Haplotype spanning TTC12 and ANKK1, flanked by the DRD2 and NCAM1 loci, is strongly associated to nicotine dependence in two distinct American populations

Abstract

Nicotine dependence (ND) is a moderately heritable trait. We ascertained a set of 1615 subjects in 632 families [319 African-American (AA) and 313 European-American (EA)] based on affected sibling pairs with cocaine or opioid dependence. Subjects were interviewed with the Semi-Structured Assessment for Drug Dependence and Alcoholism (SSADDA). Previously, we identified a modest linkage peak (LOD score =1.97) for ND in the EA part of the sample on chromosome 11q23, a region that includes the NCAM1–TTC12–ANKK1–DRD2 gene cluster. DRD2 and NCAM1 are functional candidate genes for substance dependence; the TTC12 and ANKK1 loci are not well characterized. We genotyped a set of 43 single nucleotide polymorphisms (SNPs) spanning this region, and performed family-based association and haplotype analysis. There was relatively weak evidence for association of the flanking DRD2 and NCAM1 markers to ND, but very strong evidence of association of multiple SNPs at TTC12 and ANKK1 in both populations (minimal P=0.0007 in AAs and minimal P=0.00009 in EAs), and in the pooled sample, as well as strong evidence for highly significant association of a single haplotype spanning TTC12 and ANKK1 to ND in the pooled sample (P=0.0000001). We conclude that a risk locus for ND, important both in AAs and EAs, maps to a region that spans TTC12 and ANKK1. Functional studies of these loci are warranted. These results provide additional information useful in evaluating the many earlier discrepant findings regarding association of DRD2 with substance dependence.

INTRODUCTION

Nicotine dependence (ND) is highly prevalent and highly destructive in the US and throughout the world. In 2004, it was reported that 67.3% of the US population (aged 12 and over) had ever smoked cigarettes, and 29.1% had smoked cigarettes in the past year (1). ND and many behaviors related to it, like other forms of substance dependence, have been demonstrated to be heritable. Li et al. (2) conducted a meta-analysis of twin studies for phenotypes related to smoking, and found that smoking initiation heritability could be estimated as 0.37 for males and 0.55 for females; smoking persistence heritability was estimated as 0.59 for males and 0.46 for females. Increased understanding of genetic risk factors for ND would be of great social and economic value. Identification of risk loci has been addressed so far through linkage methodology [2, review of studies up to 2004; (3–9)], and through association approaches. Several candidate gene associations have been replicated [e.g. DDC, dopa-decarboxylase (10,11)].

There is also a long history of reports of association of the D2 dopamine receptor (DRD2) locus with substance dependence. Dopamine has been implicated in the reinforcing effects of a variety of drugs of abuse, including nicotine (12). In concentrations achieved by smokers, nicotine activates and desensitizes nicotinic receptors and regulates the activity of mesolimbic dopamine neurons (13). D2 dopamine receptors are largely distributed throughout limbic structures (14), consistent with a prominent role of this receptor type in motivated behavior, including addiction. Association studies of variants that map at or near DRD2 with ND have yielded variable findings, with most failing to show association. A meta-analysis of 12 studies (2) showed that overall there was a significant association (pooled OR=1.47). Another meta-analysis of 13 studies of this association (15) showed significant heterogeneity among the studies, and no association when a random-effects model (which accounts for between-study differences) was applied. Overall, the possibility of an association at this locus has not been resolved definitively, and deserves to be addressed in an adequately powered and rigorously assessed sample.

We collected a set of small families suitable for linkage and family-based association analyses of cocaine and opioid dependence (16,17). These subjects were evaluated with the Semi-Structured Assessment for Drug Dependence and Alcoholism (SSADDA) (16,18), a polydiagnostic instrument that provides for DSM-IV ND, and in which the Fagerstrom Test for ND (FTND) (19,20) is embedded. This family set proved to be highly informative for studies based on ND and the FTND as well, although ascertainment was not conditioned on smoking (9,11).

The neural cell adhesion molecule (NCAM1) locus maps molecularly close to DRD2 (21). NCAM1 is also an attractive candidate locus; its protein product regulates the secretion of neurotransmitter as a membrane-bound glycoprotein that mediates cell–cell contact by homophilic interactions (22). NCAM1 has been associated genetically with neural tube defects, bipolar affective disorder and schizophrenia (23–25), and with adverse effects of nicotine on brain plasticity (26). On the basis of the hypothesis that NCAM1 was also a gene that could influence risk of addiction, we genotyped sets of single nucleotide polymorphism (SNP) markers at DRD2 and NCAM1. In the course of our study, it was reported that two additional genes, tetratricopeptide repeat domain 12 (TTC12) and ankyrin repeat and kinase domain containing 1 (ANKK1), map between DRD2 and NCAM1, so we extended our work to provide coverage of these loci also. Thus, to test the hypothesis that variants in NCAM1, DRD2 or genes located between them, namely TTC12 and ANKK1, are associated with smoking behavior, we evaluated 43 SNP markers spanning this region in our family sample.

Several noteworthy findings emerged from our genomewide linkage study of ND and FTND in this cohort, including a possible locus on chromosome 11 (at D11S908), where we observed a LOD score of 1.97 in the European-American (EA) part of the sample for ND (9). Under this LOD score peak is a 600 kb region including the NCAM1, TTC12, ANKK and DRD2 genes, which we will refer to as a ‘cluster’ (without intending any implication beyond their close physical localization). This finding (noted after SNP genotyping for this project had been completed) greatly increased our already-strong interest in this candidate region, in our specific clinical sample. Initially, interest was focused on NCAM1 and DRD2 as candidate loci, as discussed earlier. The protein encoded by ANKK1 is one of a family of proteins involved in signal transduction pathways. The so-called ‘DRD2 TaqI A’ system causes an amino acid substitution within the 11th ankyrin repeat of this gene (Glu713Lys), which may affect substrate-binding specificity (27). TTC12 is still undercharacterized.

Thus, the three of the four genes mapped to this cluster that are characterized (at least to some extent) functionally, may all be considered candidate loci for regulating risk for substance dependence. Prior, although ambiguous, genetic association data, added to this a priori support; and our own linkage data were also consistent with a risk locus in this region. Our analysis, however, led us to an unexpected conclusion; we discovered weak evidence of association of ND with SNPs located within DRD2 and NCAM1, but strong and consistent evidence of association of ND with markers within TTC12 and ANKK1. A specific haplotype comprising SNPs that span parts of both of these genes was strongly associated with ND in sub-samples of African-American (AA) and EA families.

RESULTS

Single SNP analysis

In general, there were many nominally-significant results for DSM-IV ND, and fewer for the FTND trait. We found a moderate correlation between the diagnosis of ND and FTND score (AAs: Spearman's ρ=0.521; EAs: ρ=0.591; combined: ρ=0.575), indicating that most of the variance in these measures is not shared. These findings are consistent with the differential association findings for these traits. For AAs, eight markers showed at least nominally-significant association to ND and two to FTND. For EAs, 15 markers showed at least nominally-significant association to ND and seven to FTND. For the entire (pooled) sample, 13 markers showed at least nominally-significant association to ND and four to FTND. All of these findings, but one in the pooled sample (P=0.04 for SNP3 with ND), were observed between SNP17 (in NCAM1) and SNP40 (in DRD2). Of these 24 SNPs, spanning 201 179 bp, all but six showed at least ‘trend-level’ (i.e. P≤0.10) significance for association to ND in the EAs, and all but eight did so in the pooled sample.

Nearly all variants that were statistically significant after correction cluster at TTC12 and ANKK1; one SNP meeting this threshold (rs6277) is located in DRD2 and is the most proximal marker to ANKK1 studied at that locus. This variant, C957T, has been shown previously to be functional (28); moreover, it has been reported to regulate the effect of nicotine on working memory (29). However, the association in our study is seen only in EAs, where rs6277 is in strong LD with the markers in TTC12 and ANKK1, which are significantly associated with ND. There is no association of this marker with phenotype in AAs, nor is the marker in LD with markers in the other genes in this population. Thus, it is possible, and more consistent with our other observations, that this association reflects LD with markers in TTC12 or ANKK1.

In EAs, the most statistically significantly associated markers (besides rs6277) map to ANKK1: rs4938012 (P=0.0003), rs4938013 (P=0.0004), rs4938015 (P=0.00009) and rs11604671 (P=0.0007). In AAs, only the first two of these reach even nominal significance, and the most statistically significantly associated marker is rs2303380 located in TTC12 (P=0.0007 in AAs and P=0.007 in EAs).

In the combined (EA plus AA) sample, numerous markers that tend to cluster at the TTC12/ANKK1 border show increased evidence of association. The most strongly associated markers include two in TTC12, rs2303380 (P=0.00002) and rs2282511 (P=0.00009); and the next three markers in ANKK1, rs4938012 (P=0.000008), rs4938013 (P=0.00003) and rs4938015 (P=0.0006). One possible explanation for this pattern, which is consistent with LD as observed in this sample, is that the findings in AAs and EAs are driven by the same functional variant or variants, and the difference in single SNP associations is due to population-based LD differences with that variant or set of variants. If this explanation is correct, the functional variant would be expected to be localized near the 3′ region of TTC12 and/or the 5′ region of ANKK1.

Haplotype association analysis

Analysis of SNPs individually indicated that the region most likely to contain an ND susceptibility locus is bounded by TTC12 rs2303380 centromerically and extends into ANKK1 but not into DRD2. Haplotype analysis was conducted to help narrow the location of a susceptibility locus, and determine whether a single functional variant could explain the pattern of association findings with individual SNPs in each population group. As a first step, we evaluated LD among the 43 SNPs in the cluster of four genes in order to reduce the number of markers that would be potentially informative for haplotype analysis. This analysis, shown in Figure 1, revealed slightly more extensive LD in EAs than in AAs, and these population-specific patterns are consistent with the LD structure reported in the HapMap database for this genetic region (30). There was a considerable amount of intergenic LD, particularly in the EAs.

Figure 1.

LD patterns in (A) EA and (B) AA subjects.

We observed a total of 10 LD blocks in EAs and nine LD blocks in AAs. The block structures are remarkably similar in these two population groups; in fact, three blocks accounting for nine SNPs are identical. Taking into account the LD block structure (Fig. 1) and the association findings with individual SNPs (Table 1), four SNPs (rs2303380, rs4938012, rs4938015 and rs11604671) were selected for haplotype analysis. These markers include four of the six responsible for the most significant results in the combined sample, and account for the potentially uniquely important information from each LD block spanning the distal portion of TTC12 and all of ANKK1 in both EAs and AAs.

Table 1.

Results of family based analysis for the single SNPs

Locus	Marker number	dbSNP rs number	AA		EA		Pooled
			TobDep	Fag_res	TobDep	Fag_res	TobDep	Fag_res
NCAM1	1	rs4547132	0.99	0.87	0.95	0.04	0.95	0.11
	2	rs2011505	0.38	1.00	0.31	0.10	0.18	0.27
	3	rs1245119	0.10	0.41	0.16	0.23	0.03	0.80
	4	rs2043602	0.77	0.46	0.37	0.63	0.40	0.83
	5	rs2117912	0.17	0.21	0.35	0.16	0.12	0.53
	6	rs1821693	0.18	0.85	0.02	0.09	0.50	0.22
	7	rs686934	0.15	0.81	0.05	0.13	0.74	0.18
	8	rs584427	0.81	0.98	0.17	0.21	0.35	0.26
	9	rs646558	0.52	0.66	0.61	0.79	0.83	0.61
	10	rs586903	0.67	0.91	0.14	0.86	0.57	0.94
	11	rs2303377	0.15	0.29	0.44	0.26	0.58	0.95
	12	rs605843	0.38	0.52	0.12	0.39	0.72	0.28
	13	rs2288158	0.61	0.57	0.53	0.88	0.42	0.82
	14	rs2298703	0.03	0.64	0.20	0.84	0.60	0.73
	15	rs598026	0.71	0.38	0.17	0.38	0.66	0.21
	16	rs2156485	0.25	0.03	0.81	0.96	0.69	0.52
	17	rs618114	0.03	0.45	0.09	0.40	0.008	0.27
	18	rs593217	0.18	0.71	0.08	0.36	0.03	0.36
	19	rs688011	0.50	0.25	0.33	0.12	0.24	0.67
	20	rs7103866	0.87	0.24	0.17	0.23	0.35	0.73
TTC12	21	rs4517559	0.15	0.28	0.02	0.04	0.58	0.48
	22	rs2236709	0.08	0.42	0.48	0.01	0.08	0.36
	23	rs1893699	0.32	0.51	0.07	0.11	0.05	0.50
	24	rs723077	0.78	0.10	0.02	0.08	0.13	0.67
	25	rs2303380	0.0007	0.62	0.007	0.04	0.00002	0.06
	26	rs948176	0.06	0.37	0.42	0.58	0.30	0.28
	27	rs2288159	0.008	0.22	0.30	0.41	0.15	0.13
	28	rs719804	0.72	0.53	0.03	0.15	0.07	0.47
	29	rs2282511	0.02	0.46	0.002	0.09	0.00009	0.07
ANKK1	30	rs4938012	0.01	0.46	0.0003	0.03	0.000008	0.03
	31	rs4938013	0.02	0.77	0.0004	0.02	0.00003	0.05
	32	rs4938015	0.48	0.54	0.00009	0.02	0.0006	0.04
	33	rs7118900	0.78	0.24	0.07	0.71	0.29	0.31
	34	rs11604671	0.93	0.95	0.0007	0.18	0.004	0.24
	35	rs4938016	0.58	0.12	0.002	0.48	0.09	0.55
	36	rs1800497	0.95	0.14	0.06	0.38	0.22	0.10
DRD2	37	rs6277	0.49	0.53	0.001	0.16	0.014	0.14
	38	rs1800498	0.41	0.76	0.004	0.17	0.05	0.20
	39	rs1079597	0.50	0.25	0.03	0.32	0.04	0.15
	40	rs1799732	0.002	0.02	0.90	0.20	0.02	0.008
	41	rs6589377	0.85	0.33	0.71	0.46	0.68	0.24
	42	rs4482060	0.50	0.87	0.81	0.25	0.81	0.45
	43	rs4938023	0.91	0.31	0.75	0.29	0.85	0.15

Please see Table 4 for SNP characteristics.

The haplotype including this SNP combination was significantly associated with ND in AAs (global P=0.0003), EAs (global P=0.0008) and the combined sample of families (global P=0.000005) (Table 2). The specific haplotype G-A-T-C was associated with ND in EAs (P=0.00006), AAs (P=0.0008) and in the pooled sample (P=0.0000001). Thus, these results showing strong evidence for association of the same haplotype to ND in two distinct populations suggest the possibility of a single causative variant. Haplotypes A-G-C-T and A-G-T-C were associated with decreased risk of ND in EAs (P=0.001) and AAs (P=0.0009), respectively. The divergent frequencies of these protective haplotypes in these two population groups may explain the observation of distinct association peaks, at rs2303380 in AAs and near rs4938013 in EAs. This SNP combination was also modestly associated with FTND score, but in EAs only (global P=0.04). This result is attributable primarily to the haplotype G-A-T-C (P=0.007), which was associated with increase in FTND score (data not shown).

Table 2.

Haplotype analysis results for the association of major haplotypes for rs2303380–rs4938012–rs4938015–rs11604671 with TobDep

Haplotype	AA (0.0003)*			EA (0.0008)*			Pooled (0.000005)*
	Frequency	Number of FAM	(Z) P-value	Frequency	Number of FAM	(Z) P-value	Frequency	Number of FAM	(Z) P-value
A-G-C-C	0.30	52	(−1.4) 0.17	0.17	33	(−0.6) 0.56	0.25	81	(−1.5) 0.14
A-G-C-T	0.07	19	(0.6) 0.58	0.45	48	(−3.2) 0.001	0.23	67	(−2.6) 0.009
G-A-T-C	0.18	46	(3.4) 0.0008	0.28	45	(4.3) 0.00006	0.22	89	(5.3) 0.0000001
A-G-T-C	0.34	51	(−3.3) 0.0009				0.20	59	(−2.8) 0.005
G-G-T-C	0.10	23	(1.2) 0.24				0.06	23	(0.9) 0.32

*Global P-value.

Correlation of trait with cocaine and opioid dependence

Since the sample was recruited for cocaine and opioid dependence, we examined the correlations between these diagnoses and ND. The correlations were low (ND–CD: Spearman's ρ=0.190; ND–OD, ρ=0.166). The modest association between ND and these other diagnoses argues against the possibility that the allelic and haplotypic associations reported here for ND can be explained on the basis of confounding by comorbid CD or OD. Further, we did not observe noteworthy evidence of association with the DSM-IV diagnoses of CD or OD in the specific region implicated for ND. However, our ability to detect association with these traits in this sample is less than that for ND, since the sample was ascertained for linkage studies [i.e. affected sibling pairs (ASPs)] with CD and OD (but not ND), whereas discordant sibling pairs are more informative for family-based association studies.

DISCUSSION

We present here strong evidence for association of markers that span portions of TTC12 and ANKK1 to ND in two distinct populations, EAs and AAs. Although the single SNP results for the two populations favor somewhat different localizations for a risk locus, with the strongest results in ANKK1 for EAs and in TTC12 for AAs, both populations show multiple at least nominally significant associations at both loci; and pooled results show multiple significant (when corrected) associations at both loci. One haplotype that comprised SNPs from both loci is significantly associated in each population group, and is highly significantly associated in the pooled population sample (P=0.0000001). All of these results derive from family-based analyses, and thus cannot easily be attributed to population stratification artifact. We therefore conclude that a variant influencing risk for ND maps to this region, probably within ANKK1 or TTC12, and probably not within DRD2 or NCAM1. This does not exclude the presence of risk loci of smaller effect in either of these flanking loci, but in this data set, there are much stronger results supporting a risk locus in ANKK1 or TTC12.

DRD2 is well known in psychiatric genetics for many early claims of association (mostly to alcohol dependence) sometimes based on small study samples. Rs1800497 is a variant that has been called historically ‘DRD2 TaqI A,’ although it is now known to map within ANKK1 and result in a non-synonymous base substitution (27). We found no evidence of association of this variant with ND (or FTND); we have previously studied this variant in other samples, and did not find association with alcohol or drug dependence (31–33) and argued previously that no functional relationship of this variant to alcohol dependence had been proved (34). However, inspection of Figure 1A reveals that this marker is in LD with other variants we did find to be associated with ND, for example, D′=0.73 with rs4938015, and D′=1.0 with rs1604671. Thus, it is apparent that previous reports of association of substance dependence traits that included this marker only could, potentially, be attributable to LD with a more centromeric marker mapping to ANKK1 or TTC12. Since this marker is also in LD with other markers at DRD2, we cannot exclude functional risk-influencing variants at that locus, nor can we exclude a variant 3′ to DRD2 that has a primary regulatory effect on that gene. There is also a previous report of very strong haplotypic association of DRD2 variants with heroin dependence (35). That study did not consider other ANKK1 or TTC12 variants, so it is unclear to what extent, if at all, variants at those loci could explain the observed association, or if, alternatively, DRD2 itself is an important locus in some populations and for some substance use disorders.

We found evidence of association to the trait of DSM-IV ND, and only weak evidence of association to the related trait, FTND score. Key elements of the DSM-IV diagnostic criteria are at least three of the following symptoms in a 12-month period: (i) tolerance; (ii) withdrawal; (iii) tobacco taken in larger amounts or over a longer period than intended; (iv) persistent desire to cut down; (v) great deal of time spent acquiring tobacco or dealing with its effects; (vi) important activities given up because of tobacco; (vii) use continued despite subject's knowledge that it is causing harm (28). FTND stresses somewhat different aspects of smoking; it queries respondents regarding, for example, how soon after waking they smoke their first cigarette, which cigarette in the day would be hardest for the respondent to give up, and smoking quantity (20); these issues are not addressed directly in the DSM-IV criteria. Although both ND and FTND scores are heritable, as they measure somewhat different aspects of nicotine use and dependence, it is not surprising that there may be different contributions to each from different risk loci. Consistent with this idea, the linkage signal we observed in this region (9) was specific to ND, and so is the observed association. Conversely, we reported previously that DDC variants are associated much more markedly to FTND than to ND (11).

Our sample was recruited for the presence of opioid and/or cocaine dependence, not for ND, although our subjects had a high prevalence of ND and the sample proved to be very informative for the purpose of ND gene mapping (many of the subjects are polysubstance abusers). Since the sample was not recruited for ND, it would be reasonable to ask how representative this sample is for the trait in general. As we noted previously (9), we have a direct way to address the generalizability of findings from this particular sample, at least with respect to one specific locus, DDC. Ma et al. (10) reported association of several ND-related traits to markers at this locus in AAs; we (11) confirmed this finding, and extended it to EAs. The clinical sample for the study by Ma et al. (10) was recruited for a genetic linkage study of ND; our study (11) was conducted in the same sample as the present study. The DDC locus contributes to ND risk in both samples. Some of the linkage peaks we observed (9) were also consistent with those observed in previous studies. Thus, we conclude that our sample is, in a number of respects, comparable with other samples used in ND studies. Even so, we do not claim that this association is specific to ND alone; since many subjects are dependent on multiple substances of abuse, there may be some more specific predisposition to a substance or a set of substances.

Our results, which were derived from Family-Based Association Test (FBAT) analysis of family samples using unaffected sibs as controls, are unlikely to be the result of stratification within a population group. Our observation of significant results in two distinct populations, also greatly supports the robustness of these results. AA populations may have >20% admixture with EA populations, with estimated EA admixture of ∼20% for a New York AA population (36), reasonably comparable geographically to the population we studied, most of which was recruited in New England (17); the pattern of the results we observed (reflected in both the statistical strength of the observed effects, and in the fact that some different SNPs were associated in the EA and AA samples) supports the conclusion that the results in AAs were not driven merely by admixture from EAs.

This is one of the most robust reports of genetic association to ND to date. We studied a genomic region now supported by our own linkage findings and by some previous association analyses; however, our strongest evidence for association did not coincide with what might have been considered the best-supported candidate loci a priori. We cannot rule out additional risk loci that map to DRD2 or NCAM, but it seems likely that, at least for the ND phenotype, the major susceptibility variant in this region is located in ANKK1 or TTC12. It seems very unlikely that these genes that either clearly (DRD2) or presumptively (ANKK1, NCAM1) act in neurotransmitter pathways related to addiction are chromosomal neighbors based on chance alone; we hypothesize that the proximity of these genes in this cluster is in some way related to their physiology. Functional studies of the loci implicated here may, we hope, provide new insights into pathophysiological mechanisms of nicotine dependence.

MATERIALS AND METHODS

Subject recruitment and assessment

ASPs with cocaine or opioid dependence defined according to DSM-IV criteria (37) were ascertained at four clinical sites in the United States: University of Connecticut Health Center (Farmington, CT), Yale University School of Medicine (APT Foundation; New Haven, CT), McLean Hospital (Harvard Medical School; Belmont, MA) and Medical University of South Carolina (Charleston, SC). Subjects diagnosed with a major psychotic illness (e.g. schizophrenia or schizoaffective disorder) were excluded. Additional siblings and parents of the ASPs were recruited whenever possible regardless of affection status. Characteristics of the sample included are described in Table 3.

Table 3.

Sample characteristics

Characteristic	AA	EA	Pooled
Number of families	319	313	632
Average genotypedmembers/family	2.7±1.0	2.4±0.8	2.6±1.0
Number of subjects	854	761	1615
Sex (% female)	504 (59.0)	366 (48.1)	870 (53.9)
Age (years)	40.9±7.3	37.8±10.2	39.4±8.9
FTND score	3.7±2.6	5.0±2.9	4.3±2.9
Cocaine dependence
Affected (%)	676 (79.2)	559 (73.5)	1235 (76.5)
Unaffected (%)	115 (13.5)	145 (19.0)	260 (16.1)
Unknown (%)	63 (7.4)	57 (7.5)	120 (7.4)
Opioid dependence
Affected (%)	214 (25.1)	252 (33.1)	666 (41.2)
Unaffected (%)	579 (67.8)	452 (59.4)	831 (51.5)
Unknown (%)	61 (7.1)	57 (7.5)	118 (7.3)
Tobacco dependence
Affected (%)	470 (55.0)	515 (67.7)	985 (61.0)
Unaffected (%)	326 (38.2)	192 (25.2)	518 (32.1)
Unknown (%)	58 (6.8)	54 (7.1)	112 (6.9)
Alcohol dependence
Affected (%)	328 (38.4)	323 (42.5)	651 (40.3)
Unaffected (%)	435 (50.9)	361 (47.4)	796 (49.3)
Unknown (%)	91 (10.7)	77 (10.1)	168 (10.4)

Subjects were interviewed using the SSADDA for psychiatric diagnosis as described previously (16,18). The reliability of the SSADDA for the diagnosis of ND was very good to excellent, with test–retest reliability κ=0.97 (based on 120 subjects) and interrater reliability κ=0.77 (based on 173 subjects) (18). The diagnosis of ND was based on application to the SSADDA data of a computer algorithm using DSM-IV diagnostic criteria. The FTND (19,20) (range of 0–10 points) is embedded in the SSADDA. Diagnostic information on tobacco or alcohol dependence was not involved in pedigree selection or extension. Subjects were classified as AA or EA based on a Bayesian model-based clustering method using genetic marker information as previously described (9). The parts of the sample recruited at the four sites were similar in terms of percent diagnosed with ND and mean FTND scores.

DNA was obtained from immortalized cell lines in most cases, but for a small number of subjects, DNA was obtained directly from blood or saliva. Subjects gave informed consent as approved by the institutional review board at each clinical site, and a certificate of confidentiality for the work was obtained from NIH (NIDA).

SNP selection and genotyping

Forty-three SNPs covering the region containing NCAM1, TTC12, ANKK1 and DRD2 were selected, most from the ABI SNPbrowser (38) and the NCBI SNP database (http://www.ncbi.nlm.nih.gov/projects/SNP/) (Table 4). Figure 2 locates this gene cluster with respect to our previously observed linkage signal on chromosome 11. Overall, we genotyped 20 markers in NCAM1, nine markers in TTC12, seven markers in ANKK1 and seven markers in DRD2. Prior published reports (for markers mapped to DRD2), information content, minor allele frequency, function potential, LD structure and validation evidence were considered in SNP selection. Detailed information about these SNPs is summarized in Table 4. SNPs were genotyped with a fluorogenic 5′ nuclease assay method, that is the TaqMan technique (39), using the ABI PRISM 7900 Sequence Detection System (ABI, Foster City, CA, USA); or using PCR-RFLP methods.

Figure 2.

Nicotine dependence linkage analysis (9), and map of SNPs genotyped.

Table 4.

SNP characteristics

Marker number	dbSNP rs number	GENE	SNP type (aka)	Chromosome position (bp)	Distance from previous marker (bp)	Alleles	MAFa (AA)	MAFa (EA)	MAFa (all)
1	rs4547132	NCAM1	Intron	112338023	…….	C/T	0.37	0.33	0.35
2	rs2011505	NCAM1	Intron	112493446	155423	A/G	0.47	0.49	0.48
3	rs1245119	NCAM1	Intron	112506871	13425	G/C	0.49(G)	0.35	0.42
4	rs2043602	NCAM1	Intron	112548936	42065	T/C	0.46	0.37	0.41
5	rs2117912	NCAM1	Intron	112565679	16743	T/C	0.07	0.21	0.15
6	rs1821693	NCAM1	Intron	112582751	17072	G/A	0.36	0.44(G)	0.47
7	rs686934	NCAM1	Intron	112590658	7907	C/T	0.36	0.42(C)	0.48
8	rs584427	NCAM1	Intron	112609206	18548	G/T	0.13	0.46	0.32
9	rs646558	NCAM1	Intron	112611117	1911	C/A	0.48(C)	0.28	0.39
10	rs586903	NCAM1	Intron	112616156	5039	A/C	0.26	0.20	0.23
11	rs2303377	NCAM1	Intron	112616711	555	T/C	0.24	0.37	0.31
12	rs605843	NCAM1	Intron	112630444	13733	T/C	0.47	0.28	0.36
13	rs2288158	NCAM1	Intron	112638886	8442	T/G	0.11	0.15	0.13
14	rs2298703	NCAM1	Intron	112639815	929	G/A	0.04	0.23	0.15
15	rs598026	NCAM1	Intron	112644460	4645	A/G	0.28	0.15	0.21
16	rs2156485	NCAM1	Intron	112648767	4307	C/T	0.05	0.26	0.17
17	rs618114	NCAM1	Intron	112650283	1516	G/A	0.24	0.39	0.33
18	rs593217	NCAM1	intergenic	112658699	8416	T/C	0.29	0.43	0.37
19	rs688011	NCAM1	intergenic	112659380	681	G/A	0.23	0.27	0.25
20	rs7103866	NCAM1	intergenic	112669978	10598	G/A	0.11	0.33	0.23
21	rs4517559	TTC12	intergenic	112685412	15434	A/G	0.50(A)	0.40	0.44
22	rs2236709	TTC12	Intron	112691986	6574	A/G	0.27	0.29	0.28
23	rs1893699	TTC12	Intron	112697734	5748	T/G	0.39	0.39	0.39
24	rs723077	TTC12	M73L	112699378	1644	A/C	0.18	0.46	0.34
25	rs2303380	TTC12	Intron	112705919	6541	A/G	0.35	0.40	0.37
26	rs948176	TTC12	Intron	112709691	3772	A/G	0.38	0.19	0.28
27	rs2288159	TTC12	Intron	112716539	6848	G/T	0.42	0.19	0.30
28	rs719804	TTC12	Intron	112739985	23446	A/G	0.23	0.24	0.23
29	rs2282511	TTC12	intergenic	112749387	9402	C/A	0.26	0.35	0.31
30	rs4938012	ANKK1	Intron	112764864	15477	G/A	0.25	0.36	0.31
31	rs4938013	ANKK1	exon (S)	112769680	4816	C/A	0.24	0.36	0.31
32	rs4938015	ANKK1	Intron	112769854	174	C/T	0.39(C)	0.39	0.49
33	rs7118900	ANKK1	A239T	112772031	2177	G/A	0.35	0.23	0.29
34	rs11604671	ANKK1	G318R	112773269	1238	C/T	0.11	0.45	0.31
35	rs4938016	ANKK1	G442R	112775225	1956	C/G	0.47(C)	0.35	0.43
36	rs1800497	ANKK1	E713K (DRD2-TaqI A)	112776038	813	G/A	0.39	0.23	0.30
37	rs6277	DRD2	exon (S) C957T	112788669	12631	A/G	0.12	0.49	0.33
38	rs1800498	DRD2	Intron (TaqI D)	112796798	8129	G/A	0.21	0.45(G)	0.40
39	rs1079597	DRD2	intron (TaqI B)	112801496	4698	G/A	0.18	0.19	0.18
40	rs1799732	DRD2	Promoter (ins/del)	112851462	49966	A/G	0.50(A)	0.15	0.31
41	rs6589377	DRD2	Intergenic	112860946	9484	A/G	0.20	0.37	0.30
42	rs4482060	DRD2	Intergenic	112863421	2475	A/T	0.32	0.42	0.38
43	rs4938023	DRD2	Intergenic	112880057	16636	G/T	0.20	0.36	0.29

^aMinor allele frequency is the frequency of the second allele unless noted in the parenthesis.

Statistical analysis

The PedCheck program was used to detect Mendelian inconsistencies in the genotype data (40). Consistency with Hardy–Weinberg equilibrium expectations (HWEE) was tested for each SNP in a set of unrelated non-smokers for each population group using a χ² test. One SNP (rs4938012) showed a significant deviation from HWEE in the EA group, and also showed numerous Mendelian errors. The Taqman genotypes were therefore discarded and genotypes were obtained instead via a PCR-RFLP assay; the new data were consistent with HWEE and generated no Mendelian errors. Subsequently, a total of 16 genotyping inconsistencies were identified out of 68 370 assays for 43 SNPs across all DNA samples (i.e. <0.02% of all assays), and these results were excluded from all subsequent statistical analyses (however, our sample had limited power to detect errors in this way). Because AA and EA samples are distinct populations with different histories, the allele frequencies in some markers differ prominently between them, and there is the potential for genetic heterogeneity, analyses were initially conducted separately for each population sample. To reduce potential confounding, the FTND score was adjusted for age and sex in each population sample by computing standardized residuals using SAS (version 9.0). Pairwise LD between all markers was assessed using the Haploview program (41) and haplotypes blocks were discerned according to the criteria of Gabriel et al. (42).

Association of individual SNPs with ND and FTND score was evaluated using the FBAT program (43) assuming an additive model under the null hypothesis of linkage and no association. The issue of testing multiple individual markers was addressed by applying a false discovery rate correction to the empirically derived P-values.The experiment-wide significance threshold required to keep type I error rate at 5% was 0.002, according to SNPSpD software (44,45). HBAT (46), the haplotype extension routine in the FBAT program, was used to evaluate haplotypic associations. Each SNP included in the haplotype analysis ‘tags’ an LD block determined empirically for both population groups separately. Overall association for a combination of SNPs was tested using a multi-allelic global test. For significantly associated combinations of SNPs (global P<0.05), haplotype-specific associations were tested using the diallelic mode in HBAT. P-values were estimated by the Monte Carlo sampling method under the null hypothesis of linkage and no association. The significance level of haplotype-specific tests was adjusted using Bonferroni correction, that is the α-level was divided by the number of major haplotypes with a frequency greater than 5.0%.

ACKNOWLEDGEMENTS

We thank the families who volunteered to participate in this research study. This work was supported by NIH (NIDA) grants R01 DA12690, R01 DA12849, K24 DA15105 and K02 DA00326; M01 RR06192; and the U.S. Department of Veterans Affairs (the VA Connecticut–Massachusetts Mental Illness Research, Education and Clinical Center [MIRECC]). Ann Marie Lacobelle, Lisa Naito, Lingjun Zuo and Yakov Lozovatsky provided excellent technical assistance. John Farrell provided excellent database support. We also thank the Rutgers University Cell and DNA Repository, the contractor for the NIDA Center for Genetic Studies, co-directed by Dr Jay Tischfield and Dr John Rice.

Conflict of Interest statement. None declared.