Background
Major psychiatric disorders, including schizophrenia (SCZ) and bipolar disorder (BPD), are complex genetic mental illnesses that affect more than 1% of people worldwide [
1]. The complexity of these disorders arises from both their heterogenetic symptoms and multifactorial: genetic, developmental, and environmental nature. Epidemiological and genetic studies have indicated substantial overlap and high relative risks among relatives of both SCZ and BPD patients [
1]. Numerous genetic variants in noncoding regions implicated by genome-wide association studies (GWAS) are enriched in regulatory domains [
2], indicating that regulatory role for nonsequence-based genomic variation in mediating associations among genetic risk burden, environmental or epigenetic risk exposure, and phenotype. Recently, the identification of nonimprinted allele-specific DNA methylation (ASM) associated with genetic variation in
cis, where the genetic variant was associated with DNA methylation of a neighboring cytosine base on the same chromosome, suggested that mapping this type of allelic asymmetry may prove useful as a post-GWAS strategy for identifying functional genetic variants [
3‐
5]. Our recent study based on monozygotic (MZ) twins discordant for SCZ or BPD revealed thousands of single-nucleotide polymorphisms (SNPs) identified with ASM imbalances and phenotypic variation-associated switching at regulatory loci and that affected the interaction among one or more transcription factors (TFs), DNA methylation levels, and likely other epigenomic mark levels [
3]. Mapping of ASM or allele-specific chromatin states has also been reported to facilitate genome-wide screening for disease-linked regulatory SNPs [
4,
6,
7], which can be prioritized for functional studies. Since there are two alleles on homologous chromosomes in an individual in the same cellular and nuclear environment, there might be heterogeneity between allelic epigenomic patterns among individuals, providing important implications for interindividual differences in disease susceptibility.
Recently, 5-hydroxymethylcytosine (5hmC), which is catalyzed from 5mC by ten eleven translocation (TET) proteins, was discovered to be a relatively abundant form of cytosine modification in embryonic stem cells (ESCs) and Purkinje neurons [
8,
9]. This discovery raised the possibility that 5hmC mark may function as an intermediate during DNA demethylation; in addition, it may be a regulatory epigenetic mark that alters chromatin structure or contributes to the recruitment or exclusion of other DNA-binding proteins that affect transcription, and therefore, the dysregulation of this mark may cause certain diseases, such as SCZ or BPD [
10,
11]. Previously, sequence-dependent allelic imbalances in the epigenome, including imbalanced DNA methylation and histone marks or open chromatin at regulatory loci, have been reported to cause disease-associated switching, providing a powerful framework for identifying disease-associated variants and genes [
3,
4,
7]. Although several groups have investigated the genomic distribution of DNA hydroxymethylation marks [
12,
13], the role and functional importance of this modification and its links to disease-associated genetic variants in SCZ or BPD are still unclear.
We recently studied MZ twin epigenetic profiles and observed epigenetic variations, such as differences in DNA methylation [
3], lncRNA [
14], and miRNA [
15] levels, in phenotypes among concordant or discordant MZ twins. Examination of the epigenetic profiles of MZ twins, particularly disease-discordant MZ twins, is a powerful strategy to gain understanding of how genetic, environmental, and stochastic factors impact epigenetic modifications and how epigenetic variations affect the acquisition of complex traits [
3,
16]. MZ twins were exclusively matched for genotype, age, sex, paternal effects, population cohort effects, and exposure to several shared environmental factors. In summary, the same sources of cells from MZ cotwins were expected to exhibit identical genetic signatures, and alterations in allele-specific epigenetic modification in MZ cotwin pairs would indicate that certain epigenetic modifications may be particularly vulnerable to environmental influences or specific TFs, especially during embryonic development, leading to individual differences in phenotype and disease susceptibility [
3,
14]. The effects of genetic variation on the stochasticity or metastability of the DNA hydroxymethylome on the underlying heterogeneity of human disease have been unexplored.
In this study, we combined the 5hmC-selective chemical labeling method (5hmC sequencing [5hmC-seq]) [
17] and whole-genome sequencing (WGS) of peripheral blood DNA obtained from MZ twins discordant for SCZ or BPD to identify allelic imbalances in hydroxymethylome maps and found mechanistic effects of allele-specific hydroxymethylation (AShM) transitions (gain or loss) at regulatory loci on the phenotypic variation of psychiatric disorders in discordant MZ twins (PDC twins). Determining the molecular pathogenic basis of these differences will contribute to our understanding of the interaction between genetic and epigenetic factors in mediating individual differences in disease susceptibility and provide a powerful strategy for prioritizing SCZ/BPD-associated genes or variants.
Discussion
Here, we performed a genome-wide examination of AShM sites from MZ twins and identified a large number of SNPs showing AShM transitions across MZ twins with discordant phenotypes. These psyAShM sites displayed epiallele-specific effects on chromatin states and TF binding, and their AShM transitions may have led to dysregulated gene expression and functions and may have eventually increased the risk of SCZ or BPD. We then employed multiple lines of data to show that competitive binding of POU3F2 on the alternative T allele at the psyAShM site rs4558409 (G/T) can enhance PLLP expression, while the hydroxymethylated alternative allele alleviating the POU3F2 binding activity at the rs4558409 site might be associated with the downregulated PLLP expression observed in patients with BPD/SCZ. This study has etiological implications for the AShM transition in patients with SCZ and BPD.
Our findings showed that sensitivity of the hydroxymethylome to genetic variation among MZ twin pairs with discordant phenotypes displayed disease-associated features through epiallele-specific effects on chromatin states, TF binding, and gene expression. Our allelic epigenome profiling of 17 MZ twin pairs revealed that 1.2% of informative SNPs showed AShM (FDR<0.1), which was a more conservative outcome than estimates for ASM (2.2%) in the same twin pairs analyses in our previous study [
3]. Some of our identified AShM sites (11.7%) displayed sequence-dependent ASM, as shown in our previous study [
3], and were located in previously reported imprinted regions. The regulatory role of AShM sites was shown by the evidence of their uneven distribution across the genome with 90% of the psyAShM sites located in noncoding regions, greater enrichment of pysAShM sites in the active chromatin state, and AShM transitions that may induce changes in chromatin state and TF binding ability, eventually inducing altered gene expression. The regulatory roles of psyAShM were further demonstrated by their significant enrichment in brain eQTLs/mQTLs, SCZ/BPD-associated DEGs, and TWAS or GWAS of SCZ/BPD. In fact, we recently reported that ASM sites are more enriched in repressed chromatin states and showed epiallelic effects on TF binding and gene expression [
3]. DNA methylation has long been associated with gene silencing, inferring that DNA demethylation can lead to gene activation. As an intermediate of DNA demethylation processing, 5hmC has also been reported to be highly enriched in the gene bodies of transcriptionally active genes, promoters, and enhancers and to undergo highly dynamic changes during development and differentiation and in the context of neuropsychiatric disorders [
10,
12]. The functional implication of sequence-dependent ASM and AShM sites examined in our study, together with mostly functional genetic variants identified in noncoding regions via GWAS [
2], provided further evidence for dynamic DNA methylation/demethylation in mediating associations among genetic risk burden, environmental or epigenetic risk exposure, and phenotype.
Our study validated the biological role of AShM transitions at the
PLLP rs4558409 locus, leading to alterations in POU3F2 binding ability and contributing to the downregulation of
PLLP, providing etiological implications for SCZ or BPD.
PLLP encodes the proteolipid plasmolipin, which is a main component of synaptic plasma membranes and myelin sheaths and is involved in intracellular transport and neurite growth [
49]. AShM transitions at the
PLLP rs4558409 locus across PDC twin pairs reduced
PLLP transcriptional activity through hydroxymethylation-mediated inhibition of active POU3F2 binding affinity at the active T allele of the rs4558409 (G/T) regulatory locus. The allele-specific regulatory role of rs4558409 in a luciferase reporter assay was consistent with the eQTL pattern. The alternative T allele of rs4558409 showing higher binding activity was demonstrated via ChIP and EMSAs, and its hyper-hydroxymethylation/methylation, which reduced binding ability, might significantly contribute to the reduced
PLLP expression observed in postmortem SCZ brain[
47], and rs4558409 disruption promotes neural development and vesicle trafficking. Similarly, reduced
PLLP has been observed in the temporal cortex of patients with MDD [
52]. In fact, rs4558409 is a subthreshold GWAS SNP associated with SCZ (
p=0.007479 in PGC3) [
36] and MDD (
p=0.026) [
40], and hyper-hydroxymethylation at the risk alternative T allele of rs4558409 (G/T), which reduced TF binding affinity and
PLLP expression, might increase disease susceptibility. PLLP plays an important role in membrane biogenesis and myelination [
49], and distorted oligodendrocyte differentiation and subsequent defects in myelination might lead to SCZ and BPD onset [
49,
53]. Intriguingly,
POU3F2 is one of the key regulators coexpressed with many genes within the SCZ-related module [
54], and
POU3F2-regulated target genes may contribute to neurodevelopment and synaptic function in various ways [
55]. In addition, POU3F2-deficient mice exhibited impaired hippocampal neurogenesis [
56].
POU3F2 expression is decreased in cerebral organoids in patients with BPD [
57], and its expression is associated with differential SCZ-associated alterations in brain tissues, including downregulation in the SZDB dataset [
58] but upregulation in the PsychENCODE dataset [
47]. In summary, involvement of AShM transitions at rs4558409 regulatory loci might contribute to disease-associated
PLLP downregulation through epiallele-specific inhibition of POU3F2 binding.
This study also has several limitations. First, although our study identified that disruption of rs4558409-contained region promotes neural development and vesicle trafficking, whether rs4558409 (G/T) display allelic effects on neural development and vesicle trafficking remains to be further examined through single-base mutation strategy in SK-N-SH cells. Second, whether KO of PLLP could restore neural development induced by rs4558409-KO may be needed to validate the role of rs4558409 in regulation of PLLP. Third, whether ERBB4, NRG1, or GAD1 except for PLLP are regulated by rs4854158 may provide further mechanism of rs4854158 on regulation of neural development since expression of those genes are significantly dysregulated in rs4558409-KO cells. Finally, despite the human neuroblastoma SK-N-SH cell lines employed in this study are the most widely applied and cited in vitro system for neurodevelopmental and neuropsychiatric studies, further studies in other appropriate test models to validate the function of rs4854158 in pathogenesis of the BPD or SCZ are needed.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.