Nominal association between a polymorphism in DGKH and bipolar disorder detected in a meta-analysis of East Asian case–control samples

Authors


Hiroaki Kawasaki, MD, PhD, Department of Neuropsychiatry, Graduate School of Medical Sciences, Kyushu University, 3-1-1, Maidashi, Higashi-ku, Fukuoka 812-8582, Japan. Email: hiroakik@npsych.med.kyushu-u.ac.jp

Abstract

Aim:  Recent genome-wide association studies (GWAS) of bipolar disorder (BD) have detected new candidate genes, including DGKH, DFNB31 and SORCS2. However, the results of these GWAS were not necessarily consistent, indicating the importance of replication studies. In this study, we tested the genetic association of DGKH, DFNB31 and SORCS2 with BD.

Methods:  We genotyped 18 single-nucleotide polymorphisms (SNP) in DGKH, DFNB31 and SORCS2 using Japanese samples (366 cases and 370 controls). We also performed a meta-analysis of four SNP in DGKH, using the previously published allele frequency data of Han-Chinese case–control samples (1139 cases and 1138 controls).

Results:  In the association analysis using Japanese samples, a SNP in SORCS2 (rs10937823) showed nominal genotypic association. However, we could not find any association in an additional analysis of tag SNP around rs10937823. In the meta-analysis of SNP in DGKH, rs9315897, which was not significantly associated with BD in the previous Chinese study, showed nominal association.

Conclusion:  Although the association was not strong, the result of this study would support the association between DGKH and BD.

FAMILY, TWIN AND adoption studies have consistently demonstrated the contribution of inherited genetic variation on risk for bipolar disorder (BD).1 Therefore, numerous genetic studies, including linkage mapping and candidate gene studies, have been carried out. However, the results of these studies have largely been inconsistent. After the era of linkage study and candidate gene approach, genome-wide association studies (GWAS), which investigate 500 000 to 1 000 000 single-nucleotide polymorphisms (SNP) throughout the genome using DNA microarray, have become popular. For BD, an increasing number of GWAS, including meta-analyses, have been conducted.2–7,16,17 Meta-analyses of GWAS data of BD and major depressive disorder were also performed.8,9 These studies identified many previously unsuspected candidate genes.

DGKH, DFNB31 and SORCS2 are included in these new candidate genes for BD, as well as other promising genes, such as ANK3, CACNA1C,3PBRM18 and so on. The association of DGKH, DFNB31 and SORCS2 with BD was identified in the first reported GWAS using a DNA pooling strategy followed by individual genotyping.2 All of these genes are expressed in the brain and have some functional implication with neuropsychiatric disorders. Diacylglycerol kinase (DGK) eta, encoded by DGKH, is a member of the DGK family that plays an important role in the inositol pathway, the putative site of action of lithium.10DFNB31 encodes whirlin, which is also known as Cip98. This protein interacts with a calmodulin-dependent serine kinase and is suggested to be involved in the formation of scaffolding protein complex and in synaptic transmission.11 A mutation in this gene is known to cause Usher syndrome,12 which is characterized by hearing impairment and progressive vision loss. One study reported frequent comorbidity of bipolar disorder or depressive disorder in Usher syndrome patients.13SORCS2, encoding a VSP10 domain containing-receptor, is expressed in both the developing and adult brain.14 While the ligand for SORCS2 is so far unknown, highly homologous other members of the SORCS VSP10 domain-containing-receptor family, SORCS1 and SORCS3, are known to bind several growth factors, such as NGF and PDGF.15 These findings further make DGKH, DFNB31 and SORCS2 good candidates for BD. However, GWAS usually investigate hundreds of thousands SNP and always involve the potential for false positive results. In fact, some inconsistent results about these genes were reported in other GWAS for BD.3–7,16,17

Several replication studies of DGKH, DFNB31 and SORCS2 were also reported and they include both positive and negative results. A study in Sardinian samples (197 cases and 300 controls) detected a haplotypic association between BD and DGKH.18 Another study that examined 36 tag SNP from DGKH in a Scandinavian population (594 cases and 1421 controls) reported no association.19 A replication study of GWAS that investigated 26 SNP, including those from DGKH, DFNB31 and SORCS2 using a Finnish family cohort (723 individuals from 180 families), reported associations of DFNB31 and SORCS2 but no association of DGKH with BD.20 Recently, Zeng et al. reported a strong haplotypic association (minimum P = 3.87 × 10−6) between DGKH and BD in a Han-Chinese case–control cohort.21

In this study, we investigated associations of DGKH, DFNB31 and SORCS2 with BD using Japanese case–control samples (366 cases and 370 controls). Furthermore, we conducted a meta-analysis of four SNP in DGKH, which are overlapped with the SNP investigated by Zeng et al. In total, 1505 cases and 1508 control samples were used for the meta-analysis.

METHODS

The Japanese case–control samples consisted of 366 patients with bipolar I disorder (BDI), bipolar II disorder (BDII) or schizoaffective disorder bipolar type (SAB) (257 BDI, 104 BDII and five SAB; 181 men and 185 women, aged 50.1 ± 13.4 years) and 370 control subjects (185 men and 185 women aged 50.6 ± 12.6 years). All subjects were unrelated and ethnically Japanese. The patients were diagnosed by at least two experienced psychiatrists according to the DSM-IV criteria on the basis of unstructured interviews and reviews of their medical records. All healthy control subjects were also psychiatrically screened on the basis of unstructured interviews. The objective of the present study was clearly explained, and written informed consent was obtained from all the participants. The characteristics of the Han-Chinese cohort used for meta-analysis were described elsewhere.21 This study was approved by the ethics committees of Kyushu University Graduate School of Medicine and RIKEN Brain Science Institute.

For SNP selection, we focused on the SNP that were reported in a previous GWAS by Baum et al. and that cause non-synonymous amino acid substitutions and possibly affect the function of encoded proteins. SNP chosen from Baum et al. included six SNP whose associations were detected in the follow-up individual genotyping (three in DGKH, one in DFNB31 and two in SORCS2, reported P-values ranging from 0.0005 to 1.5 × 10−8). Three SNP in DGKH indicated to be associated in both of the two independent pooling sample sets were additionally selected. The selected non-synonymous SNP consisted of two SNP in DGKH, five SNP in DFNB31 and two SNP in SORCS2. The total number of selected SNP was eighteen. Genotyping was performed using TaqMan assay (Applied Biosystems, Foster City, CA, USA). Differences in allele and genotype frequencies between BD and controls were evaluated with Fisher's exact test. Deviations from Hardy–Weinberg equilibrium (HWE), structures of linkage disequilibrium (LD) blocks and haplotypic associations were analyzed using Haploview version 4.1 software,22 (http://www.broadinstitute.org/scientific-community/science/programs/medical-and-population-genetics/haploview/haploview). A meta-analysis using the Mantel–Haenszel model and evaluation of sample heterogeneity were performed on Review Manager.23 The I2 statistics24,25 were used for the assessment of heterogeneity between the samples.

RESULTS

The results of genotyping in the Japanese cohort are summarized in Table 1. Three SNP (rs35776153, rs35003670 and rs34058821) registered in dbSNP (http://www.ncbi.nlm.nih.gov/projects/SNP/) were not polymorphic in our samples. Genotype frequencies of each SNP were in HWE (P > 0.05) except for rs16840892 in the BD group. No SNP displayed significant allelic association with BD. A nominally significant genotypic association was observed with rs10937823 of SORCS2. However, the over-represented allele in BD was opposite to the one that was reported in a previous GWAS (T in this study and C in Baum et al.). To investigate this locus more intensively, we selected four tag SNP in the region between rs4411993 and rs34058821 (physical chromosomal position spanning 7 517 366 to 7 717 060), using TAG SNP Selection,26http://manticore.niehs.nih.gov/), and genotyped them. However, no significant allelic or genotypic association between these four SNP and BD was found (Table 1, with ).

Table 1.  Allele and genotype frequencies of SNP in DGKH, DFNB31 and SORCS2 in the case–control sample
GeneSNP ID HWE
P-value
nAlleleP-value
(Fisher's exact test)
GenotypeP-value
(Fisher's exact test)
MAF
  • *

    P < 0.05.

  • Non-synonymous SNP.

  • Additionally examined SNP.

  • BD, bipolar disorder sample; CT, control sample; HWE, Hardy–Weinberg equilibrium; MAF, minor allele frequency; n, number of genotyped participants; SNP, single-nucleotide polymorphism.

DGKH    TC T/TT/CC/C  
rs9315885BD0.5045364333395 73187104 45.7%
 CT0.95303703364040.9166761841100.909445.4%
    AC A/AA/CC/C  
rs1012053BD0.4608364359369 8518990 49.3%
 CT0.95953693663720.916991184940.857849.6%
    AG A/AA/GG/G  
rs1170195BD0.2023360352368 8019288 51.1%
 CT0.75473703723680.601192188900.678949.7%
    TC T/TT/CC/C  
rs9525570BD0.7813360615105 262917 14.6%
 CT0.7634368645910.22012828150.452012.4%
    CT C/CC/TT/T  
rs1170191BD0.2748358373343 9218977 47.9%
 CT0.99823683643720.319090184940.443050.5%
    TC T/TT/CC/C  
rs9315897BD0.8558364623105 267898 14.4%
 CT0.6845369645930.32112818350.539912.6%
    CT C/CC/TT/T  
rs35776153BDN/A3657300 36500 0.0%
 CTN/A37074001.0000370001.00000.0%
    TC T/TT/CC/C  
rs17646069BD0.547636563496 274865 13.2%
 CT0.08583706321080.450827484120.263014.6%
DFNB31    TG T/TT/GG/G  
rs2274158BD0.8790362283441 56171135 39.1%
 CT0.52303693054330.3937661731300.652541.3%
    AG A/AA/GG/G  
rs2274159BD0.4654364336392 81174109 46.2%
 CT0.90393693473910.7536811851030.790647.0%
    AG A/AA/GG/G  
rs942519BD0.5699364320408 73174117 44.0%
 CT0.81103703294110.8747721851130.836344.5%
    TC T/TT/CC/C  
rs4978584BD0.9308365290440 58174133 39.7%
 CT0.46753703024380.7097651721330.836340.8%
    CG C/CC/GG/G  
rs35003670BDN/A3627240 36200 0.0%
 CTN/A37074001.0000370001.00000.0%
    AG A/AA/GG/G  
rs942518BP0.2645363563163 22211922 22.5%
 CT0.45263705611790.4587210141190.308724.2%
SORCS2    TC T/TT/CC/C  
rs4411993BD0.8482360227493 35157168 31.5%
 CT0.28953672215130.5703291631750.696930.1%
    GC G/GG/CC/C  
rs11736984BP0.2064364192536 30132202 26.4%
 CT0.16963691845540.5499181482030.145724.9%
    TC T/TT/CC/C  
rs10937823BD0.0500363200526 35130198 27.5%
 CT0.06813701875530.3432171532000.0175*25.3%
    GC G/GG/CC/C  
rs13139074BP0.4360366206526 32142192 28.1%
 CT0.39933671885460.2891211462000.292925.6%
    AG A/AA/GG/G  
rs11734984BP0.4737363206520 32142189 28.4%
 CT0.82563702035370.7268271491940.750127.4%
    TC T/TT/CC/C  
rs12645507BP0.0801365449281 14615762 38.5%
 CT0.93723684722640.3049151170470.268235.9%
    GA G/GG/AA/A  
rs34058821BDN/A3667320 36600 0.0%
 CTN/A36973801.0000369001.00000.0%
    TC T/TT/CC/C  
rs16840892BD0.0254364254474 54146164 34.9%
 CT0.24373702554850.8695491571640.748034.5%

In LD structure analysis, two LD blocks in DGKH and one block in each of DFNB31 and SORCS2 were detected. In haplotype analysis, a total of 14 haplotypes whose frequencies were more than 1% were estimated, but no significant haplotypic associations with BD were detected (data not shown).

The results of the meta-analysis are shown in Table 2. We found a nominal association in rs9315897 (P = 0.039), which was not significantly associated with BD in Zeng et al. No sample heterogeneity was observed in the four investigated SNP.

Table 2.  Result of the meta-analysis of four SNP in DGKH
SNP, minor allelePresent studyZeng et al.21Meta-analysis
MAFPOR
(95%CI)
MAFPOR
(95%CI)
POR
(95%CI)
Cases
(n = 366)
Controls
(n = 370)
Cases
(n = 1139)
Controls
(n = 1138)
  • *

    P < 0.05.

  • Minor alleles in the Japanese control samples.

  • Test for heterogeneity: I2 = 0%.

  • CI, confidence interval; MAF, minor allele frequency; OR, odds ratio; SNP, single-nucleotide polymorphism.

rs9315885, T0.4570.4540.91661.010.4930.4820.4531.040.4781.04
(0.83, 1.24)(0.93, 1.17)(0.94, 1.15)
rs1012053, A0.4930.4960.91690.990.5270.5070.1911.080.2581.06
(0.96, 1.17)
(0.97, 1.22)
(0.81, 1.21)
rs1170191, C0.5210.4950.31901.110.5280.5080.1791.080.0931.09
(0.90, 1.37)(0.97, 1.22)(0.99, 1.21)
rs9315897, C0.1440.1260.32111.170.1110.0950.0891.190.039*1.18
(0.87, 1.58)(0.98, 1.44)(1.01, 1.39)

DISCUSSION

In this study, we performed association analyses of DGKH, DFNB31 and SORCS2 with BD in Japanese case–control samples. We also conducted a meta-analysis of four SNP in DGKH using the data from the present Japanese cohort and previously reported Han-Chinese cohort.

The results of association analyses in the Japanese cohort were largely inconsistent with the initial study.2 Among GWAS, while partly common associations of DFNB31 (in Sklar et al.4 and WTCCC6, minimum P-value was 8.8 × 10−6) and SORCS2 (in Smith et al.,5 minimum P-value was 0.009) were found, largely inconsistent results for these genes were also reported. From the results of GWAS for BD and other complex traits, the effect of a single common variant has been recognized to be relatively low (odds ratio [OR] less than 1.5, often 1.1–1.227), with some exceptional genes containing common high-risk markers, such as APOE for Alzheimer's disease and CFH and ARMS2/HTRA1 for age-related macular degeneration. In this study, statistical powers of the Japanese cohort calculated with allele frequencies of our Japanese control samples and OR 1.3 for an alpha level of 0.05 or 0.002 (0.05 divided by the numbers of tested SNP) were 71.7 or 28.9% for DGKH, 61.3 or 19.9% for DFNB31, and 56.8 or 16.9% for SORCS2 (calculated with the Genetic Power Calculator,28http://statgen.iop.kcl.ac.uk/gpc/). Therefore, the sample number was not sufficient to detect risk-conferring markers with small effects. As a reason for inconsistent results, the adverse effect of population stratification should also be considered. In this study, all participants were recruited in the central area of Japan. Previous studies reported no substantial stratification in this population.17,29 In particular, samples in Hattori et al.17 were part of the samples used for this study, indicating that false positive results due to the effect of population stratification could be avoided.

In the meta-analysis, we found a nominal but significant association between rs9315897 in DGKH and BD. The total number of samples used for meta-analysis was larger than 3000. Thus, this result was more reliable than the result solely from the Japanese cohort. Although the association was not strong as was reported in Baum et al., and did not overcome a correction for multiple testing, this result may indicate this SNP as a risk marker common across ethnicities.

In conclusion, we found a nominal association between a polymorphism in DGKH and BD in the meta-analysis using more than 3000 samples from East Asia. However, the association was not strong and further investigations are required to obtain a conclusive result. For polymorphisms in SORCS2 and DFNB31, the statistical power obtained solely from our Japanese samples was apparently insufficient to detect risk markers with weak effect. However, this data also have a particular meaning, because they could be utilized in future meta-analysis.

ACKNOWLEDGMENTS

This study was supported in part by Health and Labour Sciences Research Grants (Research on Psychiatric and Neurological Diseases and Mental Health [2006–012] Neurophysiology, neuroimaging, and molecular biology of bipolar disorders).

Ancillary