Genome-Wide Association for Nicotine Dependence and Smoking Cessation Success in NIH Research Volunteers

Tomas Drgon,1 Ivan Montoya,1,2 Catherine Johnson,1 Qing-Rong Liu,1 Donna Walther,1 Dean Hamer,3 and George R Uhl1

1Molecular Neurobiology Branch, National Institutes of Health–Intramural Research Program (National Institute on Drug Abuse), Baltimore, Maryland, United States of America; 2Division of Pharmacotherapies and Medical Consequences of Drug Abuse, National Institute on Drug Abuse, National Institutes of Health, Rockville Maryland, United States of America; 3Section on Gene Structure and Regulation, Laboratory of Biochemistry, National Institutes of Health–Intramural Research Program (National Cancer Institute), Bethesda, Maryland, United States of America

Phenotypes related to both nicotine dependence and ability to successfully quit smoking display substantial heritabilities in classical and molecular genetic studies. Twin studies suggest that some genetic components for dependence overlap with ge- netic components of ability to quit, but that many components do not overlap. Initial genome-wide association (GWA) studies have demonstrated haplotypes that distinguish nicotine-dependent from nondependent smokers.These haplotypes overlap par- tially with those that distinguish individuals who successfully quit smoking from those who were not able to quit smoking in clinical trials for smoking cessation. We now report novel genome-wide association results from National Institutes of Health research vol- unteers who reported smoking histories, symptoms of nicotine dependence, and ability to successfully quit smoking outside the context of a clinical trial. These results buttress data from several prior GWA studies. The data from these volunteers support the idea that previously reported studies of genes associated with smoking cessation success in clinical trial participants may also apply to smokers who are more or less able to initiate and sustain abstinence outside of clinical trial settings. © 2009 The Feinstein Institute for Medical Research, www.feinsteininstitute.org Online address: http://www.molmed.org doi: 10.2119/molmed.2008.00096

INTRODUCTION tine dependence as defined by criteria well as to dependence on other addictive Vulnerability to dependence on an ad- from either the DSM (Diagnostic and Sta- substances (12–17). Comparisons of data dictive substance is a complex trait with tistical Manual) or Fagerstrom test for from smokers whose FTND scores indi- strong genetic influences that are well nicotine dependence (FTND) (6–9). cate dependence on cigarettes with data documented by data from family, adop- Twin-study data also document sub- from smokers whose FTND scores dis- tion, and twin studies (1–4). Twin studies stantial heritability of the ability to suc- play little evidence of dependence have support the view that much of the herita- cessfully abstain from smoking (10,11). revealed repeated associations of moder- ble influence on vulnerability to depen- These data suggest that some, but not ate size within a chromosome-15 cluster dence on addictive substances from dif- most, of these genetic influences are of nicotinic-receptor genes (15–18). Partial ferent pharmacological classes (for shared with the genetics of developing data from one of these studies, which also example, nicotine and stimulants) is dependence on nicotine. reported modest associations at a number shared (2,3,5). These findings have in- Genome-wide association (GWA) ap- of additional loci, are available for com- creased interest in genes that might pre- proaches of increasing sophistication have parative analyses (15). dispose individuals to dependence on been developed and used to identify the We recently reported GWA data for multiple substances, such as nicotine and specific genes and genomic variants that success in quitting smoking in each of other drugs. For smoking, substantial predispose to phenotypes related to three independent samples of carefully data also document heritabilities of nico- smoking and to smoking cessation, as monitored individuals who tried to quit smoking in the context of clinical trials that used placebo, nicotine replacement, Address correspondence and reprint requests to George R Uhl, Molecular Neurobiology, or bupropion treatments (14). These tri- Box 5180, Baltimore, MD 21224. Phone: 443-740-2799; Fax: 443-740-2122; E-mail: guhl@intra. als, however, involved highly selected nida.nih.gov. smokers who were required to fulfill Submitted September 3, 2008; Accepted for publication October 1, 2008; Epub (www. many entrance criteria. Individuals molmed.org) ahead of print October 2, 2008. whose data were analyzed were required

MOL MED 15(1-2)21-27, JANUARY-FEBRUARY 2009 | DRGON ET AL. | 21 GENES FOR SMOKING AND SMOKING CESSATION

to participate in detailed and lengthy MATERIALS AND METHODS and amplified with a GeneAmp poly- follow-up. Such strict selection criteria merase chain reaction (PCR) system 9700 might have rendered these trial partici- Samples (Applied Biosystems, Foster City, CA, pants unrepresentative of smokers in the DNA samples were obtained from in- USA) with 3 min at 94°C; 30 cycles of 30 s general population who seek to quit dividuals who were recruited at the Na- at 94°C, 45 s at 60°C, 15 s at 68°C; and a smoking. Replication of many of these tional Institutes of Health (NIH) clinical final 7-min 68°C extension. PCR products results in individuals who quit in com- center for protocols based on cancer risk were purified (MinEluteTM 96 UF kits; munity settings would add to our confi- behaviors and personality as previously Qiagen, Valencia, CA, USA) and quanti- dence in the generalizability of these described (19). DNA samples from a tated, then 40 μg of PCR product was di- findings outside of the somewhat artifi- total of 480 individuals from the largest gested for 35 min at 37°C with 0.04 unit/ cial setting of a clinical trial. self-reported ethnicity in this sample, μL DNase I to produce 30–100 bp frag- Hamer and colleagues have assembled European-Americans, were divided into ments, which were end-labeled using ter- a research volunteer population at and (a) 13 pools (n = 20 in each) of DNA from minal deoxynucleotidyl transferase and around the NIH clinical center in Be- individuals who reported never smoking biotinylated dideoxynucleotides and hy- thesda, Maryland (19). These individuals extensively or becoming nicotine depen- bridized to the appropriate 500k array provided data on their histories of smok- dent, (b) 6 pools of DNA from 120 indi- (Sty I or Nsp I arrays) (Affymetrix). Ar- ing and success at quitting smoking. We viduals who reported lifetime nicotine rays were stained and washed as de- now report 500k single-nucleotide poly- dependence and current smoking, and scribed (Affymetrix Genechip Mapping morphism (SNP) GWA studies of the (c) 5 pools of DNA from 100 individuals Assay Manual) by use of immunopure participants of the Hamer et al. research who reported having been nicotine de- strepavidin (Pierce, Milwaukee, WI, volunteer population. We have targeted pendent at some time in their lives but USA), biotinylated antistreptavidin anti- these analyses to seek overlaps between who achieved abstinence. As previously body (Vector Labs, Burlingame, CA, nominally positive results from this sam- reported, individuals in these three USA), and R-phycoerythrin strepavidin ple and positive results from other GWA groups described their smoking history (Molecular Probes, Eugene, OR, USA). study populations, including a 38,000 as (a) smoking fewer than 101 cigarettes Arrays were scanned and fluorescence SNP GWA dataset obtained for heavy in their lives, (b) starting smoking at age intensities quantitated using an smokers who displayed FTND nicotine 17 ± 4 years, smoking for 18 ± 13 years, Affymetrix array scanner as described dependence versus smokers who did not consuming 20 ± 13 cigarettes/d, and con- previously (20). display FTND dependence (Beirut et al. tinuing to smoke when interviewed, or Chromosomal positions for each SNP dataset) (15), 500–600,000 SNP GWA (c) starting smoking at age 17 ± 3 years, were sought using Build 36.1 (National datasets obtained for smokers who suc- smoking an average of 20 ± 13 cigarettes/d Center for Biotechnology Information) ceeded versus those who did not succeed for 13 ± 11 years, and subsequently main- and NETAFFYX (Affymetrix) data. Al- in at least 2 of 3 clinical trials of smoking taining abstinence for 16 ± 12 years by lele frequencies for each SNP in each cessation (Uhl et al. dataset) (14), and the time of interviews (19). DNA pool were assessed based on hy- 600,000–1 million SNP GWA datasets bridization-intensity signals from four obtained for individuals who were de- DNA Preparation and Assessment of arrays, allowing assessment of hybridiza- pendent on at least one illegal addictive Allelic Frequencies tion to the 12 “perfect-match” cells on substance compared with control indi- DNA was prepared from blood or cell each array that are complementary to viduals with little or no peak lifetime use lines (21–23) and carefully quantitated. the PCR products from alleles “A” and of any addictive substance, including to- DNAs from groups of 20 individuals of “B” for each diallelic SNP on sense and bacco (Liu/Drgon et al. dataset) (20). We the same phenotype were combined. antisense strands. We eliminated SNPs discuss the significance of the substantial Pooled genotyping reduced costs and al- on sex chromosomes to provide greater overlaps in the data that we report, as lowed us to assess high densities of power based on inclusion of both males well as the limitations of the samples and genotypes in these subjects while provid- and females, and SNPs whose chromo- datasets. These data buttress previous ing no threat of loss of genetic confiden- somal positions could not be adequately GWA results for FTND nicotine depen- tiality to these individual research volun- determined. dence. They also document, for the first teers. Hybridization probes were Each array was analyzed as described time, overlaps between the molecular ge- prepared with precautions to avoid con- (20), with background values subtracted, netics of smoking cessation in clinical tri- tamination, as previously described normalization to the highest values als versus smoking cessation in commu- [500k array set; Affymetrix, Santa Clara, noted on the array, averaging of the hy- nity-based research volunteer samples CA, USA (20)]. A total of 150 ng of bridization intensities from the array who quit outside of the context of a clini- pooled DNA was digested by use of StyI cells that corresponded to the perfect- cal trial. or NspI, ligated to appropriate adaptors, match A and B cells, calculation of A/B

22 | DRGON ET AL. | MOL MED 15(1-2)21-27, JANUARY-FEBRUARY 2009 RESEARCH ARTICLE

ratios by dividing average normalized A fied by overlapping clusters of positive vided 0.99, 0.98, 0.87, and 0.54 power to values by average normalized B values, SNPs from each of these samples. detect allele frequency differences of arctangent transformations to aid combi- We used 10,000 Monte Carlo simula- 12.5%, 10%, 7.5%, and 5%, respectively. nation of data from arrays hybridized tion trials to compare observed results to For the comparison of continuing smok- and scanned on different days, and de- those expected by chance, as previously ers versus successful quitters, the sam- termination of the average arctan value described (20). For each trial, a randomly ples provided 0.97, 0.89, 0.66, and 0.33 for each SNP from the 4 replicate arrays. selected set of SNPs from the current power to detect differences of the same A t statistic for the differences between dataset was assessed to see if it provided sizes. The allele frequency differences nicotine abusers and controls was gener- results equal to or greater than the re- used in these power calculations were ated as described (20) for each SNP. We sults that we actually observed. The comparable to the actual mean allele fre- focused on SNPs that displayed t statis- number of trials for which the randomly quency differences for the nominally sig- tics with P < 0.05 to P < 0.01 for differ- selected SNPs displayed (at least) the nificant SNPs for smoker versus non- ences between nicotine-dependent versus same features displayed by the observed smoker and continuing smokers versus control individuals who never smoked results was then tallied to generate an successful quitters in these studies (4% more than 101 cigarettes and nicotine- empirical P value. These simulations and 7%, respectively). dependent current smokers versus indi- were thus correct for the number of re- viduals who were successful in maintain- peated comparisons made in these analy- Smokers versus Nonsmokers ing abstinence. We sought evidence for ses, an important consideration in evalu- When we compared nicotine-dependent clustering of these SNPs by focusing on ating these GWA datasets. individuals from these samples to non- chromosomal regions in which at least 3 To assess the power of our current ap- dependent individuals who smoked to 4 of these outlier SNPs, assessed by at proach, we used current sample sizes fewer than 101 cigarettes in their lives, least two array types, lay within 25 kb of and standard deviations, the program 27,803 SNPs displayed allele frequency each other. We termed these clustered, PS v2.1.31 (24,25) and α = 0.05. To pro- differences with nominal P < 0.05. Of nominally positive SNPs “clustered- vide controls for the possibilities that these, 3563 nominally positive SNPs fell positive SNPs,” and focused our analy- abuser-control differences observed into 665 clusters of at least 4 SNPs in ses on regions containing these clustered- herein were attributable to occult ethnic/ which each nominally positive SNP was positive SNPs. We used different criteria racial allele frequency differences or separated by ≤25 kb from at least 1 for P values and numbers of SNPs per noisy assays, we assessed, respectively, other nominally positive SNP. In each of cluster in these different analyses so that the overlap between the results obtained these clusters, we also required that the methods would be as identical as here and the SNPs that displayed the SNPs assayed on both array types dis- possible to those used for previously re- largest allele frequency differences be- play nominally positive results. These ported analyses of the appropriate com- tween African-American versus European- results identified 289 genes (Table S1). parison Uhl et al., Liu/Drgon et al., and American control individuals and the None of the 10,000 Monte Carlo simula- Beirut et al. datasets. largest assay “noise.” tion trials identified clustering that was To seek additional support for the clus- this significant (thus P < 0.0001). For the ters of nominally positive SNPs from the All supplementary materials are available 5570 SNPs that fell into 1334 clusters of current dataset, we sought convergence online at www.molmed.org. at least 3 SNPs, Monte Carlo P values between data from these clustered nomi- were <0.0001. nally positive SNPs and clustered nomi- RESULTS nally positive SNPs, determined in the We assessed allele frequencies in mul- Current Smokers versus Quitters same way, from the datasets of Beirut et tiple pools of DNA from smokers who We also observed significant clustering al. for 38,000 SNP GWA studies of light smoked when assessed, former smokers of nominally positive SNPs when we versus heavy smokers (15), Uhl et al. for who quit prior to assessment, and indi- compared current smokers with those 500–600,000 SNP GWA studies of individ- viduals who never smoked significant who had successfully quit. There were uals who were successful versus unsuc- numbers of cigarettes. Modest variability 5408 SNPs that displayed allele fre- cessful in quitting (14), and Liu/Drgon et occurred among replicate arrays that as- quency differences with nominal P < al. for 600,000–1 million SNP GWA stud- sessed the same pool. Standard error of 0.01; 914 of these nominally positive ies of dependence on illegal addictive the mean (SEM) values were 0.039. For SNPs fell into 239 clusters of at least 3 substances (20). To provide insights into variability among the different pools nominally positive SNPs separated from some of the genes likely to harbor vari- used to assess the same phenotype, SEM each other by <100 kb. In each of these ants that contribute to individual differ- was 0.024. For the smoker versus non- clusters, we again required that SNPs as- ences in addiction vulnerability, we smoker comparison, these samples and sayed on both array types display nomi- sought candidate genes that were identi- these estimates of variance thus pro- nally positive results. These results iden-

MOL MED 15(1-2)21-27, JANUARY-FEBRUARY 2009 | DRGON ET AL. | 23 GENES FOR SMOKING AND SMOKING CESSATION

Table 1. Nicotine dependence genes.a

Clustered SNPs MNB Gene Gene description Chromosome Base pair Hamer I II Bierut

A2BP1 Ataxin 2-binding protein 1 16 6009133 6 3 13 C10orf11 Chromosome 10 open reading frame 11 10 77212525 9 3 1 CAPN13 Calpain 13 2 30799141 4 3 2 CDH13 Cadherin 13 16 81218079 14 6 3 2 CHL1 Close homolog of L1 3 213650 2 7 3 CNTN6 Contactin 6 3 1109629 9 4 6 CSMD1 CUB and Sushi multiple domains 1 8 2782789 14 15 10 12 CSMD2 CUB and Sushi multiple domains 2 1 33752196 5 4 CSRP3 Cysteine and glycine-rich protein 3 11 19160154 3 3 CTNNA3 Catenin α 3 10 67349937 4 4 3 21 CTNND2 Catenin delta 2 5 11024952 4 3 1 CXCL14 Chemokine (C-X-C motif) ligand 14 5 134934274 2 4 DDEF1 Development and differentiation enhancing factor 1 8 131133535 4 3 FHIT Fragile histidine triad gene 3 59710076 15 3 5 2 FHOD3 Formin homology 2 domain containing 3 18 32131700 8 3 FLJ45872 FLJ45872 protein 8 138890869 4 3 GRIK2 Glutamate receptor, ionotropic, kainate 2 6 101953675 4 3 1 GRM3 Glutamate receptor, metabotropic 3 7 86111166 4 3 1 HPSE2 Heparanase 2 10 100208867 9 3 1 LRP1B Low density lipoprotein-related protein 1B 2 140705466 4 5 4 LRRN6C Leucine rich repeat neuronal 6C 9 27938528 7 4 18 NME7 Nucleoside-diphosphate kinase 1 167368393 4 14 PARD3B Par-3 partitioning defective 3 homolog B 2 205118761 4 3 PBX1 Pre-B-cell leukemia transcription factor 1 1 162795561 4 5 PDE1C Phosphodiesterase 1C 7 31795772 9 3 3 PTPRM Receptor type protein tyrosine phosphatase M 18 7557817 5 3 RAD51L1 RAD51-like 1 14 67356262 5 3 RYR3 Ryanodine receptor 3 15 31390469 5 5 3 SGCZ Sarcoglycan zeta 8 13991744 6 7 THSD4 Thrombospondin type I dom 4 15 69220842 4 _ 3 aGenes that contain overlapping clusters of nominally positive SNPs in comparisons of nicotine-dependent individuals versus individuals who report use of <100 cigarettes in their lives (Hamer samples) as well as Liu/Drgon et al. and/or Beirut et al. comparisons. Columns list gene description, chromosome, base pairs for gene’s start, number of nominally positive SNPs that lay in clusters within the gene’s exons and in 10-kb genomic flanking regions, and the numbers of clustered, nominally positive SNPs that fall within overlapping clusters within the same gene from Liu/Drgon et al. The 600k (MNB I) and 1M (MNB II) data from the National Institute on Drug Abuse research volunteers dependent on at least one illegal substance and from the Bierut et al. 38k dataset (15) are listed in the last three columns. All of the genes that are identified by clustered nominally positive data from the current dataset are presented in Table S1. tified 67 genes (Table S2). Monte Carlo tween dependent and nondependent cal significance (P = 0.047). These over- P values for this degree of clustering subjects provided highly significant laps identified 30 genes (Table 1). were P < 0.0001. overlap with the subset of 38,000 SNPs that were identified as nominally signif- Overlap with the Uhl et al. Dataset Overlap with Beirut et al. and icant by Beirut et al. in comparisons of The clusters of three nominally posi- Liu/Drgon et al. Datasets dependent smokers to nondependent tive SNPs from the data that compare These data for clustered, nominally smokers (Monte Carlo P < 0.0001). current to former smokers also overlap positive SNPs from the current dataset Overlaps with genes identified in significantly with the clustered, nomi- revealed significant overlap with genes 600,000–1 million GWA for dependence nally positive results from 2 of the 3 re- identified by other relevant datasets. on at least one illegal substance versus cently reported samples of individuals The clusters of 4 nominally positive ethnically matched control individuals who were successful versus those who SNPs identified by comparisons be- provided P values that reached statisti- were unsuccessful in quitting smoking in

24 | DRGON ET AL. | MOL MED 15(1-2)21-27, JANUARY-FEBRUARY 2009 RESEARCH ARTICLE

Table 2. Quit success genes.a clinical trials. Comparisons between these data and results reported for sam- Clustered ples from Rose and colleagues and from Gene Gene description Chromosome Base pair SNPs Niarua, David, and colleagues [samples A2BP1 Ataxin 2-binding protein 1 16 6009133 3 II and III in (14)] each displayed signifi- CSMD1 CUB and Sushi multiple 8 2782789 10 cant Monte Carlo P values ≤0.0001. Com- domains 1 parison with similarly analyzed samples DSCAM Down syndrome cell 21 40306213 3 from Lerman and coworkers (14), how- adhesion molecule ever, provided only a trend toward sig- PCDH15 Protocadherin 15 10 55250866 3 nificance (P = 0.10). Genes identified by RARB Retinoic acid receptor β 3 25444758 3 clusters of nominally positive SNPs in current and prior studies of ability to aGenes that contain overlapping clusters of nominally positive SNPs in comparisons of successfully quit smoking include ataxin nicotine-dependent individuals who reported continuing smoking when interviewed 2-binding protein 1 (3 clustered nomi- versus those who reported prolonged abstinence and clusters of nominally positive SNPs nally positive SNPs in the current from studies of at least two samples of successful versus unsuccessful quitters in clinical dataset); CUB and Sushi multiple do- trials recently reported by Uhl et al. (14). Columns list gene description, chromosome, base mains (10 SNPs), Down syndrome cell pairs for gene’s start, and number of nominally positive SNPs that lay in clusters within the adhesion molecule (3 SNPs), protocad- gene’s exons and in 10-kb genomic flanking regions in the current dataset. All genes identified by clustered, nominally positive data from the current dataset are presented in herin 15 (3 SNPs), and the retinoic acid Table S2. receptor β (3 SNPs) (Table 2).

Possible Alternative Explanations for tween African-American versus Euro- search volunteers converge with findings Observed Results pean-American control individuals in from two larger research-volunteer sam- We would anticipate the observed, Baltimore. We found 57 and 90, respec- ples that were ascertained in different highly significant clustering of SNPs tively, versus 89 and 89 expected by fashions are thus reassuring. Our find- that display nominally positive results if chance for dependence analyses. We ings also support the idea that many of many of these reproducibly positive found 23 and 15, respectively, versus 29 the molecular genetic findings that have SNPs lay near and were in linkage dise- and 29 expected by chance for the quit- been previously reported are not simply quilibrium with functional allelic vari- success analyses. attributable to the methods used for as- ants that distinguished substance- certainment and definition of “cases” dependent subjects from control DISCUSSION and “controls.” It is important to note, subjects. We would not anticipate this The current results provide (a) inde- however, that this overall conclusion degree of clustering if the results were pendent support for GWA results from does not exclude sampling-related con- due to chance. The Monte Carlo P val- larger samples of individuals studied for tributions related to the observations for ues noted here are thus likely to receive other substance-dependence phenotypes, some particular genes. contributions from both the extent of (b) independent support for GWA results The observations comparing individu- linkage disequilibrium among the clus- for studies of smoking cessation, and (c) als who have smoked less than 101 ciga- tered, nominally positive SNPs and the a control for one of the potential con- rettes with a group of individuals who extent of linkage disequilibrium be- founding features of previously reported are currently nicotine dependent or were tween these SNPs and the functional smoking cessation GWA samples. The nicotine dependent in the past fit with haplotype(s) that lead to the association possibility that genetic results from many, but not all, aspects of the GWA with substance dependence. members of any sample of research vol- Beirut et al. dataset (15). The overall fit Neither control for occult stratification unteers might not represent the genetics between these datasets provides mutual nor assay variability provide convincing of members of the general population is support for a number of the genes iden- alternative explanations for most of the ever present. In studies of molecular ge- tified herein. However, we identified no data obtained here. We examined the netics in addiction, however, features positive associations for SNPs in the overlap between the 3563 and 914 clus- that are both heritable and differentially chromosome 15 nicotinic acetylcholine tered positive SNPs from nicotine- present in substance-dependent individ- receptor gene cluster. This gene cluster dependence and quit-success analyses, uals might, a priori, be considered to be was previously identified by compar- respectively, and the 2.5% of the SNPs especially likely to provide confounding isons of heavy smokers with FTND- for which the noise in validating studies influences (26). defined dependence versus smokers was highest as well as the 2.5% of SNPs Our present observations that findings who were not nicotine dependent by that displayed the largest differences be- from a community-based sample of re- FTND criteria but who were required to

MOL MED 15(1-2)21-27, JANUARY-FEBRUARY 2009 | DRGON ET AL. | 25 GENES FOR SMOKING AND SMOKING CESSATION

have smoked more than 100 cigarettes in these samples. Although there is signifi- present work provide a number of inter- their lives. Almost 25% of the members cant confidence in the overall set of con- esting views of addiction. These data re- of this “control” comparison group thus vergent positive results reported here, inforce our observations that many of displayed nicotine dependence based on tests for the significance of individual these genes are likely to contribute to DSM criteria. The current results are genes provide much more modest levels brain differences that are reflected in the consistent with those predicted from re- of statistical assurance. mnemonic aspects of addiction, and that cent reports from Thorgerison and col- Our focus on data from autosomal some of them also provide tempting tar- leagues, in which SNP allele frequencies chromosomes allowed us to combine gets for antiaddiction therapeutics. We in the nicotine-receptor chromosome 15 data from male and female subjects but discuss these ideas in more detail else- acetylcholine-receptor gene cluster were at the cost of missing potentially impor- where (26). similar in nonsmokers and heavy smok- tant contributions from sex chromo- The functional classes into which the ers (16). The significant differences that somes. Our pooling approach provided genes identified here are interesting, and emerged in their data when both groups excellent correlations between individu- too complex to be described in detail here. were compared with individuals who ally genotyped and pooled allele fre- We remain impressed by the identification smoked more than 100 cigarettes in their quency assessments in validation experi- of a number of genes whose products are lives but did not progress to heavy ments and allowed us to use these involved in cell adhesion, cell-to-cell com- smoking and FTND nicotine dependence samples without adding additional confi- munication processes that are likely to be fit with the present data. Nondependent dentiality burdens. Nevertheless, esti- key to proper establishment and plasticity individuals who smoked at low and mates of allele frequencies based on of neuronal connections. CUB and Sushi steady levels would fit into neither the pooled data represent approximations of multiple domains 1 (CSMD1), Down syn- nicotine-dependent nor the control “true” allele frequency differences that drome cell adhesion molecule (DSCAM), and group in our current study. might be determined by error-free indi- protocadherin 15 (PCDH15) are 3 of the 5 A number of limitations must be kept vidual genotyping of each participant. “quit success” genes identified in Table 2. in mind in considering the present re- There is no indication that the overall Each of these 14 of the 30 genes identified sults. The preplanned approach used positive results reported here are based in Table 1 encodes a protein implicated in here demands that multiple nominally on the SNPs whose assays provide more cell adhesion and/or extracellular matrix positive SNPs from each sample tag the noise, and no indication that occult strat- activities that are crucial to synapse forma- same genomic region within a gene. Re- ification on racial/ethnic lines con- tion and maintenance: cadherin 13 (CDH13), quirements that nominally positive tributed overall to the results that we ob- close homolog of L1 (CHL1), contactin 6 SNPs from the current dataset come tain here. However, we cannot totally (CNTN6), CUB and Sushi multiple domains 1 from each of the two 500k array types exclude contributions of occult stratifica- and 2 (CSMD1, CSMD2), catenin α 3 add a technical control. Monte Carlo ap- tion that cannot be detected by these (CTNNA3), catenin delta 2 (CTNND2), frag- proaches that do not require specifica- overall screens to findings in specific ile histidine triad gene (FHIT), heparanase 2 tion of underlying distributions can genes. The convergence of data from (HPSE2), low density lipoprotein-related readily judge the degree to which all of smoking GWA with data derived from protein 1B (LRP1B), leucine rich repeat neu- the observations made here could be at- studies of individuals with addictions to ronal 6C (LRRN6C), protein tyrosine phos- tributable to chance. Nevertheless, no substances in several different pharma- phatase, receptor type, M (PTPRM), sarcogly- unanimous criteria exist for declaring cological classes supports the idea that can zeta (SGCZ), and thrombospondin type I “replication” or “convergence” for GWA many allelic variants enhance vulnera- domain 4 (THSD4). studies, a consideration worth consider- bility to many addictions; the more The significant convergence between ing in evaluating the current results. modest statistical significance of this the “quit success” molecular genetics Another limitation is that these NIH re- overlap fits with the idea that there is identified here and data from 2 of the 3 search volunteer samples are of modest also likely to be some nicotine-selective currently reported samples supports the size. Power calculations that document genetics as well. We have no convincing idea that genetic contributions to the the modest power in European- explanation for the strong fit of our cur- ability of individuals to stop smoking in American samples reveal even more rent quit-success molecular genetic re- a community-based, often with modest modest power for the smaller numbers sults with data from 2 of the 3 previ- pharmacological and behavioral support, of individuals of other racial/ethnic an- ously reported samples but not with the are likely to overlap significantly with cestries. We have thus not analyzed third sample. genetic contributions to the ability to quit samples with non-European ancestries. Our focus is on identification of genes. smoking in research settings that fre- This modest power limits interpretation Although associations away from anno- quently manifest substantially greater of negative data and restricts inferences tated genes can also provide interesting levels of pharmacological and behavioral about genes that are not identified in results, the genes that we identify in the support. These findings are thus reassur-

26 | DRGON ET AL. | MOL MED 15(1-2)21-27, JANUARY-FEBRUARY 2009 RESEARCH ARTICLE

ing for future attempts to use genotype- ular Medicine, or other interests that 15. Bierut LJ, et al. (2007) Novel genes identified in a based data to match smokers with the might be perceived to influence the re- high-density genome wide association study for types and/or intensities of treatments sults and discussion reported in this nicotine dependence. Hum. Mol. Genet. 16:24–35. 16. Thorgeirsson TE, et al. (2008) A variant associated that provide the best and most cost- paper. with nicotine dependence, lung cancer and pe- effective smoking cessation opportunities ripheral arterial disease. Nature 452:638–42. in community settings. Our findings also REFERENCES 17. Amos CI, et al. (2008) Genome-wide association fit with the data from twin studies, be- 1. Uhl GR, Elmer GI, Labuda MC, Pickens RW. scan of tag SNPs identifies a susceptibility locus cause none of the twin studies of smok- (1995) Genetic influences in drug abuse. In: for lung cancer at 15q25.1. Nat. Genet. 40:616–22. 18. Saccone SF, et al. (2007) Cholinergic nicotinic re- ing cessation success was carried out in a Gloom FE, Kupfer DJ (eds.) Psychopharmacology: The Fourth Generation of Progress. New York: ceptor genes implicated in a nicotine dependence clinical trial setting (10,11). Raven Press. pp. 1793–2783. association study targeting 348 candidate genes It appears likely that investigations of 2. Tsuang MT, et al. (1998) Co-occurrence of abuse of with 3713 SNPs. Hum. Mol. Genet. 16:36–49. initial molecular genetic influences on different drugs in men: the role of drug-specific 19. Lueders KK, et al. (2002) Genetic and functional therapeutic responses in other diseases and shared vulnerabilities. Arch. Gen. Psychiatry analysis of single nucleotide polymorphisms in the beta2-neuronal nicotinic acetylcholine recep- may well be carried out by use of DNA 55:967–72. 3. Karkowski LM, Prescott CA, Kendler KS. (2000) tor gene (CHRNB2). Nicotine Tob. Res. 4:115–25. from participants in clinical trials, be- Multivariate assessment of factors influencing il- 20. Liu QR, et al. (2006) Addiction molecular genet- cause the intensive patient/subject moni- licit substance use in twins from female-female ics: 639,401 SNP whole genome association iden- toring that is often built into such trials pairs. Am. J. Med. Genet. 96:665–70. tifies many “cell adhesion” genes. Am. J. Med. provides reassurance that clinical out- 4. True WR, et al. (1999) Interrelationship of genetic Genet. B. Neuropsychiatr. Genet. 141B:918–25. comes are well documented in these and environmental influences on conduct disor- 21. Uhl GR, Liu QR, Walther D, Hess J, Naiman D. der and alcohol and marijuana dependence (2001) Polysubstance abuse-vulnerability genes: datasets. However, the greatest utility of symptoms. Am. J. Med. Genet. 88:391–7. genome scans for association, using 1,004 sub- the information from genetic studies of 5. Kendler KS, Karkowski LM, Neale MC, Prescott jects and 1,494 single-nucleotide polymorphisms. clinical trial participants will come when CA. (2000) Illicit psychoactive substance use, Am. J. Hum. Genet. 69:1290–300. there is additional data that can assure heavy use, abuse, and dependence in a US popu- 22. Smith SS, et al. (1992) Genetic vulnerability to us that the results can be generalized to lation-based sample of male twins. Arch. Gen. drug abuse. The D2 dopamine receptor Taq I B1 Psychiatry 57:261–9. restriction fragment length polymorphism ap- more real-world settings and to individ- 6. Fagerstrom KO. (1978) Measuring degree of pears more frequently in polysubstance abusers. uals who would be less likely to be en- physical dependence to tobacco smoking with Arch. Gen. Psychiatry 49:723–7. rolled in a clinical trial. The findings pre- reference to individualization of treatment. Ad- 23. Persico AM, Bird G, Gabbay FH, Uhl GR. (1996) sented here thus support ongoing dict. Behav. 3:235–41. D2 dopamine receptor gene TaqI A1 and B1 re- processes for comparing GWA datasets 7. Fagerstrom KO, Schneider NG. (1989) Measuring striction fragment length polymorphisms: en- hanced frequencies in psychostimulant-preferring from more intensively studied and nicotine dependence: a review of the Fagerstrom Tolerance Questionnaire. J. Behav. Med. 12:159–82. polysubstance abusers. Biol. Psychiatry 40:776–84. meticulously selected individuals with 8. Pomerleau CS, Carton SM, Lutzke ML, Flessland 24. Dupont WD, Plummer WD Jr. (1990) Power and those from community-based samples. KA, Pomerleau OF. (1994) Reliability of the sample size calculations. A review and computer For addictions, as for many complex dis- Fagerstrom Tolerance Questionnaire and the program. Control Clin. Trials 11:116–28. orders, such data provide an increasingly Fagerstrom Test for Nicotine Dependence. Addict. 25. Dupont WD, Plummer WD Jr. (1998) Power and sample size calculations for studies involving lin- rich basis for improved understanding Behav. 19:33–9. 9. Pomerleau CS, Majchrzak MJ, Pomerleau OF. ear regression. Control Clin. Trials 19:589–601. and for development of personalized (1989) Nicotine dependence and the Fagerstrom 26. Uhl GR, et al. (2008) Molecular genetics of addic- prevention and treatment strategies. Tolerance Questionnaire: a brief review. J. Subst. tion and related heritable phenotypes: genome- Abuse 1:471–7. wide association approaches identify “connectiv- ACKNOWLEDGMENTS 10. Broms U, Silventoinen K, Madden PA, Heath AC, ity constellation” and drug target genes with pleiotropic effects. Ann. N. Y. Acad. Sci. We are grateful for thoughtful advice Kaprio J. (2006) Genetic architecture of smoking behavior: a study of Finnish adult twins. Twin 1141:318–81. and discussion from Dr. J Rose and for Res. Hum. Genet. 9:64–72. fruitful collaborations with Drs. Rose, 11. Lessov CN, et al. (2004) Defining nicotine depen- Lerman, Niaura, and David. This re- dence for genetic research: evidence from Aus- search was supported financially by the tralian twins. Psychol. Med. 34:865–79. NIH Intramural Research Programs, Na- 12. Uhl GR, et al. (2008) Genome-wide association for methamphetamine dependence: convergent tional Institute on Drug Abuse, National results from 2 samples. Arch. Gen. Psychiatry Institute of Mental Health, and National 65:345–55. Cancer Institute, Department of Health 13. Uhl GR, et al. (2007) Molecular genetics of nico- and Social Services. tine dependence and abstinence: whole genome association using 520,000 SNPs. BMC Genet. 8:10. 14. Uhl GR, et al. (2008) Molecular genetics of suc- DISCLOSURE cessful smoking cessation: convergent genome- We declare that the authors have no wide association study results. Arch. Gen. Psychi- competing interests as defined by Molec- atry 65:683–93.

MOL MED 15(1-2)21-27, JANUARY-FEBRUARY 2009 | DRGON ET AL. | 27