Evolution of DNA methylation in the human brain

Abstract

DNA methylation is a critical regulatory mechanism implicated in development, learning, memory, and disease in the human brain. Here we have elucidated DNA methylation changes during recent human brain evolution. We demonstrate dynamic evolutionary trajectories of DNA methylation in cell-type and cytosine-context specific manner. Specifically, DNA methylation in non-CG context, namely CH methylation, has increased (hypermethylation) in neuronal gene bodies during human brain evolution, contributing to human-specific down-regulation of genes and co-expression modules. The effects of CH hypermethylation is particularly pronounced in early development and neuronal subtypes. In contrast, DNA methylation in CG context shows pronounced reduction (hypomethylation) in human brains, notably in cis-regulatory regions, leading to upregulation of downstream genes. We show that the majority of differential CG methylation between neurons and oligodendrocytes originated before the divergence of hominoids and catarrhine monkeys, and harbors strong signal for genetic risk for schizophrenia. Remarkably, a substantial portion of differential CG methylation between neurons and oligodendrocytes emerged in the human lineage since the divergence from the chimpanzee lineage and carries significant genetic risk for schizophrenia. Therefore, recent epigenetic evolution of human cortex has shaped the cellular regulatory landscape and contributed to the increased vulnerability to neuropsychiatric diseases.

Fig. 1. CG and CH methylation in neurons (NeuN+ cells) and oligodendrocytes (OLIG2+ cells) in human and non-human primate prefrontal cortex.

a The proportions of methylated CG and CH sites. Human and non-human primate neurons and oligodendrocytes are highly CG methylated. Human and non-human primate neurons show low levels of CH methylation and oligodendrocytes show even lower levels. CH methylation is highest in human neurons, followed by chimpanzees, rhesus macaques, and mice. b Principal component analysis of methylated cytosines in two contexts (CG and CH). The top two principal components (PCs), PC1 and PC2, distinguish cell-type and species, respectively. c CG methylation levels in neurons (left columns for each species) and oligodendrocytes (right columns for each species). A greater number of DMRs are hypermethylated in neurons (red, in the left columns) compared to oligodendrocytes (right columns). d Approximately half (45.5%) of CG DMRs differentially methylated between NeuN+ and OLIG2+ cells are conserved in all three species, with 27% conserved between humans and chimpanzees, and 27.5% specific to the human. e The absolute methylation difference of NeuN+ and OLIG2+ cells is highest for DMRs conserved in all three species (39,202 and 17,284 DMRs hypermethylated in neurons and oligodendrocytes, respectively) compared to those specific to humans (3103 and 5361 DMRs hypermethylated in neurons and oligodendrocytes, respectively) or chimpanzees (4370 and 2989 DMRs hypermethylated in neurons and oligodendrocytes, respectively). DNA methylation differences between NeuN+ and OLIG2+ cells calculated from genomic regions serving as statistical control (CTRL), with a matched number of CG and G + C nucleotide contents, are also displayed. Statistical significance was computed using two-sided Mann–Whitney U-test. Box represents a range from the first quartile to the third quartile. The line in the box indicates the median value. The minima and maxima are within 1.5 times the distance between the first and third quartiles from box. Source data are provided as a Source Data file. Cell-type images were color modified from the original image, which was created by Akiyao and available at the Wikimedia Commons under the Creative Commons license (https://creativecommons.org/licenses/by-sa/3.0/deed.en).

Fig. 2. Evolutionary changes in CG methylation.

a Heatmap representation of mean DNA methylation of all 23,703 human DMRs in the three species illustrates dramatic reduction of CG methylation in human prefrontal cortex, especially in neurons. b Numbers of DMRs in NeuN+ and OLIG2+ cells in human and chimpanzee frontal cortex. c An example of the relationship between human neuron-hypo CG DMR and other epigenetic marks in the CLUL1 locus, a gene widely expressed in the brain. This DMR overlaps with multiple other epigenetic marks of active chromatin in the human brain, including neuron-specific ATAC-Seq peak, neuron-specific H3K4me3 peak, neuron-specific H3K27ac peak. This DMR also overlaps with a human-specific brain H3K4me3 peak compared to chimpanzee and macaque. Box represents a range from the first quartile to the third quartile. The line in the box indicates the median value. The minima and maxima are within 1.5 times the distance between the first and third quartiles from box. Source data are provided as a Source Data file. Cell-type images were color modified from the original image, which was created by Akiyao and available at the Wikimedia Commons under the Creative Commons license (https://creativecommons.org/licenses/by-sa/3.0/deed.en).

Fig. 3. CH hypermethylation is significantly higher in human neurons compared to other primates.

a Differences in the proportions of sites with neuronal CH methylation between species. b Mean methylation levels of human-specific CH DMRs demonstrate pronounced hypermethylation of human neurons. c CH methylation between humans and chimpanzees strongly predicts gene expression difference. The shaded band represents the 95% confidence interval for the fitted regression line. d Gene expression fold-change between human and macaque in CH DMR genes across developmental time points (Human CH DMR genes, n = 450 and Chimpanzee CH DMR genes, n = 144). Macaque samples were age-matched to human developmental time points in a previous study. Statistical significance was computed using Kruskal–Wallis test (two-sided). e Enrichment of human and chimpanzee CH DMR genes in specific cell-types. Human CH DMR genes are enriched in inhibitory neurons, whereas chimpanzee CH DMR genes are enriched in excitatory neurons. In each gene set, genes expressed in at least 50% of the cells that are statistically significant (FDR < 0.05 and log₂FC > 0.3) are included. Cell-type data are from human medial temporal gyrus (MTG). f CH methylation of neuronal subtypes for CH DMR genes using methylation of single nuclei from the human frontal cortex. Human CH DMR genes are hypomethylated in inhibitory neurons, whereas chimpanzee CH DMR genes are hypomethylated in excitatory neurons (excitatory neurons, n = 1879 and inhibitory neurons, n = 861). Statistical significance was computed using two-sided Mann–Whitney U-test. Box represents a range from the first quartile to the third quartile. The line in the box indicates the median value. The minima and maxima are within 1.5 times the distance between the first and third quartiles from box. Source data are provided as a Source Data file. Cell-type images were color modified from the original image, which was created by Akiyao and available at the Wikimedia Commons under the Creative Commons license (https://creativecommons.org/licenses/by-sa/3.0/deed.en).

Fig. 4. Evolutionary DMRs contribute to brain disease susceptibility.

a Significance levels for the enrichment for genetic heritability in different DMRs (+/-25kb) and complex traits. Both conserved and human-specific neuronal DMRs are associated with schizophrenia. Enrichment with FDR < 0.05 are highlighted in squares. Notably, CG DMRs hypomethylated in NeuN+ cells compared to OLIG2+ cells in all three species (conserved NeuN+ hypo) are highly enriched in variants for several brain-related traits, and human-specific NeuN+ hypo shows enrichment in schizophrenia. b A sliding-window analysis further demonstrates that the aforementioned signal for schizophrenia was centered at the DMRs and did not originate from extended adjacent regions. The y-axis represents the P-values in sliding windows around DMRs classified by species (human or chimpanzee), cell-type (NeuN+ or OLIG2+), and cytosine context (mCG or mCH). Source data are provided as a Source Data file.