Biotin tagging of MeCP2 in mice reveals contextual insights into the Rett syndrome transcriptome

Abstract

Mutations in MECP2 cause Rett syndrome (RTT), an X-linked neurological disorder characterized by regressive loss of neurodevelopmental milestones and acquired psychomotor deficits. However, the cellular heterogeneity of the brain impedes an understanding of how MECP2 mutations contribute to RTT. Here we developed a Cre-inducible method for cell-type-specific biotin tagging of MeCP2 in mice. Combining this approach with an allelic series of knock-in mice carrying frequent RTT-associated mutations (encoding T158M and R106W) enabled the selective profiling of RTT-associated nuclear transcriptomes in excitatory and inhibitory cortical neurons. We found that most gene-expression changes were largely specific to each RTT-associated mutation and cell type. Lowly expressed cell-type-enriched genes were preferentially disrupted by MeCP2 mutations, with upregulated and downregulated genes reflecting distinct functional categories. Subcellular RNA analysis in MeCP2-mutant neurons further revealed reductions in the nascent transcription of long genes and uncovered widespread post-transcriptional compensation at the cellular level. Finally, we overcame X-linked cellular mosaicism in female RTT models and identified distinct gene-expression changes between neighboring wild-type and mutant neurons, providing contextual insights into RTT etiology that support personalized therapeutic interventions.

Figure 1

Utilization and characterization of Mecp2^Tavi mice and associated RTT variants. (a) Diagram of wild-type and tagged MeCP2 showing R106W or T158M missense mutations. MBD, Methyl-CpG Binding Domain; TRD, Transcriptional Repression Domain. (b) Breeding strategy to biotinylate the Tavi tag in a Cre-dependent manner. (c) Representative western blot showing the conditions in which the Tavi tag is biotinylated using whole brain nuclear extracts. Blot is probed with streptavidin for biotin detection and antibodies against MeCP2 N-terminus, Tavi tag, and NeuN. (d) Representative images showing immunofluorescent detection of biotinylated MeCP2 and mutant variants in hippocampal sections of untagged (WT) and tagged (TAVI, T158M, R106W) male mice at 6 weeks of age. Tissue is probed with streptavidin for biotin detection and antibody against the MeCP2 C-terminus. Scale bars represent 10 μm. (e) Quantification and representative western blot comparing MeCP2 protein expression levels between TAVI and mutant (T158M, R106W) male mice at 6 weeks of age. Blot is probed with antibodies against the MeCP2 C-terminus and TBP (n_replicates = 3, One-way ANOVA). (f) Quantification of salt-extracted MeCP2 from chromatin using 200mM (left) and 400mM (right) NaCl, normalized to extracts using 500mM NaCl (see Supplemental Fig. 2e; n_replicates = 4–5, One-way ANOVA). (g) Box-and-whisker plot of brain weights from untagged (WT, KO (Mecp2-null)) and tagged (TAVI, T158M, R106W) male mice at 6 weeks of age (n_WT = 20, n_TAVI = 11, n_KO = 6, n_T158M = 6, n_R106W = 12; One-way ANOVA). Box limits denote 25^th and 75^th percentiles, center line denotes median, ‘+’ denotes mean, and whiskers denote data max and min. Each genotype is indicated with a different color. (h) Body weight over postnatal age in untagged (WT, KO) and tagged (TAVI, T158M, R106W) male mice. Data points consist of at least 6 observations each. Total number of mice assessed: n_WT = 31, n_TAVI = 23, n_KO = 15, n_T158M = 14, n_R106W = 28. (i) RTT-like phenotypic score across postnatal development in untagged (WT, KO) and tagged (TAVI, T158M, R106W) male mice. Data points over time consist of at least 6 observations each. Total number of mice assessed: n_WT = 31, n_TAVI = 23, n_KO = 15, n_T158M = 14, n_R106W = 28. (j) Kaplan-Meier survival curve for untagged (WT, KO) and tagged (TAVI, T158M, R106W) male mice (n_WT = 31, n_TAVI = 23, n_KO = 17, n_T158M = 39, n_R106W = 26). *P < 0.5, **P < 0.01, ***P < 0.001, ****P < 0.0001, n.s. = not significant; all pooled data depicts mean ± SEM unless otherwise stated. See also Supplementary Figs. 1 and 2.

Figure 2

Cell type-specific transcriptional profiling of neuronal nuclei. (a) Representative images showing immunofluorescent detection of biotinylated MeCP2-Tavi protein in Cre-specified neuronal populations of the mouse hippocampus. Probed using streptavidin for biotin detection and antibody against the MeCP2 C-terminus. Scale bars represent 100μm. (b) Schematic of cortical nuclei preparation and FACS isolation. (c) FACS analysis of labeled cortical nuclei populations. Data shown is representative of nine independent experiments using NEX-Cre mice. Percentages indicate the mean distribution of neurons that are NeuN+Biotin+ (excitatory; 85.2% ± 0.35) or NeuN+Biotin− (inhibitory; 14.8% ± 0.35). (d) RT-PCR validation of FACS-isolated populations depicted in (c) (n_replicates = 3, Two-way ANOVA). (e) Pearson correlation of biological replicate nuclear RNA-seq libraries within (intra-replicate) and across (inter-replicate) FACS-isolated populations depicted in (c). Colors correspond to EXC-enriched (blue) and INH-enriched (red) genes identified through differential expression analysis of excitatory and inhibitory neurons. Note lower Pearson correlation and clear dispersal of cell type-enriched genes across FACS populations. (f) IGV browser snapshot of Dlgap1 genomic locus in excitatory and inhibitory neurons of TAVI male mice at 6 weeks of age. RefSeq and Ensembl gene annotations are both shown. *P < 0.5, **P < 0.01, ***P < 0.001, ****P < 0.0001, n.s. = not significant; all pooled data depicts mean ± SEM. See also Supplementary Figs. 3 and 4.

Figure 3

Analysis of T158M and R106W differentially expressed genes. (a) FACS isolation of cortical excitatory and inhibitory neuronal nuclei from TAVI, T158M, or R106W male mice at 6 weeks of age. (b) Total number of protein coding and non-coding differentially expressed genes (DEGs) identified in excitatory or inhibitory neurons of Mecp2 mutant mice. (c) Heatmaps display log₂ fold changes among protein-coding DEGs in excitatory and inhibitory neurons of Mecp2 mutant mice, compared across genotypes. Excitatory DEGs n_shared = 69 genes, Hypergeometric P = 3.15e⁻⁷⁷. Inhibitory DEGs n_shared = 107 genes, Hypergeometric P = 5.33e⁻¹³⁴. Boxplots compare log₂ median fold changes among overlapping DEGs between T158M and R106W neurons (One-tailed Wilcoxon Signed Rank). (d) Heatmap displaying log₂ fold changes among protein-coding DEGs in excitatory and inhibitory neurons of Mecp2 mutant mice, compared across cell types. (e) Left graph, Distribution of constitutive, EXC- or INH-enriched genes among T158M and R106W protein-coding DEGs, compared against genomic distribution (Chi-square Goodness-of-Fit). Right graph, Bar plot summarizing R106W DEGs in excitatory neurons, partitioned by cell type-enriched or constitutive genes, and which are preferentially upregulated or downregulated. Red indicates statistical significance (One-tailed Fisher’s Exact Test). (f) Enrichment map of pre-ranked Gene Set Enrichment Analysis (GSEA) functional network associations. Data represents DEGs from R106W (top) and T158M (bottom) excitatory neurons (P-value < 0.01, Q-value < 0.1). Nodes denote functional categories, colored by Normalized Enrichment Score (NES). Line weight denotes extent of gene overlap between connected nodes. Red text highlights the similarity in functional annotations between both genotypes. (g) Boxplots comparing the log₂ FPKM distribution of actively expressed genes against T158M, R106W, and shared DEGs for each cell type (Pairwise Wilcoxon Rank Sum P displayed). *P < 0.5, **P < 0.01, ***P < 0.001, ****P < 0.0001, n.s. = not significant. See also Supplementary Figs. 5 and 6.

Figure 4

Genome-wide length-dependent transcriptional changes in RTT mutant mice. (a) Genome-wide log₂ fold changes in R106W mice (n = 2) compared to TAVI mice (n = 2) at 6 weeks of age using GRO-seq. Top, Lines represent mean fold change in expression for genes binned according to gene length (200 gene bins, 40 gene step) as described in. Ribbon represents SEM of genes in each bin. Bottom, Smoothed scatterplot depicting LOESS correlation between gene length and log₂ fold change for all individual protein-coding genes detected in GROseq. Genes in red highlight R106W DEGs identified from sorted excitatory and inhibitory neuronal nuclei. (b,c) Same as in (a), but using total RNA-seq analysis of whole cell (b) or nuclear (c) RNA isolated from left or right cortex of the same mice at 6 weeks of age (n=2). (d) Top, Diagram of RNA distribution across subcellular compartments. Bottom, Area proportional Venn diagram comparing overlap in gene expression changes between nuclear RNA, whole cell RNA, and nascent RNA. (e) Cumulative distribution function of gene lengths for all upregulated and downregulated protein-coding genes among nascent, nuclear, and whole cell RNA pools (n = 10,390 genes, Kolmogorov-Smirnov). (f) Top, Boxplots depicting median log₂ fold changes in R106W mice between nascent, nuclear, and whole cell RNA pools, classified by the direction of gene misregulation (n = 10, 390 genes, Pairwise Wilcoxon Rank Sum P displayed). Gene groups are arranged by median gene length (black bar on top). Arrows highlight the percentage of 10,390 genes that display similar (38.4% of expressed genes), opposite (48%), or dynamic changes (13.6%) across subcellular RNA pools. Bottom, Heatmap displaying statistical enrichment of T158M and R106W DEGs in excitatory neurons among gene groups (One-tailed Fisher’s Exact Test). (g) DAVID Gene ontology terms (Benjamini P < 0.01, FDR < 0.05) for Group A and Group G sets of genes defined in (f). (h) Top, Diagram of RT-PCR primer design to measure mature and primary RNA transcripts. Bottom, Data shows overall trend in gene expression mean fold changes using primers against primary and mature RNA transcripts (left) or primary transcripts only (right)_across individual genes from Group A/C (n = 7 genes), Group B (n = 5 genes), and Group D (n = 5 genes) in R106W compared to TAVI mice (Two-way ANOVA). Data depicts mean ± S.D. (i) Mean log₂ fold change in 6-week R106W (red; n = 4) and T158M (orange, n = 4) sorted excitatory (left) and inhibitory neurons (right) using genes that are also detected in GRO-seq. *P < 0.5, **P < 0.01, ***P < 0.001, ****P < 0.0001, n.s. = not significant. See also Supplementary Figs. 7–9.

Figure 5

T158M and R106W differentially expressed genes in mosaic female mice. (a) RTT-like phenotypic score in TAVI (n = 12), T158M (n = 4), and R106W (n = 9) heterozygous female mice (Two-way ANOVA). Data depicts mean ± SEM. (b) FACS isolation of excitatory neuronal nuclei from the cortex of heterozygous TAVI, T158M, or R106W female mice. (c) Biotin signal intensity from FACS-isolated populations depicted in (b) (n_T158M = 4, n_R106W = 9, Two-way ANOVA). Data depicts mean ± SEM. (d) X-inactivation ratios among cortical excitatory neurons in all sorted female mice, displayed as a percentage of the FACS-sorted WT population (n_TAVI = 12, n_T158M = 4, n_R106W = 9, One-way ANOVA). Data points in red indicate samples used for RNA-seq. Data depicts mean ± SEM. (e) Bar graph showing the cell and non-cell autonomous distribution of total protein-coding DEGs identified from T158M and R106W female mice. (f) Principal component analysis of WT and MUT cell populations isolated from TAVI, T158M, and R106W female mice. (g) Heatmap displaying log₂ fold changes among the total number of protein-coding DEGs detected in both WT and MUT populations from T158M or R106W female mice. Note genes that overlap across genotype (n = 194). (h) Proportion of cell autonomous (CA) and non-cell autonomous (NCA) genes that overlap between T158M and R106W female excitatory neurons (One-tailed Fisher’s Exact Test). (i) Boxplots comparing absolute log₂ fold change between cell autonomous and non-cell autonomous shared DEGs (n = 185) between T158M and R106W female mice (One-tailed Wilcoxon Signed Rank). (j) Enrichment map of pre-ranked GSEA functional network associations (P-value < 0.01, Q-value < 0.1). Data represents DEGs that overlap between T158M and R106W mice (n = 185). Nodes denote functional categories, colored by Normalized Enrichment Score (NES). Line weight denotes extent of gene overlap between connected nodes. *P < 0.5, **P < 0.01, ***P < 0.001, ****P < 0.0001, n.s. = not significant. See also Supplementary Fig. 10.