In this study, we found a significant phenotypic association of AD with light eye colors, particularly blue eye color (Table I), significant enrichment of genetic interactions between selected eye color genes and AD genes (Fig. 2), and strong LD between pigmentation genes and AD-associated genes on chromosomes 15q12 and 11q14 (Supplementary Figures S2 and S6). The strengths of this study include 1) extensive control for potential population stratification of all samples using genome-wide SNP information, 2) leverage of genomic data to assess the extent of biologically relevant interactions between eye color genes and AD candidate genes, and 3) multilevel (i.e., population-phenotype, network and genetics) approaches to test our hypothesis.
Population stratification is a well-established source of false positive findings in association studies. To address this issue and assess the genetic homogeneity of our samples, we carefully selected only individuals who self-identified as EA and excluded admixed outliers such as individuals who were Hispanic based on self-report and our principal component analysis. These quality control procedures are likely to lead to moderately homogenous samples (Fig. 1 and Supplementary Figure S1). It should be noted that the PCA-based correction may not adequately correct for the south-north eye color cline in Europe or for the potential variation of this trait within countries of origin.
A few other lines of research support the observed AD-eye color association. Firstly, there is evidence of association between light eye color and SAD [Pacchierotti et al., 2001] (Supplementary Figure S4). SAD is often comorbid with AD [Sher, 2004]. While the relationship between eye color and SAD could plausibly be explained by varying light sensitivity, there is no readily available explanation for the association between eye color and AD. One possible physiological mechanism connecting eye color and AD is as follows: blue-eyed individuals have greater light sensitivity than brown-eyed individuals; and heightened sensitivity to varying light intensities has been associated with abnormal changes in endogenous melatonin production [Pacchierotti et al., 2001]. The latter has also been associated with SAD, which is often comorbid with AD (Supplementary Figure S4). Thus, we hypothesize that AD and eye color may have partially shared etiological factors. Terman et al. showed that light-eyed individuals were less likely to develop SAD than brown-eyed individuals during the winter [Terman and Terman, 1999]. However, this conclusion did not exclude the possibility that light-eyed individuals are at a higher risk for SAD than their dark-eyed counterparts when exposed to varying light intensities, which is known to alter endogenous levels of serotonin and melatonin in light-supersensitive individuals [Pacchierotti et al., 2001]. Furthermore, our results complement a recent paper in which sunshine was shown to influence behavior [Vyssoki et al., 2014]. This study suggested that sunshine might facilitate suicidal behavior during the 10-day period prior to suicide. Since AD is a known risk factor for suicidal behavior [Inskip et al., 1998; Sher, 2006; Wojnar et al., 2009], our results imply that individuals with light eye color might be at greater susceptibility to sunshine-triggered behavior alterations (e.g., mood, aggression and impulsiveness) than dark-eyed individuals. In sum, the inconsistent findings [Pacchierotti et al., 2001; Higuchi et al., 2007] in the literature reflect an incomplete understanding of the connection between eye color and psychiatric disorders.
Secondly, we observed strong LD blocks between eye color genes and GABA genes on chromosome 15q12. Interestingly, the 15q12 cytoband lies within the Prader-Willi syndrome (PWS) region. PWS presents with two relevant clinical features: hypopigmentation of the eyes and behavioral and psychiatric disturbances [Cassidy et al., 2011], which demonstrates that mutations in the 15q12 region can lead to both phenotypes. Similarly, we also observed strong LD between the GRM5 (AD-associated) and TYR (pigmentation-associated) genes in cytoband11q14.3. Interestingly, microdeletions in this region have been associated with leukodystrophy, a group of central nervous system disorders affecting the brain's white matter [Goizet et al., 2004]. Additionally, variation in this region, specifically in TYR, has been associated with melanin production [Beleza et al., 2013]. Overall, these observations support the idea that two independent gene regions in the human genome may be concurrently associated with pigmentation variation and brain function.
Thirdly, animal experiments have also shown that hypopigmentation may correlate with behavioral changes (e.g., in the Astyanax cavefish model [Elipot et al., 2014]). Despite alack of direct evidence, these reports support the association between blue eye color and AD in EAs (Fig. 3).
To conclude, our findings complement existing research on the connection between eye color and psychiatric illnesses, addiction, and behavioral problems. Our study is the first to report an association between blue eye color and AD in EAs using well-diagnosed subjects and a moderate sample size. Our findings indicate that the selection pressures acting on the genetics of pigmentation might not only have implications for personality features, as previously reported [Gardiner and Jackson, 2010], but also for AD susceptibility. Thus, integration of population-phenotype and gene and network analyses is helpful for the identification of risk factors in AD, and a broad range of psychiatric illnesses and addiction, in general. Although we carefully controlled for stratification, we cannot exclude underlying occult stratification as a contributor to this observation. While replication is needed, our findings suggest that eye pigmentation information may be useful in the future research of AD and related alcohol consumption behaviors. Further characterization of this association may unravel novel etiological factors in alcohol addiction.