In a recent genome-wide association study (GWAS), we identified a unique novel, functional and replicable risk genomic region for alcohol dependence. This region (Chr6: 64,066,604-64,831,120) included the protein tyrosine phosphatase type IVA 1 gene (PTP4A1), the plant homeodomain finger protein 3 gene (PHF3) and part of the eyes shut homolog gene (EYS). It was enriched with numerous replicable common risk variants (minor allele frequency (MAF) >0.05) for alcohol dependence across African–Americans, European–Americans and European–Australians. Within 90 Mb range surrounding this region in the discovery sample, all variants with P<10−4 were concentrated in this region. Most of these risk variants had significant cis-acting regulatory effects on mRNA expression. The distributions of −log(P) values for association and functional signals in this region were highly consistent across six independent populations. We thus speculated that this region might harbor a causal variant(s) for alcohol dependence.1

Rare variants, which may not be well-captured by a GWAS, may be of equal importance as common variants in the etiology of complex disorders. Recent studies have indicated that rare variants, which usually have strong effects and high penetrance, are involved in the genetic susceptibility to an increasing number of human diseases or traits; for example, plasma levels of high-density lipoprotein cholesterol2 and type 1 diabetes3 as well complex neuropsychiatric disorders such as autism.4, 5, 6 However, the associations between rare variants within the PTP4A1-PHF3-EYS region and alcohol dependence have never been tested. Furthermore, alcohol dependence has high rates of comorbidity with numerous neuropsychiatric disorders, including anxiety disorders, major depression, bipolar disorders, schizophrenia, post-traumatic stress disorder and so on. The associations between rare PTP4A1-PHF3-EYS variants and these neuropsychiatric disorders have never been tested either.

To further explore the role of PTP4A1-PHF3-EYS region in alcohol dependence, in the present study, we imputed the untyped variants in this PTP4A1-PHF3-EYS region across 21 independent cohorts with 11 different neuropsychiatric disorders (Table 1), and then examined the associations between rare PTP4A1-PHF3-EYS variants (0<MAF<0.05 in controls) and these disorders, so that we could test whether rare PTP4A1-PHF3-EYS variants were associated with alcohol dependence and whether these associations were specific for alcohol dependence. The data of these disorders were all of those with neuropsychiatric and neurological disorders available for our analysis from the dbGaP database (http://www.ncbi.nlm.nih.gov/gap/).

Table 1 Associations between rare PTP4A1-PHF3-EYS variants and different neuropsychiatric disorders
Full size table

A total of 49 268 subjects in these 21 cohorts were analyzed (Table 1), including 10 554 subjects in 3 cohorts with alcohol dependence (DSM-IV) and 38 714 subjects in other 18 non-alcoholism cohorts. The European–Australian cohort with alcohol dependence included 871 parents, 1645 affected offspring and 3922 unaffected offspring. Detailed demographic information for these samples has been published (Supplementary Table S1).1, 7, 8, 9 We used the same strategies as previously to maximize the success rate and accuracy of imputation and to stringently clean the phenotype and genotype data.8, 9 Associations between individual variants and diseases were tested using the program PLINK (for case–control data) and FBAT (for family-based data).8, 9 The experiment-wide significance levels (α) were set at 1.4 × 10−5 (in AAs) and 2.1 × 10−5 (in EAs) for these individual rare variant analyses based on the correction for the numbers of effective genetic markers (n=175, 115 in AAs and EAs, respectively; calculated from the entire marker set by the adjusted Bonferroni-type program SNPSpD) and the number of cohorts (that is, n=21). Furthermore, associations between rare variant constellations across the entire PTP4A1-PHF3-EYS region and diseases were tested using the Variable MAF Threshold (VT) test implemented in the program SCORE-Seq,9 in order to explore the synthetic effects of rare variants within the entire region. In VT test, the weight was set at 1 in the regression model, while the MAF threshold varied between 0 and 0.05. Finally, a total of 197–956 single-nucleotide polymorphisms (SNPs) with 0<MAF<0.05 (in controls) were extracted for association analysis. The MAFs and minimal P-values of the most significant risk SNPs, the SNP numbers with P<0.05 and the P-values for VT tests are shown in Table 1.

We found a rare variant constellation across the entire region that was significantly associated with alcohol dependence in European–Australians (P=4.2 × 10−3), and modestly associated with schizophrenia in European–Americans (in GAIN data: P=0.033), but not with any other neuropsychiatric disease examined (P>0.10). Additionally, single-point association analysis showed that 0–108 individual SNPs were nominally associated with these 11 diseases (P<0.05), but none of them were significant after experiment-wide correction (Table 1).

We confirmed that rare variants within the PTP4A1-PHF3-EYS region had significant synthetic effects on alcohol dependence in European–Australians, consistent with our previous findings in a GWAS. Both studies support our conclusion that this region might harbor a causal variant(s) for alcohol dependence. Additionally, the rare variant constellation was also modestly associated with schizophrenia in European–Americans, but not with any other neuropsychiatric disorder examined. This genetic commonality between alcohol dependence and schizophrenia may underlie high rate of comorbidity between them. It has been reported that over one-third of patients with schizophrenia have an alcohol use disorder.10 Our findings support us to postulate that the underlying neuropathological abnormalities of schizophrenia related to the PTP4A1-PHF3-EYS variants might potentially facilitate the positive reinforcing effects of alcohol use. Finally, we admit that, as a future work, the imputed genotype data in this study need to be validated by a real molecular experiment, more neuropsychiatric disorders comorbid with alcohol dependence need to be examined, and the current findings warrant replication in independent cohorts.