|Home | About | Journals | Submit | Contact Us | Français|
The role of the nicotinic acetylcholine receptor gene cluster on chromosome 15q24-25 in the etiology of nicotine dependence (ND) is still being defined. In this study, we included all 15 tagging single nucleotide polymorphisms (SNPs) within the CHRNA5-CHRNA3-CHRNB4 cluster and tested associations with 30 smoking-related phenotypes.
The study sample was ascertained from the Finnish Twin Cohort study. Twin pairs born 1938–1957 and concordant for a history of cigarette smoking were recruited along with their family members (mainly siblings), as part of the Nicotine Addiction Genetics consortium. The study sample consisted of 1,428 individuals (59% males) from 735 families, with mean age 55.6 years.
We detected multiple novel associations for ND. DSM-IV ND symptoms associated significantly with the proxy SNP Locus 1 (rs2036527, p = .000009) and Locus 2 (rs578776, p = .0001) and tolerance factor of the Nicotine Dependence Syndrome Scale (NDSS) showed suggestive association to rs11636753 (p = .0059), rs11634351 (p = .0069), and rs1948 (p = .0071) in CHRNB4. Furthermore, we report significant association with DSM-IV ND diagnosis (rs2036527, p = .0003) for the first time in a Caucasian population. Several SNPs indicated suggestive association for traits related to ages at smoking initiation. Also, rs11636753 in CHRNB4 showed suggestive association with regular drinking (p = .0029) and the comorbidity of depression and ND (p = .0034).
We demonstrate novel associations of DSM-IV ND symptoms and the NDSS tolerance subscale. Our results confirm and extend association findings for other ND measures. We show pleiotropic effects of this gene cluster on multiple measures of ND and also regular drinking and the comorbidity of ND and depression.
Genes are known to be involved in the etiology of nicotine dependence (ND) with heritability estimates ranging from 40% to 75% (Rose, Broms, Korhonen, Dick, & Kaprio, 2009). Similarly, moderate to high heritability estimates have been reported for smoking initiation (32%–78%; Rose et al., 2009), first smoking experiences (37%; Haberstick, Ehringer, Lessem, Hopfer, & Hewitt, 2011), cigarettes smoked per day (CPD; 40%–56%; Rose et al., 2009), and smoking cessation (11%–86%; Rose et al., 2009). However, studies on smoking and ND based on linkage scans and candidate gene approach has yielded somewhat inconsistent results (Han, Gelernter, Luo, & Yang, 2010), even though multiple pathways and neurotransmitter systems have been implicated (Wang & Li, 2010).
The most established finding involves the CHRNA5-CHRNA3-CHRNB4 nicotinic acetylcholine receptor (nAChR) gene cluster on chromosome 15q24-25. A number of studies have implicated the consistent role of this gene cluster in a wide range of smoking-related phenotypes: ever smoking (smoked 100 cigarettes or more in lifetime; Erlich et al., 2010), CPD (Berrettini et al., 2008; Caporaso et al., 2009; Erlich et al., 2010; Keskitalo et al., 2009; Li et al., 2010; Liu et al., 2010; Saccone et al., 2010; Spitz, Amos, Dong, Lin, & Wu, 2008; Stevens et al., 2008; Thorgeirsson et al., 2008, 2010; Tobacco and Genetics Consortium, 2010), persistent smoking (Bierut et al., 2008), heavy smoking (Stevens et al., 2008), number of quitting attempts (Erlich et al., 2010), age at smoking initiation (Grucza et al., 2010; Schlaepfer et al., 2008; Weiss et al., 2008), pleasurable early smoking experience (Sherva et al., 2008), physical effects reported after smoking first experimental cigarette (Hoft, Stitzel, Hutchison, & Ehringer, 2011), as well as serum cotinine levels (Keskitalo et al., 2009). The CHRNA5-CHRNA3-CHRNB4 gene cluster has been associated with ND defined by the Fagerström Test for Nicotine Dependence (FTND; Heatherton, Kozlowski, Frecker, & Fagerström, 1991) (Bierut et al., 2007; L. S. Chen et al., 2009; X. Chen et al., 2009; Erlich et al., 2010; Johnson et al., 2010; Kim et al., 2011; Li et al., 2010; Saccone et al., 2009; Saccone, Hinrichs, et al., 2007a; Spitz et al., 2008; Weiss et al., 2008; Wessel et al., 2010), as well as the Heaviness of Smoking Index (HSI) consisting of two key items of FTND (Li et al., 2010; Marques-Vidal et al., 2011). In addition, association has been found with the DSM-IV ND diagnosis (American Psychiatric Association, 1994) in a Black sample (Sherva et al., 2010) and in an Islandic sample defining ND as a score of 4 or more on the FTND or endorsement of 3 or more DSM-IV ND criteria (Thorgeirsson et al., 2008). Furthermore, haplotypes within the gene cluster were associated with the multidimensional ND scale, the Wisconsin Inventory of Smoking Dependence Motives, (WISDM-68) among individuals with early smoking initiation (Baker et al., 2009).
The CHRNA5-CHRNA3-CHRNB4 gene cluster codes for the α5, α3, and β4 nAChR subunits. Nicotinic acetylcholine receptors are the primary targets for nicotine and initiate the brain and peripheral responses to smoking. It is thus biologically plausible that genetic variants in the genes coding for the nAChR subunits influence smoking intensity and ND. Supporting this hypothesis, functional studies have shown that a common nonsynonymous variant (rs16969968) in CHRNA5 affects receptor function (Bierut et al., 2008; Kuryatov, Berrettini, & Lindstrom, 2011). Furthermore, rs588765 and correlates are associated with CHRNA5 mRNA levels in brain tissue (Wang et al., 2009). In addition, Chrna5 knockout mice have shown reduced sensitivity to nicotine-induced hypolocomotion and seizures (Salas et al., 2003).
Variation in genes coding for nAChRs has an established role in ND but may also play a role in alcohol dependence (Sherva et al., 2010; Wang et al., 2009), early alcohol use (Schlaepfer et al., 2008), cocaine dependence (Sherva et al., 2010), and opioid dependence (Erlich et al., 2010). The effects of different drugs share biological mechanisms, most importantly the increase of dopamine release from limbic brain areas. The α4α5β2 nAChR subtype is involved in nicotine-stimulated dopamine release (Salminen et al., 2004). Recent work highlights the role of the medial habenula, with high density of α4α5β2 receptors, in responses to nicotine (Fowler, Lu, Johnson, Marks, & Kenny, 2011; Salas, Sturm, Boulter, & De Biasi, 2009).
The CHRNA5-CHRNA3-CHRNB4 gene cluster exhibits extensive linkage disequilibrium (LD). CHRNA5 and CHRNA3 are oriented in opposite directions and share part of their 3′UTR; thus, coordinated expression of these two genes may occur (Solda et al., 2005). Previous studies support the existence of multiple distinct smoking behavior loci within the CHRNA5-CHRNA3-CHRNB4 region (Liu et al., 2010; Saccone et al., 2010; Thorgeirsson et al., 2010; Tobacco and Genetics Consortium, 2010). The large meta-analysis on CPD (Saccone et al., 2010) identified three statistically distinct loci: Locus 1 tagged by rs16969968, Locus 2 tagged by rs578776, and Locus 3 tagged by rs588765. For each loci, a list of so-called proxy single nucleotide polymorphisms (SNPs; Saccone et al., 2010), that is, adjacent SNPs in high LD with these variants, has been provided to ease future study comparisons and meta-analyses.
Both depression and alcohol use are known to co-occur with smoking and ND (Dani & Harris, 2005; Durazzo & Meyerhoff, 2007; Korhonen et al., 2007). Twin and family studies show that significant genetic correlations underlie this co-occurrence (Rose et al., 2009), suggesting shared genetic predisposition. As nAChRs have an important role in mediating the effect of nicotine on the dopaminergic pathway (Benowitz, 2010), it is reasonable to consider that variation in nicotine receptor genes may have pleiotropic effects and potentially associates not exclusively to ND but also to alcohol use and depression. Furthermore, the various ND measurements are suggested to capture slightly different aspects of ND, and by including several measures, we attempt to comprehensively portray the dimensions of ND. Here, the study sample utilized was specifically enriched for smoking and ND, with detailed phenotype profiles including not only assessment of CPD and ND but also age at smoking initiation, diagnoses and symptoms of depression, as well as of alcohol use, abuse, and dependence. The aim of this study was to utilize detailed phenotype information to more comprehensively clarify the involvement of the CHRNA5-CHRNA3-CHRNB4 gene cluster in the etiology of ND and other smoking related traits as well as co-occurring phenotypes.
The sample collection has been previously described in detail (Broms et al., 2007; Loukola et al., 2008; Saccone, Pergadia, et al., 2007b). Briefly, the study sample was ascertained from the Finnish Twin Cohort study consisting of adult twins born between 1938 and 1957. Based on earlier health questionnaires, the twin pairs concordant for ever smoking were identified and recruited along with their family members (mainly siblings) for the Nicotine Addiction Genetics Finland study (N = 2,265), as part of the consortium including Finland, Australia, and United States. Data collection took place between 2001 and 2005. Because of the relatively old age of the siblings, very few parents were available for the study. The study sample consisted of 1,428 individuals (59% males) from 735 families, with mean age 55.6 years, who smoked on average 19.7 (SD = 9.9) CPD. Ninety-four percent had smoked 100 or more cigarettes over lifetime. Sample characteristics are presented in Table 1. The study was approved by the Ethics committee of the Hospital District of Helsinki and Uusimaa, Finland and by the IRB of Washington University, St. Louis, Missouri, USA.
In order to gain further information on the LD structure, we utilized a total of 679 genome-wide association (GWA) samples ascertained from the population-based FinnTwin12 study (unselected for smoking; Kaprio, 2006).
Participants were telephone interviewed using the diagnostic Semi-Structured Assessment for the Genetics of Alcoholism (Bucholz et al., 1994) protocol, including an additional section on smoking behavior and ND adapted from the Composite International Diagnostic Interview (Cottler et al., 1991). The customized computer-assisted telephone interviews included more than 100 questions on smoking behavior. All participants provided written informed consent forms by mail. All phenotypes used in the analyses are based on the interview data (survey for the Nicotine Dependence Syndrome Scale [NDSS]; Shiffman, Waters, & Hickcox, 2004). Phenotype definitions are presented in Table 2. The examined binary, continuous, and categorical smoking-related phenotypes are divided into five groups: (a) amount smoked, (b) smoking initiation, (c) ND, (d) DSM-IV–based lifetime major depression, and (e) alcohol use, abuse, and dependence.
Participants were mailed a blood sample kit, which they took to the nearest health center laboratory for phlebotomy, and the venous blood samples were returned to the National Public Health Institute by mail. DNA was extracted by standard methods.
A total of 303 individuals were genotyped using Sequenom’s homogeneous Mass Extend (hME) and iPLEX Gold technology (Sequenom, San Diego, CA, USA). Tagging SNPs were selected based on the HapMap Project (http://www.hapmap.org) and NCBI (http://www.ncbi.nlm.nih.gov) databases. The selected variants were biallelic and had a minor allele frequency (MAF) for more than 10% in the Caucasian population. The ability to amplify the flanking regions of each SNP was determined by using the applications SNPper (http://www.snpper.chip.org) and RealSNP (http://www.realsnp.com), which define, respectively, the most reliable regions for designing primers and the quality of the amplicons. All tagging SNPs failing during the procedure were replaced by newly generated tagging SNPs proposed by Haploview (Barrett, Fry, Maller, & Daly, 2005). The polymerase chain reaction (PCR) and extension primers were designed using Sequenom’s MassARRAY Assay Design software (version 2.0). SNPs were genotyped in 384-well plates according to manufacturer’s instructions. For quality controls, each plate contained at least eight water controls and 22 duplicate samples. PCR reactions were performed in a total reaction volume of 5 μl using 20 ng of genomic DNA. The alleles were automatically called by Sequenom's MassARRAY Typer Analyzer software and verified by two independent persons. Further, marker-specific quality controls included a call rate more than 80% and a Hardy–Weinberg equilibrium (HWE) p value > .01 (estimated using unrelated individuals). Mendelian errors were excluded using PedCheck (O’Connell & Weeks, 1998).
For 1,125 individuals, genotypes for the same 15 SNPs were derived from GWA data. Genotyping was performed at the Welcome Trust Sanger Institute (Hinxton, UK) on the Human670-QuadCustom Illumina BeadChip (Illumina, Inc., San Diego, CA, USA). The data were checked for minor allele frequency (>1%), genotyping success rate per SNP and per individual (>95%), HWE (p > 1 × 10−6), gender, and heterozygosity. In addition, to check whether any individuals were unexpectedly related to each other, an multidimensional scaling plot (using a pairwise-IBS matrix) with only one member of each known family was created. After the pedigree was reassured to be correctly formulated, the basic filters (MAF, genotyping success, HWE) were reapplied to the data. Seven of the markers were genotyped and eight were imputed using the software IMPUTE v2.1.0 (Howie et al., 2009). The reference panel used in the imputation was HapMap rel#24 CEU—NCBI Build 36 (dbSNP b126), which is available on the IMPUTE website (https://mathgen.stats.ox.ac.uk/impute/impute_v1.html#Using_IMPUTE_with_the_HapMap_Data). The posterior probability threshold for “best-guess” imputed genotype was .9. Genotypes below the threshold were set to missing.
Marker quality controls are presented in Supplementary Table 2.
The LD between SNPs was estimated among nonrelated individuals (one per family) by using Haploview 4.2 (Barrett et al., 2005). The pairwise comparisons of markers more than 500 kb apart were excluded. Haplotype blocks were defined according to the “solid spine of LD” algorithm by using the default threshold values for block estimation (Figure 1).
The associations between discrete phenotypes and candidate SNPs (qualitative association) were estimated with pseudomarker (Goring & Terwilliger, 2000), which performs separate and joint linkage and LD analyses, testing each marker locus against a phenotype-based “pseudomarker” locus. This likelihood-based estimation method is numerically equivalent to model-free analysis and efficiently uses data on all family types. Both recessive and dominant models (default parameters) were fitted. We report uncorrected p values minimized over “LD given linkage,” “LD given no linkage,” and “LD and linkage” (joint test) as well as dominant and recessive models.
The associations between continuous phenotypes and candidate SNPs (quantitative association) were performed with quantitative transmission disequilibrium test (QTDT; Abecasis, Cardon, & Cookson, 2000). In the analysis, the proportion of alleles shared identically by descent (IBD) was estimated by multipoint computation of MERLIN (Abecasis, Cherny, Cookson, & Cardon, 2002) to extract maximal inheritance information from the pedigrees. Prior to analysis, we estimated our sample not being stratified with the appropriate “population stratification” model of QTDT, which compares the between and within family components of association (Abecasis et al., 2000). The total association model was used, allowing powerful analysis of the sample, including incomplete families. In the analysis, the variance components “polygenic,” “nonshared environment” (environmental effects unique to each family member), “common environment” (environmental effects shared by all related individuals), “nuclear family environment” (environmental effects shared by all members of a nuclear family), and “twin environment” (environmental effects shared only by twins) were used to model the phenotypic similarities between the pedigree members. In the analysis of traits related to age of initiation (age at first puff, age at first cigarette, age of onset of weekly smoking, age of onset of daily smoking), sex was included as a covariate, whereas both sex and age at recruitment were used as covariates for all other continuous traits.
To account for multiple testing we used a modified Bonferroni correction to set p value thresholds for significant and suggestive association signals. Since neither analyzed markers nor traits are independent, we utilized an established methodology to evaluate the numbers of corresponding independent markers and traits with the programs SNPSpD and matSpD, respectively (Cheverud, 2001; Li & Ji, 2005; Nyholt, 2004), and their MeffLi and VeffLi estimates (Li & Ji, 2005) were used as they were smaller than Meff and Veff, respectively, as recommended by the author (http://gump.qimr.edu.au/general/daleN/SNPSpD/). The trait “regular smoker” was not accounted for when estimating the number of independent traits, as it is the ascertainment criteria for our families. In our dataset, the number of independent markers was 6.0022 and the number of independent traits was 16.956. A p value threshold of .0005 for significant association was achieved by dividing p = .05 by the product of the number of independent markers and the number of independent traits. A p value threshold of .0083 for suggestive association was achieved by dividing p = .05 by the number of independent markers.
Nonnormally distributed continuous variables (kurtosis and/or skewness >1 or <−1) were transformed with square root or 10-base logarithm, whichever provided most normal-like distribution. Trait values above 30 years considered as outliers were removed from the analysis for phenotypes measuring the age at first puff (seven individuals), age at first cigarette (nine individuals), age of onset of weekly smoking (20 individuals), and age of onset of daily smoking (28 individuals).
To estimate effect sizes for all SNPs showing significant or suggestive p values in the genetic association analyses, we performed regression analyses with Stata 11.1 statistical software (StataCorp, 2009) by using an additive model. Odds ratios (OR) and beta coefficients (β) with 95% CIs were reported for binary and quantitative traits, respectively. Regression analyses were adjusted for sex and age. Furthermore, individuals were clustered based on the possible lack of statistical independence of observations of subjects who came from the same family (Williams, 2000).
In the study sample, the LD blocks were similar to those in the HapMap CEPH data (Figure 1) and the somewhat stronger LD between markers is in agreement with previous findings from the Finnish population (Service et al., 2006). In the HapMap CEPH data, CHRNB4 SNPs lie outside the CHRNA5–CHRNA3 LD block, whereas in the study sample, the LD block extends to rs1948 and rs12440014 in CHRNB4 (Figure 1). The LD blocks were almost identical between the study sample and the FinnTwin12 sample (data not shown).
We detected significant association with a variety of traits measuring ND, the strongest signal emerging for DSM-IV ND symptoms in Locus 1 (rs2036527 p = .000009, rs1051730 p = .00002) and Locus 2 (rs578776 p = .0001, rs667282 p = .0002). A significant association was observed with DSM-IV ND diagnosis in Locus 1 (rs2036527 p = .0003, rs1051730 p = .0004). Altogether we detected significant or suggestive p values for six traits measuring ND (DSM-IV ND, DSM-IV ND symptoms, FTND score, NDSS tolerance factor, Time to first cigarette (TTF), and DSM-IV ND + HSI; Table 3).
We detected suggestive association with the categorized variable of CPD in Locus 1 (rs1051730 p = .0077, rs2036527 p = .0073) and maximum CPD at rs12440014 (p = .0074) (Table 3). Additionally, four SNPs (rs578776 and rs667282 at Locus 2, rs684513, rs3743078) indicated suggestive association for the age of onset of both weekly and daily smoking (Table 3).
Furthermore, rs11636753 in CHRNB4 showed suggestive association with both regular drinking (p = .0029) and the comorbidity of depression and ND (p = .0034) (Table 3).
Association analysis results for all phenotypes are presented in Supplementary Table 3.
For those SNPs showing significant associations, notable effect sizes were detected (Supplementary Table 4). In Locus 1, DSM-IV ND diagnosis showed an OR of 1.25 (p = .0085), and corresponding effect size for DSM-IV ND symptoms was a beta coefficient of 0.30 (p < .0001). Similarly, Locus 1 showed a beta coefficient of 0.33 (p = .0017) for the FTND score. Locus 2 showed plausible protective effect for DSM-IV ND symptoms (β = −0.28, p = .0001). For SNPs showing suggestive association, modest effect sizes were detected. As an exception, a notable protective effect (OR = 0.76, p = .0239) was detected for rs11636753 in the comorbidity of depression and ND. The effect size for the regular drinker phenotype failed to meet statistical significance (p > .05; Supplementary Table 4).
Multiple distinct loci at the CHRNA5-CHRNA3-CHRNB4 locus on 15q24-25 appear to affect smoking behavior (Liu et al., 2010; Saccone et al., 2010; Thorgeirsson et al., 2010; Tobacco and Genetics Consortium, 2010). In a previous meta-analysis including the current study sample (Saccone et al., 2010), three SNPs in the CHRNA5-CHRNA3-CHRNB4 gene cluster (rs16969968, rs578776, and rs588765) showed association with CPD (heavy i.e., CPD > 20 vs. light i.e., CPD ≤ 10). Presuming the previous evidence as an “a priori” hypothesis, we set out to extend and deepen the analyses of these three loci by utilizing previously established proxy SNPs (Saccone et al., 2010; S. M. Hartz [personal communication, May 30, 2011]) as well as all detected tagging SNPs within this gene cluster. We performed association analyses in a Finnish family sample of twins and their siblings ascertained specifically for smoking and ND using a broad spectrum of phenotypes covering aspects of smoking behavior, ND, and depression, as well as alcohol use, abuse, and dependence.
In agreement with previous evidence from extensive studies with over 1,40,000 participants (Liu et al., 2010; Saccone et al., 2010; Thorgeirsson et al., 2008, 2010; Tobacco and Genetics Consortium, 2010), we detected some association with CPD. Larger samples may be needed to detect significant association with CPD, as shown by a study of over 13,000 Icelandic smokers (Thorgeirsson et al., 2008) in which variation in Locus 1 (rs1051730) accounted for only about 1% of the variance in CPD, the average effect per allele being about 1 CPD (Thorgeirsson et al., 2008). In agreement with this, an effect size of a beta coefficient 1.05 was detected in our sample, roughly corresponding to an increment of one cigarette per day for each allele of the locus.
The involvement of the CHRNA5-CHRNA3-CHRNB4 gene cluster in FTND has been well documented in the literature, whereas studies reporting findings for DSM-IV ND are sparse. It is unclear whether this is due to lack of positive findings, publication bias, challenges in reliably creating DSM-IV ND diagnosis, or the predominance of FTND in measuring ND, possible due to ease of assessment. Here, we report a significant novel association with the number of DSM-IV ND symptoms, with Locus 1 conferring risk (β = 0.30) and Locus 2 showing plausible protective effect (β = −0.28), in agreement with a previous study reporting similar effects for CPD (Saccone et al., 2010). We also report the first significant association with DSM-IV ND diagnosis in a Caucasian sample, replicating previous findings in Blacks (Sherva et al., 2010). Further, we replicate association with the FTND as a continuous trait (X. Chen et al., 2009; Erlich et al., 2010; Kim et al., 2011; Wessel et al., 2010). These two widely used ND measures are suggested to extract partly different dimensions of ND (Moolchan et al., 2002; Piper, McCarthy, & Baker, 2006; Piper et al., 2008). DSM-IV predominantly measures loss of control in terms of smoking behavior, including cognitive, behavioral, and physiological symptoms, resulting in tolerance, withdrawal, and compulsive drug-taking behavior (American Psychiatric Association, 1994). FTND aims to measure the construct of physical dependence, associating with cessation outcome, predicting smoking relapse, and having the key component of difficulty to stand reduced nicotine levels (Haddock, Lando, Klesges, Talcott, & Renaud, 1999; Heatherton et al., 1991).
We demonstrate the first evidence of genetic association of the NDSS tolerance factor, TTF, as well as the combination of DSM-IV ND diagnosis and HSI (a combination of CPD and TTF) in the CHRNA5-CHRNA3-CHRNB4 locus. Functional evidence for the involvement of the CHRNA5-CHRNA3-CHRNB4 gene cluster in tolerance includes knockout mice showing association with alpha4beta2 nAChR subunits and the development of tolerance (McCallum, Collins, Paylor, & Marks, 2006; Tapper, McKinney, Marks, & Lester, 2007). Further, chronically treated mice lacking the beta4 subunit show an increased tolerance to an acute dose of nicotine (E. E. Meyers and M. J. Marks [personal communication, May 31, 2011]). Tolerance to nicotine is defined by the ability to smoke increased amounts of cigarettes without experiencing toxic effects (Piper et al., 2006). Repeated exposure to nicotine leads to tolerance (neuroadaptation), and as neuroadaptation develops, the number of binding sites on the nAChRs in the brain increases, probably in response to nicotine-mediated desensitization of receptors (Benowitz, 2010).
The gene cluster has previously been associated with the WISDM (Piper et al., 2004) tolerance factor among 886 early-onset smokers from the United States (Baker et al., 2009). NDSS and WISDM are measuring somewhat different aspects of ND and have a slightly different focus on the tolerance dimension as well (Piper et al., 2008). The use of multidimensional scales such as NDSS and WISDM potentially aids in deciphering the construct and nature of ND (Piper et al., 2006). Our results confirm the CHRNA5-CHRNA3-CHRNB4 association to tolerance, one of the dimensions of ND, which supposedly also is embedded within the traditional unidimensional scales, DSM-IV and FTND.
Twin studies suggest that genetic influences on age at smoking initiation are correlated with the amount smoked (Broms, Silventoinen, Madden, Heath, & Kaprio, 2006; Morley et al., 2007). Several SNPs within the CHRNA5-CHRNA3-CHRNB4 locus associate with multiple phenotypes measuring age at onset of smoking, consistent with previous evidence (Schlaepfer et al., 2008). Previous results have been inconsistent, with findings supporting CHRNA5-CHRNA3-CHRNB4 variation underlying ND in individuals with early age at smoking initiation (≤16 years; Weiss et al., 2008) and CHRNA5 association with later age at smoking initiation (≥17 years; Grucza et al., 2010). Adolescence clearly is a crucial period of vulnerability for the development of substance use disorders (Chambers, Taylor, & Potenza, 2003; Crews, He, & Hodge, 2007), and also in the current study, we found that later starters were less dependent (see Supplementary Table 1) as previously reported (Breslau & Peterson, 1996; Broms, Silventoinen, Lahelma, Koskenvuo, & Kaprio, 2004; Chassin, Presson, Rose, & Sherman, 1996; J. Chen & Millar, 1998; Kandel, Hu, Griesler, & Schaffran, 2007; Lando et al., 1999). This study covered four stages of smoking initiation, allowing a comprehensive evaluation of the genetic factors contributing to this process.
Depression commonly co-occurs with smoking and ND (Chaiton, Cohen, O’Loughlin, & Rehm, 2009; Korhonen et al., 2007; Morrell & Cohen, 2006; Rose et al., 2009), recent evidence showing that ND predicts depression (Boden, Fergusson, & Horwood, 2010) but the causal nature of this association is unknown. Under certain conditions, nicotine can act as an antidepressant (Picciotto, Brunzell, & Caldarone, 2002), whereas chronic nicotine use can lead to neuroadaptation resulting in increase of depressed mood (Picciotto et al., 2002). On the other hand, depressed mood may promote continued nicotine intake to maintain desensitization of nAChRs (Mineur & Picciotto, 2009). So far, the effects of specific genes on the comorbidity of smoking and depression have been investigated in a limited number of studies with small sample sizes (Rose et al., 2009), mostly for dopaminergic genes (Audrain-McGovern, Lerman, Wileyto, Rodriguez, & Shields, 2004; Lerman & Niaura, 2002; Lerman et al., 1998). By stratifying on smoking status, it has been shown that genetic predisposition for depression actually run in families of smokers (Pergadia et al., 2011). The involvement of nAChRs in depression is a biologically plausible hypothesis. Preclinical studies show antidepressant-like effects in drugs targeting nAChRs for example, α4β2 nAChR modulators, such as varenicline (Philip, Carpenter, Tyrka, & Price, 2010). Interestingly, in our study, rs11636753 in CHRNB4 showed suggestive association with the comorbidity of depression and ND, to our best knowledge a novel finding. It is noteworthy that the very same rs11636753 was associated also with the NDSS tolerance subscale. Such pleiotropic effect could be explained so that tolerance encompasses neuroadaptation, that is, quantitative and qualitative changes in nAChRs, being further associated with withdrawal symptoms, such as depressed mood (Benowitz, 2010).
Comorbidities for cigarette smoking and ND are alcohol use and dependence (Dani & Harris, 2005; Durazzo & Meyerhoff, 2007), shared genetic vulnerability being supported by twin studies (Madden & Heath, 2002; True et al., 1999). Genetic variation within the CHRNA5-CHRNA3-CHRNB4 gene cluster has been associated with DSM-IV alcohol dependence (Sherva et al., 2010; Wang et al., 2009), DSM-IV alcohol symptoms (X. Chen et al., 2009), and early alcohol use onset (Schlaepfer et al., 2008). Interestingly, Locus 1 variants (rs16969968 and rs1051730) are associated with both ND and alcohol abuse, however, with opposite risk alleles (X. Chen et al., 2009), possibly reflecting the different actions of the two substances: alcohol being a depressant and nicotine a stimulant. In our sample ascertained for smoking, we detected suggestive association between rs11636753 and regular drinking. Heavy drinking or alcohol dependence did not show any association, although those traits likely are more genetically relevant extreme phenotypes. In the Finnish drinking culture, regular drinking defined as one or more drinks per week reflects a social drinking pattern, slightly surmising our association findings. Furthermore, the regular drinker phenotype showed statistically nonsignificant results (p > .05) in the effect size analysis (Supplementary Table 4).
Our study is based on a strong a priori hypothesis for the involvement of this locus in ND and smoking quantity. We set up to scrutinize a vast array of phenotypes to gain comprehensive information on the involvement of the 15q24-25 region and to study potential pleiotrophic effects in traits related to or co-occurring with ND. Although acknowledging the issue of increased number of tests, we chose to implement all relevant phenotypes within a single study rather than dividing them into separate entities. In order to account for multiple testing, we used a modified Bonferroni correction, utilizing estimated numbers of independent markers and traits, to set p value thresholds for significant and suggestive association signals. As the included markers and traits are correlated, standard procedures for the correction for multiple testing would certainly be overly conservative. The estimation of independent markers, based on LD matrixes, is rather straightforward. However, the number of independent traits is difficult to estimate and depends on what that information is used for. One criterion could be informativeness in predicting cessation or disease outcomes, which could be tested in a multivariate predictive model (logistic regression or survival models). In the current study, we used a statistical estimate based on the correlation/covariance matrix, resulting in a sample-based estimate that may fluctuate when applied to novel independent population samples. The use of estimated numbers of independent markers and traits in adjusting p value thresholds likely is quite conservative but nevertheless successful in reducing the type I error rate. We acknowledge that many of our findings are suggestive and need to be further confirmed in independent replication samples.
Although our study sample was of moderate size (n = 1,428), it harbors a number of advantages. Our sample of twins and their siblings was ascertained specifically for smoking and ND, being initially drawn from the population-based Finnish twin cohort with extensive phenotypic profiles. The population of Finland represents a well-established isolate with minuscule population admixture. In isolates, the genetic drift may lead to an overabundance of morbid alleles for particular disorders, and a high proportion of patients are likely to share these alleles IBD. Although the effect is strongest for rare disease alleles, isolates are also advantageous for genetic studies of common disorders (Peltonen, Palotie, & Lange, 2000).
The inclusion of other variants besides the well-established loci (tagged by rs16969968, rs578776, and rs588765; Saccone et al., 2010) proved beneficial. Although our strongest association signals emerged from Locus 1 (rs1051730 and rs2036527; in full LD with the functional and most well-known Locus 1 proxy SNP, rs16969968), we also detected association with SNPs not included in the proxy lists. Although we aimed to study haplotype effects, the construction of haplotypes in our family sample with virtually all parental genotypes missing proved to be a challenging task. We tested several softwares including PHASE 2.1.1 (Stephens & Scheet, 2005; Stephens, Smith, & Donnelly, 2001; disregarding family structures) and FBAT 2.0.2c (Laird, Horvath, & Xu, 2000; including nuclear family information), but these led to either very inconsistent and unreliable haplotypes or the exclusion of virtually all pedigrees due to missing founder genotypes.
To conclude, we report the novel associations of DSM-IV ND symptoms, NDSS tolerance subscale, TTF, the combined phenotype of DSM-IV ND and HSI, as well as regular drinking and the comorbidity of ND and depression to the CHRNA5-CHRNA3-CHRNB4 gene cluster. Our results confirm and extend the association findings for DSM-IV ND, FTND, CPD, and age at smoking initiation. Using detailed phenotype profiles, we show a broad involvement of this gene cluster on various traits measuring or co-occurring with ND.
This work was supported for data collection by Academy of Finland grants to JK and a National Institutes of Health (NIH) grant (grant numbers DA12854 to PAFM, DA019951 to MLP). This work was supported by Doctoral Programs of Public Health (UB), the Yrjö Jahnsson Foundation (UB), the Juho Vainio Foundation for Post-Doctoral research (UB), Academy of Finland Post-Doctoral Fellowship (AL), Finnish Cultural Foundation (TK, JW), Finnish Medical Foundation (JW), Helsinki Biomedical Graduate School (JW), a NIH grant (DA019951 to MLP), a Global Research Awards for Nicotine Dependence funded by Pfizer Inc. (to JK), and by the Academy of Finland Center of Excellence in Complex Disease Genetics (grant numbers: 213506, 129680 to JK).
Dr. Kaprio has served as a consultant to Pfizer in 2008 and 2011. Dr. Broms has served as a consultant to Pfizer in 2008. Dr. Korhonen has served as a consultant to Pfizer in 2011.
We warmly thank the participating twin pairs and their family members for their contribution. We would like to express our appreciation to the skilled study interviewers A.-M. Iivonen, K. Karhu, H.-M. Kuha, U. Kulmala-Gråhn, M. Mantere, K. Saanakorpi, M. Saarinen, R. Sipilä, L.Viljanen, and E. Voipio. M. Levander, L. Arala, O. Törnwall, and M. Jussila are acknowledged for their skillful technical assistance. E. E. Meyers, M. J. Marks, and S. M. Hartz are acknowledged for personal communication. We are ever grateful to the late Academician Leena Peltonen for her indispensable contribution throughout the years of the study.