Distribution of tag single-nucleotide polymorphisms (SNPs) in the BCL9 gene. The red numbers represent the different SNPs. The upper part of the figure shows the distribution of the 10 selected tag SNPs in the BCL9 gene. The lower part of the figure shows the linkage disequilibrium among all the SNPs. The black triangles show the blocks existing in the gene. UTR indicates untranslated region.
Multidimensional scaling plots in the third-stage samples. BPD indicates bipolar disorder; MDD, major depressive disorder, NC, normal controls; and SZ, schizophrenia.
Li J, Zhou G, Ji W, Feng G, Zhao Q, Liu J, Li T, Li Y, Chen P, Zeng Z, Wang T, Hu Z, Zheng L, Wang Y, Shen Y, He L, Shi Y. Common Variants in the BCL9 Gene Conferring Risk of Schizophrenia. Arch Gen Psychiatry. 2011;68(3):232-240. doi:10.1001/archgenpsychiatry.2011.1
Recent genome-wide association studies have revealed that common variations and rare copy-number variations contribute to the risk of mental disorders. Rare recurrent microdeletions at 1q21.1 were reported to be associated with schizophrenia, and the BCL9 gene at 1q21.1 was also a functional candidate gene for mental disorders.
To investigate and validate whether common variations exist in a functional candidate gene in the copy-number variation region, and, if so, to determine whether these variations confer risk of schizophrenia or other mental disorders.
A 3-stage case-control study.
A total of 12 229 subjects were included: 5772 normal controls, 4187 patients with schizophrenia, 1135 patients with bipolar disorder patients, and 1135 patients with major depressive disorder.
Main Outcome Measure
During the first and second stages of our study, we genotyped 10 single-nucleotide polymorphisms using the ligation detection reaction method. During the third stage of our study, all single-nucleotide polymorphisms were genotyped using TaqMan technology (Applied Biosystems, Foster City, California).
During the first stage of our study, we found that rs672607 was significantly associated with schizophrenia (P = 2.69 × 10−5). During the second stage, rs672607 was successfully replicated (P = 1.33 × 10−5), and rs9326555 (P = .002), rs1240083 (P = 1.7 × 10−4), and rs688325 (P = .006) were newly identified to be significant. During the third stage, we genotyped all single-nucleotide polymorphisms in 1135 patients with schizophrenia, 1135 patients with bipolar disorder, 1135 patients with major depressive disorder, and 1135 normal controls for further validation. Finally, when we combined all the data from the 3 stages of our schizophrenia study, we found that rs9326555 (P = 1.53 × 10−5), rs10494251 (P = .02), rs1240083 (P = 1.52 × 10−4), rs672607 (P = 1.23 × 10−11), rs688325 (P = 2.54 × 10−4), and rs3766512 (P = .01) were significant. Moreover, we found that rs672607 was significant in major depressive disorder (P = .001) and bipolar disorder (P = .03), and rs10494251 (P = .04), rs1541187 (P = .04), rs688325 (P = .02), and rs946903 (P = .006) were significant in major depressive disorder.
These findings indicate that common variations in the BCL9 gene confer risk of schizophrenia and may also be associated with bipolar disorder and major depressive disorder in the Chinese Han population.
Mental disorders are now recognized as leading causes of morbidity, and they require a great deal of long-term medical and social care.1 The results of numerous family, twin, and adoption studies conclude that genetic factors, rather than shared environment, play a major role in common mental disorders.1
Schizophrenia is characterized by psychotic features (delusions and hallucinations), disorganization, dysfunction in normal affective responses, and altered cognitive functioning, with heritability at up to 80%.1,2 Bipolar disorder and major depressive disorder are both related to distinct mood changes. Bipolar disorder is characterized by disturbances in mood ranging from extreme elation (mania) to severe depression, and it is often accompanied by psychotic features and cognitive changes, with heritability at 62% to 80%.1,3,4 Major depressive disorder includes a distinct change of mood, characterized by sadness or irritability and accompanied by at least several psychophysiological changes.5 Previous studies comparing concordance rates of major depressive disorder between monozygotic and dizygotic twins suggested a heritability of approximately 37%.6 Early-onset depression that is severe and recurrent has a higher heritability than do other forms of depression.5 Although the psychiatric etiology is poorly understood, accumulating evidence suggests that neurodevelopmental defects are involved.7,8
The B-cell CLL/lymphoma 9 gene (BCL9 ; OMIM 602597) extends to the 84.84-kb region from base 145 479 806 to base 145 564 639 at chromosome 1q21.1, including 10 exons and 9 introns. In 1998, Willis et al9 first reported the molecular clone using long-distance-inverse polymerase chain reaction (PCR) of a t(1;14)(q21;q32) in a pre-B acute lymphoblastic leukemia cell line (CEMO-1) and designated the novel BCL9 gene as the 1q21 target. The BCL9 gene encodes a 1394–amino acid protein that contains several pentapeptide repeats, a potential nuclear localization signal, and a 30–amino acid region that shares 90% homology with a Drosophila embryo expressed sequence tag clone but possesses no other recognizable domains.10 The BCL9 gene is expressed ubiquitously as 2 low-level transcripts from all tissues (a major 6.3-kb transcript and a less prominent 4.2-kb transcript), whereas a 1.6-kb transcript is present only in the spleen, thymus, and small intestine.9 The BCL9 protein is required for efficient T-cell factor–mediated transcription in the Wnt signaling pathway.11 There are several lines of evidence that the Wnt signaling pathway may be involved in the etiology of mental disorders.12 The Wnt signaling pathway influences neuroplasticity, cell survival, and adult neurogenesis. Recent studies12,13 have suggested that mental disorders may involve impairments in these functions. In 2008, Zandi et al13 first conducted a family-based association study between candidate genes in the Wnt signaling pathway containing the BCL9 gene and disease susceptibility in bipolar disorder. However, they found no association between the BCL9 gene and bipolar disorder in European samples.13
Several recent genome-wide association studies (GWASs) have identified that genetic common variants showed consistent association with schizophrenia and bipolar disorder. In 2009, the Genetic Risk and Outcome in Psychosis (GROUP), International Schizophrenia Consortium (ISC), and Molecular Genetics of Schizophrenia (MGS) cohort studies,2,14,15 using single nucleotide polymorphism (SNP)–based GWASs in large samples, suggested that individual genes associated with complex diseases may reveal underlying biological pathways, and these studies2,14,15 showed that common genetic variation may confer the risk of schizophrenia and bipolar disorder. In addition, they also speculated that schizophrenia and bipolar disorder may have overlap in pathogenesis.
In 2008, Stefansson et al16 discovered that large recurrent microdeletions were associated with schizophrenia at 1q21.1. They used a large population-based sample of 9878 transmissions from parents to offspring to identify de novo copy-number variations, after which they tested for an association in 1433 schizophrenia cases and 33 250 controls (phase I). Then they replicated the most promising variants from the first phase in a second, larger sample of 3285 cases and 7951 controls (phase II).16 They found significant association with schizophrenia and related psychoses for deletions at chromosome 1q21.1 in the phase I sample, the phase II sample, and the combined samples (uncorrected P = .02, P = 5.6 × 10−4, and P = 2.9 × 10−5, respectively).16 The ISC cohort study also found large deletions at chromosome 1q21.1 that were associated with schizophrenia.17 Both of these findings indicated that rare structural variations at 1q21.1 may play a role in the etiology of schizophrenia. We raise the question whether there are common variations that are associated with mental disorders, especially common variations in a functional candidate gene at this rare copy-number variation region.
The 1.35-Mb deleted segment of the chromosome 1q21.1 deletion contains several genes, including PRKAB2, FOM5, CHD1L, ACP6, GJA5, GJA8, GPR89B, GPR89C, NBPF11, and BCL9. The GJA8 and GJA5 genes were previously studied in association with schizophrenia in an Asian population,16 and there was no evidence that the PRKAB2, FOM5, CHD1L, ACP6, GPR89B, GPR89C, and NBPF11 genes played a role in neurodevelopment. However, for the BCL9 gene, it is required in the Wnt signaling pathway, which influences the neurodevelopment of mental disorders. Therefore, it was considered to be a functional candidate gene for mental disorders and was also considered to be located at 1q21.1. In addition, to our knowledge, there have been no studies on the association between the BCL9 gene and mental disorders in an Asian population until now.
In our study using 3 independent sample sets of the Chinese Han population, we chose 10 tag SNPs (rs671056, rs9326555, rs10494251, rs11240083, rs1541187, rs672607 rs688325, rs2236570, rs946903, and rs3766512) covering BCL9 (Figure 1) to investigate whether common variants in the BCL9 gene predispose an individual to schizophrenia. We also conducted association studies on bipolar disorder and major depressive disorder.
All cases were outpatients or stable inpatients from a mental health center in Shanghai, China. Eligible subjects had to meet all of the following inclusion criteria: (1) all subjects had to be interviewed by 2 independent psychiatrists, and their cases had to be diagnosed according to the DSM-IV criteria; (2) all subjects had to be unrelated, born in Shanghai, living in Shanghai, and their parents had to be local residents of Shanghai as well; (3) all subjects had to be from 18 to 65 years of age; (4) all subjects had to have had a 2-year history of psychiatric disease; and (5) all subjects had to provide written informed consent. Subjects were excluded if they had other diseases, such as cerebral infarction. Approval was received for our study from the local Ethics Committee of Human Genetic Resources.
The sample of schizophrenia cases consisted of 4487 unrelated participants meeting the DSM-IV criteria. All of our schizophrenia cases were paranoid schizophrenic patients and had not had a lifetime episode of mania or depression. All cases met 2 criteria: preoccupation with 1 or more delusions and frequent auditory hallucinations; however, none of the following symptoms were prominent: disorganized speech, disorganized or catatonic behavior, or flat or inappropriate affect. Subjects with an IQ of less than 70 were excluded. The schizophrenia cases in the 3 stages of our study were recruited using the same criteria. The sample of subjects from the first stage of our study consisted of 1034 patients with schizophrenia (570 men and 464 women). The mean (SD) age was 38.8 (14.1) years, and the mean (SD) onset age was 25.2 (11.1) years. The second-stage sample consisted of 2018 patients with schizophrenia (1041 men and 977 women). The mean (SD) age was 42.7 (16.1) years, and the mean (SD) onset age was 25.7 (10.1) years. The third-stage sample consisted of 1135 unrelated patients with schizophrenia (630 men and 505 women). The mean (SD) age was 35.4 (7.2) years, and the mean (SD) onset age was 24.8 (6.7) years.
The cases of bipolar disorder consisted of 1135 unrelated participants meeting the DSM-IV criteria (618 men and 517 women). The mean (SD) age of patients with bipolar disorder was 47.9 (10.4) years, and the mean (SD) onset age was 36.6 (13.8) years. Of these unrelated participants in our study, 673 (59.3%) had a lifetime experience of psychotic symptoms, and 442 (38.9%) had a lifetime experience of mood-incongruent psychotic symptoms.
All of the subjects with major depressive disorder were carefully selected on the basis of having at least 2 distinct major depressive disorder episodes and no sign of bipolar disorder symptoms during the 2 years after the onset of depression. A total of 1135 unrelated subjects with major depressive disorder (483 men and 652 women) were recruited. The mean (SD) age was 49.3 (12.8) years, and the mean (SD) onset age was 35.1 (11.6) years. Of these unrelated subjects, 583 (51.4%) had psychotic symptoms, and 563 (49.6%) had a lifetime experience of comorbid anxiety disorder (panic or social phobia).
All controls were randomly selected from the local population of Shanghai. Volunteers replied to a written invitation to have their medical history completely evaluated; along with this invitation were questions about schizophrenia, psychosis, major depressive disorder, and bipolar disorder. Practice lists were screened for potentially suitable volunteers by exclusion of subjects with major mental illness or with first-degree relatives with major mental illness. All controls in the third stage of our study were recruited using the same criteria. The first-stage sample of subjects consisted of 1034 normal controls (522 men and 512 women), and the mean (SD) age was 30.0 (8.7) years. Of these normal controls, 307 (29.7%) participated in our study. The second-stage sample of subjects consisted of 3603 normal controls (1800 men and 1803 women), and the mean (SD) age was 61.5 (9.9) years. Of these normal controls, 930 (25.8%) participated in our study. The third-stage sample of subjects consisted of 1135 normal controls (369 men and 766 women), and the mean (SD) age was 58.7 (9.9) years. Of these normal controls, 326 (28.7%) participated in our study.
Genomic DNA was prepared from peripheral blood samples obtained from the subjects using the standard phenol-chloroform extraction method. Tag SNP selection was performed using Haploview software (http://www.broad.mit.edu/mpg/haploview) (parameter setting as pairwise tagging with r2 ≥ 0.8)18,19 and a minor allele frequency of 0.05 or higher. Ten tag SNPs covering the whole BCL9 gene by linkage disequilibrium were selected: rs671056, rs9326555, rs10494251, rs11240083, rs1541187, and rs672607 in intron 1; rs688325 in intron 2; rs2236570 in intron 8; and rs946903 and rs3766512 in the 3′ untranslated region of the exon region (Figure 1). In the coding region of the BCL9 gene, there was no nonsynonymous mutation with a genotype frequency higher than 0.05 in the Chinese Han population in the National Center for Biotechnology Information database.
During the first and second stages of our study, we genotyped all 10 SNPs using the ligation detection reaction method, with technical support from the Shanghai Biowing Applied Biotechnology Company. The mean call rate for all markers was 96%. There was no significant deviation from the Hardy-Weinberg equilibrium in the controls (P > .05).
During the third stage of our study, all SNPs were genotyped on the ABI 7900 DNA detection system (Applied Biosystems, Foster City, California) using TaqMan technology. All probes were designed by Applied Biosystems. The standard 5-μL PCR was performed using TaqMan Universal PCR Master Mix reagent kits according to the guidelines provided. There was no significant deviation from the Hardy-Weinberg equilibrium in the controls (P > .05).
All calculations of allele and genotype frequencies, of the Hardy-Weinberg equilibrium, and of pairwise linkage disequilibrium were performed using the online SHEsis software (http://analysis.bio-x.cn).20,21 The significance level was set at α = .05. Power calculations were performed using the G-power software (http://www.psycho.uni-duesseldorf.de/abteilungen/aap/gpower3/).
To avoid the false-positive association caused by potential population stratification, we performed population stratification analysis on the third-stage sample set, using genotype data on 89 additional random SNPs (eTable 1). We obtained data on these 89 SNPs from 270 HapMap samples (90 SNPs for CEU [Utah residents with ancestry from northern and western Europe in the United States], 90 SNPs for YRI [Yoruba in Ibadan, Nigeria], 45 SNPs for CHB [Han Chinese in Beijing], and 45 JPT [Japanese in Tokyo]; HapMap public release 23a at http://hapmap.ncbi.nlm.nih.gov/biomart) and found that these SNPs worked well to differentiate these samples into 3 clusters by ethnicity (eFigure 1A). Structure version 2.3.1 software (http://pritch.bsd.uchicago.edu/software/structure_v.2.3.1.html) was used to track potential population structures using these SNPs.22,23 We applied an admixture model and an independent allelic frequency model and ran from K = 2 to K = 10, where K is the number of potential classifications of tested samples. For all Structure runs, we set the parameters with a burn-in of 10 000 iterations and 20 000 follow-on iterations. All results of this analysis are listed in the eAppendix.
We also calculated pairwise identities by state (IBS) for all samples, and we performed classical multidimensional scaling on the identity matrices for the total data. The genomic inflation factor (λ) was generated using Plink software (http://pngu.mgh.harvard.edu/~purcell/plink/), running the “--assoc --adjust” command line option, and an entry was generated in the log file.24 The value should be close to 1 and not greater than 1.2.
For 1034 subjects with schizophrenia and 1034 normal controls, we found that rs672607 was significantly associated with schizophrenia (P = 2.69 × 10−5) (Table 1). To validate this initial finding, we conducted the second stage of our study on a much larger sample set: 2018 subjects with schizophrenia and 3603 normal controls. We successfully replicated the significance of rs672607 (P = 1.33 × 10−5) for schizophrenia (Table 1). Moreover, we found that rs9326555 (P = .002), rs1240083 (P = 1.7 × 10−4), and rs688325 (P = .006) were also significantly associated with schizophrenia (Table 1).
During the third stage of our study, we genotyped 1135 subjects with schizophrenia, 1135 subjects with bipolar disorder, 1135 subjects with major depressive disorder, and 1135 normal controls. We still found rs672607 (P = .006) and rs688325 (P = .006) to be risk factors for schizophrenia (Table 2). We also detected the significance of rs672607 in bipolar disorder (P = .03). For major depressive disorder, rs672607 (P = .001), rs10494251 (P = .04), rs1541187 (P = .04), and rs688325 (P = .02) were found to be significant (Table 2).
We also conducted a combined study of schizophrenia, with 4192 subjects with schizophrenia and 5777 normal controls. We found rs9326555 (P = 1.53 × 10−5), rs10494251 (P = .02), rs1240083 (P = 1.52 × 10−4), rs672607 (P = 1.23 × 10−11), rs688325 (P = 2.54 × 10−4), and rs3766512 (P = .01) to be significantly associated with schizophrenia (Table 2).
Genotype distributions were in Hardy-Weinberg equilibrium for all SNPs in the controls during each stage of our study (P > .05), except for rs1541187 during the first stage of our study (P = 2.48 × 10−6) and rs9326555 (P = .001), rs10494251 (P = .007), and rs1541187 (P = 3.17 × 10−11) during the second stage. Genotype frequencies are shown in eTables 2, 3, and 4. We performed the population stratification analysis during the third stage of our study using Structure software (eAppendix). We also used Plink software for the pairwise IBS analysis and calculated the multidimensional scaling of IBS distances. The multidimensional scaling result indicated no obvious population stratification (Figure 2). The λ values were also obtained using Plink software. For the third-stage samples, λ = 1.092 (for schizophrenia, λ = 1.000; for bipolar disorder, λ = 1.068; and for major depressive disorder, λ = 1.131). These findings indicated that there was no obvious population stratification in our sample of subjects, and our results should not be considered a false-positive association that was simply caused by population stratification.
The BCL9 gene is a good candidate gene for psychiatric diseases because it is a functional candidate gene for mental disorders and is also located in a region affected by rare copy-number variations that represent a risk of schizophrenia. The BCL9 gene is required in the Wnt signaling pathway; this pathway plays a crucial role in a number of animal developmental processes and directs growth and cell development in a variety of processes (such as embryonic central nervous system segmentation).25- 27 In our study, we found that common variations in the BCL9 gene were significantly associated with schizophrenia in 3 independent sample sets, and that these variations were also moderately associated with major depressive disorder and marginally associated with bipolar disorder.
In 2009, the ISC performed a GWAS of 3322 European subjects with schizophrenia and 3587 controls. First, they suggested that the major histocompatibility complex is associated with schizophrenia. Second, they provided molecular genetic evidence for a substantial polygenic component to the risk of schizophrenia involving thousands of common alleles of very small effect. For schizophrenia, their data pointed to a genetic architecture that included many common variants of small effect.2 The MGS cohort study was a GWAS of common SNPs in the MGS case-control samples, and they performed a meta-analysis of data from the MGS, ICS, and SGENE consortium data sets.15 The GROUP cohort study also performed a genome-wide scan of 2663 subjects with schizophrenia and 13 498 controls (from 8 European locations), and then they combined the findings from their top 1500 markers with the results from both the ISC and MGS studies.14 The combined findings proved that common variants located in the major histocompatibility complex region conferred risk of schizophrenia, and there was also speculation that schizophrenia and bipolar disorder may share the same genetic factors in pathogenesis.
In 2008, Ferreira et al28 tested 1.8 million variants in 4387 cases of bipolar disorder and 6209 controls (including data from studies by the Wellcome Trust Case Control Consortium and the Systematic Treatment Enhancement Program for Bipolar Disorder and the University College London) and found that a common SNP located in the ANK3 gene was strongly associated with bipolar disorder. In 2009, Sullivan et al29 conducted a GWAS of 435 291 SNPs genotyped in 1738 cases of major depressive disorder and 1802 controls selected to be at low liability for major depressive disorder. They also detected a common SNP significantly associated with major depressive disorder.29 All these studies confirmed that common variations contributed to the risk of schizophrenia, major depressive disorder, and bipolar disorder, consistent with our findings.
In these GWASs, there was no information about whether common variants in the BCL9 gene were associated with schizophrenia, bipolar disorder, or major depressive disorder. The minor allele frequency of rs672607 varies quite a lot between the white European population and the Chinese Han population, which indicates that obvious genetic heterogeneity exists in this region. Because the minor allele A is very rare in Europeans, we thought that rs672607 might even be excluded after quality control procedures were performed in these GWASs of common variants. However, rs672607 was frequently found in African populations. Therefore, we checked the data from a GWAS on African American population–based schizophrenia from the Genetic Association Information Network (GAIN) database. In this GWAS, we found that 11 SNPs in BCL9 were associated with schizophrenia (eTable 5), in which rs672607 and rs666106 were in linkage disequilibrium (D ′ = 1.0; r2 = 0.17; logarithm of odds [LOD] = 5.45; P = .003), rs688325 and rs11240089 (P = .006) were in strong linkage disequilibrium (D ′ = 1.0; r2 = 0.82; LOD = 12.6; P = .006), and rs11240083 and rs666106 were also in linkage disequilibrium (D ′ = 1.0; r2 = 0.054; LOD = 2.26) (eFigure 2). These results further support our findings. In addition, by checking a GWAS on schizophrenia in white European subjects from the GAIN database, we found only 2 SNPs in the BCL9 gene that were marginally associated with schizophrenia (eTable 6). Such differences among populations were considered to be related to genetic heterogeneity, including differences in allele frequencies and in correlated linkage-disequilibrium patterns. Fine mapping studies of BCL9 in the white European population may help us understand this phenomena.
Power analysis revealed that the statistical power of our sample to detect a significant association (P < .05) was about 90% in allele or genotype comparisons. To test whether population stratification was significant in our samples, we collected genotyping data of another 89 SNPs (eTable 1) in the third-stage samples, which contained 1135 unrelated cases of schizophrenia, 1135 unrelated cases of major depressive disorder, 1135 unrelated cases of bipolar disorder, and 1135 normal controls. The mean call rate for these SNPs was 98.7%. By using Structure version 2.3.1, we found no stratification in our third-stage samples.22,23 However, the 270 HapMap samples could easily be divided into 3 groups by ethnicity (eFigures 1 and 3). The λ value for the third-stage samples was 1.092. Therefore, these findings indicated that there was no sign of population stratification in our subjects, and our findings should not be considered spurious associations caused by population stratification.
In conclusion, our experimental data provide strong evidence that common variations in the BCL9 gene are associated with schizophrenia. We also found common variations in the BCL9 gene to be moderately associated with major depressive disorder and marginally associated with bipolar disorder in the Chinese Han population. Further studies investigating the role of the BCL9 gene in the etiology of schizophrenia would be worthwhile, and replications of this association in further independent samples are necessary.
Correspondence: Yongyong Shi, PhD, Bio-X Center and Affiliated Changning Mental Health Center, Key Laboratory for the Genetics of Developmental and Neuropsychiatric Disorders, Ministry of Education, Shanghai Jiao Tong University, Shanghai 200030, China (email@example.com).
Submitted for Publication: January 15, 2010; final revision received October 6, 2010; accepted November 19, 2010.
Author Contributions: Dr J. Li had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Financial Disclosure: None reported.
Funding/Support: This work was supported by grants from the National 863 program (2006AA02A407 and 2009AA022701), the Shanghai Changning Health Bureau program (2008406002), and the Shanghai Municipal Health Bureau program (2008095), the Shanghai Leading Academic Discipline Project (B205), and 973 grant 2010CB529600.