Associations of rare nicotinic cholinergic receptor gene variants to nicotine and alcohol dependence

Abstract

Objective

Nicotine’s rewarding effects are mediated through distinct subunits of nAChRs, encoded by different nicotinic cholinergic receptor (CHRN) genes and expressed in discrete regions in the brain. In the present study, we aimed to test the associations between rare variants at CHRN genes and nicotine dependence (ND) and alcohol dependence (AD).

Methods

A total of 26,498 subjects with 9 different neuropsychiatric disorders in 15 independent cohorts, which were genotyped on Illumina, Affymetrix or PERLEGEN microarray platforms, were analyzed. Associations between rare variants (minor allele frequency (MAF) < 0.05) at CHRN genes and nicotine dependence and alcohol dependence were tested. The mRNA expression of all Chrn genes in whole mouse brain and ten specific brain areas was investigated.

Results

All CHRN genes except the muscle-type CHRNB1, including eight genomic regions containing 11 neuronal CHRN genes and three genomic regions containing four muscle-type CHRN genes, were significantly associated with ND and/or AD. All of these genes were expressed in the mouse brain.

Conclusion

We conclude that CHRNs are associated with ND (mainly) and AD, supporting the hypothesis that the full catalog of ND/AD risk genes may contain most neuronal nAChRs-encoding genes.

Introduction

There are two major subtypes of acetylcholine receptors (AChR) — metabotropic muscarinic AChRs and ionotopic nicotinic AChRs (nAChRs). Both are activated by the endogenous neurotransmitter acetylcholine. The nAChRs can be either neuronal or muscle-type. The former are heavily expressed in central and peripheral nervous systems and in some non-neuronal tissues such as the adrenal medulla. In contrast, the muscle-type receptor is localized primarily at neuromuscular junctions. nAChRs are pentameric ligand-gated ion channels composed of membrane-spanning subunits arranged around a cation channel. [Lindstrom 2003] Each subunit is encoded by a single nicotinic cholinergic receptor (CHRN) gene. Neuronal nAChR subunits include α2 through α10 and β2 through β4 that are encoded by CHRNAs 2 through 10 and CHRNBs 2 through 4, respectively (Supplementary Table S1). The muscle-type nAChR subunits include α1, β1, γ, δ, and ε that are encoded by CHRNA1, B1, G, D and E, respectively [Gotti and others 2007] (Supplementary Table S1). nAChRs are typically heteropentamers composed of α- and β-subunits in different ratios or homopentamers of α subunits. Two main classes of neuronal nAChRs are known: α-bungarotoxin (αBgtx)-sensitive receptors composed of heteromeric α7–α10-subunits or a homomeric form (e.g., (α7)5); and αBgtx-insensitive receptors composed of heteromeric α2–α6 and β2–β4-subunits and binding nicotine but not αBgtx (such as α4β2* and α3β4* receptors, where * indicates other subunits). Homopentameric receptors such as (α7)5 nAChRs have five identical acetylcholine-binding sites in each molecule, whereas the heteropentameric receptors have only two. In each heteromeric receptor, two α subunits (α2–α4 or α6) carry the principal acetylcholine-binding sites. The two non-α subunits (β2 or β4) carry complementary components of the binding sites. The fifth accessory subunit (α5, β2, β3, or β4) does not participate in binding.

Distinct neuronal nAchRs are distributed in different parts of the CNS and contribute to the psychoactive properties of nicotine and to the etiology of many neuropsychiatric illnesses. Here we focused on the associations of nAchRs with nicotine dependence (ND) and alcohol dependence (AD). Although the ND- or AD- related traits have been assessed using different approaches in numerous genetic association studies, they have moderate agreement among them. Variants in most of the CHRN gene regions have been related to ND-related traits (Supplementary Table S2) [Chen and others 2013; Cui and others 2013; Fowler and others 2011; Greenbaum and others 2009; Keskitalo-Vuokko and others 2011; Lou and others 2006; Pergadia and others 2009; Philibert and others 2009; Saccone and others 2010; Stephens and others 2013; Thorgeirsson and others 2008; Thorgeirsson and others 2010; Wang and others 2014; Xie and others 2011; Yang and others 2014]. The majority of studies have focused on the CHRNA5-A3-B4 cluster [Berrettini and others 2008; Bierut and others 2007; Bierut and others 2008; Caporaso and others 2009; Fowler and others 2011; Greenbaum and others 2009; Hartz and others 2012; Li 2008; Liu and others 2010; Saccone and others 2009; Saccone and others 2010; Saccone and others 2007a; Stephens and others 2013; Thorgeirsson and others 2008; Thorgeirsson and others 2010; Tobacco and Genetics Consortium 2010; Wang and others 2009a] which is most reliably associated with ND-related traits and the association has been verified by many functional studies. Some associations at this locus were highly significant with p values as low as 10⁻⁷³. Additionally, a large body of neurobiological and genetic evidence supports an important role for the CHRNA7 gene in ND [Greenbaum and others 2006; Saccone and others 2010]. Several other genome-wide association studies (GWASs) identified association peaks for ND at CHRNA4 [Bierut and others 2007; Wei and others 2012], which has been replicated in the same and numerous other studies [Bierut and others 2007; Feng and others 2004; Li and others 2005; Saccone and others 2009; Saccone and others 2010; Wei and others 2012]. Many GWASs demonstrated an association of CHRNB3–CHRNA6 with ND [Bierut and others 2007; Saccone and others 2007a; Saccone and others 2007b; Thorgeirsson and others 2010], and the findings were replicated by candidate gene studies [Culverhouse and others 2014; Hoft and others 2009; Saccone and others 2009; Saccone and others 2010; Wei and others 2012; Won and others 2014]. Evidence also supports involvement of seven other CHRN loci (except for CHRNE) in genetic susceptibility to ND, including neuronal CHRNA2 [Bergen and others 1999; Faraone and others 2004; Heitjan and others 2008; Swan and others 2006; Yang and others 2014], CHRNA9 [Ehringer and others 2010; Yang and others 2014], CHRNA10 [Pergadia and others 2009; Saccone and others 2010], CHRNB2 [Conti and others 2008; Perkins and others 2009] and muscle-type CHRNA1, CHRNB1 [Philibert and others 2009; Saccone and others 2009; Saccone and others 2010] and CHRND/G [Keskitalo-Vuokko and others 2011; Saccone and others 2009; Saccone and others 2010]. Finally, CHRNA5-A3-B4 [Wang and others 2009b], CHRNA7 [Kamens and others 2010], CHRNA4 ^{[Coon and others 2014]}, CHRNB3-A6 [Haller and others 2014] and CHRNB2 [Ehringer and others 2007] have also been associated to AD or alcohol use related traits (Supplementary Table S2), although these associations were much weaker than those with ND.

An important discovery in human genetics is that many diseases appear to be caused by constellations of multiple rare, regionally concentrated variants, rather than by common variants [Dickson and others 2010]. The combined effects of rare variants distributed across the genome may be highly significant. The main goal of the present study is to test whether or not rare variants in CHRNs are associated with both ND and AD. We are also interested in possible contributions of CHRN variants to schizophrenia [Lin and others 2014], depression [Philip and others 2010], Alzheimer’s disease [Lombardo and Maskos 2015], Parkinson’s disease [Burghaus and others 2003], amyotrophic lateral sclerosis [Sabatelli and others 2009] and ischemic stroke [Liu and others 2014].

Materials and Methods

Subjects

We initially evaluated a total of 27 independent cohorts with different neuropsychiatric disorders as available from the database of Genotypes and Phenotypes (dbGaP). In the present study, we only analyzed case-control data, and excluded six family-based datasets. Six other datasets with absolute genomic inflation factor (λ) larger than 1.1 as derived from QQ plots were also excluded (data not shown). Consequently, 15 datasets as listed in Supplementary Table S3 were included in the analyses. These 15 independent cohorts included a total of 26,498 subjects with 9 different neuropsychiatric disorders, genotyped on Illumina, Affymetrix or PERLEGEN microarray platforms. These 9 disorders included ND, AD, major depression, schizophrenia, Alzheimer’s disease, ALS, early onset stroke, ischemic stroke and Parkinson’s disease. Detailed demographics data are shown in Supplementary Table S3.

These subjects contained four cohorts with ND or AD: (1) COGEND+UW-TTURC (EA) cohort included 836 European-American cases with ND (15.4% were AD), and 80 European-American controls (17.8% were AD); (2) COGEND+UW-TTURC (AA) cohort included 392 African-American cases with ND (7.5% were AD), and 79 African-American controls (18.6% were AD); (3) SAGE+COGA (EA) cohort included 1,409 European-American cases with AD (58% were ND) and 1,518 European-American controls; and (4) SAGE+COGA (AA) cohort included 681 African-American cases with AD (66% were ND) and 508 African-American controls. Affected subjects with ND were defined by a commonly used definition of ND – a current score of 4 or more (out of a maximum score of 10) on the Fagerström Test for Nicotine Dependence (FTND). Control status was defined as an individual who smoked at least 100 cigarettes during their lifetime, yet never became dependent (lifetime FTND=0). Affected subjects with AD met DSM-IV criteria for AD [American Psychiatric Association 1994] and controls were defined as individuals who had been exposed to alcohol (and possibly to other drugs), but had never become dependent on these substances. Additionally, all cases and controls in these four cohorts were screened to exclude individuals with major axis I disorders, including schizophrenia, mood disorders, and anxiety disorders. This study was approved by the Institutional Review Board of Yale University.

Imputation and data cleaning

To make the genetic marker sets consistent across different cohorts, we imputed the untyped SNPs across each CHRN gene region (detailed in Supplementary Methods). We stringently cleaned the phenotype and genotype data within each ethnicity prior to association analysis (detailed in Supplementary Methods).

Association tests for individual rare variants

A total of 179–429 (in subjects of European descent) and 136–620 (in subjects of African descent) cleaned rare SNPs with 0<MAF<0.05 in controls were extracted for association analysis (Table 1). In each cohort, the allele frequencies of each SNP were compared between cases and controls using logistic regression as implemented in PLINK [Purcell and others 2007]. Diagnosis and alleles each served as the dependent and independent variables, with sex, age and the first 10 principal components of ancestries as covariates. In non-AD and non-ND cohorts, “alcohol drinking” and “smoking”, if available, were included as additional covariates 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. An α was set to guard against type I error with Bonferroni correction. The false discovery rate (q value) for each SNP was estimated from the p values within each cohort using the R package QVALUE [Storey and Tibshirani 2003].

Table 1.

Associations between individual rare CHRN variants and different neuropsychiatric disorders

Human Diseases	Ethnicity	Dataset name	SNP # (total)	SNP # (p<0.05)	Most significant		Minimal p value	Minor allele frequency (N)
					SNP	Gene		Affected	Unaffected
Nicotine dependence	EA	COGEND+UW-TTURC	268	8	rs60757049	CHRNA4	0.0104	0.004 (836)	0.047 (80)
Nicotine dependence	AA	COGEND+UW-TTURC	235	4	rs2231540	CHRNA10	0.0021	0.124 (392)	0.044 (79)
Alcoholism	EA	SAGE+COGA	429	15	rs1346725	CHRNA2	0.0082	0.001 (1409)	0.005 (1518)
AD+ND	EA	SAGE+COGA	413	6	rs13040031	CHRNA4	0.0175	0.032 (1409)	0.016 (1518)
Alcoholism	AA	SAGE+COGA	620	15	rs76706995	CHRNA9	5.4×10−4	0.031 (681)	0.010 (508)
Schizophrenia	EA	GAIN	203	4	rs2231547	CHRNA10	0.0244	0.031 (1351)	0.044 (1378)
Schizophrenia	AA	GAIN	596	10	rs79576636	CHRNG	8.4×10−4	0.008 (1195)	0.0006 (954)
Schizophrenia	EA	MGS_nonGAIN	212	7	rs1399190	CHRNA7	0.0182	0.006 (1437)	0.002 (1347)
Major Depression	CA	PRSC	412	5	rs6011783	CHRNA4	0.0221	0.036 (1805)	0.026 (1820)
Alzheimer’s disease	EA	GenADA	226	4	rs8834	CHRNE	0.0018	0.054 (806)	0.022 (782)
Parkinson’s Disease	CA	NGRC	398	26	rs62518191	CHRNB3	8.5×10−4	0.024 (2000)	0.014 (1986)
Parkinson’s Disease	CA	lng_coriell_pd	388	8	rs75628551	CHRNB3	9.8×10−4	0.0006 (940)	0.009 (801)
ALS	CA	GRU	297	9	rs66850163	CHRNA7	0.0076	0.041 (261)	0.007 (246)
Ischemic Stroke	CA	ISGS	302	23	rs16956147	CHRNA7	0.0075	0.014 (219)	0.045 (266)
Early Onset Stroke	EA	GEOS × 3	321	3	rs2273504	CHRNA4	0.0040	0.074 (372)	0.028 (430)
Early Onset Stroke	AA	GEOS × 3	538	8	rs28454587	CHRNA9	0.0253	0.041 (309)	0.018 (290)

Dataset names refer to dbGaP. Only the most significant risk markers with minimal p values are listed; “minimal p-value” is the smallest p-value among all the SNPs in all genes studied for a particular dataset. ALS, Amyotrophic Lateral Sclerosis; AA, African-American; EA, European-American; CA, Caucasian; N, sample size of subjects. 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].

Association tests for family-wide rare variant constellations with diseases

Sixteen CHRN genes in 12 regions (Supplementary Table S1) have similar sequences and the subunits encoded by these genes have similar structures and might share similar biological features. Five subunits comprise one nAch receptor; that is, five subunits integrate to exert the receptor function. Thus, global synthetic effects of rare variant constellations across the whole family of CHRNs might exist. We tested these effects using a score-type program, SCORE-Seq [Lin and Tang 2011]. The mutation information was aggregated by virtue of a weighted linear combination across all rare variants of all CHRN genes/regions, and then related to each disease phenotype using regression models. Sex, age and the first 10 principal components of ancestries, as estimated using the program EIGENSTRAT [Price and others 2006], served as the covariates in the regression models. The first 10 principal components explained more than 95% of genetic variance in our samples. In non-AD and non-ND cohorts, “alcohol drinking” and “smoking”, if available, were included as additional covariates in the models.

We first analyzed the associations between diseases and rare variant constellations where the rare variants came with 0<MAF<0.01 (T1 test). If it was significant, we then tested those rare variants with 0<MAF<0.05 (T5 test) to validate the T1 test results. We also employed the Fp test to validate the findings: (1) In the T1 and T5 tests, the MAF upper bound thresholds were fixed at 0.01 and 0.05, respectively, and the weight was fixed at 1; and (2) In the Fp tests, the MAF upper bound threshold was fixed at 0.05, and the weight was set as 1/sqrt(p(1−p)) where p was the estimated MAF with pseudo counts in the pooled sample. Statistical significance was assessed using a bootstrap procedure with 1 million times of resampling [Lin and Tang 2011] and corrected for the number of effective datasets (n=15). All association analyses were performed within the same ethnicity.

Association tests for single-locus rare variant constellations with diseases

After significant associations between family-wide rare variant constellations and diseases in some cohorts were identified, we decomposed these significant associations to identify the significant single-locus rare variant constellations across each gene region. The same analytic procedure was performed as the above and the results were examined with statistical correction for the number of gene regions (n=12). Finally, we meta-analyzed the two datasets with ND, two datasets with AD, and two sub-datasets with ND+AD, respectively, using the program MetaP (http://igm.cumc.columbia.edu/MetaP/metap.php).

Bioinformatics analysis

To explore the potential functions of the individual rare variants included in the rare variant constellations, we predicted their functions using a series of bioinformatics analyses. We used UCSC Genome Browser data or other bioinformatics analysis software packages (e.g., FuncPred [Xu and Taylor 2009] or VE!P [McLaren and others 2010]) to see whether the risk SNPs are located within methylated CpG islands, within copy number variations (CNVs) or in exonic splicing silencers (ESS) or enhancers (ESE). Additionally, Polyphen [Adzhubei and others 2013] and SIFT [Ng and Henikoff 2003] were applied to predict the pathogenicity in order to see whether these rare variants affect protein function or structure. The conservation of these SNPs across 17 species was also predicted [King and others 2005].

Detection of CHRN mRNA expression in mouse brains

The mRNA expression of all Chrn genes in whole mouse brain and ten specific brain areas was investigated (detailed in Supplementary Methods).

Results

No individual rare variant was significantly associated with diseases (Table 1)

Among the rare variants, 0–26 variants were each nominally associated with diseases in different cohorts (p<0.05). Minimal p value in each cohort was between 5.4×10⁻⁴ and 0.134. However, after Bonferroni correction, no variants were significantly associated with diseases (p>α). The q values for all SNPs were > 0.05.

There were significant associations between family-wide rare variant constellations and ND and AD in four cohorts (Table 2)

Table 2.

p values for associations between global CHRN rare variant constellation and diseases

Diseases	Dataset	T1	T5	Fp
Nicotine dependence	COGEND+UW-TTURC (EA)	5.0×10−18	8.4×10−54	1.5×10−58
Nicotine dependence	COGEND+UW-TTURC (AA)	1.3×10−4	1.9×10−35	3.9×10−33
Alcoholism	SAGE+COGA (EA)	0.544	0.213	0.267
AD+ND	SAGE+COGA (EA)	1.5×10−8	8.6×10−17	1.0×10−17
Alcoholism	SAGE+COGA (AA)	7.9×10−4	3.5×10−6	1.7×10−6
AD+ND	SAGE+COGA (AA)	0.987	0.569	0.626
Schizophrenia	GAIN (EA)	0.665	0.523	0.435
Schizophrenia	GAIN (AA)	0.302	0.033	0.024
Schizophrenia	MGS_nonGAIN (EA)	0.996	0.295	0.274
Major Depression	PRSC	0.127	0.198	0.197
Alzheimer’s disease	GenADA	0.711	0.539	0.400
Parkinson’s Disease	NGRC	0.669	0.197	0.206
Parkinson’s Disease	lng_coriell_pd	0.177	7.1×10−3	8.0×10−3
ALS	GRU	0.523	0.149	0.152
Ischemic Stroke	ISGS	0.427	0.324	0.405
Early Onset Stroke	GEOS × 3	0.311	0.054	0.061
Early Onset Stroke	GEOS × 3	0.371	0.477	0.390

EA, European-American; AA, African-American. Dataset names correspond to Table 1. T1, T5 and Fp, association tests using SCORE-Seq: (1) In the T1 and T5 tests, the MAF upper bound thresholds were fixed at 0.01 and 0.05, respectively, and the weight was fixed at 1; (2) 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. α=0.05/15=3.3×10⁻³ for T1 test.

There are no significant associations between global rare variant constellations across the whole family of CHRN gene regions and major depression, schizophrenia, Alzheimer’s disease, amyotrophic lateral sclerosis, early onset stroke, ischemic stroke and Parkinson’s disease (p>0.05/15=3.3×10⁻³).

However, there are significant associations between these family-wide rare variant constellations and ND in EAs and AAs, and AD in AAs (1.5×10⁻⁵⁸≤p≤7.9×10⁻⁴). The T5 and Fp tests validated the T1 test very well, with more significant results than the latter.

Although there was no significant association between family-wide rare variant constellations and AD in EAs, there was significant association with nicotine and alcohol co-dependence (ND+AD) in the subset of this EA sample (p values: T1=1.6×10⁻⁸; T5=8.6×10⁻¹⁷ and Fp=1.1×10⁻¹⁷).

We found that although family-wide rare variant constellations were associated with AD comorbid with ND (e.g., 1.7×10⁻⁶≤p≤7.9×10⁻⁴ in AAs that included 66% ND, and 1.1×10⁻¹⁷≤p≤1.6×10⁻⁸ for AD+ND in EAs that was ND too), associations were much more significantly associated with ND-alone (e.g., 1.9×10⁻³⁵≤p≤1.3×10⁻⁴ for ND in AAs, and 1.5×10⁻⁵⁸≤p≤5.0×10⁻¹⁸ for ND in EAs).

There were significant associations between eleven single-locus rare variant constellations across each CHRN gene/region and ND or AD in the above four cohorts (Tables 3 and S4)

Table 3.

p values for associations between rare CHRN variant constellation at each gene/region and diseases

Gene/region	Chr	Length (bp)	Number of rare SNPs	Nicotine dependence in 1st EAs			Nicotine dependence in 1st AAs			ND+AD in 2nd EAs.			Alcohol dependence in 2nd AAs
				T1	T5	Fp	T1	T5	Fp	T1	T5	Fp	T1	T5	Fp
CHRNB2	1	12098	38	N/A	1.3×10−18	4.9×10−18	N/A	2.4×10−6	9.5×10−12	0.580	5.1×10−5	3.0×10−3	0.132	1.9×10−3	1.2×10−3
CHRNA1	2	16881	112				0.030	2.5×10−11	1.2×10−14
CHRND-CHRNG	2	20169	82							5.9×10−5	1.5×10−7	9.4×10−8
CHRNA9	4	19889	167	0.034	5.0×10−19	6.1×10−30				0.020	6.5×10−8	1.9×10−8	0.039	4.0×10−6	3.5×10−6
CHRNA2	8	19536	120	2.1×10−4	5.2×10−13	3.3×10−12
CHRNB3-CHRNA6	8	71368	329	1.3×10−4	1.5×10−15	3.8×10−16
CHRNA10	11	7622	56	4.3×10−6	3.4×10−29	1.2×10−25							7.5×10−3	9.1×10−6	1.4×10−6
CHRNA7	15	36080	565	1.1×10−6	1.5×10−19	7.6×10−22	4.9×10−3	2.4×10−25	6.1×10−27	6.5×10−3	8.9×10−9	3.5×10−9	4.5×10−3	5.1×10−4	4.0×10−4
CHRNA5-CHRNA3-CHRNB4	15	90067	374	2.2×10−4	1.9×10−25	4.6×10−32	2.6×10−3	1.5×10−16	1.3×10−22	8.0×10−4	1.1×10−9	9.5×10−11
CHRNE	17	5360	35	5.3×10−4	4.9×10−15	9.1×10−14
CHRNA4	20	18087	175	8.3×10−3	9.8×10−39	1.9×10−39				9.9×10−3	2.7×10−12	2.3×10−10

T1, T5 and Fp, association tests using SCORE-Seq. EA, European-American; AA, African-American; ND+AD, nicotine and alcohol co-dependence. N/A, not available because of lack of rare variants with MAF<0.01. α=0.05/12=0.004 in which “12” is the number of gene regions. P values larger than α are not listed in the table.

CHRNA10 (T1=7.5×10⁻³, T5=9.1×10⁻⁶ and Fp=1.4×10⁻⁶), CHRNA7 (T1=4.5×10⁻³, T5=5.1×10⁻⁴, and Fp=4.0×10⁻⁴) and CHRNA9 (T1=0.039, T5=4.0×10⁻⁶ and Fp=3.5×10⁻⁶) were nominally significantly associated with AD in AAs. Similarly, CHRNA4, CHRNA5-CHRNA3-CHRNB4, CHRNA7, CHRNA9 and CHRND-CHRNG were nominally significantly associated with AD+ND in EAs. CHRNA1, CHRNA5-CHRNA3-CHRNB4 and CHRNA7 were nominally significantly associated with ND in AAs. CHRNA10, CHRNA2, CHRNA4, CHRNA5-CHRNA3-CHRNB4, CHRNA7, CHRNA9, CHRNB3-CHRNA6 and CHRNE were nominally significantly associated with ND in EAs. Among these associations, CHRNA7 was nominally significantly and replicably associated with ND, ND+AD or AD across four cohorts (Table 3). All of these associations remained to be nominally significant in the meta-analysis of EAs and AAs (Table S4). After correction for 12 regions (α=4.2×10⁻³ for T1 test), CHRNA5-CHRNA3-CHRNB4 and CHRND-CHRNG were significantly associated with AD+ND in EAs; CHRNA5-CHRNA3-CHRNB4 was significantly associated with ND in AAs; CHRNA10, CHRNA2, CHRNA5-CHRNA3-CHRNB4, CHRNA7, CHRNB3-CHRNA6 and CHRNE were significantly associated with ND in EAs. All of these associations remained significant in the meta-analysis of EAs and AAs (Table S4). Among these associations, CHRNA5-CHRNA3-CHRNB4 was significantly and replicably associated with ND in EAs and AAs and with AD+ND in EAs. The T5 and Fp tests for these associations were all highly significant and verified the T1 test.

Bioinformatics analysis (Table 4)

Table 4.

Bioinformatics analysis for the missense rare variants

SNP	Gene	AAs	Bioinformatics analysis
rs61737716	CHRNA1	E/D	probably damaging
rs2289080	CHRNG	A/T	splicing (ESE or ESS)
rs35327613	CHRNB3	K/E
rs2231546	CHRNA10	G/S
rs2231547	CHRNA10	E/A	deleterious; conservative (1.000)
rs55719530	CHRNA10	T/N	deleterious; probably damaging; splicing (ESE or ESS)
rs16969968	CHRNA5	D/N
rs79109919	CHRNA5	L/Q	deleterious; probably damaging
rs79835149	CHRNA5	Y/*	stop gained
rs12914008	CHRNB4	T/I	deleterious; possibly damaging; splicing (ESE or ESS); conservative (0.941)
rs56095004	CHRNB4	R/Q	splicing (ESE or ESS)
rs61737499	CHRNB4	T/I
rs76927517	CHRNB1	I/T	deleterious; possibly damaging; 538bp-CpG
rs4790235	CHRNE	G/V	probably damaging; splicing (ESE or ESS)

AAs, amino acid changes; splicing (ESE or ESS), these SNPs are located in exonic splicing silencers or enhancers; deleterious/probably damaging, these SNPs are predicted by SIFT/Polyphen to significantly affect protein function or structure; conservative (score), these SNPs are predicted to be conservative across species; CpG, this SNP is located within a 538bp methylated CpG island;

^*stop gained, this SNP results in a premature stop codon, leading to a shortened transcript.

Almost all variants with potential functions were missense variants located in exons. A total of 14 potentially functional SNPs were identified (Table 4). Among them, rs76927517 at CHRNB1 is located within a 538bp CpG island, whose methylation status may affect the expression of CHRNB1. Five SNPs are located in exonic splicing silencers or enhancers and thus might influence the splicing activity. rs2231547 at CHRNA10 and rs12914008 at CHRNB4 are predicted to be conservative across species. Six SNPs are predicted to be damaging and five SNPs are predicted to be deleterious, suggesting that they may significantly affect the protein function or structure. Finally, rs79835149 at CHRNA5 may result in a premature stop codon, leading to a shortened transcript.

All Chrn genes were expressed in mouse brain, in distinct areas and at different levels (Table 5)

Table 5.

Chrna gene expression at whole brain and different brain areas of BXD mice

Gene	Location (Chr, Mb)	Whole brain	Cortex	Striatum	NAc	Hypothalamus	Hippocampus	Amygdala	Mid-brain	VTA	Cerebellum	Pituitary
Chrnb2	Chr3: 89.561004	8.39	9.54	8.65	7.93	11.15	10.22	10.66	9.98	7.90	7.94	10.60
Chrna1	Chr2: 73.403930	7.10	7.10		7.01	7.08	7.90				7.54	7.90
Chrnd	Chr1: 89.095164	9.15			8.58		7.54			8.54	7.70
Chrng	Chr1: 89.107206	7.55		7.29		7.49	8.62	7.43			8.00	7.44
Chrna9	Chr5: 66.367985	7.35	7.30			7.03	7.50	7.34	7.90		7.08	7.55
Chrna2	Chr14: 66.771218	7.43	7.14	7.33			7.85	7.59	8.87
Chrnb3	Chr8: 28.504645	7.85	7.11			7.11	7.25			7.65		9.03
Chrna6	Chr8: 28.513939	8.73			7.79	7.05	7.13		10.37	8.97	7.23	8.68
Chrna10	Chr7: 109.259780		7.08				7.43	7.44
Chrna7	Chr7: 70.243861	7.98	7.19	9.33	7.49	9.78	10.88	9.95	8.28	7.53	7.16	7.50
Chrna5	Chr9: 54.852890		7.22				8.44	7.45
Chrna3	Chr9: 54.860390	9.35			7.22	8.20	8.32			7.60	7.61
Chrnb4	Chr9: 54.877893	7.23		7.58		8.26	8.65	8.18	8.23			8.17
Chrnb1	Chr11: 69.597667			7.01			7.50					8.11
Chrne	Chr11: 70.428505	8.13	7.05		7.12	7.32	7.63	7.34				7.20
Chrna4	Chr2: 180.759407	9.48	10.05	9.02	8.29	11.09	9.95	10.73	9.64	8.71	8.30	9.23

The order of gene list corresponds to Table 3. Only the expression values > 7 were listed.

We found that all Chrn genes were expressed in mouse brain at different levels. All of these genes were expressed in the hippocampus, in which the genes with most highly abundant expression [standardized expression value (SEV) > 9] were Chrna7 (SEV=10.88), Chrnb2 (SEV=10.22) and Chrna4 (SEV=9.95). These three were also modestly to highly abundantly expressed in other brain areas examined (SEV=7.03–11.15), with 1.0–17.8-fold changes (FC) in mRNA level compared to the expression base, among which Chrna7 was most abundantly expressed in the hippocampus (SEV=10.88) and Chrnb2 and Chrna4 were most abundant in the hypothalamus (SEV=11.15 and 11.09, respectively).

Chrna5, Chrna3 and Chrnb4 were expressed in multiple brain areas (SEV=7.22–9.35), with a 1.2–5.1-fold change in mRNA expression levels compared to the expression base (SEV=7). Chrna6 and Chrnb3 were expressed in several areas (SEV=7.05–10.37), among which a6 was abundantly expressed in the midbrain (FC=10.4) and Chrnb3 abundantly in the pituitary (FC=4.1). Chrna2 was modestly to abundantly expressed in multiple areas (SEV=7.14–8.87) with 1.1–3.7-fold changes too. Both Chrna9 and Chrna10 were modestly expressed in the brain (SEV<8) that included the cortex. Interestingly, the muscle-type Chrn genes including Chrna1, Chrnb1, Chrnd, Chrne and Chrng were expressed at a low level in most brain areas (SEV<8), whereas Chrnb1 in the pituitary, Chrnd in the whole brain, ventral tegmental area (VTA) and nucleus accumbens (NAc), Chrne in the whole brain and Chrng in the hippocampus and cerebellum were expressed at abundant levels (SEV=8.00–9.15).

Discussion

In the present study, we investigated a total of 16 CHRN genes in 12 genomic regions, including 11 neuronal CHRN genes in 8 genomic regions and 5 muscle-type CHRN genes in 4 genomic regions (Supplementary Table S1). We found that all CHRN genes but the muscle-type CHRNB1 were significantly associated with ND and/or AD. Further, all of these genes were expressed in the mouse brain, which supports previous findings and hypothesis that the development of ND and AD involves more than one nAChR subunits.

Nicotine’s psychoactive effects appear to be mediated through distinct subunits of nAChRs, encoded by different genes and expressed in discrete regions in the brain [Lindstrom 2003]. Adult midbrain neurons express a variety of nAChR subunit mRNAs, including α2, 4, 5, 6, 7, 9, β2 and β4, as confirmed by this study, although the α5 expression in the midbrain is very low in our study (SEV=6.6; data not shown). Because the midbrain, including the VTA, is widely implicated in drug dependence [Subramaniyan and Dani 2015], these receptor subunits are all plausible candidates for a functional role in mediating nicotine effects on addiction. For instance, a large portfolio of different nAChR subtypes expressed in the VTA may define dopaminergic neuronal subpopulations with different implications for the reinforcing effects of nicotine [Faure and others 2014]. In addition, the number of binding sites per pentamer varies from two to five, depending on its composition, and the non-equivalent binding sites formed by different subunits have different affinities for agonists and antagonists. For example, five functional nAChR subtypes in dopaminergic terminals in the striatum have different affinities for nicotine; that is, α4α6β2β3, α6β2β3 and α6β2 have the highest sensitivity to nicotine whereas α4β2 and α4β2α5 have lower affinity for nicotine [Champtiaux and others 2002; Drenan and others 2008; Grady and others 2007]. The expression of α2, 4 and 7 and β1, 2 and 4 in mouse striatum was confirmed by our study.

The most robust associations, which were highly significant (even after correction for multiple comparisons) and replicable, were found between CHRNA5-A3-B4 and ND/AD, consistent with a previous report [Greenbaum and Lerer 2009]. We found that this gene region was significantly associated with ND in EAs and AAs, and with AD+ND in EAs, suggesting shared ND and AD susceptibility on this region. In support, variants in this gene cluster have been reported to influence and promote the early use of tobacco and alcohol [Schlaepfer and others 2008]. Further, a partial agonist (CP-601932) for the α3β4* nAChRs has been used as smoking cessation therapy and may have utility too in decreasing alcohol consumption and seeking in cigarette smokers [Chatterjee and others 2011; McKee and others 2009].

CHRNA5 and CHRNA3 are in high linkage disequilibrium (LD), but they are in lower LD with CHRNB4 [Berrettini and others 2008]. In humans, mRNA encoding the α5 subunit has been detected in the nervous system (including the cortex) but appears to be mainly expressed in the cerebellum, medial habenula, and the autonomic ganglia [Flora and others 2000]. In particular, α4β2α5-containing receptors are expressed on dopaminergic neurons in rat and mouse striatum and are involved in nicotine-elicited dopamine release [Salminen and others 2004; Zoli and others 2002]. In addition, α5-containing receptors are also expressed on GABAergic neurons in the striatum and VTA. α3-containing nAChRs (mainly αBgtx-insensitive α3β4* receptor) were found in several regions of the rodent brain, including the VTA, pineal gland, medial habenula, interpeduncular nucleus (IPN), brainstem nuclei, hippocampus and cerebellum [Gahring and others 2004; Perry and others 2002; Turner and Kellar 2005]. Both CHRNA3 and CHRNA5 genes are expressed in the neural circuits of reward processing and learning as well as cognitive and affective control, involving the NAc, entorhinal cortex, cerebellum and amygdala. Our study confirmed that these two genes were expressed in the hippocampus, VTA, NAc, cerebellum, amygdala, hypothalamus and cortex in mice. Unfortunately, we did not examine the expression of CHRNA5, CHRNA3 and CHRNB4 in medial habenula where these genes were reported to be highly expressed. Many of these brain regions in humans are implicated in the etiological process of ND. For instance, nicotine activates the habenulo-interpeduncular pathway through α5-containing nAChRs, which can trigger an inhibitory motivational signal that acts to limit nicotine intake [Fowler and others 2011]. Future research combining optical imaging and mouse genetic models may significantly advance our understanding of the roles of these genes and the molecular mechanisms underlying ND [Pistillo and others 2015].

The nicotinic β4-subunit is thought to be widely expressed in the mouse peripheral nerve system, with a restricted expression pattern in the central nervous system (CNS). In CNS, it was previously observed at high levels only in the medial habenula and IPN [Klink and others 2001; Quick and others 1999; Salas and others 2004]. Functional data have shown changes in nicotine intake following manipulations of β4 in medial habenula, supporting its role in the reinforcing effect of nicotine [Rose 2007]. It was observed at relatively low densities in the rat and mouse substantia nigra and VTA [Klink and others 2001; Salas and others 2003] but not in NAc, cerebellum or cortex, in accord with with our findings.

α3 is a principal subunit that could harbor a binding site. α5 and β4 subunits do not have acetylcholine binding sites and cannot form functional nAChRs by themselves. They may contribute to receptor targeting and localization in the neuronal membrane, but do not play a direct role in receptor–ligand binding, suggesting their auxiliary roles when expressed with other α- and β-subunits in conferring risk for ND.

Another robust association, which was highly significant and replicable across all four cohorts, was also found between CHRNA7 and ND/AD. We found that this gene was significantly associated with ND both in EAs and AAs, with ND+AD in EAs, and with AD in AAs. ND and AD share rare risk variant constellations at CHRNA7. The homomeric α7-receptors were previously reported to be widely expressed within the CNS [Brody and others 2006], including the VTA, prefrontal and motor cortex, basal ganglia, hippocampus, cerebellum, and lateral and medial geniculate bodies, and with lower expression in the thalamus [Conejero-Goldberg and others 2008]. Upregulation of α7-subunit containing nAChRs after chronic nicotine treatment has been reported in mouse striatum, midbrain and cortex [Nuutinen and others 2005; Pakkanen and others 2005], in some but not all studies. Our findings confirmed that α7 was abundant in the hippocampus, midbrain, VTA, NAc, cerebellum, striatum, amygdala, hypothalamus, pituitary and cortex. The hippocampus and neocortex are involved in learning, memory and emotion regulation [Gotti and others 2007; Klink and others 2001] and the VTA–NAc pathway is critical in drug-induced rewards [Benowitz 2008], supporting a role of α7 in ND and AD.

We found that CHRNA4 was significantly associated with ND and ND+AD in EAs. The α4 subunit usually combines the β2 subunit to form a neuronal α4β2* nAchR. This receptor is αBgtx-insensitive and the predominant receptor subtype in the brain of humans and other mammalian species. Unlike that the expression pattern of other subunits is much more restricted to particular brain regions [Gotti and Clementi 2004; Gotti and others 2006; Wilens and Decker 2007], the α4β2* receptor is widely expressed within the CNS, [Brody and others 2006] including the mesolimbic dopamine system and the VTA–NAc pathway. We confirmed in mice that both α4 and β2 were expressed across all eleven brain regions examined and were highly abundant in hippocampus, midbrain, amygdala, hypothalamus, pituitary and cortex (SEV>9). Previous evidence suggested highest expression of α4, and β2 in the thalamus, [Picciotto and others 1998] which we unfortunately did not examine in this study. The α4 is a principal subunit with an acetylcholine binding site; thus, the α4β2* receptor has high affinity for nicotine [Hogg and others 2003] and is sufficient to mediate nicotine reward, tolerance and sensitization [Tapper and others 2004]. This receptor is upregulated in the brains of smokers after chronic nicotine exposure [Breese and others 1997]. The α4β2* receptor may co-assemble with additional units to form more complex subtypes such as (α4β2)2α3, (α4β2)2α6, (α4β2)2α5 or (α4β2)2β3 [Gotti and others 2006; Quik and others 2000]. Different α4/β2 ratios in the receptor can lead to different pharmacological and functional properties. For example, (α4β2)2α5 acetylcholine receptor may play a role in conferring the aversive experience of nicotine use, limiting the drive to smoke.

CHRNB3-CHRNA6 was significantly associated with ND in EAs. Long-term self-administration of nicotine has been found to upregulate α6- and β3-containing receptors in rat brain [Parker and others 2004]. The α6- and β3-subunit upregulation is of special interest because these subunits are co-expressed as α6β2β3* or α4α6β2β3 nAChRs [Champtiaux and others 2003; Gotti and others 2007; Salminen and others 2007] in midbrain dopamine neurons and modulate dopamine release [Champtiaux and others 2002; Exley and others 2008; Meyer and others 2008]. In rat and mouse striatum, α6β2* and α4(non-α6)β2* are the major nAChR subtype populations, differentially expressed by dopaminergic and non-dopaminergic neurons and involved in mediating dopamine release [Champtiaux and others 2003; Gotti and others 2005; Le Novere and others 1996; Zoli and others 2002]. An auxiliary subunit, β3 assembles with other neuronal nAChR α- and β-subunits, and supports the stability and transport of α6* nAChRs in dopaminergic neurons [Gaimarri and others 2007; Gotti and others 2005]. β3-containing nAChRs have a physiologically significant role in dopaminergic neurotransmission and many nAChRs expressed on mice substantia nigra-VTA dopaminergic neurons are β3-dependent [Cui and others 2003]. α4α6β2β3 receptor possesses the highest affinity for nicotine among all nicotinic receptor subtypes reported thus far [Champtiaux and others 2002; Drenan and others 2008; Grady and others 2007], and may play an important role in regulating dopamine-related behaviors, including mediating the reinforcing effects of nicotine. Furthermore, α6 and β3 may also co-express in the VTA [Moretti and others 2004], as confirmed by our study.

Of lesser statistical strength were the findings of three other neuronal nAChR loci (CHRNA2, A9 and A10) in association with ND. CHRNA2 was significantly associated with ND in EAs. α2-mRNA was reported to be expressed in the midbrain as well as in the cerebral cortex [Ishii and others 2005]. α2-containing receptors were highly expressed in monkeys and humans, suggesting that the expression of this subunit in the brain is species specific. We found that α2 was expressed in midbrain, hippocampus, amygdala, striatum and cortex. Many studies have demonstrated that this is a functional gene for nicotine response [Lotfipour and others 2013; Picciotto and others 1998]. Some functional rare variants have been identified in this gene [Dash and Li 2014; Dash and others 2014]. Nicotine influences the local balance between neuronal excitation and inhibition, gates long-term potentiation, and directs information flow through the hippocampal circuits via the activation of α2*nAChRs, providing a cellular basis of nicotine-mediated cognitive enhancement [Nakauchi and others 2007]. Additionally, we found that both CHRNA9 and CHRNA10 were significantly associated with ND in EAs and with AD in AAs. CHRNA9 was also significantly associated with ND+AD in EAs. ND and AD share risk rare variant constellations at the CHRNA9 and CHRNA10. α9-subunit was reported to be partially expressed in the pituitary [Lips and others 2002], which was confirmed by our study. We also found α9 in hippocampus, midbrain, cerebellum, amygdala, hypothalamus and cortex and α10 in hippocampus, amygdala and cortex, but not in other brain regions, consistent with previous reports [Gotti and others 2006]. Functionally, α9 subunit can form a homopentamer or heterooligomeric cation-selective channels in conjunction with nAChR α10 [Sgard and others 2002], which suggests the potential of CHRNA9 as a functional gene.

Five nicotinic receptor subunits encoded by CHRNA1, CHRNB1, CHRNG, CHRND and CHRNE are abundantly expressed in the neuromuscular junctions but not in the brain, according to earlier reports. However, we detected these subunits in all ten mouse brain regions but the midbrain in this study, consistent with the findings that CHRNA1, CHRNG, CHRND and CHRNE were associated with ND/ND+AD, and earlier findings that CHRNA1, CHRNB1, CHRNG and CHRND were associated with ND, as discussed above. On the other hand, these associations in our study were identified only in one cohort but not replicated in others. For example, CHRNA1 was significantly associated with ND in AAs; CHRNE was significantly associated with ND in EAs; and CHRNG/CHRND was significantly associated with ND+AD in EAs. Thus, these findings and their significance need to be evaluated again in the future.

We did not find significant association of CHRNB2 (neuronal nAChR) with ND/AD in the T1 test, even in the meta-analysis (p>0.05; Tables 3 and S4). However, in the T5 and Fp tests, it is significantly associated with ND/AD across four cohorts (p<α; Table 3) and in the meta-analysis (Table S4). Further, we found that β2 was highly expressed in all brain regions examined, which may explain the important role of β2 in the ND-related nAChRs as presented in the above.

In summary, we found all Chrn genes were expressed in mouse brain and most CHRN gene regions were involved in ND/AD. First, except for the muscle-type CHRNB1, variants in almost all other 11 nAChR subunit-encoding gene regions (containing 15 genes) were associated with ND, among which, CHRNA7, CHRNA9 and CHRNA10 were also associated with AD. Second, although ND and AD shared some susceptibility genes, associations with ND are much stronger than AD. Although there was no significant association between family-wide rare variant constellations and AD in an EA sample, there was significant association with ND+AD in the subset of this EA sample. While the sample size of this subset became smaller, the association became stronger, which is probably because CHRNs are mainly associated with ND rather than AD. Third, associations between ND/AD and both CHRNA5-A3-B4 and CHRNA7 are the most robust in that they were significant, replicable across more than three cohorts, and consistent with previous reports and expression findings. Fourth, many findings were highly significant with p values as low as 10⁻³⁹ (for T5 and Fp tests). When multiple rare variants are significantly associated with ND, rare variant constellation analysis highlights their cumulating effects. The more positive variants are included, the more significant the associations of the constellations would be. This is why associations by T5 test were much more significant than T1 test and why the p values for T5 test could be so small. Fifth, bioinformatics analysis indicates that a number of potentially functional individual SNPs might underlie the risk rare variant constellations. They might influence the methylation, splicing or shortening of the risk genes; or they might alter the protein function or structure directly. We conclude that CHRNs are associated with ND (mainly) and AD, consistent with previous findings and the hypothesis that “the full catalog of ND genes may contain all the neuronal nAChRs-encoding genes, each responsible for a particular intermediate phenotype, such as reinforcement, withdrawal, tolerance, and sensitization” [Greenbaum and Lerer 2009].