Rare SERINC2 variants are specific for alcohol dependence in subjects of European descent

Abstract

Objectives

We previously reported a top-ranked risk gene [i.e., serine incorporator 2 gene (SERINC2)] for alcohol dependence in the subjects of European descent by analyzing the common variants in a genome-wide association study. In the present study, we comprehensively examined the rare variants [minor allele frequency (MAF) < 0.05] in the NKAIN1-SERINC2 region, in order to confirm our previous finding.

Methods

A discovery sample (1,409 European-American cases with alcohol dependence and 1,518 European-American controls) and a replication sample (6,438 European-Australian family subjects with 1,645 alcohol dependent probands) underwent association analysis. A total of 39,903 subjects from 19 other cohorts with 11 different neuropsychiatric and neurological disorders served as contrast groups. The entire NKAIN1-SERINC2 region was imputed in all cohorts using the same reference panels of genotypes that included rare variants from the whole-genome sequencing data. We stringently cleaned the phenotype and genotype data, and obtained a total of about 220 SNPs in the subjects with European descent and about 450 SNPs in the subjects with African descent with 0<MAF<0.05 for association analysis.

Results

Using a weighted regression analysis implemented in the program SCORE-Seq, we found a rare variant constellation across the entire NKAIN1-SERINC2 region that was associated with alcohol dependence in European-Americans (Fp: overall p=1.8×10⁻⁴; VT: overall p=1.4×10⁻⁴; Collapsing p=6.5×10⁻⁵) and European-Australians (Fp: overall p=0.028; Collapsing p=0.025), but not African-Americans, and not associated with any other disorder examined. Association signals in this region came mainly from SERINC2, a gene that codes for an activity-regulated protein expressed in brain that incorporates serine into lipids. Additionally, 26 individual rare variants were nominally associated with alcohol dependence in European-Americans (p<0.05). The associations of 5 of these rare variants that lay within SERINC2 exhibited region-wide significance (p<α=0.0006); and 25 associations survived correction for false discovery rate (q<0.05). The associations of 2 rare variants at SERINC2 were replicated in European-Australians (p<0.05).

Conclusion

We concluded that SERINC2 was a replicable and significant risk gene specific for alcohol dependence in the subjects of European descent.

Introduction

Individuals with alcohol dependence continue to use alcohol despite adverse consequences in health, job and family functions. Several lines of evidence demonstrated a substantial genetic component in the risk of developing alcohol dependence. Siblings of alcoholic probands had a 3–8 fold increase in the risk of developing alcohol dependence [1]. The heritability of risk for alcohol dependence was estimated to be ~39% by studies of the adopted-away offspring of affected and unaffected parents [2] and as high as 60% by twin studies [3]. These studies provided evidence that genetic factors constitute a significant cause of alcohol dependence.

Recently, a gene at 6q21, i.e., Na+/K+ transporting ATPase interacting 2 gene (NKAIN2) (alias: TCBA1), was found to have replicable associations with alcohol dependence in both the COGA family-based Caucasian (p<10⁻³) and European-Australian samples (p=5.1×10⁻⁷) [4]. This gene is highly conserved among species, transcribed in different splice variants, specific to the central nervous system [5], and critical for neuronal functions [6]. It has also been associated with neuroticism in a genome-wide associations study (GWAS) [7], a complex neurological phenotype [5], and a developmental delay with recurrent infections [8]. This gene encodes a member protein, i.e., NKAIN2, of a mammalian protein family that contains four members, i.e., NKAINs 1–4, with similar conservation, distributions and functions. All four proteins interact with the beta subunit of Na,K-ATPase (ATP1B1) that belongs to the family of Na+/K+ and H+/K+ ATPases beta chain proteins, and to the subfamily of Na+/K+ -ATPases. Na+/K+ -ATPase is an integral membrane protein responsible for establishing and maintaining the electrochemical gradients of Na and K ions across the plasma membrane. These electrochemical gradients are essential for osmoregulation, sodium-coupled transport of a variety of organic and inorganic molecules, and electrical excitability of nerve and muscle.

Interestingly, the NKAIN1 gene at 1p35.2, which encodes the member protein NKAIN1 from the same family as NKAIN2, is closely located to the serine incorporator 2 gene (SERINC2) that was a top-ranked risk gene (p=2.3×10⁻⁷) for alcohol dependence reported by one of our recent GWASs [9]. In the present study, we comprehensively examined the rare variants [minor allele frequency (MAF) < 0.05] in the NKAIN1-SERINC2 region, in order to confirm our previous GWAS finding that was based on common variant analysis (MAF > 0.05).

The variants with low allele frequencies are proposed to be the key for "missing" heritability. An increasing number of human diseases have been found to be caused by a constellation of multiple rare, regionally concentrated, variants. Some association signals credited to common variants may be synthetic associations resulting from the contributions of multiple rare variants within a genomic region [10]. In these cases, the synthetic effects of region-wide rare variant constellations may be more significant than individual rare variants. In this study, based on our prior work, we comprehensively examined the associations between rare NKAIN1-SERINC2 variants [MAF < 0.05] and alcohol dependence in a European-American discovery cohort and a European-Australian replication cohort.

It has been reported that alcohol dependence has a high rate of comorbidity with numerous neuropsychiatric conditions including anxiety disorders, major depression, bipolar disorders, schizophrenia and PTSD [11–13]. It has also been reported that many genes have pleiotropic effects on alcohol dependence and other neuropsychiatric conditions; e.g., the autism susceptibility candidate gene 2 (AUTS2) was reported to be a risk gene for autism [14], alcohol consumption (by a GWAS) [15], mental retardation [16] and heroin dependence [17]. It is known that alcohol dependence and these other neuropsychiatric conditions shared etiologies that implicate monoaminergic, cholinergic, GABAergic and glutamatergic, neurotransmission. NKAIN1-SERINC2 region might be related to these neurotransmission systems [18]. Thus, in this study, we also examined the associations between the NKAIN1-SERINC2 variants and several other neuropsychiatric and neurological disorders available from the dbGaP database, in order to examine if the NKAIN1-SERINC2 variants are specific to alcohol dependence.

Materials and Methods

Subjects

The discovery cohort included 1,409 European-American cases with alcohol dependence (DSM-IV) (38.3±10.2 years) and 1,518 European-American controls (38.4±10.4 years). The replication cohort included 2,252 European-Australian small nuclear families with ≤ 2 generations. These families had a total of 6,438 subjects including 4,342 founders and 2,096 nonfounders; the latter included 1,645 alcohol dependent probands. The average family size was 2.97 subjects. These families included 3,818 sib-pairs, 20 half-sib pairs and 4,192 parent-child pairs.

The discovery cohort came from the merged SAGE and COGA datasets (dbGaP access number: phs000092.v1.p1 and phs000125.v1.p1), and the replication cohort came from the OZ-ALC dataset (dbGaP access number: phs000181.v1.p1). Affected subjects met lifetime DSM-IV criteria for alcohol dependence [19]. The control subjects were defined as individuals who had been exposed to alcohol (and possibly to other drugs), but never met the criteria for alcohol or substance use disorder (lifetime diagnosis). All subjects were interviewed using the Semi-Structured Assessment for the Genetics of Alcoholism (SSAGA) [20]. A total of 39,903 subjects of European or African descent from 19 other dbGaP cohorts with 11 different neuropsychiatric and neurological disorders served as the contrast groups. These different neuropsychiatric and neurological disorders included alcohol dependence (in African-Americans), major depression, bipolar disorder, schizophrenia, autism, attention deficit hyperactivity disorder (ADHD), Alzheimer’s disease, amyotrophic lateral sclerosis (ALS), early onset stroke, ischemic stroke, and Parkinson’s disease. Diagnoses, ethnicities, study designs, and dataset names for these cohorts are shown in Table 1. More detailed demographic information for these samples including dbGaP accession numbers, genotyping platforms, sample sizes, sex and age structures and mean ages were published previously [21–24]. These 21 cohorts included case-control and family-based samples, genotyped on ILLUMINA (Illumina, Inc., San Diego, CA, USA), AFFYMETRIX (Affymetrix, Inc., Santa Clara, CA, USA) or PERLEGEN (Perlegen Sciences, Inc., Mountain View, CA, USA) microarray platforms. The discovery sample was genotyped on the Illumina Human 1M and the replication sample was genotyped on Illumina Human CNV370v1.

Table 1.

Associations between NKAIN1-SERINC2 region and different neuropsychiatric or neurological disorders

			SNP #	SNP #	SNP #	SNP #	Collapsing	Minimal	Most sig.		Minor allele frequency (N)
Human Diseases	Ethnicity	Dataset name	(total)	(p<0.05)	(p<α)	(q<0.05)	p value	p value	SNP	Gene	Affected	Unaffected
Alcoholism	EA (CC)	SAGE+COGA	196	26	5	25	6.5×10−5	4.1×10−5	rs35961897	SERINC2	0.059 (1409)	0.038 (1518)
Alcoholism	EAu (Fam)	OZ-ALC	185	9	0	0	0.025	0.0196	rs77840364	SERINC2	0.015 (1645)	0.012 (4793)
Alcoholism	AA (CC)	SAGE+COGA	450	4	0	0	0.534	0.0141	rs16834507	SERINC2	0.040 (681)	0.021 (508)
ADHD	CA (Fam)	IMAGE	163	12	0	0	0.429	0.0017	rs114467377	NKAIN1	0.015 (924)	0.014 (1833)
Autism	EA (Fam)	AGP	189	13	0	0	0.977	0.0013	rs114336824	SERINC2	0.010 (1330)	0.010 (2745)
Major Depression	CA (CC)	PRSC	162	12	0	0	0.089	0.0010	rs116080631	SERINC2	0.028 (1805)	0.015 (1820)
Bipolar Disorder	EA (CC)	BDO+GRU	136	0	0	0	0.735	0.0902	rs7541681	SERINC2	0.125 (368)	0.036 (1034)
Bipolar Disorder	EA (CC)	BARD+GRU	138	5	0	0	0.714	0.0048	rs55781513	NKAIN1	0.026 (653)	0.044 (1034)
Bipolar Disorder	AA (CC)	BARD+GRU	351	7	0	0	0.312	0.0033	rs114478713	SERINC2	0.019 (141)	0.001 (671)
Schizophrenia	EA (CC)	GAIN	180	1	0	0	0.692	0.0080	rs6659255	SERINC2	0.073 (1351)	0.046 (1378)
Schizophrenia	AA (CC)	GAIN	441	20	0	0	0.816	0.0147	rs80029070	NKAIN1	0.025 (1195)	0.038 (954)
Schizophrenia	EA (CC)	MGS_nonGAIN	144	12	0	0	0.145	0.0024	rs74872508	SNRNP40	0.009 (1437)	0.003 (1347)
Alzheimer's Disease	CA (Fam)	LOAD × 4	191	16	0	0	0.556	0.0057	rs7417775	SERINC2	0.052 (2298)	0.037 (2921)
Alzheimer's disease	EA (CC)	GenADA	113	10	0	0	0.514	0.0064	rs76859788	NKAIN1	0.012 (806)	0.025 (782)
ALS	CA (CC)	GRU	125	9	0	0	0.111	0.0111	rs12024466	ZCCHC17	0.004 (261)	0.025 (246)
Early Onset Stroke	EA (CC)	GEOS × 3	144	1	0	0	0.246	0.0447	rs13376139	SNRNP40	0.008 (372)	0.034 (430)
Early Onset Stroke	AA (CC)	GEOS × 3	431	54	0	0	0.080	0.0008	rs56095638	SERINC2	0.087 (309)	0.038 (290)
Ischemic Stroke	CA (CC)	ISGS	132	11	0	0	0.512	0.0041	rs116007405	SERINC2	0.028 (219)	0.004 (266)
Parkinson's Disease	CA (CC)	NGRC	187	3	0	0	0.745	0.0348	rs114215404	NKAIN1	0.002 (2000)	0.005 (1986)
Parkinson's Disease	CA (CC)	PDRD+GRU	142	0	0	0	0.088	0.0694	rs75239059	NKAIN1	0.003 (900)	0.008 (867)
Parkinson's Disease	CA (CC)	lng_coriell_pd	171	5	0	0	0.079	0.0195	rs12564915	SNRNP40	0.015 (940)	0.039 (801)

Only the most significant risk markers with minimal p values are listed. ADHD, Attention deficit hyperactivity disorder; ALS, Amyotrophic Lateral Sclerosis; AA, African-American; EA, European-American; EAu, European-Australian; CA, Caucasian; CC, case-control design; Fam, family-based design; N, sample size. Dataset names refer to dbGaP and references [GenADA: Li et al. Arch Neurol. 2008;65(1):45–53; Filippini et al. Neuroimage. 2009;44(3):724–728. AGP: The AGP Consortium. Nature. 2010;466(7204):368–372; Human Molecular Genetics. 2010;19(20):4072–4082; Nature Genetics. 2007;39(3):319–328]. The significance level (α) is corrected for the numbers of effective genetic markers (calculated by SNPSpD). Collapsing p values for entire NKAIN1-SERINC2 region were calculated using the program ARIEL.

Imputation

The NKAIN1-SERINC2 region includes NKAIN1, SNRNP40, ZCCHC17, FABP3 and SERINC2. We imputed the missing SNPs across the entire NKAIN1-SERINC2 region from Chr1:31,425,179 to Chr1:31,732,987 using the same reference panels, to render the genetic marker sets highly consistent across different cohorts. The reference CEU panel YRI panels from two databases (i.e., 1,000 genome project and HapMap 3) were adopted in the imputation for the samples of European descent and African descent, respectively. All cohorts were imputed using both programs IMPUTE2 [25] and BEAGLE [26]. We maximized the success rate and accuracy of imputation and minimized the false-positives during imputation. Only the genotypes that were imputed consistently between the two independent reference databases and consistently both by IMPUTE2 and BEAGLE were selected for analysis. The uncertainty rate of inference for missing genotypes was controlled at less than 1%. Furthermore, only the SNPs that had similar minor allele frequencies (with frequency difference < 0.2%) in the healthy controls across different cohorts and HapMap database within the same ethnicity were selected for analysis. With these selection criteria, we were highly confident in the quality of these imputed genotype data. Checking the imputed genotypes in all of our four family-based cohorts, we did not find any one individual (considering all SNPs tested) or any one SNP (considering all individuals tested) with more than 0.1% Mendelian inconsistency.

Data cleaning

We stringently cleaned the phenotype and genotype data within each ethnicity before association analysis. Excluded were subjects with poor genotypic data, subjects with allele discordance, sample relatedness, subjects with a mismatch between self-identified and genetically-inferred ethnicity, and subjects with a missing genotype call rate ≥2% across all SNPs. Furthermore, we filtered out the monomorphic SNPs and the SNPs with allele discordance, Mendelian errors (in family samples), an overall missing genotype call rate ≥2%, and MAFs >0.05. We also filtered out the SNPs with missing rate differences > 2% between two samples that had the same phenotype and microarray platform. As a result, a total of ~220 (in the subjects of European descent) and ~450 (in the subjects of African descent) SNPs with 0<MAF<0.05 in either cases or controls were extracted for association analysis. The cleaned sample sizes and SNP numbers of all cohorts are shown in Table 1.

Association tests for region-wide rare variant constellations

We initially tested associations between rare variant constellations and alcohol dependence using a score-type program, SCORE-Seq [27]. The mutation information was aggregated by virtue of a weighted linear combination across all rare variants of the entire NKAIN1-SERINC2 region or across each gene within NKAIN1-SERINC2 region, and then related to alcohol dependence using regression models. Sex, age and the first 10 principal components served as the covariates in the regression models. The principal component scores of our samples were derived from all common autosomal SNPs across the genome using principal component analysis implemented in the software package EIGENSOFT [28]. Each individual received scores on each principal component. These principal components reflected the population structure of our samples, with the first 10 principal component scores accounting for >95% of variance. As covariates in the regression model, these principal components controlled for population stratification and admixture effects on association analysis. The same analytic procedures were applied to the other 10 neuropsychiatric and neurological disorders. For the association analyses on major depression, bipolar disorder, schizophrenia and Alzheimer’s disease, “alcohol drinking” was also included as an additional covariate in the regression to control for the potential confounding effects of “alcohol drinking behaviors”. This covariate was not assessed in ADHD, autism, ALS, Early Onset Stroke, Ischemic Stroke and Parkinson’s disease.

Two tests, Fp and VT, were performed to derive the overall p values. (1) In the Fp tests, the MAF upper bound threshold was fixed at 0.05, but the weight was 1/sqrt(p(1-p)) where p was the estimated MAF with pseudo counts in the pooled sample. (2) In the VT test, the weight was fixed at 1, while the threshold varied between 0 and 0.05. Statistical significance was assessed using a bootstrap procedure with 1 million times of resampling. Finally, these tests were confirmed by another program ARIEL [29] that used a regression-based collapsing approach.

Association tests for individual rare variants (exploratory)

For case-control samples in the discovery cohort, the allele frequencies of each SNP were compared between cases and controls using logistic regression as implemented in PLINK [30]. Diagnosis and alleles each served as the dependent and independent variables, with sex, age and the first 10 principal components as covariates. For family samples in the replication cohort, we tested the allele-disease associations using the program FBAT [31], assuming an additive genetic model under the null hypothesis of no linkage and association, biallelic mode, minimum number of informative families of 10 for each analysis and offset of zero. The same analytic procedures were applied to the other 10 neuropsychiatric and neurological disorders, with “alcohol drinking”, if available, as an additional covariate in the models. Different cohorts were analyzed independently. The MAFs and the minimal p values of the most significant risk SNPs and the numbers of the nominally-significant risk SNPs (p<0.05) in all cohorts are shown in Table 1. Finally, the cumulative Positive Predictive Values (PV⁺) and the cumulative Positive Likelihood Ratios (LR⁺) of the significant (p<α) and independent (r²<0.2) risk alleles across the NKAIN1-SERINC2 region were calculated using Bayesian formula.

Correction for multiple testing in single-point association tests

The experiment-wide significance levels (α) were corrected for the numbers of effective markers that were calculated by the Bonferroni-type program SNPSpD [32], accounting for linkage disequilibrium (LD). Approximately 80 and 120 effective SNPs captured most of the information content of all rare variants across the entire NKAIN1-SERINC2 region in the subjects of European and African descent, respectively. Thus, the corrected significance levels (α) for single-point association tests were set at 0.0006 in the subjects of European descent and 0.0004 in the subjects of African descent, respectively. The numbers of the statistically-significant (i.e., p<α) risk SNPs in all cohorts are shown in Table 1. The false discovery rate (q value) for each SNP was estimated from the p values within each disease group using the R package QVALUE [33]. Finally, for those associations replicated in the European-Australian cohort, the α was set at 0.05.

Results

Rare variant constellation across the entire NKAIN1-SERINC2 region was associated with alcohol dependence in European-Americans (Fp: overall p=1.8×10⁻⁴; VT: overall p=1.4×10⁻⁴; Collapsing p=6.5×10⁻⁵) and European-Australians (Fp: overall p=0.028; Collapsing p=0.025), but not African-Americans, and not associated with any other disease examined (Collapsing p>0.05). When the rare variant constellation within each gene region were tested, SERINC2 variant constellation was significantly associated with alcohol dependence in European-Americans (Fp: p=2.7×10⁻⁴; VT: p=1.6×10⁻⁴; Collapsing p=8.0×10⁻⁵) and suggestively in European-Australians (VT: p=0.028; Collapsing p=0.030) (corrected α=0.01 for five genes within NKAIN1-SERINC2). The other four genes were suggestively (i.e., p close to 0.05) or modestly (0.01<p<0.05) associated with alcohol dependence in European-Americans (Tables 1 and 2).

Table 2.

p values for associations between rare variant constellations and alcohol dependence

	European-Americans						European-Australians
Tests	Whole region	NKAIN1	SNRNP40	ZCCHC17	FABP3	SERINC2	Whole region	SERINC2
Fp	1.8×10−4	0.039	0.020	0.025	0.041	2.7×10−4	0.028	0.028
VT	1.4×10−4	-	0.022	0.022	0.066	1.6×10−4	-	-
Collapsing	6.5×10−5	0.088	4.7×10−3	8.0×10−3	0.045	8.0×10−5	0.025	0.030

Fp and VT, association tests using SCORE-Seq. Collapsing, association test using ARIEL.

Single-point association analysis showed that, among a total of 196 individual rare variants in European-Americans, 26 SNPs were nominally associated with alcohol dependence (p<0.05). Twenty-five associations survived correction for false discovery rate (q<0.05) and the associations of 5 SERINC2 variants survived Bonferroni correction (p<α=0.0006) (Table 1). If further corrected by the number of cohorts examined (i.e., n=21), 2 variants (i.e., rs35961897 and rs4949405) of these 5 SNPs remained suggestively significant (α=2.9×10⁻⁵). Two independent SNPs, i.e., rs34278290 and rs7417775, can tag these 5 variants (Table 3). The cumulative PV⁺ of these two markers was 0.0791 when we used 3.81% as the one-year prevalence rate of alcohol dependence; it was 0.2366 when we used 12.5% as the lifetime prevalence rate of alcohol dependence. Furthermore, the cumulative LR⁺ of these two markers was 2.169. Two other associations at SERINC2 were replicable (p<0.05) between European-Americans and European-Australians (Table 4). Among these SNPs, rs7417775 at the 3’-UTR of SERINC2 had significant cis-acting regulatory effects on SERINC2 mRNA expression (p=0.024; n=45 unrelated HapMap individuals) [34], and was predicted to affect miRNA binding site activity; this SNP was also the most significant risk variant for Alzheimer’s disease in one Caucasian sample (Tables 1 and 3). Additionally, rs34278290 at intron 2 of SERINC2 was located in a transcript factor binding site. Finally, no significant individual rare variant was associated with any other disease including alcohol dependence in African-Americans (p>α), although this African-American cohort had a 66% power to detect the most significant risk variant, i.e., rs35961897.

Table 3.

Top-ranked risk SNPs (p<α) for alcohol dependence in SERINC2 in European Americans

			MAF		European-Americans
SNP	Gene	Location	Cases	Controls	OR	p-value	q-value
rs34278290◄	SERINC2	Intron 2	0.061	0.044	1.52	5.9×10−4	0.0043
rs2275437	SERINC2	Exon 7	0.057	0.039	1.60	2.0×10−4	0.0025
rs35961897	SERINC2	Intron 10	0.059	0.038	1.69	4.1×10−5	0.0008
rs4949405	SERINC2	Intron 12	0.050	0.030	1.79	4.2×10−5	0.0008
rs7417775●#	SERINC2	3’ UTR	0.033	0.021	1.79	5.3×10−4	0.0043

^◄, located in transcription factor binding sites;

^●, affacting miRNA binding site activity;

^#, having significant cis-acting regulatory effects on SERINC2 mRNA expression (p=0.024).

Table 4.

Replicable risk SNPs (p<0.05) for alcohol dependence in SERINC2 in subjects of European descent

			MAF		European-Americans			Australians		Meta-analysis
SNP	Gene	Location	Cases	Controls	OR	p-value	q-value	OR	p-value	Z-score	P-value
rs77840364	SERINC2	Intron 10	0.030	0.022	1.45	0.029	0.0433	8.00	0.020	2.348	0.0189
rs115360541	SERINC2	Intron 12	0.065	0.043	1.50	0.008	0.0256	2.00	0.019	2.815	0.0049

Furthermore, we did transcriptome-wide mRNA expression correlation analysis in 93 European brain tissues and 80 European Peripheral Blood Mononuclear Cell (PBMC) samples [35]. The expression of NKAIN1-SERINC2 transcripts was significantly correlated with the expression of numerous alcoholism-related genes, mostly from the dopaminergic, serotoninergic, cholinergic, GABAergic, glutamatergic, histaminergic, endocannabinoid, metabolic, neuropeptide and opoidergic systems [18].

Discussion

Mainly from the association tests for the region-wide rare variant constellations, we drew the conclusion that SERINC2 was a replicable and significant risk gene specific for alcohol dependence in the subjects of European descent. Results from individual variant analysis supported this conclusion. Rare SERINC2 variants may contribute a small increase to the risk for alcohol dependence based on their cumulative PV⁺ and LR⁺. Based on these results, we postulate that SERINC2 may harbor a causal variant(s) for alcohol dependence. Our study provides an additional example to support the hypothesis that region-wide rare variant constellations could have significant synthetic effects on disease phenotypes, even though the effects of individual variants might be not significant. Rare variant constellation analysis is an important tool in genetic association studies.

SERINC2 encodes serine incorporator 2 (Serinc2). Serinc2 is highly expressed in neurons of the hippocampus and cerebral cortex [36]. It is an effector in endoplasmic reticulum membranes that incorporates serine into membranes and facilitates the synthesis of phosphatidylserine and sphingolipids [37]. Phosphatidylserine is specifically distributed in the brain. Consumption of phosphatidylserine supplement has been reported to reduce the risk of dementia and cognitive dysfunction in the elderly [38, 39], and thus has been used to treat memory deficit disorders such as Alzheimer’s disease and other forms of dementia, to support cognitive functions during aging, and to remediate cognitive deficits as a result of heavy drinking and cigarette smoking. Also specifically expressed in the brain [37], sphingolipids play a functional role in neural plasticity, signaling and axonal guidance [40–42]. Activity of the sphingolipid metabolism enzyme, i.e., acid sphingomyelinase (ASM), has been reported to be increased in alcohol-dependent patients [43]. Alcohol consumption can increase sphingosine levels in the rat brains [44]. Additionally, there are numerous functional variants in SERINC2 including rare variants (Table 3), common variants and frameshift variants such as Indels and CNVs (see NCBI dbSNPs). The function of Serinc2 altered by the alleles of these functional SERINC2 variants may be implicated in the synthesis of phosphatidylserine and sphingolipids and thus relevant for the development of alcohol dependence. Alternatively, correlation between the expression of NKAIN1-SERINC2 transcripts and other genes suggested that NKAIN1-SERINC2 may contribute to alcohol dependence via other neurotransmitter or metabolic pathways [18]. For example, the glutaminergic pathway is known to play important roles in alcohol intoxication and withdrawal [45]. Within the hippocampus, Serinc2 expression is increased following seizures induced by kainite, a glutamate agonist [37]. A drug that blocks kainite glutamate receptor function appears to decrease drinking [46]. This evidence supports the glutaminergic pathway hypothesis underlying the connection between Serinc2 and alcohol dependence.

A few limitations need to be considered in the current study. The imputed genotypes were not directly observed from molecular experiments, even though their error rates and uncertainty were extremely low. Future work is warranted to verify these results by directly sequencing the samples. Additionally, because not all neuropsychiatric and neurological disorders comorbid with alcohol dependence were exhaustively examined in the present study, we could not completely exclude the possibility that the other neuropsychiatric and neurological disorders not examined might share this SERINC2 risk gene with alcohol dependence. Furthermore, “alcohol drinking behaviors” were not assessed in some of the neurological disorders in the present study. Their potential confounding effects on the association analysis of these disorders need to be assessed. Finally, more independent cohorts with alcohol dependence to replicate our findings in the future are warranted too.