Cell-type-specific DNA methylation dynamics in the prenatal and postnatal human cortex

Abstract

The human cortex undergoes extensive epigenetic remodeling during development, although the precise temporal and cell-type-specific dynamics of DNA methylation remain incompletely understood. In this study, we profiled genome-wide DNA methylation across human cortex tissue from donors aged 6 post-conception weeks to 108 years of age. We observed widespread, developmentally regulated changes in DNA methylation, with pronounced shifts occurring during early- and mid-gestation that were distinct from age-associated modifications in the postnatal cortex. Using fluorescence-activated nuclei sorting, we optimized a protocol for the isolation of SATB2-positive neuronal nuclei, enabling the identification of cell-type-specific DNA methylation trajectories in the developing cortex. Developmentally dynamic DNA methylation sites were significantly enriched near genes implicated in autism and schizophrenia, supporting a role for epigenetic dysregulation in neurodevelopmental conditions. Our findings underscore the prenatal period as a critical window of epigenomic plasticity in the brain with important implications for understanding the genetic basis of neurodevelopmental phenotypes.

Keywords: DNA methylation; aging; autism; brain; cortex; development; epigenetics; fetal; neuronal; schizophrenia.

Graphical abstract

Figure 1

Development-associated changes in DNA methylation in the human fetal cortex (A and B) The top-ranked dDMPs demonstrating (A) hypermethylation and (B) hypomethylation over fetal cortex development. cg08125539, annotated to IGF2BP1 on chromosome 17, is characterized by a significant increase in DNA methylation across prenatal cortex development (% increase in DNA methylation = 4.5% per week, p = 1.16 × 10⁻⁴⁰). cg11884704, annotated to SLC25A25 on chromosome 9, is characterized by a significant decrease in DNA methylation across prenatal cortex development (% decrease in DNA methylation = 2.7% per week, p = 8.63 × 10⁻⁴⁰). Of note, the dramatic changes in DNA methylation at these sites are specific to the prenatal period (see also Figure S11). Shaded region indicates 95% confidence interval. (C) dDMPs cluster into differentially methylated regions (DMRs) associated with cortex development (see also Table S3). The PAX6 gene on chromosome 11 contains three independent DMRs (yellow shaded regions), including the top-ranked DMR (mean change in DNA methylation % per week across region = 6.50, p < 1 × 10⁻³²⁰) overlapping an intragenic CpG island (green). Blue, dDMP with a positive effect size; red, dDMP with a negative effect size; black, non-significant site. Dotted line indicates genome-wide significance. (D) Relative enrichment (red) and depletion (blue) of dDMPs across different CGI and genic features. Compared to the frequency of dDMPs among all sites profiled in this study (white bar), there was a significant enrichment of dDMPs in CGI shores and gene bodies but a significant depletion in other regions, most notably CGIs (see also Table S7). (E) Proportion of hypermethylated (blue) and hypomethylated (red) dDMPs across CGI and genic features. Despite the overall enrichment of hypomethylated dDMPs across all sites tested, there are substantial differences across genomic features with CGIs and CGI shores characterized by an enrichment of hypermethylated dDMPs (see also Table S8). (F) Enrichment of cortex dDMPs in cell-type-specific regions of open chromatin identified by scATAC-seq. Shown is a volcano plot of the relative effect size versus p value from a logistic regression testing for an enrichment of dDMPs within cell-type-specific scATAC-seq peaks for 54 human fetal cell types. Peaks for 31 fetal cell types were significantly enriched for dDMPs, highlighting a general overlap with transcriptionally active regions of the genome during development (red, significant enrichment; blue, significant depletion; gray, non-significant). The greatest enrichment was observed within ATAC-seq peaks specific to fetal excitatory neurons (effect size = 1.27, odds ratio = 3.56, corrected p = 1.17 × 10⁻¹³⁹) and astrocytes (effect size = 0.75, odds ratio = 2.12, corrected p = 6.74 × 10⁻⁸⁰).

Figure 2

Nonlinear trajectories of DNA methylation in the developing cortex Nonlinear DNA methylation sites identified through Gaussian process regression modeling were clustered into distinct modules using WGCNA (STAR Methods). For each module, the eigengene (first PC) is plotted against fetal age (pcw). The number of nonlinear sites assigned to each module is indicated in the subtitles, with hub DNA methylation sites for each module shown in Figure S29.

Figure 3

The majority of dDMPs are not characterized by age-associated changes in DNA methylation in the postnatal cortex (A) DNA methylation at dDMPs was dramatically more variable in fetal cortex samples than adult cortex samples. Shown is the distribution of variance in DNA methylation across the 10,000 most significant dDMPs in fetal samples (pink) (mean variance = 120, mean SD = 11% DNA methylation) compared to the variance in DNA methylation at the same sites in postnatal samples (blue) (mean variance = 41.7, mean SD = 6.46% DNA methylation). (B–F) (B) Comparison of age effect sizes between prenatal and postnatal cortex for the 41,518 dDMPs also measured in the postnatal samples. Each dot represents a cortex dDMP with significance for age in the postnatal cortex indicated by color (gray, not significantly associated with age in postnatal cortex). Examples of dDMPs showing (C) a consistent age-associated increase in DNA methylation in both prenatal and postnatal cortex, (D) a consistent age-associated decrease in DNA methylation in both fetal and postnatal cortex, (E) a developmentally associated decrease in DNA methylation in the prenatal cortex followed by an age-associated increase in DNA methylation in the postnatal cortex, and (F) a developmentally associated increase in DNA methylation in the prenatal cortex followed by an age-associated decrease in DNA methylation in the postnatal cortex. (G) Comparisons of mean DNA methylation at the 10,000 top-ranked dDMPs between the earliest fetal samples (n = 21, age range = 6–9 pcw), the eldest mid-fetal samples (n = 12, age range = 20–23 pcw), and adult samples (n = 661, age range = 25–104 years). Mean DNA methylation of the mid-fetal samples is most strongly correlated to postnatal samples (corr = 0.65) (right) than to early-fetal samples (corr = 0.36) (left). Color bar indicates number of DNA methylation sites per bin.

Figure 4

Cell-type-specific changes in DNA methylation across cortex development and postnatal life (A) Principal component analysis (PCA) on the top 10,000 most variable autosomal DNA methylation sites across FANS-isolated nuclei populations from prenatal and postnatal cortex samples (n = 259 donors). The top four principal components (PCs) correlated against age, cell type, and sex (∗p < 0.05, ∗∗∗p < 0.001). (B) Heatmap of DNA methylation in purified nuclei populations isolated from adult cortex at the 6,531 DMPs between SATB2+ and SATB2− nuclei identified in fetal cortex. (C) PCA on the top 10,000 most variable autosomal sites in FANS-isolated nuclei populations from adult cortex (n = 212 donors). Using the adult PCA loadings, fetal samples were projected onto the same PCA space, showing a clear but attenuated difference between cell types in fetal cortex. Colors indicate the antibody labels used to isolate cortical nuclei. Colors are grouped into neuron-enriched (purple shades) and neuron-depleted (green shades) cell types. (D) cg16424082, annotated to METTL7A on chromosome 12, is characterized by developmental-stage-specific differences in DNA methylation in neuron-enriched but not neuron-depleted nuclei populations. (E) DNA methylation at cg04831806 annotated to SDK1 on chromosome 7, a gene with a key role in neurodevelopment, is characterized by a highly significant interaction between cell-type and developmental age (SATB2+: % change in DNA methylation per week = −4.25, p = 6.76 × 10⁻¹³). (F) Enrichment of SATB2+ dDMPs in cell-type-specific regions of open chromatin in the developing cortex identified by scATAC-seq. SATB2+-specific autosomal dDMPs (n = 1,596) are most strongly enriched in peaks specific to excitatory neurons. (G) Enrichment of SATB2− dDMPs in cell-type-specific regions of open chromatin in the developing cortex identified by scATAC-seq. SATB2−-specific autosomal dDMPs (n = 548) are most strongly enriched in peaks specific to astrocytes. Color indicates direction of effect: red, significant enrichment; blue, significant depletion; gray, non-significant.