and peripheral tissues, which may respond differently to Mecp2 overexpression and warrant further investigation. In addition, our experiments are conducted in wild-type cells, which may not fully capture the effects of supraphysiological Mecp2 levels in mutant backgrounds. However, it is likely that Mecp2 behaves similarly in disease-relevant contexts, as multiple studies have shown that restoring its expression, either genetically or through gene therapy, in Mecp2-deficient mouse models leads to robust phenotypic and molecular rescue^{18,20,21,77–81}. This suggests that Mecp2 retains its ability to properly engage with its chromatin targets even in pathological conditions.

Our findings significantly advance our understanding of Mecp2 function and have important implications for gene therapy strategies for Rett syndrome. We show that Mecp2 overexpression is not intrinsically harmful; conversely, its impact is highly context-dependent. Neurons, specifically, appear capable of tolerating moderate Mecp2 increase (3-4-fold) without evident detrimental effects. This aligns with preclinical and clinical gene therapy studies, where even high AAV doses rarely caused Mecp2-related toxicity and adverse events were primarily linked to immune responses against the viral vector^20–22. Altogether, our data challenge the notion that Mecp2 dosage must be narrowly constrained, supporting the possibility of more flexible and potentially higher dosing strategies in therapeutic settings. In conclusion, our study reveals an additional layer of complexity in Mecp2 biology, showing how dosage and timing intersect with cellular identity to shape neurodevelopmental processes. At the same time, our findings provide critical insights that may inform the design of safer and more effective gene therapy approaches for RTT and related disorders.

Methods

Ethics

All procedures were performed according to the Declaration of Helsinki. The hiPSC lines were generated within the framework of a research project that received approval from the Institutional Ethics Committee of IRCCS Ospedale San Raffaele (# Metabolic-StemCells).

All animal-related procedures received ethical approval for animal experiments (n. 93/2022-PR) from the institutional IRCCS Ospedale San Raffaele IACUC (#1245) and was reported to the Italian Ministry of Health in accordance with European Communities Council Directive 2010/63/EU. C57BL/6 mice were housed at the Institutional Animal Facility of the San Raffaele Scientific Institute (Milan, Italy) under specific pathogen-free (SPF) conditions in micro-isolator cages, maintained at 18–23 °C with 40–60% humidity and a 12 h light–dark cycle. Mice were fed a standard laboratory chow diet with ad libitum access to food and water unless otherwise specified. All animals were euthanized in accordance with institutional and national ethical guidelines using approved methods designed to minimize suffering.

Plasmids

For the generation of lentiviral plasmids containing the mouse or human Mecp2/MECP2 gene, the CDS (e1 isoform) was PCR amplified in order to add the V5 tag at the 5’ of the coding sequence and inserted in the LV-Ef1a-EGFP removing the EGFP coding sequence. For generation of AAV plasmids containing the mouse Mecp2 gene, the CDS was PCR amplified in order to add the V5 tag at the 5’ of the coding sequence and inserted in the single-strand AAV-Ef1a-EGFP removing the EGFP coding sequence.

Primary cell cultures

Primary neural progenitor cell cultures

Neural Progenitors Cells (NPCs) were derived from telencephalic cortex of C57BL/6 N embryos at 13.5 days of gestation as described in ref. ⁸². Briefly, embryonic cortices were dissociated, fragmented in Hank’s Balanced Salt Solution (HBSS, Life Technologies) with 1% Penicillin/Streptomycin (Sigma- Aldrich) and digested with papain (10 U/ml, Worthington Biochemical) and cysteine (1 mM, Sigma-Aldrich) in HBSS with 0.5 mM EDTA at 37 °C. Cells were then cultured in adhesion seeded in flask coated with 1:100 Matrigel, in Neural inducing medium (NIM) composed by: DMEM/F12 (Sigma-Aldrich) supplemented with Hormon Mix (DMEM/F12, 0.6% Glucose (Sigma-Aldrich) (30%), Insulin (Sigma-Aldrich) 250 µg/ml, putrescin powder (Sigma-Aldrich) 97 µg/ml, apotransferrin powder (Sigma Aldrich), sodium selenite 0.3 µM, progesterone 0.2 µM), 1 mg/ml penicillin/streptomycin (Sigma-Aldrich), 2 mM glutamine (Sigma-Aldrich), 0.66% Glucose (30% in phosphate buffer salt (PBS) (Euroclone), Heparin 4 µg/ml, 10 ng/ml bFGF (basic fibroblast growth factor) (ThermoFisher Scientific) and 10 ng/ml EGF (epithelial growth factor) (10 ng/ml) (ThermoFisher Scientific).

For immunofluorescence analysis (Fig. S1b), 5 × 10⁴ cells were plated on coverslips in a 24-well plate. The following day, cells were transduced with increasing MOIs of Lv-Mecp2 (1, 5, 10, 20), and 2 days post-transduction, they were fixed. To study premature NPC differentiation after Mecp2 overexpression (Fig. 1c, d), EGF and bFGF were withdrawn from the NIM medium, and cells were fixed 5 days after transduction.

For Western blot and RNA-seq analyses (Fig. 1), 2.5 × 10⁵ cells were plated in 6-well plates. The following day, cells were transduced with Lv-Mecp2 at a selected MOI of 10, and 5 days post-transduction, they were lysed.

Primary neuronal cell cultures

Primary Neuronal culture were prepared at embryonic day 18.5 (E18.5) from wild-type mouse embryos as described in Luoni et al.²¹. Briefly, after dissection, cortexes were enzymatically digested with 0.025% trypsin (GIBCO) in HBSS (Euroclone) for 20 min at 37 °C and then mechanically dissociated with a P1000-pipette to obtain a homogeneous cell suspension. Neuronal cells were then plated on poly-L- lysine (PLL) (0.1 mg/ml) coated plates and maintained in Neurobasal medium (TermoFisher Scientific) supplemented with 0,6% glucose (Sigma-Aldrich), 0,2%penicillin/streptomycin (Sigma-Aldrich), 0,25% L-glutamine (Sigma-Aldrich) and 1% B27 (TermoFisher Scientific). For immunofluorescence analysis (Fig. S1c), 2 × 10⁵ cells were plated on coverslips in 24-well plates. The following day, cells were transduced with increasing MOIs (1, 5, 10, 20) of Lv-Mecp2, and 2 days post-transduction, they were fixed.

For Western blot and RNA-seq analyses, 1 × 10⁶ cells were plated in 6-well plates. The following day, cells were transduced with Lv-Mecp2 at a selected MOI of 10, and 5 days post-transduction, they were lysed.

Lentiviral vector production and transduction

Lentiviral replication-incompetent, VSVg-coated lentiviral particles were packaged in 293 T cells as described in Luoni et al.²¹. Briefly, cells were transfected with 30 μg of vector and packaging constructs using a standard CaCl₂ transfection protocol. After 30 h, the medium was collected, filtered through a 0.44 μm cellulose acetate membrane, and centrifuged at 68,320 × g for 2 h at 20 °C to concentrate the virus. Viral titers were determined using the Lentivirus Titration Kit (qPCR-based, Gentaur) according to the manufacturer’s instructions. For transduction, cells were infected with Lv-Mecp2 at the indicated multiplicities of infection (MOIs), as specified in each experiment.

AAV-PHP.eB vector production and systemic injection

AAV replication-incompetent, recombinant viral particles were produced 293 T cells as described in Luoni et al.²¹. In brief, cells were cultured in DMEM—high glucose (Sigma-Aldrich) supplemented with 10% fetal bovine serum (Sigma-Aldrich), 1% non-essential amino acids (Gibco), 1% sodium pyruvate (Sigma-Aldrich), 1% glutamine (Sigma-Aldrich) and 1% penicillin/streptomycin (Sigma-Aldrich). Cells were transfected with polyethylenimine (PEI) (Polyscience) using three different plasmids: a transgene-containing plasmid, a packaging plasmid encoding the rep and cap genes (AAV-PHP.eB), and pHelper (Agilent) for adenoviral helper gene expression. After 120 h, both cells and supernatant were harvested. Cells were lysed in a hypertonic buffer (40 mM Tris, 500 mM NaCl, 2 mM MgCl₂, pH 8) supplemented with 100 U/ml Salt Active Nuclease (SAN) (ArcticZymes) and incubated for 1 h at 37 °C. Meanwhile, viral particles in the supernatant were concentrated by precipitation with 8% PEG8000 (Sigma-Aldrich) and incubated for an additional 30 min at 37 °C. The viral fraction was purified using an iodixanol step gradient (15, 25, 40, and 60% OptiPrep, Sigma-Aldrich), with the virus collected from the 40% fraction and further concentrated in PBS using a 100 K cut-off concentrator (Amicon Ultra-15, MERCK-Millipore). Viral titers were quantified using the AAVpro© Titration Kit Ver2 (TaKaRa).

Wild-type juvenile (P25-30) C57BL/6 N mice were then injected in the tail vein with 1*10^11 vg/mouse of AAV-PHP.eB carrying EGFP or Mecp2 transgene.

Gene expression analysis

Total RNA was isolated from primary Ns, NPCs, and human-derived NPCs and Ns using Trizol (Sigma-Aldrich). The amount of RNA samples was quantified using a Nanodrop spectrophotometer (Thermo Scientific). Libraries and paired-end 51 × 51 run sequencing was performed by Genewiz (www.genewiz.com).

Immunofluorescence

For immunostaining, mNPCs, mNs, hNPCs and hNs were fixed with ice-cold 4% paraformaldehyde (PFA) for 30 min at 4 °C, washed with PBS (3×) and incubated with 10% donkey serum and 3% Triton X-100 for 1 h at RT to saturate the unspecific binding site before the overnight incubation at 4 °C with the primary antibody: mouse anti-V5 (1:500; ThermoFisher Scientific, #R960); chicken anti-EGFP (1:1000; ThermoFisher Scientific, #A10262); rabbit anti-Ki67 (1:500; ThermoFisher Scientific, # MAB-90948) and rabbit anti-Tubb3 (1:500, Covance, # MMS-435P).

Upon wash with PBS (3×), cells were incubated for 1 h at RT in blocking solution with Hoechst and with Alexa Fluor-488 and Alexa Fluor-594 anti-rabbit or anti-mouse or anti-chicken secondary antibodies. After PBS washes (3×), cells were mounted with fluorescent mounting medium (Dako).

Embryos at E18.5 were dissected to extract the brain, which was post-fixed in 4% PFA for 12 h. The following day, brains were immersed in a cryoprotective solution (30% sucrose in PBS) to prevent tissue damage during freezing.

For adult mice, animals were anesthetized with a ketamine/xylazine mix and transcardially perfused with 0.1 M phosphate buffer (PB) at room temperature (RT, pH 7.4). Brains and livers were then post-fixed in 4% PFA for 2 days before being transferred into a cryoprotective solution (30% sucrose in PBS).

Tissues were embedded in optimal cutting temperature (OCT) compound and frozen on dry ice before sectioning with a cryostat. Free-floating 50 μm-thick coronal sections were rinsed in PBS and incubated in a blocking solution containing 10% donkey serum (Sigma-Aldrich) and 3% Triton X-100 (Sigma-Aldrich) for 1 h at RT to prevent non-specific antibody binding. Sections were then incubated overnight at 4 °C with the primary antibody, diluted in the blocking solution.

Primary antibodies for the following epitopes were used: rabbit anti-MeCP2 (1:500; Cell Signaling Technology, #3456), mouse anti-V5 (1:500; ThermoFisher Scientific, #R960), chicken anti-EGFP (1:1000; ThermoFisher Scientific, #A10262); rabbit anti-Pax6 (1:500; Covance, #PRB-278P) and rabbit anti-Tubb3 (1:500, Covance, #MMS-435P). Upon wash with PBS (3×), sections were incubated for 1 h at RT in blocking solution with Hoechst (1:1000, Sigma-Aldrich) and with Alexa Fluor-488 and Alexa Fluor-594 anti-rabbit, anti-mouse, or anti-chicken secondary antibodies (1:1000, ThermoFisher Scientific). After PBS washes (3×), sections were mounted with fluorescent mounting medium (Dako). Images in Figs. S1a, b and 5a, b were acquired using a Leica SP8 confocal microscope with identical settings for the V5 channel (gain: 938; laser power: 35%; scan speed: 100 Hz). Images in Figs. S3d and S4a, b were acquired using a Nikon Eclipse 600 fluorescence microscope with identical exposure settings for EGFP (300 ms) and V5 (1 s).

Western blot

Protein extracts were prepared using RIPA buffer (10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EGTA, 0.5% Triton X-100) supplemented with 1% complete protease and phosphatase inhibitor mixture (Roche Diagnostics).

For Western blot analysis, 50 μg of protein lysates were resolved on an 8% polyacrylamide gel and transferred onto PVDF membranes. Membranes were then incubated overnight at 4 °C with the following primary antibodies, diluted in 1× PBST containing 5% (w/v) nonfat dry milk: Rabbit anti MeCP2 (1:1000, Sigma, #M9317), Mouse anti-V5 (1:1000, Thermo Fisher Scientific, #R960); Mouse anti-FLAG (1:1000, Sigma); Phospho-MECP2 Ser421 (1:500, Thermo Fisher Scientific; Cat#PA5-35396); Phospho-MECP2 Ser80 (1:500, Thermo Fisher Scientific, #PA5-104679), Mouse anti-β-Actin (1:50000, Sigma, #A5228). The next day, membranes were incubated with the appropriate HRP-conjugated secondary antibodies (1:10000, Dako). The signal was visualized using an ECL chemiluminescent reagent (RPN2232, GE Healthcare) and detected with the ChemiDoc imaging system (Bio-Rad). Signal intensities were quantified using ImageLab software and normalized to β-actin levels or total Mecp2 protein levels. For presentation of full scan blots, see the Source Data file.

Dot blot assay for 5-mCpA detection

Dot blot analysis was performed on positively charged nylon membranes (0.45 µm; Hybond-N⁺ or Nytran SuPerCharge) to detect 5-mCpA using the monoclonal anti-5-mCpA antibody 2C8H8A6 (Abcam, #ab307565). Approximately 20–30 ng of DNA samples (quantified using the Qubit dsDNA Broad Sensitivity Assay, Thermo Fisher Scientific) or probes were spotted on the membrane. Genomic DNA or control oligonucleotides were denatured in 0.1 M NaOH and 1 mM EDTA for 10 min at 95 °C, neutralized with 1 M Tris-HCl, pH 7.4, and chilled on ice. Denatured were spotted (5–10 µL) onto nylon membranes and air-dried for 20–30 min. DNA was crosslinked to the membrane by UV irradiation (120 mJ/cm²). After blocking with 3% BSA in TBS-T (0.1% Tween-20) for 1 h at RT, membranes were incubated overnight at 4 °C with anti-5-mCpA antibody (clone 2C8H8A6, 1 µg/mL in TBS-T + 3% BSA). Following four 5-min washes in TBS-T, membranes were incubated for 1 h at RT with HRP-conjugated anti-mouse IgG (1:10,000, Dako), washed again, and developed using ECL chemiluminescent reagent (RPN2232, GE Healthcare) and detected with the ChemiDoc imaging system (Bio-Rad). Signal intensities were quantified using ImageLab software. and normalized to β-actin levels.

As controls, three synthetic double-stranded oligonucleotide probes were used:

• mCG probe: G/5MedC/GATAGCTGATGCACAGTA/5MedC/GATC

• non-methylated probe: GCGATAGCTGATGCACAGTACGATC

• 5mCA probe: GCGATAGCTGATG/5MedC/A/5MedC/AGTACGATC

CUT&Tag

Three independent biological replicates of NPCs and Ns for each group (EGFP or Mecp2) were harvested and the cell number was determined by Countess Automated Cell Counter (Invitrogen). Cells were centrifuged and processed for CUT&Tag following the protocol described in ref. ^26,83. Briefly, 100,000 nuclei were immobilized on Concanavalin A-coated beads (company name) and incubated overnight at 4 °C with the following primary antibodies: Anti-MeCP2 (Diagenode, #C15410052); Anti-V5 (Thermo Fisher Scientific, #R960); Anti-Smarcb1 (Abcam #ab307985); Rabbit IgG (Cell Signaling, Cat#2729). The next day, nuclei were incubated with an unconjugated secondary antibody (Goat anti-rabbit IgG antibody, Thermo Fisher Scientific, Cat#A27033; Goat anti-mouse IgG antibody, Thermo Fisher Scientific, Cat#31160) before treatment with pA-Tn5 transposase (Diagenode). Finally, amplicon libraries were generated using the Nextera® DNA Library Prep Kit (Illumina), following the manufacturer’s instructions. Barcoded libraries were pooled and sequenced on the NovaSeq 6000 (Illumina) platform.

CO-IP assay

HEK293 cells were lipofected using Lipofectamine LTX (ThermoFisher) with pMECP2-EGFP (Addgene #181904) and Ef1a-SMARCB1 (Addgene #144634), either alone or in combination. The following day, cells were lysed in IP buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100). To remove non-specific binding, preclearing was performed with magnetic beads (Protein G Dynabeads, Life Technologies) for 2 h at 4 °C on a rotating wheel. A portion of each sample was saved as input before antibody incubation. Samples were then incubated overnight with the required antibody (V5, Thermo Fisher Scientific, or EGFP, Thermo Fisher Scientific) while rotating. The following day, immunoprecipitation (IP) was performed using Protein G-coated magnetic beads for 2 h at 4 °C. After three washes, the samples were boiled at 70 °C. Western blot (WB) analysis was performed as described in the Western Blot section.

4f-SAMMY-seq

The 4f-SAMMY-seq was performed on 2 × 10⁵ NPCs at 90% confluency and 2 × 10⁵ Ns using 2 units of DNase I (Invitrogen, AM2222) as described in Lucini et al.⁵¹ Libraries were generated from each sample using NEBNext Ultra II DNA Library Prep Kit for Illumina (NEB, E7645L) and Unique Dual Index NEBNextMultiplex Oligos for Illumina (NEB, E6440S). Libraries were then qualitatively and quantitatively checked on and run on the TapeStation System. Libraries with distinct adapter indexes were normalized to a concentration of 2 nM, equimolarly, pooled, and then loaded onto the Illumina NextSeq 2000 instrument. The sequencing was performed with a minimal target of 15 million reads for 100 bases in single-end mode on the Illumina NextSeq 2000 instrument.

Salt-solution extraction assay

NPCs and Ns transduced with EGFP or Mecp2 were lysed on ice for 5 min in a lysis buffer composed of 300 mM sucrose, 2 mM MgCl₂, 3 mM CaCl₂, 10 mM Tris-HCl (pH 8), 0.1% Triton X-100, and 0.5 mM DTT, supplemented with 1% complete protease and phosphatase inhibitor mixture (Roche Diagnostics).

Following lysis, the samples were centrifuged at 720 × g for 5 min at 4 °C. The supernatant was carefully removed, and the pellet was resuspended in extraction buffer 1 (10 mM Tris-HCl, pH 7.4, 200 mM NaCl, and protease and phosphatase inhibitors) and incubated for 20 min.

To perform sequential extractions, the pellet was subsequently resuspended in extraction buffer 2 (400 mM NaCl), extraction buffer 3 (600 mM NaCl), and finally extraction buffer 4 (800 mM NaCl), ensuring that at each step, the pellet was thoroughly resuspended before centrifugation.

Finally, all fractions were separated on polyacrylamide gels, as described in the Western blot section.

Cycloheximide protein stability assay

NPCs and neurons (Ns) transduced with EGFP or Mecp2 were incubated with CHX (50 µM). Cells were collected at 0 and 8 h after treatment for protein extraction. Lysates were then prepared for Western blot analysis as described below.

In utero electroporation

In utero electroporation was performed as described in Zaghi et al.⁸⁴. Briefly, E14.5 pregnant C57BL/6 N females were anesthetized with Avertin (312 mg/kg), and a midline laparotomy was performed to expose the uterus. Two separate plasmids encoding either EGFP or Mecp2 were mixed with fast green dye in PBS and injected together (for Mecp2 overexpression) or EGFP alone into the telencephalic vesicle of using a pulled glass micropipette inserted through the uterine wall and amniotic sac. Platinum tweezer-style electrodes (7 mm) were positioned outside the uterus over the telencephalon at a 40–95° angle to target the somatosensory cortex⁸⁵, and five square-wave pulses (40 V, 50 ms each) were delivered at 950 ms intervals using a BTX electroporator. After electroporation, the uterus was returned to the abdominal cavity, which was then filled with warm sterile PBS. The abdominal muscle and skin incisions were closed with silk sutures. All procedures were approved by the Italian Ministry of Health and the San Raffaele Scientific Institute Animal Care and Use Committee, following the relevant guidelines and regulations.

Electrophysiological recording

For single-cell electrophysiological recordings, coronal brain slices were obtained from C57BL/6 mice of the 4 experimental groups (Embryo EGFP; Embryo Mecp2; Adult EGFP; Adult Mecp2) at the age of P45-55. Briefly, mice were euthanized following deep isoflurane anesthesia, and brains were isolated in ice-cold choline-based solution, containing (in mM): 110 CholineCl, 25 NaHCO3, 11 D-glucose, 11.6 NaAscorbate, 3.1 NaPyruvate, 2.5 KCl, 1.25 NaH2PO4, 0.5 CaCl2 and 7 MgCl2, bubbled with 95% O2 and 5% CO2 (pH 7.4), adapted from Chiu et al.⁸⁶. For mice electroporated in utero, the presence of a fluorescent spot in the correct area (S1 cortex) of the targeted hemisphere was confirmed under a stereomicroscope. 350μm-thick coronal sections were cut using a Leica VT 1200 vibratome. The slices were allowed to recover in the cutting solution at 32 °C for 15 min, then at TR for 15 more minutes before being moved in a chamber filled with the recording solution (ACSF), containing (in mM): 125 NaCl, 25 NaHCO3, 10 D-glucose, 2.5 KCl, 1.25 NaH2PO4, 2 CaCl2 and 1 MgCl2, (pH 7.4), maintained at TR under continuous carbogenation. For recordings, S1 cortical slices were placed in the recording chamber beneath a 40X water immersion lens and visualized using infrared differential interference contrast (IR-DIC) video microscopy. Whole-cell patch-clamp recordings were performed in continuous superfusion with carbogenated ACSF, maintained at 30–32 °C. Spontaneous excitatory postsynaptic currents (sEPSC) were recorded in voltage clamp (VC) configuration from L2/3 pyramidal neurons, identified by their typical drop-shaped somata, using glass pipettes (3–4 MΩ) filled with a cesium-based intracellular solution, containing (in mM): 125 Cs-CH3SO3, 5 NaCl, 10 HEPES, 5 EGTA, 2 MgCl2, 2 Mg-ATP, 0.2 Na-GTP, adjusted to pH 7.25 with CsOH (modified from Valassina et al.⁸⁷). sEPSC were recorded in the absence of blockers, isolated by holding the membrane potential at −65 mV (the reversal potential for GABAergic chloride currents).

All recordings were acquired using a Multiclamp 700B Amplifier (Molecular Devices), low-pass filtered at 2 kHz, and digitized at 10 kHz using a Digidata 1550 D/A converter (Molecular Devices) controlled by the pCLAMP11 software (Molecular Devices). Access resistance (Ra) was monitored continuously during the recording, and cells with a Ra > 25 MΩ, or with a variation greater than 20%, were excluded from analysis.

sEPSC were recorded for at least 3 min after dialization with the intracellular solution, and 120 consecutive seconds were analysed off-line using the event detection tool Mini Analysis Program (Synaptosoft). Traces were low-pass filtered at 1 kHz to improve the signal-to-noise ratio, and the threshold amplitude for event detection was adjusted to −10 pA, above the double of the root mean square noise level. Events were subsequently checked manually for accuracy.

Differentiation of hiPSCs into NPCs and neurons

Human NPCs and cortical neurons were differentiated as described in Iannielli et al.⁸⁸ Briefly, iPSCs were dissociated in cell clusters using Accutase (Sigma-Aldrich) and seeded onto low-adhesion plates in mTeSR1 supplemented with N2 (1:200, Thermo Fisher Scientific), Pen/Strept (1%, Sigma-Aldrich), human Noggin (0.5 μg/ml, R&D System), SB431542 (5 μM, Sigma-Aldrich), and Y27632 (10 μM, Selleckchem). 10 days later, embryoid bodies were seeded onto matrigel-coated plates (1:100, matrigel growth factor reduced, Corning) in DMEM/F12 (Sigma-Aldrich) supplemented with N2 (1:100), non-essential amino acids (1%, MEM NEAA, Thermo Fisher Scientific), and Pen/Strept. After 10 days, rosettes were dissociated with Accutase and plated onto matrigel-coated-flasks in NPC media containing DMEM/F12, N2 (1:200), B27 (1:100, Thermo Fisher Scientific), Pen/Strept (1%) and FGF2 (20 ng/ml, Thermo Fisher Scientific).

For immunofluorescence experiments (Fig. S5a–c), 1 × 10⁵ cells were plated on coverslips in 24-well plates, and different lentiviral MOIs (1, 5, 10, and 20) of Lv-MeCP2 were tested as indicated in the figure. For Western blot and RNA-seq analyses (Figs. 7d and S5d), 5 × 10⁵ cells were plated in 6-well plates, transduced with the selected MOI 10 (Lv-MeCP2 and Lv-EGFP as control), and lysed 5 days after transduction.

For differentiation of cortical neurons, NPCs were dissociated with Accutase (Sigma-Aldrich) and plated on matrigel-coated 6-well plates (3 × 10⁵ cells per well) in NPC medium. Two days after, the medium was changed with the differentiation medium containing Neurobasal (ThermoFisher Scientific), 1% Pen/Strep (Sigma-Aldrich), 1% Glutamine (Sigma-Aldrich), 1:50 B27 minus vitamin A (ThermoFisher Scientific), 5 μM XAV939 (Sigma-Aldrich), 10 μM SU5402 (Sigma-Aldrich), 8 μM PD0325901 (Tocris Bioscience), and 10 μM DAPT (Sigma-Aldrich) was added and kept for 3 days. After 3 days, the cells were dissociated with Accutase (Sigma-Aldrich) and plated on poly-L-lysine (Sigma-Aldrich)/laminin (Sigma-Aldrich)-coated 6-well plates (4 × 10⁵ cells per well) and 24-well plates with coverslip (1 × 10⁵ cells per well) in maturation medium containing Neurobasal (ThermoFisher Scientific), 1% Pen/Strep (Sigma-Aldrich), 1% Glutamine (Sigma-Aldrich), 1:50 B27 minus vitamin A (ThermoFisher Scientific), 25 ng/ml human BDNF (PeproTech), 20 μM Ascorbic Acid (Sigma-Aldrich), 250 μM Dibutyryl cAMP (Sigma-Aldrich), 10 μM DAPT (Sigma-Aldrich) and Laminin for terminal differentiation. At this stage half of the medium was changed every 2–3 days. Lentiviral particles were directly added to neuronal cultures after 6 weeks of differentiation, testing different MOIs of Lv-MeCP2 for immunofluorescence analysis (Fig. S5b, c). Protein and RNA extractions for Western blot and RNA-seq analyses (Figs. S5c and 7d) were performed 5 days after transduction with the selected MOI 20 (Lv-MeCP2 and Lv-EGFP as control).

Computational analysis

CUT&Tag data processing and analysis

Raw CUT&Tag sequencing reads for MeCP2 (endogenous and exogenous conditions in Neural Stem Cells (NSC), Neurons (NEU)) and control (IgM) samples underwent quality control using FastQC (0.12.1). Adapters and low-quality bases were removed using Trim Galore (0.6.10). Trimmed reads were aligned to the mouse reference genome (mm10) using Bowtie2 (2.5.4) with parameters optimized for CUT&Tag (--local --very-sensitive-local --no-mixed --no-discordant --maxins 1000). Alignments were filtered for proper pairing, mapping quality (MAPQ ≥ 20), and to exclude unmapped, non-primary, low-quality, and supplementary alignments. PCR duplicates were identified and marked using samtools markdup (1.22.1). Alignment statistics (samtools flagstat, idxstats) and library complexity (preseq 3.2.0) were assessed. Peak calling was performed using MACS2 (2.2.7.1) with parameters optimized for CUT&Tag data (-f BAMPE --nomodel --keep-dup all) for both narrow and broad peak detection, using IgM as the control. Peaks were called at an FDR threshold of q < 0.05 and filtered to remove ENCODE blacklist regions. Replicate peaks were combined using a custom interval-tree-based approach requiring peaks to be present in at least 2 out of 3 replicates. Peaks within 500 bp of each other were considered matching and merged. Normalized signal tracks in bigWig format were generated using deepTools bamCoverage with RPKM normalization and 10 bp bin size. Peaks were annotated using the ChIPseeker R package with the TxDb.Mmusculus.UCSC.mm10.knownGene transcript database and org.Mm.eg.db gene information to identify nearest genes, distance to TSS (using ±3 kb TSS regions), and genomic feature distribution. CpG island analysis was performed using CpG island coordinates from a pre-compiled BED file (cpg_islands.bed). Peaks overlapping promoter regions (defined asymmetrically as −2500 bp upstream and +500 bp downstream of TSS for plus-strand genes, reversed for minus-strand) and CpG islands (within 500 bp) were identified. CpG islands within 100 bp of each other were merged. Peaks were categorized as “exo_only”, “endo_only”, “common”, or “none” based on presence/absence in exogenous and endogenous samples, with categorization determined by peak count > 0 and mean signal > 0. Two versions of promoter-CpG analysis were performed: one using only primary TSS and another using all alternative TSSs from GENCODE vM10 annotation. Metaprofiles visualizing MeCP2 signal distribution at CpG islands and TSS regions were generated using deepTools computeMatrix reference-point (center-point for CpG islands, TSS for transcription start sites) with ±3 kb flanking regions, followed by plotProfile for average profile generation.

CpG island enrichment analysis was performed using a custom multi-step workflow. First, CpG islands overlapping with called peaks from at least 2 out of 3 replicates were identified. For each CpG island with overlapping peaks, the outer boundaries of all overlapping peaks across both exogenous and endogenous samples were determined. This merged region was used for signal quantification. Mean signal intensities were extracted from bigWig files (RPKM-normalized, 10 bp bins) for each replicate using pyBigWig. Enrichment scores were calculated as the ratio of the mean exogenous to the mean endogenous signal. Regions were categorized as up-enriched (exo/endo signal ratio > 1.0, 1.5, or 2.0), down-enriched (ratio < 1.0, 0.8, or 0.5), exo-only (signal only in exogenous samples), and endo-only (signal only in endogenous samples). Peak localization relative to genomic features was performed using ChIPseeker. The distribution of each enrichment category was analyzed across promoters, gene bodies, CpG islands, and intergenic regions. Integration with RNA-seq data was performed by mapping CpG-enriched regions to their nearest genes (within ±2 kb). For genes with multiple associated CpG regions, only the closest region was retained. Genes were categorized by DESeq2 results: up-regulated (log2 fold change > 0.5, adjusted p-value < 0.05), down-regulated (log2 fold change < −0.5, adjusted p-value < 0.05), or non-deregulated ( | log2FC | ≤ 0.5 or padj ≥ 0.05). Enrichment score distributions were compared across expression categories.

MeCP2 binding-expression correlation analysis was performed to assess the relationship between MeCP2 enrichment at peaks and differential gene expression. Peaks were assigned to the nearest gene and categorized by genomic location (promoter, 5’UTR, exon, intron, 3’UTR, downstream, distal intergenic). Enrichment ratios were calculated as log2(exogenous signal/endogenous signal). Peak-gene associations were integrated with RNA-seq differential expression analysis results (DESeq2). Only differentially expressed genes (from filtered DEA results) with overlapping peaks were retained. For genes with multiple overlapping peaks, signals were aggregated by genomic region type. Spearman’s rank correlation was calculated between peak enrichment ratios log2(Exo/Endo) and absolute log2 fold changes in gene expression, separately for each genomic region category. Only region categories with ≥50 genes were analyzed. The code used to analyze the data presented in this article has been posted on Zenodo [10.5281/zenodo.18257503].

RNA-seq data analysis and integration

RNA sequencing and differential expression analysis were performed using a standard pipeline. Raw paired-end FASTQ files were quality-assessed using FastQC. Reads were aligned to the human reference genome (GRCh38/hg38) using STAR (v2.7.x) with GENCODE v44 annotation. STAR was run in two-pass mode with the following key parameters: --outFilterMultimapNmax 20, --alignSJoverhangMin 8, --outFilterMismatchNoverReadLmax 0.04, --alignIntronMin 20, --alignIntronMax 1,000,000, --quantMode GeneCounts, and --outSAMtype BAM SortedByCoordinate. Gene-level read quantification was performed using featureCounts (Subread package) with parameters: -p (paired-end), -B (count only properly paired reads), -C (exclude chimeric fragments), -s 0 (unstranded), counting reads mapping to exons and summarizing at the gene level (-g gene_id -t exon). Differential gene expression analysis between Control and Mutant conditions was performed using DESeq2. Raw count matrices were filtered to retain genes with at least 10 total reads across all samples. DESeq2 normalization and dispersion estimation were performed using default parameters. Differentially expressed genes were identified using the Wald test with a design formula ~condition. Genes with adjusted p-value (Benjamini-Hochberg FDR) < 0.05 were considered significantly differentially expressed. An additional stringent filtering criterion (adjusted p-value < 0.05 and |log2 fold change | > 1) was applied to identify genes with both statistical significance and substantial expression changes. For visualization, a variance-stabilizing transformation (VST) was applied to normalized counts. Principal component analysis (PCA) was performed to assess sample clustering and batch effects. Gene Ontology (GO) and KEGG pathway enrichment analyses were performed using clusterProfiler with org.Hs.eg.db annotation. Enrichment was tested separately for upregulated genes (log2FC > 1, padj < 0.05) and downregulated genes (log2FC < −1, padj < 0.05).

Smarcb1 binding analysis and integration

Initial quality control of the raw data for SMARCB1 (samples BG1, BG2, BG3, BM3) was performed using FastQC. Adapters and low-quality bases were trimmed using Trimmomatic with parameters (TruSeq3-PE.fa, LEADING:20, TRAILING:20, SLIDINGWINDOW:4:20, MINLEN:36). Trimmed reads were aligned to the mm10 reference genome using Bowtie2 with stringent parameters (--very-sensitive, -f 2, -F 1804 -q 30, --maxins 1000). Read groups were added using samtools, and PCR duplicates were optionally marked and removed using Picard MarkDuplicates. Normalized signal coverage tracks were generated using deepTools bamCoverage, producing both RPKM and CPM normalized BigWig files with a bin size of 10 bp and read extension enabled, ignoring duplicates. Peak calling to identify Smarcb1 binding sites was performed using MACS2 with BAMPE format, mm10 genome size (-g mm), a q-value threshold of 0.05, and retaining duplicates (--keep-dup all). Downstream analysis focused on characterizing SMARCB1 binding relative to genomic features and specific gene sets. Peaks were annotated using the ChIPseeker R package with the TxDb.Mmusculus.UCSC.mm10.knownGene transcript database and org.Mm.eg.db gene information to determine genomic feature distribution (promoters, introns, exons, etc.). Promoter regions were defined as 2000 bp upstream to 500 bp downstream of TSSs based on GENCODE vM10 annotation. Specific analyses focused on Smarcb1 binding within promoter regions overlapping CpG islands (using a predefined CpG island BED file). Read counts within these defined promoter or CpG-promoter regions were calculated for comparative analysis between conditions (e.g., BM3 vs BG samples). Metaprofiles and heatmaps visualizing Smarcb1 binding patterns around TSSs or other features were generated using deepTools computeMatrix, plotProfile, and plotHeatmap, as well as custom R scripts. These visualizations compared binding signals between different conditions (BM3 vs BG), across different cell types (NEU, NSC), and gene categories. Comparative analyses investigated differential Smarcb1 binding patterns, correlating binding intensity (coverage score) and width-weighted signal with gene features such as bivalency status (comparing SMARCB1 binding at bivalent vs non-bivalent genes) and Mecp2 targeting status. Statistical tests, such as the Mann-Whitney U test, were applied to assess significant differences in binding between compared groups.

SAMMY-seq bioinformatics analysis

Raw SAMMY-seq FASTQ data, representing different chromatin fractions (S2S, S3) from various conditions (e.g., NPCs_EGFP, NPCs_Mecp2, Ns_EGFP, Ns_Mecp2) and replicates, were processed

following established practices and standards from the literature^27,51,89 and in accordance with the SAMMY-seq workflow as described in the publicly available repository (https://github.com/Clockris/SAMMY-method_4f). Initial quality control and adapter trimming were performed using Trimmomatic (TruSeq3 adapter file, parameters: ILLUMINACLIP:2:30:10, SLIDINGWINDOW:4:15, MINLEN:36). Trimmed reads were aligned to the mm10 reference genome using BWA mem. SAM files were converted to BAM, coordinate-sorted, and filtered for mapping quality (MAPQ ≥ 1) using samtools. PCR duplicates were marked and removed using Picard MarkDuplicates. Replicate BAM files for the same condition and fraction were merged using samtools merge. Normalized coverage tracks in BigWig format were generated for each fraction (S2S, S3, etc.) per condition using deepTools bamCoverage. Parameters included RPKM normalization (--normalizeUsing RPKM), read extension (--extendReads 250), binning (--binSize 50), and filtering against genomic blacklist regions. To determine chromatin states, the log2 ratio of normalized coverage between accessible (S2S) and inaccessible (S3) fractions was calculated genome-wide using deepTools bigwigCompare (--operation log2 --pseudocount 1). Based on the final normalized log2(S2S/S3) ratios, genomic regions were classified into chromatin states using defined thresholds: euchromatin (log2 ratio > 0.1) and heterochromatin (log2 ratio < −0.1). The results were saved as BED files for each state and condition. Differential chromatin state analysis between conditions (e.g., NSC vs Neuron; EGFP vs M2) was performed to identify regions with significant changes. This involved calculating Z-scores for the difference in normalized log2 ratios between conditions at each genomic bin, performing Z-tests, and correcting for multiple testing to identify significantly differential regions. Bedtools intersect was also used to find regions common to or specific to certain conditions. Gene-centric analyses were performed by intersecting chromatin state BED files (euchromatin/heterochromatin) with promoter regions (defined as 2 kb upstream of TSSs from a GTF annotation file) using bedtools intersect. This allowed calculation of the proportion of different chromatin states within promoters for specific gene lists and comparison of state transitions between conditions (e.g., NSC vs Ns, EGFP vs Mecp2). Metaprofile heatmaps were generated using Python scripts to visualize the distribution of the SAMMY-seq log2(S2S/S3) ratio signal across gene bodies (from 5 kb of TSS to +5 kb of TES) for different gene sets (Mecp2 targets vs non-targets, genes changing state) and conditions.

Statistical analysis

Values are expressed as mean ± standard deviation as indicated. All statistical analysis was carried out in GraphPad Prism 8.0, using t-Test or ANOVA-one way with Tukey’s post hoc test. Number of replicates for statistical testing were indicated in corresponding figure legends.

For electrophysiological recordings, data are presented as box plots displaying mean (+), median (internal horizontal line), first and third quartiles (upper and lower box edges), and minimal and max values (whiskers) of the data distribution. Circles represent individual data points from each cell. Statistical analysis was performed using GraphPad Prism 8.0 and is reported in each figure legend, together with the number of cells and animals. Normal distribution of experimental data was assessed using D’Agostino-Pearson’s normality test, while the ROUT test was used to detect outliers. Unpaired t test followed by Welch’s correction was used to compare means of normally distributed sample groups. When the data were not normally distributed, Mann–Whitney non-parametric test was used for mean ranks comparison. Statistical significance was reached when p < 0.05.

For bioinformatics pipelines specific statistical tests were employed included: Wald tests for differential expression (RNA-seq) and differential binding (CUT&Tag/ChIP-seq) analyses, implemented via packages such as DESeq2 or DiffBind; Z-tests for identifying differential chromatin states in SAMMY-seq analysis; Mann–Whitney U tests for comparing distributions between groups (e.g., for CpG or Smarcb1 enrichment scores); and Spearman correlation tests for assessing relationships between continuous variables (e.g., methylation scores vs. Smarcb1 enrichment). Multiple testing correction for genome-wide analyses was performed using the Benjamini-Hochberg method to control the False Discovery Rate (FDR).

Author contributions

Conceptualization, V.B. and M.L.; methodology, S.G., F.M., C.D.B., and A.I.; investigation, S.G., F.M., C.D.B., A.I., and A.S.; formal analysis, M.L., A.S., G.C., and C.L.; computational analysis, M.K. and E.D.P.S.; data curation, M.L., C.L., and V.B.; writing—original draft preparation, V.B. and M.L. writing—review and editing, S.G., F.M., C.D.B., A.I., A.S., G.C., and C.L.; visualization preparation, V.B. and M.L.; supervision, V.B.; project administration, V.B.; funding acquisition, V.B. and C.L.

Peer review

Peer review information

Nature Communications thanks James Ellis and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All datasets generated in this study have been deposited at the NCBI Gene Expression Omnibus (GEO) database under accession code: GSE299957. In addition, previously published datasets were used for some analyses and are accessible at GEO under accession numbers GSE104585²⁹, GSE93011³⁸ and https://mecp2-neuroatlas.wi.mit.edu¹⁶. All graph and raw data generated are provided in the Supplementary Information or the Source Data file. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-71432-w.