Abstract

Introduction

Increased knowledge of the genetic influences on the development and manifestation of nicotine dependence would have both applied and theoretical benefits. Such knowledge could benefit tobacco control efforts by revealing who is at elevated risk for dependence and also could clarify mechanisms of dependence (e.g., McClure & Swan, 2006).

Genetic mapping of nicotine dependence will likely prove to be quite complex (Lessov, Swan, Ring, Khroyan, & Lerman, 2004). For example, different genes may affect different stages in the development of dependence (e.g., Munafó, Johnstone, Murphy, & Walton, 2001). Further, the mature nicotine dependence phenotype may have multiple correlated but distinct features (Baker, Moffit, Caspi, & Conti, in press; Hudmon et al., 2003; Piper, McCarthy, & Baker, 2006; Piper et al., 2004; Shiffman, Waters, & Hickcox, 2004). Thus, a given genetic variant might be related only to particular features of the complex phenotype (e.g., Pergadia, Heath, Martin, & Madden, 2006).

The complexity and multidimensionality of nicotine dependence argues for relating targeted genetic variants with strategically selected subphenotypes of dependence (Baker et al., in press) to reveal the breadth or specificity of the effects of genetic variants. For instance, research (e.g., Cannon et al., 2005) demonstrated a relationship between the PTC gene and the Wisconsin Inventory of Smoking Dependence Motives (WISDM-68; Piper et al., 2004) Taste/Sensory subscale for nicotine dependence but not other dependence indices.

Baker et al. (in press) proposed a model of the nicotine dependence phenotype that provides an empirical and theoretical basis for organizing dependence phenotypes for genetic studies. The model comprises three core nicotine dependence criteria: (a) a pattern of heavy, pervasive, and automatic tobacco use; (b) the tendency to relapse after a quit attempt; and (c) withdrawal severity. The model is based on analyses of behavioral genetic data and on covariance structure modeling of nicotine dependence phenotype measures. Each core criterion has been linked theoretically and empirically to a global dependence construct (e.g., Baker et al., in press; Piper et al., 2006), but the evidence also shows only modest overlap among these features. For instance, measures of heavy, pervasive tobacco use, and withdrawal severity account for orthogonal variance in relapse likelihood (e.g., Baker et al., in press; Kenford et al., 2002) and tend to load on different factors (e.g., Baker et al., in press; Lessov, Martin, et al., 2004).

Our research group reported major susceptibility and protective haplotypes for nicotine dependence in the CHRNA5-A3-B4 subunit cluster on chromosome 15 (Weiss et al., 2008). Following a single nucleotide polymorphism (SNP) discovery effort involving resequencing human neuronal acetylcholine receptor (nAChR) genes (CHRNA7 was only partially surveyed), individual SNPs in nAChR genes were associated with nicotine dependence. Haplotypes located in the CHRNA5-A3-B4 subunit cluster were characterized and correlated to the Fagerström Test for Nicotine Dependence (FTND; Heatherton, Kozlowski, Frecker, & Fagerström, 1991). Specifically, the major haplotypes A and C (H_A, H_C) were associated with high and low FTND scores, respectively (Weiss et al., 2008). Haplotype A includes the risk alleles T of rs1051730 and A of rs16969968 that have been associated with nicotine dependence in prior studies (Saccone et al., 2007; Thorgeirsson et al., 2008).

Several features of the Weiss et al. (2008) work are noteworthy. First, this research demonstrates the presence of a significant gene × environment interaction. Specifically, strong associations (p=2.0×10⁻⁵) between CHRNA5-A3-B4 variants and nicotine dependence were seen only among smokers who began daily smoking relatively early in life (e.g., at age 16 or younger; early-onset smokers). No associations (p=.38) with haplotypes in the CHRNA5-A3-B4 cluster were seen in smokers initiating daily smoking at age 17 or later. Second, the obtained associations were of substantial magnitude (odds ratio [OR]=1.82, 95% CI=1.39–2.39). Third, the research is notable because the same pattern of associations was observed in three large (n = 2,210) independent samples.

Credence in the Weiss et al. (2008) results is bolstered by other research that found signals for nicotine dependence in the same region. In a small study of Israeli women, Greenbaum et al. (2006) obtained a suggestive association between their measure of dependence and a CHRNA3 SNP, rs1051730 (see also Thorgeirsson et al., 2008). Sherva et al. (2008) have reported a significant association between rs16969968 and smoking status as well as experiencing a “pleasurable buzz” in response to early experimentation with smoking. Studies examining the CHRNA5-A3-B4 cluster as candidate genes (e.g., Berrettini et al., 2008; Saccone et al., 2007) found evidence for associations between common variants in the cluster and their respective phenotypes. Three new genome-wide association studies provide strong evidence for an association between CHRNA5-A3-B4 variants and lung cancer, but they differ on whether the connection is direct or mediated by smoking behavior (Amos et al., 2008; Hung et al., 2008; Thorgeirsson et al., 2008).

Previous CHRNA5-A3-B4 research has been based on a limited assessment of nicotine dependence. Several of the earlier studies related the CHRNA5-A3-B4 variants to the FTND, either to the total score or to individual items (e.g., Greenbaum et al., 2006; Saccone et al., 2007; Thorgeirsson et al., 2008; Weiss et al., 2008). Factor analyses typically show that the FTND items reflect only a smoking heaviness or compulsive smoking factor and a secondary morning smoking factor (Baker et al., in press; Breteler, Hilberink, Zeeman, & Lammers, 2004; Nonnemaker & Homsi, 2007; Radzius et al., 2003; Richardson & Ratner, 2005), and FTND items tend not to be highly correlated with other key features of nicotine dependence (i.e., withdrawal severity and the tendency to relapse to tobacco use; Baker et al., in press; Baker et al., 2007; Payne, Smith, McCracken, McSherry, & Antony, 1994; Piper et al., 2004, 2006; Pomerleau, Carton, Lutzke, Flessland, & Pomerleau, 1994; Pomerleau et al., 2005). Thorgeirsson et al. (2008) also related CHRNA3 SNP rs1051730 to DSM-IV nicotine dependence criteria (American Psychiatric Association, 1994). Although the DSM criteria elicit information about withdrawal severity and relapse likelihood, little evidence indicates that the criteria measure these dimensions accurately (Breslau & Johnson, 2000; Piper et al., 2006). Finally, Thorgeirsson et al. (2008) related the CHRNA3 SNP rs1051730 to unconfirmed self-reports of whether a person was a current or an ex-smoker. These authors found that individuals possessing the risk allele were more likely to be current smokers (OR=1.07, p=.015). These data do not reveal whether those who possessed the risk alleles were less likely to have made quit attempts or less successful in quit attempts. Therefore, existing data do not clearly show the relationship between the CHRNA5-A3-B4 variants and two important indices of nicotine dependence: withdrawal severity and the tendency to relapse to tobacco use (Baker et al., in press; Piper et al., 2006).

This paper relates CHRNA5-A3-B4 cluster haplotypes and diplotypes with three different phenotypic measures. One measure is a rating of withdrawal symptoms made in real time by subjects enrolled in a cessation study (McCarthy, Piasecki, Fiore, & Baker, 2006). The second is biochemically confirmed relapse in a randomized clinical trial. The third is a new measure of nicotine dependence that reflects heavy, automatic, and pervasive tobacco use that appears to capture a core element of nicotine dependence. This new measure was the mean of four scales of the WISDM-68 nicotine dependence measure (Baker et al., 2007; Cannon et al., 2005; Piper et al., 2004, 2006). Studies show that the WISDM-68 Craving, Tolerance, Automaticity, and Loss of Control scales, hereafter labeled as the Primary Dependence Motives (PDM) scale, have especially strong relationships with important dependence criteria and appear to index necessary and sufficient features of nicotine dependence (Baker et al., 2007; Piper et al., in press). We hypothesized that the CHRNA5-A3-B4 cluster variants would be especially highly related to the PDM scale, compared with the mean of the other nine WISDM-68 scales (labeled the Secondary Dependence Motives [SDM] scale). The SDM scales tap a variety of dependence motives that typically involve smoking for some instrumental purpose (e.g., smoking to reduce negative affect, to enhance positive affect, and to control weight), and factor analyses show that they constitute a different factor than do the PDM scales (Piper et al., in press). To the extent that the CHRNA5-A3-B4 cluster variants index risk for a central or core element of nicotine dependence, they should show especially strong relationships with the PDM scales.

This research adds to the existing literature by relating CHRNA5-A3-B4 cluster haplotypes and diplotypes with strategically selected phenotypes that were not used in previous research. The present study sought to determine whether the CHRNA5-A3-B4 cluster variants interacted with age at onset of daily smoking in the prediction of phenotypic variance, as was found in Weiss et al. (2008). This research uses the same population of smokers that Weiss et al. used and therefore constitutes an extension of that work, not an independent replication.

Methods

Genotyping

Haplotypes were based on five tagging SNPs (rs680244, rs569207, rs16969968, rs578776, and rs1051730) in the CHRNA5-A3-B4 region. These five tagging SNPs were chosen from a larger genotyping study that included 11 SNPs from the CHRNA5-A3-B4 region identified by resequencing a subset of individuals from these study populations (Weiss et al., 2008). Genotyping was performed using the SNPlex method, a multiplexed oligonucleotide ligation, polymerase chain reaction assay (Applied Biosystems, Foster City, CA). Individual haplotype and diplotype assignments were inferred from genotypic data using fastPHASE and independently evaluated using the EM algorithm implemented in SNPHAP (Weiss et al. 2008).

The haplotype counts and frequencies observed in the 886 individuals were as follows: H_A (CCAGA, 699, 39.4%), H_B (TCGGG, 670, 37.8%), H_C (CTGAG, 337, 19.0%), and H_D (TCGAG, 66, 3.7%). A linkage disequilibrium (LD) plot of the CHRNA5-A3 region defined by the five tagging SNPs is shown in Figure 1A, along with the observed r² LD statistic between pairs of SNP loci. Haplotype sequences reported are from the chromosome 15 (+) strand, and the structure of the haplotypes derived from the five tagging SNPs is shown in Figure 1B.

Open in a separate window

Figure 1.

Linkage disequilibrium (LD) structure and haplotype frequency in the CHRNA5-A3 region using five haplotype tagging single nucleotide polymorphisms (SNPs). (A) The LD plot between pairs of SNP markers from this chromosome 15 region is based on the measured r². Pairwise LD values (r²) are indicated at the intercept of the two SNP markers on the matrix. Each square represents the magnitude of LD for a single pair of markers, with black color indicating high r². (B) Haplotypes spanning this LD block are shown along with their frequency in the combined Utah and Wisconsin cohorts. Chromosome 15 coordinates are from NCBI Build 36.1 (hg18).

Results

Heavy, automatic, and pervasive smoking

Analyses of PDM indicated that H_A was associated with heavier smoking than was H_C in early-onset, but not late-onset, smokers (Table 3). With dichotomous PDM as the dependent variable, the interaction between age-at-onset condition and the H_A versus H_C contrast was significant, p=.04, OR=0.57 (95% CI=0.34–0.97). As shown in Figure 2, in the early-onset condition, H_A was associated with a higher percentage of PDM scores above the median than was H_C, and we found a nonsignificant trend for H_A to be associated with more PDM scores above the median than was H_B, p=.08 (see Table 3). We found no association between haplotypes and dichotomous PDM scores in late-onset smokers (see Figure 2 and Table 3). These results with dichotomous PDM scoring were consistent with findings obtained with analyses of continuous scores within age-at-onset condition. In the early-onset condition, there was a significant haplotype effect, F(2, 823)=3.88, p=.02, and pairwise comparisons were significant only for the H_A versus H_C test, p=.02. H_A was associated with higher PDM scores (M=5.1, SD=1.3) than was H_C (M=4.7, SD=1.3). The haplotype effect on continuous PDM scores was not significant in late-onset smokers.

Table 3.

Logistic regression results for the association between haplotypes and dichotomized PDM, WDR, and relapse during the first 90 days postquit

	PDM-early		PDM-late		WDR		90-Day relapse
Effect	OR (95% CI)	p value	OR (95% CI)	p value	OR (95% CI)	p value	OR (95% CI)	p value
A–B	0.75 (0.55–1.03)	.08	0.98 (0.73–1.33)	.92	1.20 (0.87–1.64)	.26	0.74 (0.53–1.02)	.07
A–C	0.56 (0.38–0.82)	.003	0.98 (0.68–1.42)	.92	1.57 (1.05–2.34)	.03	0.59 (0.39–0.89)	.01

Open in a separate window

Note. OR, odds ratio. Primary Dependence Motives (PDM) tests were computed by age at onset of daily smoking condition (PDM-early = daily smoking at 16 or younger; PDM-late = daily smoking at 17 or older), whereas withdrawal (WDR) and 90-day relapse associations were tested across both age-at-onset conditions combined. The PDM tests were based on the Utah and Wisconsin (WI) cohorts combined, whereas WDR and 90-day relapse were based on WI participants only. Effect = the specific haplotype contrast tested, that is, A–B = H_A vs. H_B and A–C = H_A vs. H_C.

Open in a separate window

Figure 2.

Dichotomous phenotypes by haplotype (top) and diplotype (bottom). PDM, Primary Dependence Motives; Early, daily smoking at age 16 or younger; Late, daily smoking at age 17 or older; and WDR, withdrawal. High Dependence = percentage in the dichotomous category indicative of a higher level of nicotine dependence, that is, percentage above the median for PDM and WDR and the percentage who relapsed for 90-day relapse.

Analyses of the relationship between PDM and the diplotypes revealed that diplotype AA was a risk factor relative to diplotype BC in early-onset but not late-onset smokers. With the binary PDM variable, the AA versus BC × age-at-onset interaction was significant, p=.02, OR=0.32 (95% CI=0.12–0.83). Within the early-onset condition, there was an AA versus BC effect, p<.01, OR=0.38 (95% CI=0.19–0.77), but there was no significant diplotype effect in the late-onset condition (see Figure 2). A post-hoc analysis in the early-onset condition, in which diplotype AB was the reference condition, indicated a significant AB versus BC effect, p=.001, OR=0.37 (95% CI=0.20–0.67). With continuous scoring of PDM, there was a diplotype effect in early-onset smokers, F(4, 376)=2.36, p=.05. There were nonsignificant trends in the early-onset condition for diplotype BC (M=4.6, SD=1.5) to be associated with lower PDM scores than either AA (M=5.2, SD=1.3), p=.06, or AB (M=5.1, SD=1.3), p=.07. There was no diplotype effect in late-onset smokers. Overall, these findings suggest that H_A and H_C exert relative risk and protective effects, respectively, on heavy, automatic, and pervasive smoking in early-onset smokers.

We also examined the associations of individual PDM scales with genetic variants to shed light on the features of heavy, pervasive smoking that are most sensitive to the CHRNA5-A3-B4 genetic signal. We analyzed the associations between dichotomous scoring of each of the four constituent PDM scales and both haplotypes and diplotypes among early-onset smokers. The strongest haplotype signals were obtained with the Tolerance scale (see Figure 3). H_A was associated with higher Tolerance scores than either H_B, p<.01, OR=0.65 (95% CI=0.48–0.89), or H_C, p=9×10⁻⁵, OR=0.47 (95% CI=0.32–0.68). Similar but slightly weaker signals were obtained with Craving: H_A versus H_B, p<.04, OR=0.71 (95% CI=0.52–0.96); and H_A versus H_C, p<.03, OR=0.64 (95% CI=0.44–0.93). Although there was a trend for H_A to be associated with higher Automaticity and Loss of Control scores than was H_C, we found no significant haplotype effects for either of these scales. The Tolerance scale also was strongly associated with diplotypes (see Figure 3). Diplotype AA was associated with higher Tolerance scores than either BB, p=0.03, OR=0.44 (95% CI=0.21–0.94), or BC, p=.001, OR=0.29 (95% CI=0.14–0.60). For the Craving scale, AA was associated with higher scores than BB, p=.04, OR=0.46 (95% CI=0.22–0.95), and there was a nonsignificant trend for AA to be associated with higher scores than BC, p=.07, OR=52 (95% CI=0.26–−1.04). Consistent with the findings for Tolerance and Craving, diplotype AA was associated with higher Loss of Control scores than was BB, p=.004, OR=0.35 (95% CI=0.17–0.71). Diplotype AA also was associated with higher scores on Loss of Control than was AB, p=.05, OR=0.53 (95% CI=0.28–1.00). Automaticity was not associated with the diplotypes.

Open in a separate window

Figure 3.

Dichotomous Primary Dependence Motives scales by haplotype (top) and diplotype (bottom) in early-onset smokers. The scales were scored dichotomously based on the median for both age-at-onset conditions.

To assess the robustness of the PDM findings across cohorts, haplotype associations with dichotomized PDM and Tolerance (i.e., the subscale with the strongest signal across cohorts) scales were tested by age-at-onset condition within each cohort (UT; WI) separately. The H_A versus H_C contrast was significant in early-onset smokers in both cohorts for Tolerance: UT, p<.01, OR=0.50 (95% CI=0.29–0.87); WI, p<.002, OR=0.45 (95% CI=0.27–0.75), and for PDM: UT, p=.05, OR=0.58 (95% CI=0.33–1.01); WI, p<.02, OR=0.54 (95% CI=0.32–0.90). The H_A versus H_B contrast for Tolerance was significant among early-onset smokers only in the UT cohort, p<.05, OR=0.60 (95% CI=0.38–0.97). No significant haplotype effects were observed among late-onset smokers in either cohort. Thus, the finding in the combined sample that H_A versus H_C effects were stronger in early-onset than late-onset smokers was obtained in the two different cohorts for both Tolerance and the PDM composite.

Smoking cessation

In a logistic regression analysis in which the dependent variable was relapse versus no relapse over the first 90 days postquit, we found no haplotype × age-at-onset interaction. Thus, further tests of smoking cessation were made across the combined age-at-onset conditions. As shown in Figure 2 and Table 3, H_A was associated with a higher relapse rate than H_C, and the H_A versus H_B contrast approached significance. Diplotype analyses, also shown in Figure 2, indicated that diplotype AA had a higher 90-day relapse rate than every other diplotype, p’s=.006–.03, OR’s<0.45 (95% CI’s=0.15–0.93).

Abstinence rates, expressed as Kaplan–Meier survival functions, over the first 90 days postquit are shown as a function of haplotype and diplotype in Figure 4. A survival analysis revealed a significant haplotype effect, χ²(2, n=767)=7.45, p=.02. Pairwise comparisons were significant for both the H_A versus H_B test, χ²(1, n=629)=5.41, p=.02, and the H_A versus H_C test, χ²(1, n=638)=4.58, p<.04. The H_B versus H_C comparison was not significant. We also found a significant diplotype effect across all five diplotypes, χ²(4, n=629)=5.41, p=.02. Pairwise comparisons indicated that the relapse rate was higher for diplotype AA than for each of the other four diplotypes, p’s=.005–.03 (see Figure 4). Entering drug (bupropion vs. placebo) as a covariate in the haplotype or diplotype survival analyses did not alter any of these findings nor were there any significant interaction effects with drug.

Open in a separate window

Figure 4.

Abstinence following a smoking cessation trial by haplotype (top, n=767) and diplotype (bottom, n=356). Abstinence is shown as the Kaplan–Meier probability of remaining abstinent for the first 90 days postquit. Data are based on the age-at-onset conditions combined.

Discussion

The present results with the PDM expand our previous findings using the FTND in the same samples (Weiss et al., 2008). The analyses reported here suggest that H_A and H_C of the CHRNA5-A3-B4 cluster are relative risk and protective factors, respectively, for heavy, pervasive, and automatic smoking among individuals who began smoking early in life (i.e., daily smoking at age 16 or younger). The interaction between age-at-onset condition and the H_A versus H_C contrast with PDM as the dependent variable is consistent with our previous finding with FTND as the dependent variable. Further, although the SNPs and haplotypes are associated similarly with the phenotypes (see Table 4), the haplotypes provide information germane to the underlying genetic architecture (see Figure 1). Finally, the present results expand our knowledge by linking CHRNA5-A3-B4 haplotypes with withdrawal and smoking cessation phenotypes. This set of genetic variants has broad implications for the manifestation of nicotine dependence.

Several findings of this research suggest that the CHRNA5-A3-B4 haplotypes confer important information about dependence risk. First, the association between these haplotypes and PDM was found in two separate samples. Second, the magnitude of the associations was meaningful. For example, persons with diplotype AA relapsed at about twice the rate of other individuals. Third, the haplotypes were predictive of risk across multiple phenotypes: pervasive, heavy smoking, withdrawal magnitude, and relapse. The relationships suggested that the nature of the genetic risk differed for heavy smoking and relapse on the one hand versus withdrawal magnitude on the other hand (see Discussion). Fourth, the general correspondence in the results of continuous and binary scorings of the PDM suggests that the obtained results are not an artifact of incorrectly modeling the relationship between this dependence index and the genotypes.

We found evidence that H_A is a risk factor and H_C a protective factor for both smoking heaviness and short-term relapse or cessation failure. We also found evidence that, for withdrawal severity, H_A is a relative protective factor and H_C a relative risk factor. The smoking heaviness and withdrawal severity phenotypes were not highly associated with one another in this research (Table 2) or in other studies (e.g., Baker et al., in press; Piper et al., 2006). Therefore, they may reflect different genetic underpinnings. It remains unclear, though, why the same cluster of genes impinges on the phenotypes in such different ways. One possibility is that H_A has a biological effect that fosters or permits smoking heaviness, whereas H_C increases withdrawal severity independently. A second possibility is that H_A, relative to H_C, may decrease the effects of nicotine, resulting in greater smoking but weaker withdrawal. The latter possibility is suggested by a recent functional study that reports diminished response to the nicotine agonist epibatidine associated with a variant in the α5 nAChR subunit (Bierut et al., 2008).

The associations with withdrawal and cessation were not moderated by age at onset of daily smoking. It is possible that, in fact, the age-at-onset moderation of nicotine dependence is specific to the smoking heaviness subphenotype. However, the withdrawal and cessation phenotypes were available in only one cohort, which reduced the power for testing moderation effects. Further research is needed to address the moderation of the CHRNA5-A3-B4 effect by age.

As noted earlier, the present results suggest a gene × environment interaction (see Caspi et al., 2003; Moffitt, 2005; Swan & Lessov, 2004) such that early exposure to nicotine is needed for the association of the risk for smoking heaviness with H_A. Confidence in this finding is bolstered somewhat by independent evidence that early exposure is associated with either higher levels of nicotine self-administration or greater neurobiological impact of nicotine. Research with smokers suggests that exposure to nicotine relatively early in life is related to heightened consumption of cigarettes in adulthood (Breslau & Peterson, 1996; Broms, Silventoinen, Lahelma, Koskenvuo, & Kaprio, 2004), a relative inability to quit smoking (Breslau & Peterson, 1996; Chen, Stanton, Shankaran, & Li, 2006; Hymowitz et al., 1997; John, Meyer, Hapke, & Rumpf, 2004), and more severe nicotine dependence (Grant, 1998; Pergadia, Heath, Agrawal, et al., 2006; Xian et al., 2007). Research with rodents shows that nicotine exposure during periadolescence induces long-lasting neurochemical, anatomical, physiological, and behavioral changes that differ markedly from those seen with initial exposure in adulthood (Adriani et al., 2003; Cruz, Delucia, & Planeta, 2005; Schochet, Kelley, & Landry, 2004; Slotkin, 2002). Thus, research suggests that early nicotine exposure may produce a more severe level of nicotine dependence, and the present results suggest that in humans this effect depends on CHRNA5-A3-B4 status.

Analyses in which the WISDM-68 PDM and SDM scales served as covariates for one another suggested that the H_A effect can be linked more directly to the particular motives indexed by PDM versus the smoking motives reflected by SDM. SDM was related to the haplotypes only to the extent that it was correlated with PDM. Recent research suggests that the PDM scales may constitute necessary and sufficient features of dependence (Piper et al., in press). That is, smokers do not show significant dependence on a variety of other measures unless these scales are elevated, and smokers can be significantly dependent if only these scales are elevated. The present findings suggest that the CHRNA5-A3-B4 variants are related to a core dimension of dependence.

In theory, the present results suggest a causal model in which the CHRNA5-A3-B4 cluster variants increase the likelihood that a person can or will smoke very heavily, and such heavy smoking then produces a highly automatic smoking ritual that can be elicited with little awareness, is hard to control, and the interruption of which occasions strong craving (Curtin, McCarthy, Piper, & Baker, 2006; McCarthy, Curtin, Piper, & Baker, in press; Tiffany, 1990). These features of heavy, automatic smoking with accompanying craving could then increase the likelihood of relapse via various routes. For example, strong craving might precipitate relapse, and so on.

Some research suggests why SDM does not reflect a core dimension of dependence. Many of the individual SDM scales elicit information on instrumental uses of smoking, for example, smoking to control weight and negative moods or to experience pleasure. If entrenched smoking reflects habit learning (Everitt & Robbins, 2005), it makes sense that instrumental dependence motives would be less highly associated with important markers of severe dependence than the features tapped by PDM.

In conclusion, the present research shows that major CHRNA5-A3-B4 haplotypes identify countervailing susceptibility (H_A) and protective (H_C) influences on nicotine dependence as it is indexed by a subphenotype that reflects self-report of heavy, automatic smoking that defies conscious control. The assessed haplotypes showed significant relationships with this subphenotype only among smokers who began daily smoking relatively early in life, suggesting the expression of genetic risk occurs only among those with early tobacco exposure. This effect could have public health implications because interventions that reduce youth tobacco use (e.g., excise taxes) could be used to mitigate the expression of genetic risk. The research also shows that the assessed CHRNA5-A3-B4 haplotypes were significantly associated with two other major nicotine dependence subphenotypes: withdrawal severity and an inability to stop smoking. Although H_A was associated with a heightened risk of relapse relative to H_C, H_C was associated with more severe withdrawal relative to H_A. The results show that the assessed haplotypes are associated with all three of the major nicotine dependence subphenotypes, and they may confer risk via multiple routes.

Funding

This research was supported, in part, by the National Institutes of Health grants National Institute on Drug Abuse/National Heart, Lung, and Blood Institute (NHLBI) P01-HL72903;, NHLBI/N01-HR46014;, CA/DA P50-84178;, P50-CA84724;, and P50-DA0197.

Declaration of Interests

TBB has served as an investigator on research projects sponsored by pharmaceutical companies, including Pfizer, GlaxoSmithKline, Sanofi, and Nabi. Over the past 3 years, MCF has served as an investigator in research studies at the University of Wisconsin that were funded by Pfizer, GlaxoSmithKline, and Nabi Biopharmaceuticals. In 1998, the University of Wisconsin appointed Fiore to a named chair funded by an unrestricted gift to the university from Glaxo Wellcome.

Supplementary Material

Acknowledgments

References