|Home | About | Journals | Submit | Contact Us | Français|
The chromosomal region, 15q13-q14, including the α7 nicotinic acetylcholine receptor gene, CHRNA7, is a replicated region for schizophrenia. This study fine-mapped genes at 15q13-q14 to determine whether the association is unique to CHRNA7.
Family-based and case–control association studies were performed on Caucasian-non-Hispanic and African-American individuals from 120 families as well as 468 individual patients with schizophrenia and 144 well-characterized controls. Single-nucleotide polymorphism (SNP) markers were genotyped, and association analyses carried out for the outcomes of schizophrenia, smoking, and smoking in schizophrenia.
Three genes were associated with schizophrenia in both ethnic populations: TRPM1, KLF13, and RYR3. Two SNPs in CHRNA7 were associated with schizophrenia in African-Americans, and a second SNP in CHRNA7 was significant for an association with smoking and smoking in schizophrenia in Caucasians.
Results of these studies support association of the 15q13-q14 region with schizophrenia. The broad positive association suggests that more than one 15q gene may be contributing to the disorder, either in combination or through a regulatory mechanism. Psychiatr Genet 22:1–14 © 2012 Wolters Kluwer Health | Lippincott Williams & Wilkins.
Schizophrenia is a common mental illness with a large genetic component (Kendler 1988; Tsuang and Faraone, 1994; Freedman et al., 2001b; Owen et al., 2004). Although heritability is estimated at approximately 80% (Sullivan et al., 2003), the disorder is thought to be multigenic with a significant environmental component (Tsuang, 2000; Petronis et al., 2003; Howes et al., 2004). Significant or suggestive evidence of linkage to schizophrenia has implicated 18 chromosomal regions (Baron, 2001; Harrison and Owen, 2003). Twelve of these chromosomal regions have replicated linkage, include one or more polymorphisms associated with schizophrenia, and contain biologically plausible candidate genes showing altered expression in the disorder (Harrison and Weinberger, 2005). The 15q13-q14 locus is one such chromosomal region, with linkage replicated in pedigrees from the National Institute of Mental Health (NIMH) Schizophrenia Genetics Initiative (Kaufmann et al., 1998; Leonard etal., 1998; Freedman and Leonard, 2001; Freedman et al., 2001b), and in other sample populations (Coon et al., 1994; Riley et al., 2000; Stöber et al., 2000; Gejman et al., 2001; Liu et al., 2001; Tsuang et al., 2001; Xuet al., 2001a; Stöber et al., 2002; Fallin et al., 2004). Nominal association has also been found for CHRNA7 in several large studies (Sanders et al., 2008; Stefansson et al., 2008). The locus is, thus, linked to schizophrenia across ethnicities and in multiple independent studies.
Genome-wide nonparametric linkage (NPL) analysis first identified the 15q13-q14 locus through linkage to the P50 deficit, a dominantly inherited sensory gating deficit found in most patients with schizophrenia and half of their first degree relatives (Freedman et al., 1997; Waldo et al., 2000). The P50 deficit was linked (LOD = 5.3) to a microsatellite marker (D15S1360), located in intron 2 of the CHRNA7 gene. This same study found positive, but less strong linkage to schizophrenia at D15S1360 (LOD = 1.3) (Freedman et al., 1997). Sibpair analysis in the NIMH pedigrees revealed a significant excess of alleles identical by descent (P < 0.0024) (Leonard et al., 1998). In addition, a transmission disequilibrium test was used for an association study of the 15q13-q14 locus in parent–child trios in the same cohort. Significant genotype-wise disequilibrium (P < 0.007) was found at a microsatellite marker, D15S165, located within one megabase of the CHRNA7 gene (Freedman et al., 2001a). Biological evidence has firmly implicated the α7 receptor in a pathway regulating the P50 auditory evoked response. Nicotine transiently normalizes the P50 deficit in schizophrenics and their first-degree relatives, consistent with the involvement of a nicotinic receptor in the disease (Adler et al., 1993; Leonard et al., 2001). Animal models of the P50 deficit further support a role for the α7 receptor. Antagonists of the α7 receptor disrupt sensory processing in rats (Luntz-Leybman et al., 1992), and agonists of the α7 receptor normalize aberrant sensory processing in mice (Stevens et al., 1998). The effects of nicotine administration and α7 stimuli on patient and animal models of the P50 deficit supports the hypothesis that smoking may be a form of ‘self-medication' for schizophrenics (Adler et al., 1998; Kumari and Postma, 2005; Leonard et al., 2007). This hypothesis is reinforced by the prevalence of smoking in schizophrenics, which is almost three times as high as in the general population (De Leon et al., 1995; Diwan et al., 1998; Leonard et al., 2001; De Leon et al., 2002; De Leon and Diaz, 2005). A significant association of the D15S1360 microsatellite marker in intron 2 of CHRNA7 to smoking in schizophrenia was also reported (De Luca et al., 2004). Recently, several neuronal nicotinic acetylcholine receptor genes, including CHRNA2, CHRNA5, CHRNB3, and CHRNA7, have been associated with nicotine addiction or to heavy smoking (Berrettini et al., 2008; Bierut et al., 2008; Weiss et al., 2008; Furberg et al., 2010; Thorgeirsson et al., 2010).
Ligand binding with an α7 receptor antagonist [125I]-α-bungarotoxin, showed 50% fewer receptors in the postmortem hippocampus of patients with schizophrenia than in nonmentally ill controls (Freedman et al., 1995). These results were confirmed in cortex (Guan et al., 1999; Marutle et al., 2001) and in the reticular nucleus of the thalamus (Court et al., 1999). Decreased levels of α7 receptors suggest that cognition is likely to be impaired (Rezvani and Levin, 2001; Young et al., 2004; Sacco et al., 2005; Levin et al., 2006). The α7 receptor is the most promising molecular target for development of drugs for cognitive impairments in schizophrenia (Psychiatric News, 2006). Phase I and II trials of an α7 receptor partial agonist, 3-[2,4-dimethoxybenzylidene] anabaseine, resulted in improvement in both the P50 gating deficit and in attention in schizophrenic nonsmokers (Olincy et al., 2006; Freedman et al., 2008). Recently single nucleotide polymorphism (SNP) rs3087454 in the 5′ upstream regulatory region of CHRNA7 was associated with schizophrenia (Stephens et al., 2009) and also with improvement in functional MRI response to 3-[2,4-dimethoxybenzylidene] anabaseine (Tregellas et al., 2010).
The 15q13-q14 region is characterized by large copy number variations (CNV), including both deletions and duplications that include the CHRNA7 gene. Although the CNVs are rare, they are associated with schizophrenia (Stefansson et al., 2008; Stone et al., 2008). Genetic characterization of CHRNA7 revealed the gene to be partially duplicated (exons 5–10), with both loci mapping to the 15q13-q14 region (Gault et al., 1998; Riley et al., 2002). The duplicated gene, CHRFAM7A, is transcribed in both human brain and in the periphery, and is unique to humans (Locke et al., 2003). The CHRFAM7A gene product was recently found to be a dominant negative regulator of α7 receptor function (Araud et al., 2011; De Lucas-Cerrillo et al., 2011).
Duplicated sequences from the CHRFAM7A gene are greater than 99% homologous to sequences from exons 5–10 of the full-length locus and therefore mutation screening of both genes required the use of mRNA (Gault et al., 1998). Results of mutation screening of CHRNA7 and its partial duplication, CHRFAM7A, identified 33 polymorphisms located in the coding region and intron/exon borders of the genes (Gault et al., 2003). A 2bp deletion in exon 6 of the duplicated gene, CHRFAM7A, is associated with schizophrenia (Sinkus et al., 2009). The 2 bp deletion is found in an inverted copy of CHRFAM7A (Flomen et al., 2008), suggesting that the inversion itself may be pathogenic. In the full-length CHRNA7 gene, three polymorphisms were nonsynonymous and mapped to the full length CHRNA7 coding region. These polymorphisms, however, were rare and not associated with schizophrenia or the P50 gating deficit (Gault et al., 2003). Mutation screening also revealed a large number of mutations in the 5′ upstream regulatory region of CHRNA7, which is not duplicated. Functional analysis of proximal promoter mutations identified polymorphisms that decreased transcription in vitro. These functional promoter mutations were statistically more prevalent as a group in schizophrenics than in control participants (P = 0.007), and the presence of a promoter polymorphism was associated with the P50 gating deficit (P < 0.0001) (Leonard et al., 2002). In addition, we reported that a SNP rs3087454, located at position −1831 bp upstream of exon 1 in CHRNA7, is significantly associated with schizophrenia in African-American and Caucasian-non-Hispanic case–control sample populations after multiple testing correction (P = 0.0009, African-American; P = 0.013, Caucasian-non-Hispanic) (Stephens et al., 2009).
Because the CHRNA7 and CHRFAM7A genes show homology, SNP analysis in other genes in this region might confirm this gene cluster is important for schizophrenia. This study was designed to fine map genes in the 15q13-q14 linkage peak region to determine whether genetic variance in the CHRNA7 gene is associated with schizophrenia or if variation in flanking genes contributes to the disorder. An informative set of SNPs was selected in each of 16 genes in the region. Ten microsatellite markers were also queried in the region of interest. SNP genotype data were analyzed for association with schizophrenia in Caucasian-non-Hispanic and African-American participants. A portion of this cohort consisted of families from the NIMH Human Genetics Initiative for Schizophrenia which had been previously linked to the 15q13-14 region (Kaufmann et al., 1998; Leonard et al., 1998; Freedman and Leonard, 2001; Freedman et al., 2001b). Smoking history was considered as a secondary outcome.
Results indicate that three genes were associated with schizophrenia in both ethnic populations after correcting for multiple comparisons: TRPM1, KLF13, and RYR3. In addition, two SNPs in CHRNA7 were associated with schizophrenia in African-Americans. One SNP in CHRNA7 was significant for an association with smoking and smoking in schizophrenia in Caucasians. A SNP in RYR3 was associated with smoking in African-Americans. Thus, broad association with schizophrenia in the 15q13-14 region was confirmed.
Characteristics of study participants are shown in Tables 1 and and2.2. DNA samples for case–control association studies included 612 African-American and Caucasian-non-Hispanic participants collected in our laboratory and one randomly chosen individual with schizophrenia per family, from families within the NIMH Human Genetics Initiative for Schizophrenia. SAS version 9 (SAS Institute Inc., Cary, North Carolina, USA) was utilized for random selection from 14 trios, 34 sibling-pairs, and 25 multiplex families in the NIMH Caucasian-non-Hispanic sample and three trios, 15 sibling-pairs, and 28 multiplex families from African-Americans in the collection. Local participants were consented for genetic studies, utilizing a consent form approved by the Colorado Multiple Institutional Review Board. DNA samples contain no identifying information. Ethnicities of case–control participants were recorded from self-report or family interview. The Caucasian-non-Hispanic case–control sample comprised 363 patients with schizophrenia (307 adult onset and 56 childhood onset schizophrenics) and 99 well-characterized controls. The African-American case–control sample included 105 individuals with schizophrenia (103 adult onset and two childhood onset), and 45 controls. Adult case individuals were chosen based on a diagnosis of schizophrenia, utilizing the Structured Clinical Interview for Diagnostic and Statistical Manual of Disorders, 4th edition Axis I disorders (First et al., 1996b). Control participants were carefully selected with a screen for evidence of current or past psychosis using a Structured Clinical Interview for Diagnostic and Statistical Manual of Disorders, 4th edition Axis I disorders for nonpatients (First et al., 1996a). Childhood onset schizophrenia (COS) cases were diagnosed utilizing the K-SADS-PL (Kaufman et al., 1997). Smoking history was determined on local individuals, utilizing a detailed questionnaire (Breese et al., 1997; Breese et al., 2000).
For the family-based study, a total of 329 African-American (47 nuclear families) and Caucasian-non-Hispanic participants (73 nuclear families) from the NIMH Schizophrenia Genetics Initiative were chosen based on a diagnosis of schizophrenia. This cohort has previously shown positive LOD scores and association to markers at the 15q13-q14 locus in both African-American and Caucasian individuals (Kaufmann et al., 1998; Leonard et al., 1998; Freedman et al., 2001b). The structure of the NIMH families included sibpairs (approximately 53% of the sample), as well as trios (approximately 22%), and extended family members (approximately 25%). Detailed information on the NIMH family structure and diagnosis is available from the web site (http://www.nimhgenetics.org). Smoking history on the NIMH patients was determined using the Fagerstrom test for nicotine dependence (Fagerstrom and Schneider, 1989) as part of the Diagnostic Interview for Genetic Studies (Nurnberger et al., 1994).
To delineate a peak region for schizophrenia at the 15q13-q14 locus, NPL scores from published studies were plotted for microsatellite markers from positive linkage reports (Fig. 1). A region was selected between CHRFAM7A and ACTC, where markers in the region had NPL scores greater than 2.0. Genes in the region between CHRFAM7A (28,440,735) and ACTC (32,869,723) (March 2006 NCBI36/hg18) are listed in Tables 3–5. SNPs in CHRFAM7A, the partial duplication of CHRNA7, could not be surveyed for this study as the duplication prevents determination of heterozygotic and homozygotic status for polymorphisms (Gault et al., 1998; Riley et al., 2002). Telomeric to ACTC, the linkage signal declines. B2M, a gene telomeric to the 15q13-q14 linkage peak region was included in the fine-mapping study because its expression is elevated in schizophrenic nonsmokers and normalized in schizophrenic smokers (Mexal et al., 2005).
SNP selection within the 16 genes localized to the 15q13-q14 linkage peak region was performed with the Tagger algorithm as implemented in Haploview (Barrett et al., 2005). TagSNPs were chosen with pairwise analyses, requiring settings r2 ≥ 0.8 between each ungenotyped SNP and at least one TagSNP, and a minimum minor allele frequency of 5%. TagSNPs from both the Yoruban African and CEPH (Utah residents with ancestry from northern and western Europe) populations were utilized in an attempt to capture all common genetic variation in both the Caucasian-non-Hispanic and African-American study participants. Exons 5–10 of the CHRNA7 gene are duplicated (CHRFAM7A); therefore, only SNPs from exons 1 to 4 of the full-length gene were identified using HapMap.
Three markers were selected within the region from previous studies. D15S128 and D15S1360 both provided convincing evidence for linkage in replicated reports (Freedman et al., 1997; Kaufmann et al., 1998; Liu et al., 2001; Fallin et al., 2003). Marker D15S659 was included even though it showed evidence of linkage in only one report (Gejman et al., 2001); it is the closest significant microsatellite marker to the B2M gene. On the basis of highly significant P values in the region centromeric to CHRNA7, seven additional microsatellite markers were chosen to expand the region between marker D15S128 and CHRNA7. These markers included D15S113 (23,665,494), D15S1002 (25,423,217), D15S1019 (27,350,717), D15S1043 (27,925,207), D15S165 (28,947,842), D15S976 (29,239,627), and D15S1031 (29,775,416).
Nucleated cells were obtained from anticoagulated blood (EDTA) by means of lysis with a high sucrose solution (0.3 mol/l of sucrose, 0.01 mol/l of Tris-HCl, pH 7.5, 0.005 mol/l of MgCl2, 1% Triton X-100) and subsequent centrifugation. DNA was then isolated from these cells as previously described (Miller et al., 1988). DNA samples for the NIMH Genetics Initiative participants were obtained from the NIMH collection at Rutgers (http://www.rucdr.org/ASHG/Schizophrenia.htm).
TagSNP genotyping was done using the Applied Biosystems SNPlex assay on an ABI 3730 DNA Analyzer (Applied Biosystems, Foster City, California, USA). This method assays genotypes in up to 48 SNP PCR multiplex reactions using an oligonucleotide ligation procedure (Tobler et al., 2005). Data analyses were carried out with the Applied Biosystems Genemapper software v4.0.
Twenty-seven probe pools were designed for this study, each containing approximately 47–48 SNP primers. All genotypes called by the Genemapper software were manually reviewed and rescored when necessary. If less than 95% of genotypes were called on any given pool, the entire assay was repeated only for samples in that pool missing at least 10 genotype calls. For a probe/pool exceeding 95% of genotypes called, the assay was repeated only on samples for which no genotypes were obtained. SNPs for which the allele clusters were indistinguishable were excluded from analysis, as were SNPs not conforming to Hardy–Weinberg Equilibrium (HWE) in the control sample (P < 0.01).
Microsatellite marker loci were amplified by PCR using fluorescently (6-FAM) labeled primers and AmpliTaq Gold. The GeneAmp PCR System 9600 (Perkin-Elmer, Foster City, California, USA) was utilized for fragment amplification; thermal cycler conditions were unique to each amplicon (Supplemental Table 1 http://links.lww.com/PG/A24). Fragments were characterized on an Applied Biosystems 3100 Avant Genetic Analyzer (Applied Biosystems, Foster City, California, USA). Genotypes were determined using the Genemapper 3.5 software (Applied Biosystems).
The case group of individuals with schizophrenia included one randomly chosen schizophrenic individual per family with SAS version 9.1 software (SAS Institute Inc.). For each marker, allele frequencies were estimated and conformance of genotype frequencies with HWE expected proportions was tested in the control participants by means of a 1 d.f. χ2 goodness-of-fit test using Haploview (Barrett et al., 2005). For microsatellite loci, the exact test for deviations from HWE was calculated using PEDSTATS (Wigginton and Abecasis, 2005). Linkage disequilibrium parameters D1 and r2 were calculated for each pair of SNP markers, and tests for differences in allele frequencies between cases and controls were computed using a χ2 test in Haploview. Chi-squared tests of association for each SNP and microsatellite marker under a genotypic model were calculated using the UNPHASED software version 3.0.13 (Frank Dudbridge, MRC Biostatistics Unit, Cambridge, UK) (Dudbridge, 2003). In addition, tests for differences in allele frequencies between cases and controls for each microsatellite marker were conducted in UNPHASED. To avoid spurious genetic associations because of multiple testing, a Bonferroni correction for the number of markers tested was applied, which is the most stringent correction for regional significance. We report nominal and corrected P values; corrected P values are indicated with an (*).
Odds ratio (OR) estimates and 95% confidence intervals (CI) for comparing relevant genotype or allele classes were computed with SPSS version 16.0 software (SPSS Inc., Chicago, Illinois, USA).
Power estimates, for case–control analyses at a significance level of 5%, were obtained with the Quanto software (Irvine, California, USA) (Gauderman, 2002a, 2002b). The Caucasian-non-Hispanic schizophrenic/control sample size had power between 0.74 and 0.99 to detect a genotype relative risk (GRR) greater than or equal to 2.0 over a range of marker allele frequencies between 0.10 and 0.40. At GRRs less than 2.0, power was substantially lower (0.12–0.88) for most allele frequencies. The African-American schizophrenic/control sample size had less power to detect a GRR greater than or equal to 2.0 (0.38–0.87) over a range of marker allele frequencies between 0.10 and 0.40. Similarly, at GRRs less than 2.0, power was considerably reduced (0.07–0.49) for most allele frequencies.
As a secondary analysis, a permutation procedure implemented with set-based tests in PLINK (Purcell et al., 2007) was utilized to obtain empirical P values for each gene (corrected for all SNPs in a gene with 10000 permutations) with the a priori outcome of schizophrenia. Genotype and phenotype information for SNP rs3087454 was included in this analysis as this SNP is located in the upstream regulatory region of CHRNA7, and is significantly associated with schizophrenia in African-American and Caucasian-non-Hispanic case–control sample populations (Stephens et al., 2009). Case–control samples of both ethnicities, including with schizophrenia patients from the NIMH collection, were evaluated for the presence of population stratification in a previous study from our laboratory (Stephens et al., 2009). These analyses indicated no significant differences in average admixture proportions between cases and controls within each ethnic group.
Before data analysis, the PEDCHECK software (Pittsburgh, Pennsylvania, USA) was utilized to detect errors in Mendelian inheritance (O'Connell and Weeks, 1999). Pedigree disequilibrium tests of association at SNP and microsatellite markers were implemented in the UN-PHASED software version 3.0.13 (Dudbridge, 2003). Similar to the case–control analyses, a Bonferroni correction was used to generate P values corrected for the number of markers tested.
Approximately 98.4% of the genotypes were collected for each of the 27 SNP probe pools. Of the 1264 SNPs surveyed in 16 genes (Table 3), 36 (2.8%) were eliminated from statistical analysis because their allele clusters were indistinguishable from one another. Of the remaining 1228 SNPs, all but 31 (2.5%) were consistent with HWE in one or more ethnic populations. Figure 2 shows the proportion of SNPs in Hardy–Weinberg disequilibrium, in each gene, across the region of interest. Several of these nonconforming SNPs lie in regions with CNVs associated with schizophrenia (Stefansson et al.,., 2008; Stone et al., 2008).
Results for single marker analyses under a genotypic model of association at a significance level of 5%, with the primary a priori outcome of schizophrenia as well as smoking and smoking in schizophrenia are shown in Table 6, and Fig. 3a and b. P-values were corrected for multiple testing utilizing a Bonferroni correction within ethnic group. A correction of 560 tests was applied to SNPs in the Caucasian-non-Hispanic sample as only 560 of the 1228 SNPs were polymorphic in Caucasians. A correction of 1043 tests was applied in the African-American sample. Figure 3a and b, displays – log P-value scatter plots for the outcome of schizophrenia. All positive associations were for SNPs in the intronic regions of the genes.
In the Caucasian sample population, five SNPs were significantly associated with schizophrenia under a genotypic model of association after a Bonferroni adjustment of the P values, designated with an (*): rs7171212, MTMR15; rs1035706, TRPM1; rs10162727, TRPM1; rs12439853, KLF13; and rs16957135, RYR3.
In the African-American sample population, nine SNPs were significantly associated with schizophrenia, under a genotypic model after Bonferroni correction: rs10162727, TRPM1; rs16956707, KLF13; rs11071532, CHRNA7; rs2651415, CHRNA7; rs11071897, SGNE1; rs16973259, GREM1; rs7173813, RYR3; rs12917360, RYR3; and rs527834, AVEN. Of the 15 SNPs associated with schizophrenia in one or more ethnic populations, six are located in genes centromeric to CHRNA7 (MTMR15, TRPM1 and KLF13), two reside in the CHRNA7 gene, and seven are located in genes telomeric to CHRNA7(SGNE1, GREM1, RYR3 and AVEN). Results of allelic tests of association supported results obtained under a genotypic model and are shown in Supplementary Tables 2(A) http://links.lww.com/PG/A25, and 2(B) http://links.lww.com/PG/A26. Additional exploratory analyses were included under the assumption that COS schizophrenics may be biologically distinct from non-COS schizophrenics. To explore this hypothesis, the a-priori analysis was repeated with the COS cases excluded to determine whether results changed considerably. P values were very similar to those obtained with the combined groups.
Utilizing smoking as an outcome with the Caucasian-non-Hispanic sample, one SNP located in CHRNA7 (rs8028396) had a significant P value (P = 0.006*) under an allelic model of association and a nominally significant P value (P = 0.0002, 0.112*) with a genotypic model. This same SNP was associated with smoking in schizophrenia (P = 0.0045*) with an allelic model, and suggestive (P = 0.056*) with a genotypic model. In the African-American sample, with the outcome of smoking, a SNP located in RYR3 (rs8035184) had a significant P value (P = 0.041*) under a genotypic model of association. This SNP was nominally significant (P = 0.034) for an association with smoking in schizophrenia.
Table 6 shows the ORs for a primary outcome of schizophrenia for SNPs significant in the case–control analysis. In the Caucasian-non-Hispanic sample, ORs range from 1.72 (95% CI = 1.31–2.81) for SNP rs7171212 (MTMR15), to 3.97 (95% CI = 2.38–5.88) for rs10162727 (TRPM1). In the African-American sample population, ORs range from 1.66 (95% CI = 1.09–2.53) for rs11071532 (CHRNA7) to 11.89 (95% CI = 5.2–27.16) for rs16973259 (GREM1). 95% CI are much narrower for SNPs significant in the Caucasian-non-Hispanic population because of the larger sample size.
Results of a secondary analysis, utilizing a gene (set)-based permutation procedure (PLINK) (Purcell et al., 2007) are shown in Tables 7 and and8.8. The gene-based empirical P value (column 5) was calculated from the mean of the single SNP statistics, permuted 10000 times. In the Caucasian-non-Hispanic sample, gene-based empirical P values were significant for TRPM1, KLF13, RYR3, and suggestive for CHRNA7 (empirical P = 0.05). In the African-American sample, empirical P values were significant for TRPM1, KLF13, CHRNA7, SGNE1, GREM, and RYR3.
In Caucasian-non-Hispanic families, three SNPs were associated with schizophrenia (rs2911855 P = 0.01*, rs2288242 P = 0.039*, rs4396507 P = 0.002*). Two of these SNPs are in TRPM1 (rs2911855 and rs2288242) and one in SGNE1 (rs4396507). In African-American family members there were no SNPs significant for an association with schizophrenia, however, the P value for rs2911855 (TRPM1) was nominally significant (P = 0.0009, P = 0.94*). No significant P values were obtained for the outcomes of smoking, smoking in schizophrenia. Tables 9 and and1010 show the results of family-based association studies with single SNPs.
In the case–control sample, there were no significant associations with the nine microsatellite markers included in the statistical analysis after adjusting for multiple comparisons. Results are shown in Supplementary Tables 3(A) http://links.lww.com/PG/A27, 3(B) http://links.lww.com/PG/A28, and 3(C) http://links.lww.com/PG/A29.
Family-based-association tests with microsatellite markers found one significant association after adjusting for multiple comparisons. D15S1043 was associated with smoking (P = 0.032*) and nominally associated with smoking in schizophrenia (P = 0.01) in Caucasian-non-Hispanic participants. This microsatellite marker, which lies upstream of the CHRNA7 gene, has been implicated in schizophrenia in two previous linkage studies (Riley et al., 2000; Liu et al., 2001).
In this study, genetic variation was surveyed in genes mapping to a region on chromosome 15 with consistent evidence for linkage to schizophrenia. A portion of this sample consisted of families from the NIMH Human Genetics Initiative for Schizophrenia which had been previously linked to the 15q13-14 region (Kaufmann et al., 1998; Leonard et al., 1998; and Leonard, 2001; Freedman et al., 2001b). Fine mapping of genes in the region and association analyses of case–control participants continue to support the CHRNA7 gene as a candidate for schizophrenia. These results are consistent with a recent large association study by Sanders et al. (2008) in which investigators found two SNPs in CHRNA7 nominally positive for schizophrenia. Although the same SNPs were not examined in this study, we found other positive SNPs within 50kb of those reported. Stefansson et al. (2008) also found nominally significant association with SNPs in the same region.
A limitation in this study, and other genome-wide-association studies, is the incomplete interrogation of CHRNA7. Exons 5–10 and intervening introns in CHRNA7 are duplicated, forming the chimeric gene, CHRFAM7A. The homology of the duplicated CHRNA7 sequence is nearly 100% (Gault et al., 1998). Thus, not all SNPs in CHRNA7 can be evaluated, possibly resulting in an underestimate of association. Although next generation sequencing of the CHRNA7 gene cluster may be possible in the future, current sequencing technologies would require cloning of the duplicated regions from each participant. There is still considerable evidence for CHRNA7 as a candidate gene for schizophrenia, including biological studies in humans and animals, as well as linkage and association studies. We have mapped polymorphisms unique to the coding region of the full length gene and its duplication (Gault et al., 2003); however, most of these SNPs are rare and as a consequence have limited power to detect association with schizophrenia.
Two SNPs were associated with smoking and smoking in schizophrenia. One SNP in CHRNA7 (rs8028396) was significant for association with smoking and smoking in schizophrenia in the Caucasian-non-Hispanic population, under an allelic model. Overall, SNPs associated with smoking and smoking in schizophrenia were not associated with an outcome of schizophrenia. These results indicate that significant associations found in these studies are to schizophrenia and not to smoking in schizophrenia.
Results of secondary analysis with a gene-based set test supported results of single SNP analysis of genes in this region. This analysis included SNP rs3087454 (CHRNA7 5′ UTR) in the derivation of the gene-based empirical P value. This SNP, in a nonduplicated region, has previously been associated with schizophrenia (Stephens et al., 2009), and with improvement in a functional MRI measure in the disorder after administration of an α7 receptor agonist (Tregellas et al., 2010). Although single SNP analysis in the Caucasian-non-Hispanic sample found no significant associations, inclusion of rs3087454 resulted in a suggestive empirical P value for CHRNA7 in this same sample.
The number of informative families in the family-based sample limited positive findings in this part of the study. Although there were 73 Caucasian-non-Hispanic families, the number available for family-based association tests in UNPHASED averaged 15. Similarly, there was an average of only 10 informative African-American families out of 47. Significant P values were obtained in Caucasian-non-Hispanic family members at two SNPs in the TRPM1 gene (rs2911855 and rs2288242). There were more informative families for SNPs rs2911855 and rs2288242 (16 and 18, respectively) compared with SNP rs10162727 (12 informative families in Caucasian-non-Hispanics). However, SNP rs10162727 was still significant in both Caucasian-non-Hispanic and African-American case–controls. Although no SNPs in African-American family members passed the threshold for significance, a nominally significant P value at rs2911855 (TRPM1) with schizophrenia lent further support to this gene. Thus, family-based association studies supported some results obtained in the case– control samples.
Evidence of association for three additional genes was seen in both ethnicities. KLF13 is a zinc–finger protein belonging to a large superfamily of transcription factors (Berg, 1990; Klug and Schwabe, 1995). The SP/XKLF family is a subgroup of this superfamily and is composed of factors that bind to GC-rich and related GT or CACCC boxes (Hagen et al., 1992; Sogawa et al., 1993; Cook et al., 1998). KLF13 is conserved across species and is expressed mainly in the heart and skeletal muscle, but also in brain, liver, and the epithelial layer of the dermis (Scohy et al., 2000; Lavallee et al., 2006). The CHRNA7 promoter is GC rich with multiple Sp-family binding sites (Leonard et al., 2002). Thus, it is possible that both CHRNA7 and KLF13 may confer susceptibility to schizophrenia.
TRPM1 functions to mediate Ca++ entry into the cell (Xu et al., 2001b). When TRPM1 is expressed in HEK293 cells, it increases the cytoplasmic-free Ca++ concentration (Miller et al., 2004). The gene is expressed in brain at low levels (Massullo et al., 2006). Although the expression profile of TRPM1 has been established, the cellular, biochemical, and electrophysiological characteristics of the channel are not well understood.
Ryanodine receptors, such as RYR3, are intracellular ion channels that release Ca++ from intracellular stores after transduction of external stimuli (Hakamata et al., 1992). RYR3 is abundantly expressed in brain and skeletal muscle and faintly expressed in the heart (Nakashima et al., 1997). Ryanodine receptors, including RYR3, have been localized to rat hippocampus, specifically in CA1 pyramidal neurons known to be involved in long-term potentiation (Van de Vrede et al., 2007). Furthermore, enhancement of long-term potentiation by nicotine activation of α7 receptors is dependent on ryanodine receptor release of calcium from intracellular stores in the rat dentate gyrus (Welsby et al., 2006). It is possible that a dysfunctional or nonfunctional RYR3 gene may blunt or alter the proper downstream signaling of activated α7 receptors.
These three genes are located both centromeric (TRPM1 and KLF13) and telomeric (RYR3) to the CHRNA7 candidate gene. The rs10162727 SNP in TRPM1 was of particular interest as it was highly significant in both ethnic populations. SNP rs10162727 (C/T) is located in intron 16 of the TRPM1 gene. The C/C genotype was significantly more prevalent in schizophrenics in both ethnicities. In African-Americans, the association signal seems to be driven by a lack of heterozygotes in the case group. According to the HapMap database (www.hapmap.org), the region surrounding this polymorphism has multiple CNVs. Large recurrent copy number deletions in genes TRPM1 and KLF13 were positively associated with schizophrenia in the study by Stefansson et al. (2008). It is reasonable to propose that the association in African-Americans may represent individuals with deletion variations.
We found a disproportionately high number of SNPs not in conformance with HWE in TRPM1, KLF13, CHRNA7, and RYR3 as compared with the region telomeric of these genes. In human populations, HWE is common and deviations are most likely because of genotyping error or population stratification (Wigginton et al., 2005). Genotyping error seems unlikely to be a large contributing factor as quality control checks, and pedigree checks for Mendelian inconsistencies in family member data were performed. There was no strong stratification in case– control populations in this dataset, based on a screen of 176 ancestry informative markers (Stephens et al., 2009).
The 15q11-q14 region is one of the most dynamic regions of the human genome, containing a wide range of clinically recognized rearrangements including deletions, duplications, inversions, and translocations (Gault et al., 1998; Locke et al., 2004; Flomen et al., 2008). An intriguing explanation for the lack of HWE in this region is the surprising number of CNVs and in particular, common deletion variants (Gault et al., 1998; Miller et al., 2009; Van Bon et al., 2009; Szafranski et al., 2010). CNVs, genomic sequences which can encompass entire genes, are uncommon (< 3%), and their occurrence varies between populations (Redon et al., 2006). Regions flanking segmental duplications (blocks of DNA ranging from 1 to 400 kb which occur at more than one site in the genome and share > 90% sequence identity) are susceptible to rearrangement by nonallelic homologous recombination. These regions represent hotspots of genomic instability which are prone to CNV (Sharp et al., 2008; Szafranski et al., 2010). A segregating deletion may leave ‘footprints' in SNP genotype data by causing runs of homozygosity, Mendelian inconsistencies, and Hardy–Weinberg disequilibrium (McCarroll et al., 2006).
Although small duplications or deletions and associated CNVs are primary mechanisms for the creation of new genes, such as CHRFAM7A (Gault et al., 1998), recent data links them with susceptibility to common diseases including autoimmune disorders (Fanciulli et al., 2007; Yang et al., 2007), mental retardation (Sharp et al., 2008), cancer (Lieberfarb et al., 2003; Zhao et al., 2004; Zhao et al., 2005) and schizophrenia (Sutrala et al., 2007; Stefansson et al., 2008; Stone et al., 2008; Sinkus et al., 2009). Importantly, genes associated with schizophrenia in this study and with a disproportionately large number of SNPs not in HWE are located in this region of genomic instability. Within these larger regions of instability, the smaller (300 kb) partial duplication of CHRNA7 occurred (Gault et al., 1998; Riley et al., 2002) at some time during human divergence (Flomen et al., 2008). An even smaller 2 bp deletion in exon 6 of the duplicated gene, CHRFAM7A, is associated with schizophrenia (Sinkus et al., 2009) and is in linkage disequilibrium with an inversion (Flomen et al., 2008). These reports lend further evidence to the possibility that increased risk of schizophrenia, associated with this region in this study, is partially a consequence of its chromosomal instability. Our current finding that multiple genes in this region are strongly associated with schizophrenia suggests that a combination of genetic mutations in this region and CNV may contribute to the development of mental illness.
We are grateful to patients and controls for their participation in the study and to Bernadette Sullivan for her help in recruiting individuals and obtaining blood samples.
NIH Grants DA009457 and MH081177 to S.L., MH068582 to R.F., and the Veterans Affairs Medical Research Service to S.L. and R.F.
Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's website (www.psychgenetics.com).
Conflicts of Interest: There are no conflicts of interest.