Family clusters of vitiligo cases are not uncommon, occurring in a non-Mendelian pattern suggestive of polygenic, multifactorial inheritance (Spritz, 2007, 2008). Genetic linkage and association studies have implicated a number of genes in vitiligo pathogenesis, especially genes involved in immune function (Spritz, 2007, 2008). However, these loci account for a relatively small fraction of total disease liability.

Genetically isolated “founder populations” afford special opportunities to identify genes involved in susceptibility to disease, as founder populations may have elevated prevalence of specific diseases and reduced heterogeneity of causal genetic and environmental risk factors compared with more outbred populations (Wright et al., 1999). Accordingly, susceptibility alleles that represent relatively minor genetic risk factors for complex diseases in the general population may become amplified and constitute major risk alleles in a founder population, and thus may be localized using less dense maps and smaller sample sizes than similar studies conducted in more outbred populations (Wittke-Thompson et al., 2007). Although complex disease alleles detected as major signals in a unique genetic isolate may be problematic to replicate in the wider, outbred population, they nevertheless can represent important candidate genes that identify potential novel pathways of disease.

We have studied a geographically and genetically isolated community in a remote region of northern Romania, in which the prevalence of generalized vitiligo is 2.9%, at least eight times higher than in other European Caucasian groups (Birlea et al., 2008). Similarly, the prevalence of other autoimmune diseases that are epidemiologically associated with vitiligo (autoimmune thyroid disease, rheumatoid arthritis, adultonset type I diabetes mellitus) are also elevated in this village, both in vitiligo patients and in their first-degree relatives. This village was founded approximately 400 years ago by only three families, and all affected individuals derive from a relatively small number of common ancestors (Birlea et al., 2008); this village thus constitutes a unique founder population for generalized vitiligo.

Following genetic matching to minimize population stratification, we compared SNP allele frequencies in cases versus controls using Fisher single-marker exact tests implemented in PLINK, version 1.05 (Purcell et al., 2007) and by allelic association tests that accounted for relatedness among cases using CCREL, version 0.3 (Browning et al., 2005). Quantile–quantile (Q–Q) analyses of the observed −log₁₀(P-values) from the Fisher exact tests (Figure 1a) and the observed χ²-test statistics from CCREL (Figure 1b) showed modest residual genomic inflation (Fisher’s exact test genomic inflation factor λ=1.06; CCREL genomic inflation factor λ=1.10). Residual genomic inflation was effectively removed by adjustment of the test statistics for the corresponding inflation factor. After all corrections, we observed greater association than expected by chance (Table 1). As shown in Figure 2 and Supplementary Table S1, the most significant association signal was for SNP rs13208776 (nominal Fisher exact test P-value=3.13×10⁻⁸), which surpassed a strict Bonferroni-corrected criterion for genomewide significance (P<1.68×10⁻⁷; 0.05 divided by 297,342 SNPs; Ioannidis et al., 2009), based on the P-values from both the Fisher exact test (adjusted P=8.51×10⁻⁸) and from the CCREL likelihood ratio χ² test (adjusted P=9.71×10⁻⁸). The adjusted Fisher exact P-value for rs13208776 exceeded the 2.1% quantile of minimum P-values from 1,000 permutations of the phenotype labels, and was the only P-value outside the 95% confidence limits that were estimated empirically for each of the 20 top-ranking GWAS signals. Together, these analyses show that rs13208776 surpasses a conservative Bonferroni-corrected significance threshold, even after adjustment for relatedness among cases and for genomic inflation.

SNP rs13208776 is located on chromosome 6q27, within intron 4 of the SMOC2 gene (Figure 3). Of the 10 SNPs showing the strongest association within the SMOC2 region, most were contained in two adjacent linkage disequilibrium blocks, the first comprising seven SNPs (average D′ =0.97; r²=0.40), including rs13208776, and the second comprising two SNPs (D′ =1.00; r²=0.04), rs1402|rs214479; SNP rs2144749 had an adjusted P-value 1.75×10⁻⁴ and is substantially correlated (D′ =0.72; r²=0.49) with SNP rs13208776. We identified several haplotypes in this region with P-values slightly improved over SNP rs13208776, although none were significantly better. Of 297,342 SNPs tested, the 20 SNPs with the most extreme P-values associated with vitiligo are listed in Supplementary Table S1.

SMOC2, which is the only gene in this region of 6q27, is located in close proximity to IDDM8 (http://www.t1dbase.org), a genetic locus that has both linkage and association with type I diabetes (Luo et al., 1995; Davies et al., 1996; Owerbach, 2000; Cox et al., 2001) and rheumatoid arthritis (Myerscough et al., 2000). Type I diabetes and rheumatoid arthritis are autoimmune diseases that are epidemiologically associated with generalized vitiligo, both in the outbred CEU population in general (Alkhateeb et al., 2003) and in this Romanian founder population specifically (Birlea et al., 2008). Recently, SNPs within intervening sequence 4 of SMOC2 have been genetically associated with measures of pulmonary function (Wilk et al., 2007), although none of these associations remain significant after multiple-testing correction. Our findings indicate that SMOC2 is an important novel candidate gene for susceptibility to generalized vitiligo, and perhaps to other autoimmune diseases, in the isolated Romanian founder population of this study. This finding may be difficult to verify in the broader CEU population, as genetic purification in this unique population isolate has led to a very high population attributable risk for this SMOC2 susceptibility allele that may be much smaller in the outbred CEU population, even for the same variant. SMOC2 encodes a widely expressed, SPARC (BM40)-related glycoprotein that contains two thyroglobulin type-I domains, two EF-hand calcium-binding domains, a follistatin-like domain, and a putative signal peptide (Nishimoto et al., 2002; Vannahame et al., 2003). The specific function of the SMOC-2 protein remains unknown, though roles have been suggested in angiogenesis (Rocnik et al., 2006), cell cycle regulation (Liu et al., 2009) and mitogenesis (Liu et al., 2009). Although defective calcium transport has been reported in vitiligo melanocytes and keratinocytes (Schallreuter-Wood et al., 1996), specific involvement of SMOC-2 in calcium homeostasis remains unproven.

Supplement