In this study we investigated linkage and association of Syn2 gene polymorphisms with schizophrenia using four microsatellite markers and twenty SNPs. We failed to identify significant (LOD≥3.0) or suggestive (LOD≥2.0) linkage to the area in this sample. However, family-based association analysis of the transmission of Syn2 SNPs individually and as haplotypes from parents to affected offspring and transmission disequilibrium between affected and unaffected siblings provided evidence that Syn2 is associated with the development and/or pathogenesis of this debilitating disorder.
Analysis of twenty SNPs spanning 117 kb in the Syn2 gene revealed a complex LD/haplotype structure of the gene. We observed four groups of SNPs in tight LD blocks with D′=r2=1. Physically, these blocks ranged in size with the largest spreading over 20.6 kb and covering almost 10% of the gene (Block 1), and the smallest encompassing only 2.4 kb (Block 4). While in our study Block 4 appeared as a discreet haplotype unit, the other blocks interdigitated significantly. The largest block (Block 1) encompassed the proximal 96.5% of the 9.6 kb Block 2 and the proximal 40.7% of the 8.8 kb Block 3, while the distal 40.9% of Block 2 overlapped with the proximal 44.6% of Block 3. We demonstrated that strong LD encompassed almost the entire gene, with the majority (39 of 55) of non-redundant marker pairs being in complete LD (D′=1), a large number (14 pairs) demonstrating very strong LD (D′ from 0.76 to 0.92) and only two marker pairs with D′<0.6 (D′=0.45 for Block 2 vs Block 4 and D′=0.53 for rs17035945 vs Block 4). Public LD databases, such as that provided by Perlegen Sciences and Hapmap, reveal a similarly complex pattern of LD in this region, although with differences in the details of the LD/haplotype structures observed, presumably due to differences in the markers and populations used for the different analyses.
Out of eleven SNPs and SNP block markers analyzed for association with schizophrenia, four showed disequilibrium of transmission to affected subjects and their unaffected siblings and five demonstrated deviation of transmission from parents to affected offspring. Although these results of association were obtained in an area without significant evidence of linkage to schizophrenia in this sample, in light of the mildly positive lod scores (maximum HLOD of 1.9) we wished to explore the possible effect of linkage on the tests of association. A simulation study to assess the effect of linkage on the results of the tests of association revealed that the PDTPHASE and TRANSMIT statistics appear to be independent of the effect of linkage for this sample. While TRANSMIT appeared to be anticonservative when calculating p-values by its internal bootstrap procedure, our simulation analysis demonstrated that the observed associations with schizophrenia are still highly significant. According to the most conservative scenario observed in our simulation results, an empiric p-value of 0.05 is equivalent to a nominal p-value of 0.02 in TRANSMIT and of 0.06 in PDTPHASE. Five out of eleven SNP markers reached this level of empiric significance for each test; markers rs598747, Block 2, Block 3, and ss35528972 were empirically significant in both tests, while markers rs17035945 and rs3817004 were empirically significant in only PDTPHASE or TRANSMIT, respectively.
While three of the SNPs tested produced nominal p-values <0.05 with both PDTPHASE and TRANSMIT, more SNPs were detected as associated using TRANSMIT than PDTPHASE, and the significant p-values obtained with TRANSMIT were one to five orders of magnitude smaller than the PTDPHASE p-values for the same markers. Part of this difference appears to be due to the increasingly large distortion of the TRANSMIT nominal p-values for smaller values. So while, under the most conservative scenario, a nominal p-value of 0.05 occurred in 8.5% of the replicates by chance, a nominal p-value of 0.001 occurred 1.2% of the time, and a nominal p-value of 0.0001 occurred 0.7% of the time. However, even accounting for this inflation does not appear to fully explain the differences in the results. Further differences can be explained by the different statistical approaches of the two methods. TRANSMIT, a TDT based method, evaluates the transmission of markers from parents to affected offspring. It calculates a score vector averaged over all possible configurations of parental haplotypes and transmissions. Data from unaffected siblings and siblings with unknown phenotype is used only to narrow the possible parental genotypes and haplotype configurations. PDTPHASE is based on the PDT test (Martin et al. 2000
) that uses a broader range of information from an extended pedigree by including unaffected siblings into the statistical analysis of association. For the PDT, a measure of linkage disequilibrium is defined for each triad and each discordant sib pair within a pedigree, and an average is determined for each pedigree. The difference in what is considered the fundamental unit for each test (heterozygous parents for TDT, triads and discordant sib pairs for the PDT) can also lead to differences in the test statistics. For example, in the case of triads with two heterozygous parents, the test statistic for the TDT can be double that of the PDT in the situation where the allelic transmission of the two heterozygous parents is concordant (Martin et al. 2000
). We assume that Syn2
is a susceptibility gene that is neither necessary nor sufficient for the development of schizophrenia. Individuals inheriting certain alleles of this gene are at increased risk for developing the schizophrenia phenotype. Environmental factors or susceptibility alleles at other genes may also be required before a sufficient threshold is reached for the phenotype to manifest, resulting in apparent reduced penetrance of the Syn2
disease-associated allele. The use of risk allele carriers who do not manifest the illness by the PDT-based tests, but not the TDT-based test, could also contribute to a weaker association result for the PDT in situations with greatly reduced penetrance. While TRANSMIT appears to have greater power to detect the association present between Syn2
and schizophrenia in this sample, small p-values generated by this program must be evaluated with great caution due to the anticonservative nature of the internal bootstrap procedure.
Overall, 11 markers were tested for association. Given the strong marker to marker association in this region, considering these as 11 independent tests would be overly conservative. While it unclear if the PDTPHASED results would reach significance if corrected for multiple testing, our simulations do indicate that the TRANSMIT results are significant. Applying the overly conservative Bonferroni correction would require a p-value of 0.0045 to reach study-wide significance. Our simulation results suggest that to obtain an empiric TRANSMIT p-value of 0.005 would require a nominal p-value of 0.00001. Thus even with a Bonferroni correction, the Block 2 nominal p-value of 0.0000005 would reach study-wide significance.
Haplotype analysis demonstrated that simply combining all markers with evidence of association under the single marker tests was not helpful in identifying a haplotype that is likely to harbor schizophrenia associated polymorphism within the Syn2 gene. One explanation could be that the associated SNPs are in LD with more than one susceptibility allele, so combining them all into a single analysis with markers spanning over 90 kb was not beneficial. Most of the associated SNP makers were located in introns 5 and 6 which are known for their evolutionary conservation due to the presence of the TIMP4 gene in intron 6 of Syn2. A positional haplotype analysis approach that densely covered a 14.4 kb interval from this area identified seven haplotypes consisting of markers from introns 5 and 6 of Syn2, with one of these haplotypes significantly overtransmitted to affected individuals.
Although the majority of the schizophrenia associated SNPs and haplotypes within the Syn2
gene fall into the area of the TIMP4
seems to be an unlikely functional candidate for schizophrenia. The expression profile of the TIMP family is diverse and includes multiple tissues, including the central nervous system. The literature on this family is limited, and fails to report a significant association of TIMP genes with schizophrenia on genetic or functional levels (Hung et al. 2001
). Moreover, the review of the clinical manifestations caused by alterations within the TIMP family (TIMP1 [MIM305370], TIMP2 [MIM188825], TIMP3 [MIM188826], TIMP4 [MIM601915]) reveals a single established human phenotype, Sorsby fundus dystrophy, that does not resemble schizophrenia or any other psychotic illnesses. However, TIMP4
cannot be ruled out as a susceptibility gene on the basis of our association data. The nested relationship of the TIMP and synapsin family members will likely necessitate a functional approach to differentiate the possible role of Syn2
in the etiology of schizophrenia.
This sample represents the first population of Northern European descent to exhibit an association between Syn2
and schizophrenia. Recently, evidence of schizophrenia associated polymorphisms in case-control and family based studies were found in the Han Chinese (Chen et al. 2004
, Chen et al. 2004
) and Korean (Lee et al. 2005
) populations. Further association studies may provide a more precise location of the schizophrenia associated polymorphism(s) within the Syn2
gene. Functional studies of polymorphisms identified through association studies should aid in the identification of the schizophrenia risk-allele within this region. Interestingly, samples from both Northern European descent (Brzustowicz et al. 2004
) and Han Chinese (Zheng et al. 2005
) populations have also demonstrated association between schizophrenia and NOS1AP
, a gene that is functionally linked with Syn2
(Jaffrey et al. 2002
). A combined influence of NOS1AP
on schizophrenia susceptibility is plausible and warrants further investigation.