Association and linkage analysis of RGS4 polymorphisms with schizophrenia and bipolar disorder in Brazil

Authors


H. Vallada, Instituto de Psiquiatria do Hospital das Clínicas, Faculdade de Medicina da Universidade de São Paulo, CEP: 05403-010, Sao Paulo-SP, Brazil. E-mail: hvallada@usp.br

Abstract

Linkage and association studies in five independently ascertained samples have suggested that polymorphisms of the regulator of G-protein signaling 4 (RGS4) may confer risk for schizophrenia (SCZ). Suggestive evidence for association with bipolar disorder (BD) has also been presented. However, the associated alleles and haplotypes have differed among the samples. Data from other independent samples may clarify the putative associations. Hence, we investigated an independent, ethnically diverse Brazilian population comprising patients with SCZ (n = 271) or BD1 (n = 306), who were contrasted with 576 community-based controls. Parents of 49 SCZ cases and 44 BD cases were available for transmission disequilibrium tests (TDTs). Four RGS4 single-nucleotide polymorphisms (SNPs) 1, 4, 7 and 18 putatively associated with SCZ were investigated. In the SCZ samples, significant case–control differences were not observed for individual SNPs or haplotypes, though the TDT suggested transmission distortion similar to that observed in the initial report. For the BD sample, case–control comparisons revealed no significant differences for individual SNPs, but an omnibus test suggested differences in the overall distribution of haplotypes bearing all four SNPs (SNP-EM Omnibus likelihood ratio test; P = 0.003). The TDT revealed over-transmission of allele A at SNP7 (P = 0.016), as well as haplotypes incorporating this allele. However, global tests incorporating all haplotypes yielded only suggestive trends for association (P = 0.19). In conclusion, association with SCZ was not detected in the present analyses. The failure to detect an association may be related to inadequate power or to confounds related to ethnic admixture. Suggestive associations with BD detected here require further investigation in a larger sample.

Regulators of G-protein signaling (RGS) are potent GTPase activators and thereby modulate G-protein function (De Vries et al. 2000; Druey et al. 1996; Zheng & Elston 1999). Given their pivotal role in receptor transduction processes, variation in RGS function may be related to the genesis of human diseases (Zhong & Neubig 2001). Recent postmortem analyses reveal that the expression of RGS4 is reduced across the cerebral cortex among individuals with schizophrenia (SCZ), in comparison with matched controls (Mirnics et al. 2001). The changes are likely to be specific for RGS4, as case–control differences were not observed for 10 other RGS proteins and 274 other genes involved in G-protein signaling. Intriguingly, RGS4 is localized to chromosome 1q21-q22, a region previously linked to SCZ (Brzustowicz et al. 2000; Gurling et al. 2001). Post-mortem analyses did not reveal expression differences for other genes localized to this region, further highlighting the specificity in relation to RGS4 (Mirnics et al. 2001). These converging lines of evidence support a role for RGS4 in the pathogenesis of SCZ (Harrison & Owen 2003).

To evaluate the role of RGS4 in SCZ genesis, we initially conducted linkage and association studies using 13 RGS4 single-nucleotide polymorphisms (SNPs), identified initially by screening individuals of Caucasian and African-American ethnicity. The SNPs were assayed among three samples ascertained independently at Pittsburgh, USA, New Delhi, India, and by the National Institute of Mental Health (NIMH) Collaborative Genetics Initiative (Chowdari et al. 2002). Family-based association analyses using the transmission disequilibirum test (TDT) revealed transmission distortion for individual SNPs as well as haplotypes encompassing four non-coding SNPs (named SNPs 1, 4, 7 and 18) in both the Pittsburgh (n = 93 trios) and NIMH (n = 39 trios) samples. Haplotypes bearing the G allele at SNPs 1, 4, 7 and 18 were most likely to be transmitted to patients from heterozygous parents in the Pittsburgh sample (G-G-G-G haplotype). Two samples of unrelated Caucasian controls were available for the Pittsburgh cohort (n = 85 and 89, respectively). No significant case–control differences were observed. Unlike the Pittsburgh sample, the haplotype with increased transmission in the NIMH sample incorporated alleles A-T-A-A at these SNPs. Trends for transmission distortion were also observed in the Indian sample (n = 269 trios), with the A-T-A-A haplotype being most likely to be over-transmitted to patients.

The TDT results, though suggestive of a genetic association at RGS4, could also be explained by stochastic variation or by other phenomena such as meiotic drive that are unrelated to SCZ. Two recent studies involving cases and unrelated controls of Caucasian ethnicity also revealed associations with these SNPs, suggesting a role for RGS4 in SCZ susceptibility. Among 709 cases and 710 controls from Wales, UK, significant differences were detected at SNPs 4 and 18. Associations were also detected for haplotypes incorporating SNPs 1 and 4 (Williams et al. 2004). These haplotypes correspond to the A-T-A-A haplotype that was over-transmitted in the NIMH sample. Independent analysis of SNPs 1, 4, 7 and 18 among patients with SCZ, schizoaffective disorder (SZAD) (n = 249), and unrelated controls (n = 231) from Dublin, Ireland, did not reveal significant associations (Morris et al. 2004). Among the subgroups of patients with SCZ (n = 196), but not those with SZAD, associations at conventional levels of statistical significance (P < 0.05) were detected at SNPs 1 and 7; however, the effect sizes were modest (OR = 1.32). In this sample, the associated haplotype corresponded to the G-G-G-G haplotype over-transmitted in the Pittsburgh sample.

If an association is indeed present with RGS4 alleles, the variation in associated alleles/haplotypes among the different samples needs to be explained. The G-G-G-G and the A-T-A-A haplotypes (or corresponding alleles at these SNPs) together account for approximately 75–85% of Caucasian control samples in the published studies. It is conceivable that an unidentified, ancient mutation occurs against the background of both these common haplotypes. It is also possible that chance could account for one or both sets of associations, particularly given the alleleic heterogeneity and modest effect sizes reported. Hence, replicate analyses are warranted.

The initial association studies also suggested transmission distortion among patients with bipolar disorder (BD) (n = 101) (Chowdari et al. 2002). Among the BD patients and their available parents, trends for global transmission distortion of haplotypes were observed using SNPs 1, 4, 7 and 18 (transmit program; χ2 = 16.99, P = 0.108, df = 11). Significant excess transmission of individual SNPs or haplotypes was not observed, although some SNPs and haplotypes were significantly under-transmitted. No significant case–control differences were observed. These suggestive findings merit further investigation, particularly in light of recent suggestions that RGS proteins may be implicated in the genesis of BD (Gould & Manji 2002).

In view of these reports, we undertook the present study in a Brazilian sample composed of patients with SCZ and BD as well as their parents. A set of unrelated controls enabled simultaneous case–control analyses.

Materials and methods

Clinical data

Patients

Participants were recruited from inpatient and outpatient services at the Institute of Psychiatry of the Hospital das Clínicas, University of São Paulo Medical School. Individuals diagnosed with SCZ or BD participated in the study, but individuals with SZAD were not included. The patients were interviewed clinically by two independent psychiatrists. The patients only participated in the study if the diagnostic of SCZ or BD was consensus according to the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria. The SCZ sample comprised 271 individuals. Parents of 49 SCZ probands were also available; this group comprised the ‘case–parent trios’. The BD sample consisted of 306 individuals, of whom 44 probands participated as case–parent trios.

Parents of patients

When feasible, participation by parents of the probands was sought. Parents were not included if they had been diagnosed with a psychiatric illness. Parents were also submitted to clinical interviews with two independent psychiatrists. If a parent was diagnosed with a psychiatric disorder, for at least one psychiatrist, he/she was excluded from the study. From those parents who agreed to participate in the study and were submitted to the clinical interviews, two mothers and one father (mothers: two unipolar depressions; father: BD) were excluded from the SCZ sample. In the BD, two mothers and one father (mothers: two unipolar depressions; father: unspecified psychosis) were also excluded due to a history of psychiatric disorder. The rest of the parents of our unitary patients were not able to be contacted.

Both parents were available for 49 patients with SCZ and 44 patients with BD.

Controls

Adults participating in the Blood Donation Service at the Hospital das Clínicas, University of São Paulo School of Medicine, were asked to participate. Such individuals undergo a physical examination, but psychiatric evaluations were not completed (n = 576). Ethnicity for cases and controls was determined by two independent raters and was based on skin pigmentation in the medial part of the arm, hair color and texture and the shape of the nose and lips. The participants were classified as Caucasian, Black, Mulatto, Asiatic, Brazilian Indian or Unknown (Table 1).

Table 1.  Distribution of ethnic subgroups
EthnicitySCZ patientsBD patientsControls
  1. BD, bipolar disorder; SCZ, schizophrenia.

Caucasian53101327
Black3640
Mulatto2143198
Asiatic7310
Brazilian Indian011
Unknown1871520
Total271306576

All participants provided written informed consent. Ethical approval for the study was obtained from the Ethics Committee at the Hospital das Clínicas, University of São Paulo Medical School (CAPPesq).

Laboratory

Genomic DNA was extracted from venous blood samples using the phenol–chloroform method. SNPs 1, 4 and 7 are located in the RGS4 putative promoter region while SNP18 is located in intron 1 of the gene. SNPs 1, 4, 7 and 18 were assayed using polymerase chain reaction (PCR)-based protocols as described (Chowdari et al. 2002) (http://www.pitt.edu/AFShome/n/i/nimga/public/html/research/RGS4). Briefly, PCR assays included primers (5 pmol) with 200 µM dNTP, 1.5 mM MgCl2, 0.5 U AmpliTaq Gold Polymerase (PE Biosystems, Foster City, CA), ×1 buffer and 60 ng of genomic DNA in 10 µl reactions. The PCR conditions were 95 °C for 10 min followed by 35 cycles (94 °C for 45 seconds, 60 °C for 45 seconds and 72 °C for 1 min) and a final extension at 72 °C for 7 min. Samples were genotyped using a fluorescence polarization-based assay (Chen & Kwok 1997). All genotypes were read independently by two investigators. In case of disagreement, assays were repeated.

Statistical analysis

We used pedcheck software (O'Connell & Weeks 1998) to test for Mendelian inconsistencies and the web-based genepop program to test for Hardy–Weinberg equilibrium (http://www.wbiomed.curtin.edu.au/genepop). We used the snpem software, which utilizes the expectation maximization algorithm to estimate haplotype frequencies. Further, the software uses permutation-based methods to estimate the statistical significance of differences in haplotype frequencies among cases and controls (Fallin et al. 2001). We employed the TDT (genehunter software) to assess transmission distortions from parents to affected probands for single SNPs and haplotypes (Kruglyak et al. 1996). We also used transmit software for global tests of association involving multiple haplotypes (Clayton 1999; Clayton & Jones 1999). Linkage disequilibrium (LD) was evaluated for all pairwise SNP combinations using published software (Zhao et al. 2000). An α-level of 0.05 was used to assess statistical significance in all analyses.

Results

Schizophrenia

Case–control comparisons

The distribution of ethnicity among cases is listed in Table 1. The distribution of genotypes for all polymorphisms was in Hardy–Weinberg equilibrium among the cases as well as the controls. No significant case–control differences were found with respect to genotypes for any of the individual SNPs or the estimated haplotype frequencies (Tables 2 and 3).

Table 2.  Distribution of genotypes and allele frequencies
 Genotype counts Allele frequencies
  1. SNP, single-nucleotide polymorphism.

SNP1GGGAAAGA
 sz_case85140450.5740.426
 bpd_case92135590.5580.442
 control1792931010.5680.432
SNP4TTTGGGTG
 sz_case86142400.5860.414
 bpd_case99144570.5700.430
 control188299830.5920.408
SNP7GGGAAAGA
 sz_case83136440.5740.426
 bpd_case94151550.5650.435
 control172292970.5670.433
SNP18AAAGGGAG
 sz_case79139490.5560.444
 bpd_case90145640.5430.457
 control176305890.5760.424
Table 3.  Estimated commonest haplotype frequencies among cases and unrelated controls
Haplotypes (SNPs 1-4-7-18)SCZ casesBD1 casesControls
  1. BD, bipolar disorder; SCZ, schizophrenia; SNP, single-nucleotide polymorphism.

A-T-A-A0.4020.4060.416
G-T-G-A0.1510.1170.156
G-G-G-G0.4100.4270.402

Family-based analysis

When the TDT was conducted using individual SNPs, a trend for increased transmission was observed for allele G at SNP18 (transmitted: 26 alleles; not transmitted: 14 alleles; χ2 = 3.60, P = 0.058, df = 1) (Table 4). Following ‘sliding window’ analysis of contiguous SNPs, a nominally significant transmission distortion of a two-SNP haplotype composed of the G-G alleles for SNPs 7 and 18 was observed (transmitted: 19 alleles; not transmitted: eight alleles; χ2 = 4.48, P = 0.03, df = 1). A trend for over-transmission of the G-G-G-G haplotype was also noted when all four SNPs were analyzed (transmitted: 16 alleles; not transmitted: eight alleles; χ2 = 2.67, P = 0.1, df = 1). These analyses entailed comparison of transmission of individual haplotypes with respect to all others. To enable an understanding of overall transmission distortion, a global test of association for all haplotypes encompassing SNPs 1, 4, 7 and 18 was conducted using the transmit program. No significant global transmission distortion was observed when haplotypes encompassing all four SNPs were analyzed (χ2 = 5.13, P = 0.40, df = 5).

Table 4.  Transmission disequilibrium test analysis of RGS4 single-nucleotide polymorphisms (SNPs) investigated
 SchizophreniaBipolar disorder
SNPsAllelesT/NTχ2P-valueAllelesT/NTχ2P-value
 1G24/171.200.274A23/132.310.128
 4G22/132.310.128T23/161.260.262
 7G24/171.200.274A21/85.830.015
18G26/143.600.057A25/161.980.159
1-4-7-18G-G-G-G16/82.670.102ATAA14/37.120.007

Bipolar disorder

Case–control comparisons

The ethnic distribution among cases is listed in Table 1. The distribution of genotypes for all polymorphisms was in Hardy–Weinberg equilibrium among the cases as well as the controls. No significant case–control differences were found with respect to genotypes for any of the individual SNPs (Table 2). Using snpem, frequencies of haplotypes incorporating SNPs 1, 4, 7 and 18 were estimated and compared among the cases and controls. Significant differences were observed overall (SNP-EM Omnibus likelihood ratio test; P = 0.003). The associated haplotypes differed from those observed for the US schizophrenia TDT analyses (Chowdari et al. 2002) (Table 4).

Family-based analysis

There was a significant increase in the transmission of A allele at SNP7 to patients (transmitted: 21 alleles; not transmitted: eight alleles, χ2 = 5.83, P = 0.016) (Table 4). Consistent with this result, there was significant over-transmission of the haplotypes bearing alleles A-T-A-A at SNPs 1, 4, 7 and 18, respectively (transmitted: 14 alleles; not transmitted: three alleles, χ2 = 14.72, P = 0.008, df = 1). Over-transmission of two- or three-SNP haplotypes bearing allele A at SNP7 was also observed (results not shown). Only a trend for transmission distortion was observed when a global test was conducted for the four-SNP haplotypes using the transmit program (P = 0.19).

Linkage disequilibrium analyses

Inspection of estimated haplotype frequencies for the cases and controls suggests significant LD among the SNPs, consistent with their relative physical proximity (the physical distance between SNP1 and SNP18 is estimated at 7.04 kb). In agreement with the results from the Caucasian samples, haplotypes G-G-G-G and A-T-A-A for all four SNPs accounted for over 81% of all estimated haplotypes among the controls. Two-point LD analysis showed that D′-values exceeded 0.85 for all combinations of SNPs among the cases and controls (Table 5).

Table 5.  Pairwise linkage disequilibrium calculations among cases, parents and unrelated controls
SNPsCasesParentsControls
  1. SNP, single-nucleotide polymorphism.

1-41.0000.9831.000
1-70.9920.9880.996
1-180.8610.8540.929
4-70.9881.0001.000
4-180.9800.9500.981
7-180.8850.8880.928

Discussion

The present study was designed to test a putative association between SNPs at RGS4 and SCZ, as well as BD. A large sample of SCZ and BD patients as well as community-based controls facilitated association analyses. A smaller sample of case–parent trios also enabled the TDT, which is not prone to artifacts introduced by population substructure following ethnic admixture (Spielman & Ewens 1996).

Significant evidence for an association with SCZ was not detected when the Brazilian cases were compared with community-based controls. These results differ from five published samples, each of which provided evidence for association. There are several possible reasons for the failure to detect association in the present sample, in addition to the null hypothesis of an absence of a ‘true’ genetic association. A prime concern is statistical power. Analyses conducted by Williams et al. (2004) suggested that a sample consisting of 709 cases and 710 controls had only power of approximately 0.7 to detect their modest effect size for SNP4 in a sample of all Caucasians (OR ≈ 1.2). The present sample, which includes a substantial number of individuals of Caucasian ethnicity lacks adequate power if the UK power estimates are applied. Power to detect association using case–control samples is also reduced in the presence of ethnic admixture, which almost certainly pertains to the present sample (McKeigue et al. 2000; Spielman & Ewens 1996). Association analyses were not conducted separately among the Caucasian individuals in this sample due to the anticipated loss of power in the reduced sample and due to the presence of a significant number of individuals with unknown ethnicity.

Two estimates for power are available for the TDT analyses. The analysis using families recruited by the NIMH suggest that 69 parent–child trios would be required to replicate an association with SCZ with 80% power and a critical P-value of 0.05 (Chowdari et al. 2002). The results from analysis of our Pittsburgh sample, however, suggested a larger sample of 130 trios would be required (Chowdari et al. 2002). Thus, the present sample had insufficient power for replication using either estimate. Nevertheless, a trend for transmission distortion was detected using the TDT. It is also instructive to note that the haplotype with a trend for increased transmission is consistent with the G-G-G-G haplotype over-transmitted in the Pittsburgh sample and the haplotype associated with differences between SCZ cases and controls in the Dublin sample (Chowdari et al. 2002; Morris et al. 2004). The need for the present analyses may be questioned, in view of the relatively low power of the present sample. It should be noted that the present sample is among the largest available for such analyses. Though lacking power, the present results could usefully contribute to meta-analyses designed to test the putative association and to resolve the question of the haplotype(s) that may confer susceptibility for SCZ. In particular, they may enable a better understanding of the evolution of the haplotypes with the putative associations (Seltman et al. 2003).

Analysis of the BD sample also yielded intriguing results, though definitive conclusions can only be drawn using a larger sample. While case–control analyses did not reveal associations with individual SNPs, comparison of estimated haplotype frequencies using SNPs 1, 4, 7 and 18 support an association at this locus. The TDT also provides suggestive evidence for an association in the BD sample. However, the associated haplotypes from the TDT differ from those detected using the case–control samples. The discordance may reflect inherent statistical difficulties in detecting associated haplotype(s) when phase is uncertain. The TDT analysis revealed over-transmission of A-T-A-A, the same haplotype that was associated with SCZ in the NIMH sample (Chowdari et al. 2002). The present analyses raise the intriguing possibility that the G-G-G-G haplotype confers susceptibility for SCZ, whereas the A-T-A-A haplotype increases the risk for BD.

In conclusion, association analyses using cases and community-based controls from Brazil did not support a putative association between RGS4 and SCZ, though separate analyses using family-based controls provided suggestive evidence. Parallel investigations using BD probands also provided suggestive evidence for an association between RGS4 polymorphisms and BD. All these analyses may have been hampered by relatively low power to detect associations. Nevertheless, these data will usefully contribute to planned meta-analyses.

Acknowledgments

We thank the participants for their help with this study. Supported in part by grants from the National Institute of Mental Health (MH63420, MH63480, MH56242 and MH66263 to VLN).

Ancillary