Key PointsQuestion
What is the genetic architecture of bipolar disorder (BD) in the Han Chinese population?
Findings
In this genome-wide association study of 6472 individuals of Han Chinese ancestry (1822 cases and 4650 controls), several novel risk loci for BD were found, and trans-ancestry genetic correlation estimation and polygenic risk score analyses of Han Chinese and European individuals suggested a shared genetic risk of BD.
Meaning
Findings of this study highlighted novel genome-wide significant risk loci for BD that can provide insight into the genetic architecture of this disorder.
Importance
The genetic basis of bipolar disorder (BD) in Han Chinese individuals is not fully understood.
Objective
To explore the genetic basis of BD in the Han Chinese population.
Design, Setting, and Participants
A genome-wide association study (GWAS), followed by independent replication, was conducted to identify BD risk loci in Han Chinese individuals. Individuals with BD were diagnosed based on DSM-IV criteria and had no history of schizophrenia, mental retardation, or substance dependence; individuals without any personal or family history of mental illnesses, including BD, were included as control participants. In total, discovery samples from 1822 patients and 4650 control participants passed quality control for the GWAS analysis. Replication analyses of samples from 958 patients and 2050 control participants were conducted. Summary statistics from the European Psychiatric Genomics Consortium 2 (PGC2) BD GWAS (20 352 cases and 31 358 controls) were used for the trans-ancestry genetic correlation analysis, polygenetic risk score analysis, and meta-analysis to compare BD genetic risk between Han Chinese and European individuals. The study was performed in February 2020.
Main Outcomes and Measures
Single-nucleotide variations with P < 5.00 × 10−8 were considered to show genome-wide significance of statistical association.
Results
The Han Chinese discovery GWAS sample included 1822 cases (mean [SD] age, 35.43 [14.12] years; 838 [46%] male) and 4650 controls (mean [SD] age, 27.48 [5.97] years; 2465 [53%] male), and the replication sample included 958 cases (mean [SD] age, 37.82 [15.54] years; 412 [43%] male) and 2050 controls (mean [SD] age, 27.50 [6.00] years; 1189 [58%] male). A novel BD risk locus in Han Chinese individuals was found near the gene encoding transmembrane protein 108 (TMEM108, rs9863544; P = 2.49 × 10−8; odds ratio [OR], 0.650; 95% CI, 0.559-0.756), which is required for dendritic spine development and glutamatergic transmission in the dentate gyrus. Trans-ancestry genetic correlation estimation (ρge = 0.652, SE = 0.106; P = 7.30 × 10−10) and polygenetic risk score analyses (maximum liability-scaled Nagelkerke pseudo R2 = 1.27%; P = 1.30 × 10−19) showed evidence of shared BD genetic risk between Han Chinese and European populations, and meta-analysis identified 2 new GWAS risk loci near VRK2 (rs41335055; P = 4.98 × 10−9; OR, 0.849; 95% CI, 0.804-0.897) and RHEBL1 (rs7969091; P = 3.12 × 10−8; OR, 0.932; 95% CI, 0.909-0.956).
Conclusions and Relevance
This GWAS study identified several loci and genes involved in the heritable risk of BD, providing insights into its genetic architecture and biological basis.
Bipolar disorder (BD) is a severe psychiatric illness characterized by recurrent episodes of mania or hypomania and depression.1,2 The World Health Organization estimated that the lifetime prevalence is 0.6% for bipolar I disorder (BD-I) and 0.4% for bipolar II disorder (BD-II) in 11 countries across the Americas, Europe, and Asia.3 According to earlier studies, the lifetime prevalence of BD is approximately 5% to 10% in first-degree relatives of patients and approximately 40% to 70% in monozygotic co-twins.4 Therefore, heritable factors likely contribute to this disorder, and genetic analyses could help to disentangle its mechanisms and to facilitate the discovery of therapeutic targets.5,6 A recent genome-wide association study (GWAS) estimated that approximately 23% of BD heritability was attributed to common single-nucleotide variations (SNVs) and provided implications for its pathology.7
Although GWASs have increased our knowledge of BD in Europeans, genetic heterogeneity between continental populations exists and may result in uncertainty when generalizing these discoveries across different populations. For example, a recent Japanese BD GWAS8 (including 2964 cases and 61 887 controls) reported on genome-wide risk loci that are either shared among distinct populations or are specific to East Asian individuals.9,10 Because most of the BD GWASs to date have been performed in European populations, further analyses of the genetic architecture of BD in other populations are needed. A previous Han Chinese BD GWAS11 (including 1000 cases and 1000 controls) identified no statistically significant loci, probably because of the limited sample size. Therefore, we conducted a BD GWAS in a larger independent sample (1822 cases and 4650 controls) of Han Chinese ancestry, followed by replication in additional Han Chinese individuals (958 cases and 2050 controls), as well as a trans-ancestry meta-analysis combining these results with summary statistics from the European Psychiatric Genomics Consortium 2 (PGC2) BD GWAS7 (20 352 cases and 31 358 controls).
The study protocol for this GWAS and trans-ancestry meta-analysis was approved by the institutional review board of the Kunming Institute of Zoology, Chinese Academy of Sciences, as well as ethics committees of all participating hospitals and universities (provided in the eMethods in the Supplement). All participants provided written informed consent before any study-related procedures were performed. This study followed the Strengthening the Reporting of Genetic Association Studies (STREGA) reporting guideline. The study was performed in February 2020.
In total, 6472 Han Chinese individuals (1822 BD cases and 4650 controls) were recruited in mainland China for the discovery GWAS. A unique sample of 958 patients with BD and 2050 control participants of Han Chinese ancestry in mainland China were included for the replication analysis. In both the discovery GWAS and replication stages, patients with BD were diagnosed through the use of an extensive clinical interview and the Structured Clinical Interview for DSM-IV Axis I Disorders–Patient Version. The control participants had no BD and no history of any mental illness. Detailed descriptions of the sample are provided in the eMethods in the Supplement.
Genotyping in the discovery stage was performed with either the Illumina Infinium Global Screening Array (GSA) chip or the Illumina Genome-Wide Asian Screening Array (ASA) chip (Beijing Guoke Biotechnology Co, Ltd). Quality control (QC) analyses were performed using the pipeline suggested by Anderson et al.12 After QC, the autosomal biallelic SNVs on different GWAS platforms underwent genotype imputation using the prephasing imputation stepwise approach in SHAPEIT and IMPUTE2 software programs,13,14 and the imputation reference set was obtained from phase 3 of the 1000 Genomes Project.15
In each GWAS cohort, logistic regression of BD diagnosis on imputed hard-called genotypes (with posterior probability >.95) was performed,16 during which the associations of the top 20 principal components with BD diagnosis were evaluated, and principal components associated with diagnostic status (P ≤ .05) were included as covariates to control for population stratification.17 The statistics in each GWAS cohort were then combined for an inverse variance–weighted meta-analysis using random-effects or fixed-effects models (referred to as the discovery GWAS). Linkage disequilibrium score regression (LDSC) was applied to assess potential population stratification and to estimate SNV heritability.18,19 Single-nucleotide variations with 2-sided P < 5.00 × 10−8 were considered to show genome-wide significance.
Two expression quantitative trait loci (eQTL) data sets (CommonMind Consortium20 and BrainSeq Phase 221) of the dorsolateral prefrontal cortex (DLPFC) were obtained for the summary data–based mendelian randomization22 and transcriptome-wide association (TWAS)23 analyses. The sample sizes of the RNA sequencing–based eQTL data sets were 467 and 397, respectively.
Given the shared clinical manifestations between different psychiatric disorders and traits,24,25 we examined the genetic correlations of BD with other psychiatric disorders (eg, schizophrenia26,27 and depression28,29) and traits (cognitive performance,30 intelligence,31 and educational attainment30) using LDSC (for analyses within the same population)18,19 or Popcorn, version 1.0 (Brielin C. Brown [https://github.com/brielin/Popcorn]) (for trans-ancestry analyses)32 based on the GWAS summary statistics. The proportion of BD variance explained by risk SNVs identified in GWASs of those phenotypes was also estimated using polygenic risk scores (PRSs).8,33 Fifteen pairs of PRS analyses were conducted in our study; hence, P < .0033 was considered statistically significant after multiple correction (approximately 0.05 ÷ 15). Details of these GWAS data sets are provided in the eMethods in the Supplement.
We examined the messenger RNA (mRNA) expression patterns of the risk genes identified by GWAS in human tissues using GTEx and BrainSpan data sets.34,35 We also used hypergeometric testing in the web-based platform FUMA36 to examine the tissue expression enrichment of the GWAS risk loci.
GWAS of BD in the Han Chinese Population
We conducted a meta-analysis of 2 BD GWAS Han Chinese cohorts, including 1822 cases (mean [SD] age, 35.43 [14.12] years; 838 [46%] male and 984 [54%] female) and 4650 controls (mean [SD] age, 27.48 [5.97] years; 2465 [53%] male and 2185 [47%] female) (referred to as the discovery GWAS). After systematic QC analysis and imputation using phase 3 of the 1000 Genomes Project,15 we assessed the associations of 4 499 546 autosomal biallelic SNVs with imputation quality score (INFO) greater than 0.8, minor allele frequency greater than 1%, call rate greater than 95%, and Hardy-Weinberg equilibrium P > 1.00 × 10−5. Population substructures of these samples were examined through a principal components analysis (eFigure 1 in the Supplement). The genomic inflation λ of the discovery GWAS was 1.038, and the λ1000 (a scaled value to 1000 cases and 1000 controls) was 1.015. We then conducted LDSC analysis to estimate BD polygenicity in these samples based on precomputed linkage disequilibrium (LD) scores in HapMap3 for East Asian individuals.18,19 The mean (SE) LDSC intercept was 1.005 (0.008), and the mean (SE) attenuation ratio was 0.077 (0.132), confirming polygenicity of BD in the discovery GWAS and suggesting that only approximately 8% of the observed genomic inflation in test statistics was attributed to population stratification.18,19 The LDSC estimated that the mean (SE) SNV heritability in the discovery GWAS was 0.220 (0.043) to approximately 0.310 (0.059) on the liability scale, assuming that the population prevalence of BD was 0.5% to approximately 2%.18,19
Manhattan and quantile-quantile plots for the Han Chinese discovery GWAS are shown in Figure 1A and in eFigure 2 in the Supplement, respectively. The discovery GWAS (1822 cases and 4650 controls) identified a single locus reaching genome-wide significance, which is located at 3q22.1 in the 5′ upstream region of TMEM108 (OMIM 617361) and the 3′ downstream region of the noncoding RNA NPHP3-AS1 (Gene ID 348808) (rs9863544; P = 5.00 × 10−8; odds ratio [OR], 0.590; 95% CI, 0.488-0.713) (Figure 2A).15,37 In addition, the discovery GWAS identified 22 SNVs with P values lower than the threshold of suggestive significance (ie, P = 5.00 × 10−6) (eTable 1 in the Supplement). These SNVs appeared to represent 4 physically distinct regions after LD pruning at r2 = 0.1 (within 500 kilobase [kb]). To replicate these results, we tested the top 4 SNVs from these distinct regions in an independent sample of Han Chinese individuals, including 958 cases (mean [SD] age, 37.82 [15.54] years; 412 [43%] male and 546 [57%] female) and 2050 controls (mean [SD] age, 27.50 [6.00] years; 1189 [58%] male and 861 [42%] female). We confirmed that rs9863544 also showed nominal significance (defined as P < .05) (P = .04; OR, 0.771; 95% CI, 0.600-0.991) (Table 1). Detailed results obtained in the replication samples are provided in eResults 1, eFigure 3, and eTable 2 in the Supplement.
Meta-analysis of the discovery GWAS and replication samples in Han Chinese individuals demonstrated that rs9863544 had genome-wide significance (P = 2.49 × 10−8; OR, 0.650; 95% CI, 0.559-0.756) (Table 1). We also explored the mRNA expression patterns of the 2 genes (TMEM108 and NPHP3-AS1) near rs9863544 in public RNA sequencing resources. In the GTEx data set,34 the mRNA of NPHP3-AS1 was barely detectable in most human organs, including the brain, whereas TMEM108 was widely expressed in the human brain (eFigure 4 in the Supplement). Further analyses of their temporal expression patterns in human brain in the BrainSpan data set35 revealed statistically significantly higher levels of TMEM108 mRNA in prenatal stages, which declined after birth; the mRNA expression of NPHP3-AS1 remained low in human brain regardless of the developmental stage (eFigure 5 in the Supplement). We also examined whether the genomic loci reaching the threshold of suggestive significance (P ≤ 5.00 × 10−6) in previous East Asian BD GWASs8,11 were statistically significant in our Han Chinese sample. We found that rs7221716 (P = 5.60 × 10−7; OR, 1.170; 95% CI, 1.100-1.244 in the prior Japanese BD GWAS8) near the PFAS (OMIM 602133) gene was nominally significant in our Han Chinese discovery GWAS (P = .01; OR, 1.115; 95% CI, 1.023-1.216) and had genome-wide significance in a meta-analysis combining the Han Chinese discovery GWAS and the previous Japanese GWAS (P = 2.02 × 10−8; OR, 1.152; 95% CI, 1.096-1.210) (detailed results are provided in eResults 2 and eTable 3 in the Supplement).
Trans-Ancestry Genetic Correlation and Meta-analysis of BD in Han Chinese and European Populations
The association statistics of SNVs from the Han Chinese discovery GWAS and the European PGC2 BD GWAS,7 as well as precomputed LD scores for European and East Asian individuals in the 1000 Genomes Project,15 were obtained to estimate the trans-ancestry genetic correlations of BD between Han Chinese and European individuals using Popcorn, version 1.0. That analysis revealed a statistically significant trans-ancestry genetic effect correlation between the Han Chinese discovery GWAS and the European PGC2 BD GWAS (mean [SE] ρ for genetic effect [ρge] = 0.652 [0.106]; P = 7.30 × 10−10), as well as a population genetic impact correlation accounting for the different SNV allele frequencies between populations (mean [SE] ρ for genetic impact [ρgi] = 0.651 [0.111]; P = 4.50 × 10−9).
We then conducted a trans-ancestry meta-analysis of our discovery GWAS and the European PGC2 BD GWAS. A total of 3 742 365 autosomal biallelic SNVs with INFO greater than 0.8 and minor allele frequency greater than 1% in both Han Chinese and European individuals were included in the trans-ancestry meta-analysis. Of these SNVs, 46 441 SNVs (approximately 1.2% of the total SNVs) showed pronounced heterogeneity (I2 >75%) and were thus meta-analyzed using a random-effects model; the other 3 695 924 SNVs were meta-analyzed using a fixed-effects model given their nonsignificant heterogeneity (I2 ≤75%). The genomic inflation λ of the trans-ancestry meta-analysis was 1.355, and the λ1000 was 1.013. The mean (SE) LDSC intercept (based on precomputed LD scores for European populations) was 1.023 (0.011), and the mean (SE) attenuation ratio was 0.054 (0.025), indicating polygenicity rather than population stratification.18,19 The mean (SE) LDSC SNV heritability estimate for BD was 0.160 (0.008) to approximately 0.220 (0.011) on the liability scale, assuming that the population prevalence of BD was 0.5% to approximately 2%.18,19
Manhattan and quantile-quantile plots for the trans-ancestry meta-analysis are shown in Figure 1B and eFigure 6 in the Supplement, respectively. A total of 191 SNVs reached the genome-wide significance threshold (P ≤ 5.00 × 10−8) (eTable 4 in the Supplement). We then combined the SNVs with r2 <0.1 within 500 kb based on European LD panels and noted that they mapped to 23 physically distinct genomic regions (Figure 1B). The top SNVs in each of these GWAS loci are listed in Table 2. Further detailed characterization of these 23 GWAS loci suggested that 21 of them had genome-wide significance in either the GWAS stage or the GWAS plus replication stages of the European PGC2 BD GWAS. The trans-ancestry meta-analysis herein identified 2 novel loci (VRK2 [OMIM 602169] and RHEBL1 [OMIM 618956]) that were not genome-wide significant in the European PGC2 BD GWAS. Specifically, the European PGC2 BD GWAS SNVs reaching the threshold of suggestive significance in the 5′ upstream region of the VRK2 gene showed nominal significance in our Han Chinese discovery GWAS and showed genome-wide significance in the trans-ancestry meta-analysis (eg, rs41335055; P = 9.85 × 10−8; OR, 0.854; 95% CI, 0.806-0.905 in the European PGC2 BD GWAS; P = .01; OR, 0.808; 95% CI, 0.683-0.956 in the Han Chinese discovery GWAS; and P = 4.98 × 10−9; OR, 0.849; 95% CI, 0.804-0.897 in the trans-ancestry meta-analysis (Figure 2B). Similarly, the European PGC2 BD GWAS SNVs reaching the threshold of statistical significance in the RHEBL1 gene showed nominal significance in our Han Chinese discovery GWAS and showed genome-wide significance in the trans-ancestry meta-analysis (eg, rs7969091; P = 3.25 × 10−7; OR, 0.933; 95% CI, 0.909-0.958 in the European PGC2 BD GWAS; P = .03; OR, 0.918; 95% CI, 0.848-0.993 in the Han Chinese discovery GWAS; and P = 3.12 × 10−8; OR, 0.932; 95% CI, 0.909-0.956 in the trans-ancestry meta-analysis) (Figure 2C). Herein, we refer to the novel risk loci by the names of their closest genes, without suggesting that a causal association between these genes and BD; the previously implicated loci are still referred to by the European PGC2 BD GWAS names.
In addition, we examined the 30 GWAS loci identified in the European PGC2 BD GWAS in our trans-ancestry meta-analysis (eTable 5 in the Supplement). We found that 18 of them had genome-wide significance, including the previously known loci at CACNA1C [OMIM 114205], TRANK1 [Gene ID 9881], ITIH1 [OMIM 147270], ANK3 [OMIM 600465], NCAN [Gene ID 1463], SCN2A [OMIM 182390], and POU3F2 [OMIM 600494]. The top SNVs or the high LD SNVs in another 8 loci (PLEKHO1, ADCY2, RPS6KA2, SRPK2, MRPS33, FADS2, SHANK2, and STARD9) identified in the European PGC2 BD GWAS were not genotyped or imputed in our discovery GWAS sample, so these loci were not included in the trans-ancestry meta-analysis. The other 4 loci (LMAN2L, FSTL5, THSD7A, and PC) from the European PGC2 BD GWAS were not statistically significant in the trans-ancestry meta-analysis because their allelic effect directions in the Han Chinese discovery GWAS were the opposite of those in the European PGC2 BD GWAS.
Tissue Expression Enrichment, Biological Processes, and In Silico Functional Analyses
To prioritize potential BD risk genes, we integrated the GWAS summary statistics of the trans-ancestry meta-analysis with the DLPFC eQTL data from both the CommonMind Consortium20 and the BrainSeq Phase 221 data sets through summary data–based mendelian randomization22 and TWAS23 analyses. Summary data–based mendelian randomization identified a single gene (NEK4 [OMIM 601959]) that had a statistically significant association with BD after multiple testing correction (P ≤ 1.00 × 10−5) in both DLPFC eQTL data sets, without evidence of heterogeneity between GWAS and eQTL association signals (eTable 6 in the Supplement). Transcriptome-wide association identified 3 genes (NEK4 [OMIM 601959], GLT8D1 [OMIM 618399], and MCM3AP [OMIM 603294]) that had statistically significant associations with BD after multiple correction in both DLPFC eQTL data sets (P ≤ 2.50 × 10−5) (eTable 7 in the Supplement).
Hypergeometric testing using the web-based platform FUMA36 was performed to examine tissue expression enrichment (in 54 subdivided types of tissues in the GTEx data set34) of the risk loci in our trans-ancestry meta-analysis. Although the cerebellum had the strongest enrichment of these genes (P = 1.78 × 10−12; false discovery rate [FDR], 5.31 × 10−11) (eFigure 7 in the Supplement), they were also statistically significantly enriched in multiple other brain tissues, such as the frontal cortex, anterior cingulate cortex, nucleus accumbens, hippocampus, amygdala, and caudate (FDR, ≤1.00 × 10−5). We then performed an enrichment analysis using Multimarker Analysis of Genomic Annotation (MAGMA)38 to examine biological processes and pathways underlying BD genetic risk identified in the trans-ancestry meta-analysis. One pathway (regulation of insulin secretion) was statistically significantly enriched for genes with BD associations after multiple correction (P = 4.83 × 10−6; FDR, 0.035) (eTable 8 in the Supplement).
PRS Analysis of BD Across Han Chinese and European Populations
We analyzed the polygenic architecture of BD by performing PRS analysis. The GAS GWAS sample was first used as the training data set to examine whether BD cases had a higher PRS than controls in the ASA GWAS sample. This procedure was then repeated with the training and target data sets swapped. Both the training and target data sets could be used to predict the risk of BD, and the maximum measures of the explained variance (ie, liability-scaled Nagelkerke pseudo R2) were approximately 2.42% when using GSA GWAS to predict ASA GWAS and approximately 2.10% when using ASA GWAS to predict GSA GWAS (P < 1.00 × 10−15) (Figure 3). Assuming that the population prevalence of BD was 0.01, the liability-scaled Nagelkerke pseudo R2 was calculated to estimate the variance of the disorder explained by the SNPs (eMethods in the Supplement). We also examined the polygenic risk of BD across Han Chinese and European populations using the European PGC2 BD GWAS as the training data set and our total discovery GWAS samples as the target data set. That analysis revealed that individuals with BD had a statistically significantly higher PRS than control participants in the target data set of samples from Han Chinese individuals (maximum liability-scaled Nagelkerke pseudo R2 = 1.27%; P = 1.30 × 10−19) (Figure 3).
Shared Genetic Risk of BD and Other Psychiatric Disorders or Traits
We conducted LDSC analysis18,19 to ascertain whether there was a genetic correlation between the Han Chinese BD discovery GWAS and the East Asian schizophrenia GWAS,26 as well as the Han Chinese depression GWAS28 (eTable 9 in the Supplement). That analysis revealed statistically significant genetic correlations between BD and schizophrenia (mean [SE] r for genetic [rg] = 0.535 [0.090]; LDSC P = 3.31 × 10−9) and between BD and depression (mean [SE] rg = 0.392 [0.153]; LDSC P = .0110). In both Han Chinese and European populations, we also found statistically significant trans-ancestry genetic effect correlations and population genetic impact correlations between BD in Han Chinese individuals and other psychiatric disorders and relevant traits in European individuals (eTable 9 in the Supplement), including schizophrenia (mean [SE] ρge = 0.503 [0.074]; P = 1.25 × 10−11; mean [SE] ρgi = 0.486 [0.078]; P = 5.15 × 10−10),27 cognitive performance (mean [SE] ρge = −0.284 [0.057]; P = 5.53 × 10−7; mean [SE] ρgi = −0.291 [0.058]; P = 4.78 × 10−7),30 intelligence (mean [SE] ρge = −0.257 [0.054]; P = 2.17 × 10−6; mean [SE] ρgi = −0.262 [0.055]; P = 1.85 × 10−6),31 and educational attainment (mean [SE] ρge = −0.178 [0.051]; P = 4.31 × 10−4; mean [SE] ρgi = −0.182 [0.050]; P = 2.40 × 10−4).30 The estimation of shared polygenic risk using PRS analysis yielded consistent results, and details are shown in Figure 3 (eResults 3, eDiscussion, and eFigure 8 in the Supplement).
This Han Chinese BD GWAS revealed genome-wide significant association between the TMEM108 locus and BD. Intriguingly, tmem108-deficient neurons in mice have fewer and smaller spines, reduced neurogenesis, and decreased excitatory postsynaptic currents,39 and tmem108-deficient mice have impaired sensorimotor gating and cognitive function.40 Therefore, TMEM108-correlated physiological processes likely contribute to BD pathogenesis. However, the Han Chinese genome-wide significant SNV rs9863544 and its surrounding variations did not show evidence of association with BD in Europeans (P = .23; OR, 1.016; 95% CI, 0.990-1.043) (eFigure 9 in the Supplement),7 suggesting that it may be a Chinese-specific BD risk locus. The T allele frequency of rs9863544 is 0.057 in Han Chinese and 0.439 in Europeans according to the 1000 Genomes Project,15 and LD structural differences in this locus between the 2 populations are also evident (ie, SNVs around rs9863544 exhibit stronger LD in Han Chinese than in Europeans) (eFigure 10 in the Supplement). Therefore, differences in both allele frequencies and LD structures implicate potential genetic heterogeneity of this locus between continental populations, which likely resulted from their different population histories and specific environmental adaptations.41
In the trans-ancestry meta-analysis, we identified novel risk loci (eg, VRK2 and RHEBL1) that did not reach genome-wide significance in the European PGC2 BD GWAS. Indeed, studies42,43 have reported preliminary evidence that VRK2 may alter neuronal proliferation and migration, as well as microglia-mediated synapse elimination. Common variations near VRK2 have also shown genome-wide significant associations with schizophrenia26,27,43-45 and depression,29 supporting the putative involvement of VRK2 in multiple psychiatric disorders.46 Another novel RHEBL1 locus in the present trans-ancestry meta-analysis, although it did not show genome-wide significance in the European PGC2 BD GWAS, was previously implicated in a smaller BD GWAS of Europeans.47 Despite the unclear function of RHEBL1, this gene encodes a brain-enriched G-protein activator of the mechanistic target of rapamycin (mTOR) pathway and thus likely participates in neurodevelopmental and neurodegenerative disorders.48,49
In the post-GWAS analysis based on the trans-ancestry meta-analysis results, we identified 3 genes (NEK4, GLT8D1, and MCM3AP) having statistically significant brain eQTL associations with genetic risk using at least one approach. It has been previously shown that NEK4 and GLT8D1 can alter dendritic spine development and synaptic transmission,50,51 which is in line with the pathological hypothesis of BD.52-54 However, the function of MCM3AP in the brain and in BD pathogenesis is less clear. Further investigations of these genes in BD-relevant physiological and behavioral abnormalities using animal models are necessary. Although some previously identified BD risk genes (eg, CACNA1C, ANK3, NCAN, SCN2A, and POU3F2) were not highlighted in the present post-GWAS analysis, these genes are still worth investigating because their associations with BD genetic risk and pathophysiology have been confirmed from multiple perspectives.5 The involvement of CACNA1C and ANK3 in BD has been extensively described in studies using the approaches of functional genomics, transcriptomics, and physiology.55-65 Similarly, ncan knockout (ncan−/−) mice exhibited mania-like behavioral abnormalities but normalized after lithium administration,66 SCN2A encodes the sodium voltage-gated channel alpha subunit 2 that changes neurophysiology and cognitive processes,67,68 and the protein encoded by POU3F2 alters the differentiation and proliferation of neural progenitor cells.69
This study has some limitations. First, the post-GWAS analyses were primarily conducted using European-based eQTL data or European LD reference panels, which would impact the prioritization of risk genes and variants given the genetic heterogeneity between Han Chinese and European individuals. Further analyses in Han Chinese individuals using such resources are necessary. Second, the control participants in the present study were recruited based on their self-reported health status rather than screening by professionals. Therefore, potential “contamination” of the controls by individuals having undiagnosed psychiatric disorders may need to be addressed.70 However, the consequences of such contamination, if any, are likely minimal because the lifetime prevalence of BD is only approximately 1% in the general population.3
This study describes several novel risk loci for BD and a shared genetic basis for BD across Han Chinese and European populations. Further investigations are warranted to illuminate the underlying pathological mechanisms.
Accepted for Publication: September 16, 2020.
Published Online: December 2, 2020. doi:10.1001/jamapsychiatry.2020.3738
Corresponding Authors: Yiru Fang, MD, PhD, Clinical Research Center and Division of Mood Disorders, Shanghai Mental Health Center, Shanghai Jiao Tong University School of Medicine, 600 S Wanping Rd, Shanghai 200030, China (yirufang@aliyun.com); Ming Li, PhD, Key Laboratory of Animal Models and Human Disease Mechanisms of the Chinese Academy of Sciences and Yunnan Province, Kunming Institute of Zoology, Chinese Academy of Sciences, No. 32 Jiao-Chang Donglu, Kunming, Yunnan 650223, China (limingkiz@mail.kiz.ac.cn).
Author Contributions: Dr Ming Li had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Ms Hui-Juan Li and Drs Chen Zhang and Li Hui are co–first authors.
Concept and design: H-J. Li, Zhong, Lv, Yue, Luo, Xiao, Fang, M. Li.
Acquisition, analysis, or interpretation of data: H-J. Li, C. Zhang, Hui, D-S. Zhou, Y. Li, C-Y. Zhang, C. Wang, L. Wang, W. Li, Y. Yang, Qu, J. Tang, He, J. Zhou, Z. Yang, Xingxing Li, J. Cai, L. Yang, Jun Chen, Fan, Wei Tang, Wenxin Tang, Jia, W. Liu, Zhuo, Song, F. Liu, Bai, Zhong, S-F. Zhang, Jing Chen, Xia, Z. Liu, Hu, X-Y. Li, J-Y. Liu, X. Cai, Yao, Y. Zhang, Yan, Chang, Zhao, Yue, Luo, X. Chen, Xiao, M. Li.
Drafting of the manuscript: H-J. Li, C. Zhang, J. Cai, Fan, Zhuo, Zhong, Luo, Xiao, M. Li.
Critical revision of the manuscript for important intellectual content: Hui, D-S. Zhou, Y. Li, C-Y. Zhang, C. Wang, L. Wang, W. Li, Y. Yang, Qu, J. Tang, He, J. Zhou, Z. Yang, Xingxing Li, L. Yang, Jun Chen, Wei Tang, Wenxin Tang, Jia, W. Liu, Song, F. Liu, Bai, S-F. Zhang, Jing Chen, Xia, Lv, Z. Liu, Hu, X-Y. Li, J-Y. Liu, X. Cai, Yao, Y. Zhang, Yan, Chang, Zhao, Yue, Luo, X. Chen, Xiao, Fang, M. Li.
Statistical analysis: H-J. Li, C. Zhang, C-Y. Zhang, J. Cai, Fan, Luo, M. Li.
Obtained funding: Hui, D-S. Zhou, W. Li, Qu, W. Liu, Lv, Yao, Luo, Xiao, Fang, M. Li.
Administrative, technical, or material support: Hui, D-S. Zhou, Y. Li, C-Y. Zhang, C. Wang, L. Wang, W. Li, Y. Yang, Qu, J. Tang, He, J. Zhou, Z. Yang, Xingxing Li, Jun Chen, Wei Tang, Wenxin Tang, Jia, W. Liu, Song, F. Liu, Bai, S-F. Zhang, Jing Chen, Xia, Z. Liu, Hu, X-Y. Li, J-Y. Liu, X. Cai, Yan, Chang, Zhao, Yue, Luo, X. Chen, M. Li.
Supervision: Y. Li, Lv, Yue, Luo, Xiao, Fang, M. Li.
Conflict of Interest Disclosures: None reported.
Funding/Support: This work was supported by grants from the following: the National Natural Science Foundation of China (81722019 and 81971259 to Dr Ming Li, 81930033 to Dr Yiru Fang, 81771450 to Dr Chen Zhang, 81771439 to Dr Li Hui, 81671330 and 81971252 to Dr Luxian Lv, 81860253 to Dr Weiqing Liu, and 81971271 to Dr Shaohua Hu); the Innovative Research Team of Science and Technology Department of Yunnan Province (2019HC004); the National Key Research and Development Program of China (2016YFC1307100 to Dr Yiru Fang, 2018YFC1314302 to Dr Chen Zhang, and 2018YFC1314600 to Dr Zhongchun Liu); the Medical and Health Science and Technology Project in Zhejiang (2018KY721 to Dr Dong-Sheng Zhou); the Science and Technology Project of Henan Province (192102310086 to Dr Wenqiang Li); the High Scientific and Technological Research Fund of Xinxiang Medical University (2017ZDCG-04 to Dr Luxian Lv); the Hubei Province Health and Family Planning Scientific Research Project (WJ2015Q033 to Dr Na Qu); the Population and Family Planning Commission of Wuhan (WX14B34 to Dr Na Qu); the Health Science and Technology Plan Projects in Yunnan Province (2017NS028 to Dr Wenqiang Li); the Bureau of Frontier Sciences and Education (QYZDJ-SSW-SMC005 to Dr Yong-Gang Yao); and the Strategic Priority Research Program (B) of the Chinese Academy of Sciences (XDB02020003 to Dr Yong-Gang Yao). Dr Xiao Xiao was supported by the Chinese Academy of Sciences Western Light Program and the Youth Innovation Promotion Association, Chinese Academy of Sciences. Dr Ming Li was also supported by the Chinese Academy of Sciences Pioneer Hundred Talents Program and the 1000 Young Talents Program. One of the brain eQTL datasets used in this study were generated as part of the CommonMind Consortium supported by funding from Takeda Pharmaceuticals Company Limited, F. Hoffman-La Roche Ltd and NIH grants R01MH085542, R01MH093725, P50MH066392, P50MH080405, R01MH097276, RO1-MH-075916, P50M096891, P50MH084053S1, R37MH057881 and R37MH057881S1, HHSN271201300031C, AG02219, AG05138 and MH06692. Brain tissue for the study was obtained from the following brain bank collections: the Mount Sinai NIH Brain and Tissue Repository, the University of Pennsylvania Alzheimer’s Disease Core Center, the University of Pittsburgh NeuroBioBank and Brain and Tissue Repositories and the NIMH Human Brain Collection Core. CMC Leadership: Pamela Sklar, Joseph Buxbaum (Icahn School of Medicine at Mount Sinai), Bernie Devlin, David Lewis (University of Pittsburgh), Raquel Gur, Chang-Gyu Hahn (University of Pennsylvania), Keisuke Hirai, Hiroyoshi Toyoshiba (Takeda Pharmaceuticals Company Limited), Enrico Domenici, Laurent Essioux (F. Hoffman-La Roche Ltd), Lara Mangravite, Mette Peters (Sage Bionetworks), Thomas Lehner, Barbara Lipska (NIMH).
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Group Information: The GeseDNA Research Team members are the following individuals affiliated with Beijing Gese Technology Co, Ltd, Beijing, China: Zenan Dou, MS; Shan Guan, PhD; Tingting Guo, MS; Hoyin Lo, MS; Qian Sun, BS; and Leilei Zhang, MS.
5.Zhang
C, Xiao
X, Li
T, Li
M. Translational genomics and beyond in bipolar disorder.
Mol Psychiatry. Published online May 18, 2020.
PubMedGoogle Scholar 7.Stahl
EA, Breen
G, Forstner
AJ,
et al; eQTLGen Consortium; BIOS Consortium; Bipolar Disorder Working Group of the Psychiatric Genomics Consortium. Genome-wide association study identifies 30 loci associated with bipolar disorder.
Nat Genet. 2019;51(5):793-803. doi:
10.1038/s41588-019-0397-8
PubMedGoogle ScholarCrossref 8.Ikeda
M, Takahashi
A, Kamatani
Y,
et al. A genome-wide association study identifies two novel susceptibility loci and trans population polygenicity associated with bipolar disorder.
Mol Psychiatry. 2018;23(3):639-647. doi:
10.1038/mp.2016.259
PubMedGoogle ScholarCrossref 10.Li
W, Cai
X, Li
HJ,
et al. Independent replications and integrative analyses confirm
TRANK1 as a susceptibility gene for bipolar disorder.
Neuropsychopharmacology. Published online August 13, 2020. doi:
10.1038/s41386-020-00788-4
PubMedGoogle Scholar 18.Bulik-Sullivan
B, Finucane
HK, Anttila
V,
et al; ReproGen Consortium; Psychiatric Genomics Consortium; Genetic Consortium for Anorexia Nervosa of the Wellcome Trust Case Control Consortium 3. An atlas of genetic correlations across human diseases and traits.
Nat Genet. 2015;47(11):1236-1241. doi:
10.1038/ng.3406
PubMedGoogle ScholarCrossref 19.Bulik-Sullivan
BK, Loh
PR, Finucane
HK,
et al; Schizophrenia Working Group of the Psychiatric Genomics Consortium. LD Score regression distinguishes confounding from polygenicity in genome-wide association studies.
Nat Genet. 2015;47(3):291-295. doi:
10.1038/ng.3211
PubMedGoogle ScholarCrossref 21.Collado-Torres
L, Burke
EE, Peterson
A,
et al; BrainSeq Consortium. Regional heterogeneity in gene expression, regulation, and coherence in the frontal cortex and hippocampus across development and schizophrenia.
Neuron. 2019;103(2):203-216.e8. doi:
10.1016/j.neuron.2019.05.013
PubMedGoogle ScholarCrossref 24.Millan
MJ, Agid
Y, Brüne
M,
et al. Cognitive dysfunction in psychiatric disorders: characteristics, causes and the quest for improved therapy.
Nat Rev Drug Discov. 2012;11(2):141-168. doi:
10.1038/nrd3628
PubMedGoogle ScholarCrossref 26.Lam
M, Chen
CY, Li
Z,
et al; Schizophrenia Working Group of the Psychiatric Genomics Consortium; Indonesia Schizophrenia Consortium; Genetic Research on Schizophrenia Network–China and the Netherlands (GREAT-CN). Comparative genetic architectures of schizophrenia in East Asian and European populations.
Nat Genet. 2019;51(12):1670-1678. doi:
10.1038/s41588-019-0512-x
PubMedGoogle ScholarCrossref 27.Pardiñas
AF, Holmans
P, Pocklington
AJ,
et al; GERAD1 Consortium; CRESTAR Consortium. Common schizophrenia alleles are enriched in mutation-intolerant genes and in regions under strong background selection.
Nat Genet. 2018;50(3):381-389. doi:
10.1038/s41588-018-0059-2
PubMedGoogle ScholarCrossref 29.Howard
DM, Adams
MJ, Clarke
TK,
et al; 23andMe Research Team; Major Depressive Disorder Working Group of the Psychiatric Genomics Consortium. Genome-wide meta-analysis of depression identifies 102 independent variants and highlights the importance of the prefrontal brain regions.
Nat Neurosci. 2019;22(3):343-352. doi:
10.1038/s41593-018-0326-7
PubMedGoogle ScholarCrossref 30.Lee
JJ, Wedow
R, Okbay
A,
et al; 23andMe Research Team; COGENT (Cognitive Genomics Consortium); Social Science Genetic Association Consortium. Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals.
Nat Genet. 2018;50(8):1112-1121. doi:
10.1038/s41588-018-0147-3
PubMedGoogle ScholarCrossref 33.Li
H, Zhang
C, Cai
X,
et al. Genome-wide association study of creativity reveals genetic overlap with psychiatric disorders, risk tolerance, and risky behaviors.
Schizophr Bull. 2020;46(5):1317-1326. doi:
10.1093/schbul/sbaa025
PubMedGoogle Scholar 42.Lee
J, Lee
S, Ryu
YJ,
et al. Vaccinia-related kinase 2 plays a critical role in microglia-mediated synapse elimination during neurodevelopment.
Glia. 2019;67(9):1667-1679. doi:
10.1002/glia.23638
PubMedGoogle Scholar 43.Yu
H, Yan
H, Li
J,
et al; Chinese Schizophrenia Collaboration Group. Common variants on 2p16.1, 6p22.1 and 10q24.32 are associated with schizophrenia in Han Chinese population.
Mol Psychiatry. 2017;22(7):954-960. doi:
10.1038/mp.2016.212
PubMedGoogle ScholarCrossref 44.Steinberg
S, de Jong
S, Andreassen
OA,
et al; Irish Schizophrenia Genomics Consortium; GROUP; Wellcome Trust Case Control Consortium 2. Common variants at
VRK2 and
TCF4 conferring risk of schizophrenia.
Hum Mol Genet. 2011;20(20):4076-4081. doi:
10.1093/hmg/ddr325
PubMedGoogle ScholarCrossref 47.Green
EK, Hamshere
M, Forty
L,
et al; WTCCC. Replication of bipolar disorder susceptibility alleles and identification of two novel genome-wide significant associations in a new bipolar disorder case-control sample.
Mol Psychiatry. 2013;18(12):1302-1307. doi:
10.1038/mp.2012.142
PubMedGoogle ScholarCrossref 50.Yang
Z, Zhou
D, Li
H,
et al. The genome-wide risk alleles for psychiatric disorders at 3p21.1 show convergent effects on mRNA expression, cognitive function, and mushroom dendritic spine.
Mol Psychiatry. 2020;25(1):48-66. doi:
10.1038/s41380-019-0592-0
PubMedGoogle ScholarCrossref 55.Gershon
ES, Grennan
K, Busnello
J,
et al. A rare mutation of
CACNA1C in a patient with bipolar disorder, and decreased gene expression associated with a bipolar-associated common SNP of
CACNA1C in brain.
Mol Psychiatry. 2014;19(8):890-894. doi:
10.1038/mp.2013.107
PubMedGoogle ScholarCrossref 59.Rueckert
EH, Barker
D, Ruderfer
D,
et al.
Cis-acting regulation of brain-specific
ANK3 gene expression by a genetic variant associated with bipolar disorder.
Mol Psychiatry. 2013;18(8):922-929. doi:
10.1038/mp.2012.104
PubMedGoogle ScholarCrossref 60.Kabir
ZD, Che
A, Fischer
DK,
et al. Rescue of impaired sociability and anxiety-like behavior in adult
cacna1c-deficient mice by pharmacologically targeting eIF2α.
Mol Psychiatry. 2017;22(8):1096-1109. doi:
10.1038/mp.2017.124
PubMedGoogle ScholarCrossref 61.Clark
MB, Wrzesinski
T, Garcia
AB,
et al. Long-read sequencing reveals the complex splicing profile of the psychiatric risk gene
CACNA1C in human brain.
Mol Psychiatry. 2020;25(1):37-47.
PubMedGoogle ScholarCrossref 63.Nelson
AD, Caballero-Florán
RN, Rodríguez Díaz
JC,
et al. Ankyrin-G regulates forebrain connectivity and network synchronization via interaction with GABARAP.
Mol Psychiatry. Published online November 30, 2018.
PubMedGoogle Scholar 65.Zhu
S, Cordner
ZA, Xiong
J,
et al. Genetic disruption of ankyrin-G in adult mouse forebrain causes cortical synapse alteration and behavior reminiscent of bipolar disorder.
Proc Natl Acad Sci U S A. 2017;114(39):10479-10484. doi:
10.1073/pnas.1700689114
PubMedGoogle ScholarCrossref 68.Shin
W, Kweon
H, Kang
R,
et al.
Scn2a haploinsufficiency in mice suppresses hippocampal neuronal excitability, excitatory synaptic drive, and long-term potentiation, and spatial learning and memory.
Front Mol Neurosci. 2019;12:145. doi:
10.3389/fnmol.2019.00145
PubMedGoogle ScholarCrossref 69.Chen
C, Meng
Q, Xia
Y,
et al. The transcription factor POU3F2 regulates a gene coexpression network in brain tissue from patients with psychiatric disorders.
Sci Transl Med. 2018;10(472):eaat8178. doi:
10.1126/scitranslmed.aat8178
PubMedGoogle Scholar