|Home | About | Journals | Submit | Contact Us | Français|
Dysbindin (DTNBP1) has been identified as a susceptibility gene for schizophrenia (SZ) through a positional approach. However, a variety of single nucleotide polymorphisms (SNPs) and haplotypes, in different parts of the gene, have been reported to be associated in different samples, and a precise molecular mechanism of disease remains to be defined. We have performed an association study with two well-characterized family samples not previously investigated at the DTNBP1 locus.
We examined 646 subjects in 136 families with SZ, largely of European ancestry (EA), genotyping 26 SNPs in DTNBP1.
Three correlated markers (rs875462, rs760666, and rs7758659) at the 3′ region of DTNBP1 showed evidence for association to SZ (p = 0.004), observed in both the EA (p = 0.031) and the African American (AA) subset (p = 0.045) with the same over-transmitted allele. The most significant haplotype in our study was rs7758659-rs3213207 (global p = 0.0015), with rs3213207 being the most frequently reported associated marker in previous studies. A non-conservative missense variant (Pro272Ser) in the 3′ region of DTNBP1 that may impair DTNBP1 function was more common in SZ probands (8.2%) than in founders (5%) and in dbSNP (2.1%), but did not reach statistical significance.
Our results provide evidence for an association of SZ with SNPs at the 3′ end of DTNBP1 in the samples studied.
A locus for SZ in chromosome 6p24-p22 has been supported by some [1,2,3,4] but not all linkage studies [5,6,7]. Straub et al. reported four SNPs from exons 1–5 of the 140 kb DTNBP1 (at 6p22.3), and several 3-marker haplotypes (p = 0.008–0.0001), to be associated with SZ . In vitro functional studies of DTNBP1 have suggested that DTNBP1 may influence exocytotic glutamate release presynaptically . Patients with SZ were reported to have decreased DTNBP1 in the glutamatergic terminal of the hippocampus and dorsolateral prefrontal cortex , and risk haplotypes for SZ were also shown to be associated with reduced DTNBP1 expression in human cerebral cortex . Convergent effects of several putative SZ susceptibility genes (including DTNBP1) upon glutamate synapses have been hypothesized as a possible model for the pathogenesis of SZ .
So far (November 2006), 19 association studies of DTNBP1 and SZ have been published, of 25 independent samples in a variety of populations (table (table1)1) [8, 9,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29]. Although a number of studies showed supportive evidence for the association between DTNBP1 and SZ, these reports also showed considerable disagreement between the associated markers, alleles, and haplotypes. It has been suggested that the inconsistencies across studies were population dependent . However, a recent analysis of the reported associated haplotypes concluded that the conflicting results across EA studies could not be attributed to population differences because the genetic architecture at DTNBP1 locus is similar in EA samples .
Another explanation of the inconsistent association reports (table (table1)1) is that multiple variants in DTNBP1 might confer susceptibility to SZ [30, 32]. Most replication studies tested predominantly markers in the 5′ region of DTNBP1 (table (table1);1); testing SZ association using a set of markers that also has good coverage at the 3′ end of the gene may help to explain or unify the divergent evidence for association. Furthermore, testing coding variants, especially missense SNPs, may also add valuable information. In a mutation scan of exons of DTNBP1 in 50 Chinese Han SZ patients and of the promoter region in 94 such patients, no mutations were detected . Another mutation scan of exons and promoters in 14 familial EA SZ patients found 2 untranslated exonic SNPs, along with 9 SNPs in putative promoter regions, but no mutations in the coding sequences; these 11 SNPs were not associated (nominal p >0.05) with SZ in 552 cases and 552 controls . When all 15 alternatively spliced mRNA transcripts from AceView (www.ncbi.nlm.nih.gov/IEB/Research/Acembly/; update November 2004) are considered (fig. (fig.1),1), 7 missense SNPs can be defined from a total of 10 coding SNPs available in a search of dbSNP (www.ncbi.nlm.nih.gov/projects/SNP/; build 123) and the literature (table (table2).2). None of these coding variants have so far been tested in previous association studies.
We present here a family based association study of DTNBP1 and SZ in two independently collected US SZ data sets. We have tested 26 SNPs, including 9 previously reported associated SNPs, 10 other SNPs to increase map coverage (particularly at the 3′ portion of the gene), and 7 additional coding SNPs. We found evidence for association with SZ at the 3′ end of DTNBP1.
Our total sample is comprised of 136 families ascertained from all over the US of EA (72%) and AA (18%) descent (online suppl. table 1, www.karger.com/doi10.1159/000101961). This sample was approximately equally derived from two collections, the NIMH-IRP (Intramural Research Program, also known as the Clinical Neurogenetics or CNG sample) [34, 35] and the NIMH-GI (Genetics Initiative – Part I) . This combined set contained 319 genotyped individuals diagnosed as SZ (n = 273) or schizoaffective disorder (n = 46) via the criteria of the Diagnostic and Statistical Manual of Mental Disorders, third edition, revised (DSM-IIIR) . Besides the affected probands (n = 136), most of the additional genotyped individuals were siblings (n = 258), parents (n = 207), or other relatives (n = 45) of the probands; genotyping these additional individuals, 183 of whom were themselves affected, made the sample more informative for the family-based association testing. Our group and others have used these families in previous linkage studies, although no suggestive or significant linkage evidence was found for 6p [34, 38, 39]. There is no overlap with any other US sample used in previous association studies with DTNBP1 and SZ. The Institutional Review Board of the Evanston Northwestern Healthcare Research Institute approved the study.
As estimated by TDT-PC (transmission disequilibrium test – power calculator) software , our sample has about 87% power to detect a locus with a relative risk (RR) of 2 at a significance level of 5%; but power drops to 44% for an RR of 1.4 .
A total of 26 SNPs in DTNBP1 were genotyped by either TaqMan or SNPlex. Out of 10 previously reported as associated SNPs (table (table1),1), 9 SNPs were successfully genotyped (rs742105, rs2619539, rs3213207, rs1011313, rs2619528, rs2619522, rs760761, rs1018381, and rs2619538); rs2005976 was not selected since it is tagged by the genotyped rs2619522 with r2 = 0.94 . To increase map coverage, we typed 3 previously known (but not associated) SNPs (rs742106, rs760666, and rs909706), and 7 other previously untested SNPs (rs742102, rs875462, rs10949305, rs2743553, rs7758659, rs2743865, and rs742208). We also genotyped 7 out of 10 coding SNPs (3 failed assay designs; SNP-flanking DNA sequences were drawn from the UCSC Genome draft, genome.ucsc.edu; May 2004 freeze) (table (table2).2). Thirteen SNPs (rs17470454, rs1094305, rs742105, rs7758659, rs1018381, rs16876589, rs6926401, rs16876573, rs16876571, rs16876569, A874G, rs760761, and rs2619538) were genotyped with TaqMan on an Applied Biosystems (ABI) Prism 7900 instrument. Thirteen other SNPs (rs742102, rs742106, rs875462, rs2743553, rs760666, rs2619539, rs2743865, rs3213207, rs1011313, rs2619528, rs2619522, rs909706, and rs742208) were genotyped with SNPlex following standard procedures.
Genotypes were checked for Mendelian inconsistencies and unlikely recombinants via MERLIN . All Mendelian inconsistencies were removed by zeroing out the genotypes for the involved individuals in that family for that SNP. For the genotypes that were flagged as ‘unlikely’ recombinants (p < 0.01), we manually checked the raw genotyping traces to search for questionable genotypes (e.g., obviously external to the genotype clusters, very low signal intensity, etc.) and questionable genotypes were zeroed. The average genotyping completion rate was 98.4% (95.7–99.4%). The total genotyping error rate was 0.54%, including 15 Mendelian errors and 55 unlikely recombinations out of 12,911 non-zero genotypes. Detected error rates for individual SNPs ranged from 0.16 to 1.30%. All the genotyping errors were blanked. Given our high genotyping completion rate and low genotyping error rates, we did not attempt second-pass genotyping for these zeroed errors. All the markers were in Hardy-Weinberg equilibrium.
LD between SNPs was estimated with Haploview 3.0  using the genotypes from unrelated founders. The standard D’, LOD, and r2 were derived in the EA and AA subsets separately. HapMap genotype data were downloaded from HapMap (www.hapmap.org).
For association analyses, we used FBAT v1.5.5 [44, 45]. Alleles and haplotypes were tested for association if there were at least 5 informative families; in our data this corresponds to testing alleles and haplotypes with frequencies over 3%. We limited the number of multi-locus systems tested by using a stepwise procedure, and limited the number of multi-locus tests to the combinations including the SNP with highest single Z score value.
Initially, 20 SNPs in −140 kb of the genomic sequence of DTNBP1 were analyzed (all the missense SNPs, except for rs17470454, were analyzed separately because of their low minor allele frequency, MAF). Consistent with previous studies [8, 13, 17], our LD analyses in EA founders showed only one long LD block (defined by D′ >0.80 and LOD >2, about 122 kb from rs875462 to rs909706) spanning almost the whole DTNBP1 gene. The r2 estimates were variable (fig. (fig.2);2); grouping SNPs with r2 >0.8  yielded 7 bins (A–G)  of high LD, where any SNP can be a proxy (r2 >0.8) of all remaining SNPs within the bin (note that one SNP that is not tagged by other SNPs constitutes a one SNP bin). The bins often overlapped (fig. (fig.2b).2b). Bin B spans approximately 110 kb, encompassing rs1094305, rs2743553, rs2743865, and rs1018381. In total, 12 out of the 20 analyzed SNPs were tag SNPs.
The LD patterns and allele frequencies of the tested markers were similar in EA founders and in AA founders, although LD was weaker in AAs (data not shown). We have performed association analyses in the EA and AA subsets separately, in addition to a combined analysis. Single-locus FBAT results for the 12 tag SNPs (r2 <0.8) are summarized in table table3.3. We found evidence for association with allele G of rs7758659 on the 3′ end of DTNBP1 in both the EA subset (p = 0.031) and the smaller AA subset (p = 0.045). Since the over-transmitted alleles for all the tested SNPs were the same in EA and AA, we also performed association tests in the combined sample. Only rs7758659 remained associated after Bonferroni correction (p = 0.004; p = 0.048 when corrected by 12, the number of tag SNPs). Because rs7758659 tags two other SNPs, rs875462 and rs760666 (bin A with r2 >0.94; fig. fig.2b),2b), the association is also present (and with similar significance) with each of these SNPs (online suppl. table 2).
We performed haplotypic analyses using the 12 tag SNPs, starting by analyzing all two-marker systems anchored with rs7758659 (table (table4).4). In the EA subset, 3 out of the 11 two-marker combinations gave global p values <0.05. The most significant two-marker system was rs7758659-rs2619522 (global p = 0.016). In the AA subset, only one two-marker system, rs7758659-rs3213207, showed nominal significance (global p = 0.02; the corresponding global p value in the EA families was 0.03). In the combined dataset, 9 of the 11 two-marker combinations gave global p values <0.05. The most significant two-marker haplotype was found with rs7758659-rs3213207 (global p = 0.0015; the only two-marker combination with higher significance than rs7758659, p = 0.004, by itself). rs3213207 (also known as P1635 ) is the most frequently reported associated SNP, albeit with the associated alleles being different across those studies (table (table1).1). Haplotypic analyses with a three-marker system anchored to rs7758659-rs3213207 did not increase significance (data not shown). We have also extended haplotypic analysis to 7 markers from rs875462 to rs909706 (tag SNPs: rs7758659, rs2619539, rs2743865, rs3213207, rs1011313, rs2619522, and rs909706; each representing one bin, A-G). Ninety-eight percent of the haplotypes were represented by only 6 common (frequency >5%) haplotypes. None of those 6 common haplotypes yielded a global p value or haplotypic p value <0.05, with haplotype G-C-C-A-A-T-A (frequency = 20%) having the lowest p value of 0.08.
Finally, we tested association with the coding SNPs in DTNBP1 (not included in haplotype analysis due to low MAF). None of the 7 genotyped coding SNPs showed association with SZ (p values >0.09; table table2)2) or co-segregated with disease. We also searched for aggregations of missense SNPs within families, but found no evidence of this. However, it is worth noting that the MAF of rs17470454 (C814T; Pro272Ser) was higher in founders in our samples (5%) than that in dbSNP (2%); interestingly, the MAF in EA probands was found to be even higher (8.2%) than in founders, though no transmission distortion was found (p = 0.53; table table22).
We have studied 26 SNPs in DTNBP1, including 6 missense variants. Our results support an association of SZ with three highly correlated markers in the 3′ region of DTNBP1 (rs875462, rs760666, and rs7758659). The association was observed in both EA and AA subsets with the same over-transmitted allele (G) at rs7758659. rs875462 and rs7758659 were previously untested, while rs760666 (or P1287 ) had been previously tested, but no evidence for association was reported (table (table1).1). Interestingly, rs7758659 is near the boundary of exon 7 and intron 7 (being 49 bp away from exon 7), suggesting the possibility that it might affect splicing.
The four initially reported associated markers and an associated haplotype that spans a promoter SNP, rs2619538 (SNP A) , are all located in the 5′ region of the DTNBP1. All other replication studies limited their tests to regions in the 5′ portion of the gene, as it has been speculated that the potential causative polymorphism(s) lie in the 5′ region of DTNBP1. With a more comprehensive set of markers than in previous studies, we did not observe evidence of association with SNPs located in the 5′ region of the gene; instead, we found evidence for association of SZ with SNPs located in the 3′ region of DTNBP1. However, it may not be possible to define the location of a true disease variant in a long region with strong LD (as in DTNBP1) . One could argue that rs875462, rs760666, and rs7758659 that were found associated to SZ in our study are, in fact, in LD with some other untested variants in the 5′ region of the gene. On the other hand, it is also conceivable that associated markers located in the 5′ region of the gene may actually reflect associations with causative loci in the 3′ region. Indeed, the LD pattern of DTNBP1 allows for this possibility (fig. (fig.2b).2b). For instance, rs1018381 in the 5′region of the gene has been reported to be associated with SZ [15, 20], but this SNP is highly correlated (r2 >0.90) with three other SNPs in bin B (fig. (fig.2b),2b), two of which (rs1094305 and rs2743553) are in the 3′ region of DTNBP1.
We have tested 7 DTNBP1 coding variants, but none of them showed association with SZ (table (table2).2). It is nonetheless worth noting that some of these missense SNPs are in LD with some previously reported associated SNPs: rs3213207 (in bin D) was found in our data moderately correlated (r2 = 0.52) to rs17470454 (C814T; Pro272Ser; not tested in any of the previous studies). The MAF of rs17470454 was higher in SZ probands than in founders and dbSNP, albeit not reaching statistical significance. The substitution of a proline for a serine at rs17470454 is predicted by PolyPhen (tux.embl-heidelberg.de/ramensky/index.shtml) to be a non-conservative change with potential for impairing DTNBP1 function . Therefore, rs17470454 might (together with other still unrecognized functional SNPs) underlie the previously observed association between DTNBP1 and SZ [8, 9, 13, 17, 18, 23].
In conclusion, our results provide evidence for an association of SNPs at the 3′ end of DTNBP1 and SZ in both EA and AA samples, but not at the 5′ region of DTNBP1. However, our sample has limited statistical power and some true loci might not have been detected. Furthermore, in the absence of a molecular hypothesis that can provide a clear mechanistic explanation, association tests should be considered ‘indirect’, and this severely constrains the statistical power. Though we genotyped a denser marker map than other studies (table (table1),1), the SNP set we have used only captures 76% of all the common SNPs of DTNBP1 (MAF >0.05) currently available in HapMap Phase II (release #20) at r2 >0.8 (data not shown). A total of 42 SNPs (based on all the known variants in DTNBP1) would be needed to capture all alleles at DTNBP1 locus . A critical experiment would require a large sample with sufficient statistical power and genotyping of a comprehensive and sufficiently dense SNP set.
Subjects and Phenotyping
FBAT Results for All 20 Markers of DTNBP1a
The National Institute of Mental Health (NIMH) Schizophrenia Genetics Initiative families, data and biomaterials were collected in three projects. From 1991–1997, the Principal Investigators and Co-Investigators were: Harvard University, Boston, Mass., USA, U01 MH46318, Ming T. Tsuang, MD, PhD, DSc, Stephen Faraone, PhD, and John Pepple, PhD; Washington University, St. Louis, Mo., USA, U01 MH46276, C. Robert Cloninger, MD, Theodore Reich, MD, and Dragan Svrakic, MD; Columbia University, New York, N.Y., USA, U01 MH46289, Charles Kaufmann, MD, Dolores Malaspina, MD, and Jill Harkavy Friedman, PhD. Research Career Development Awards (to J.D. and A.R.S.) at the Evanston Northwestern Healthcare Research Institute supported this work.
Supplementary information is available at Human Heredity's website.