|Home | About | Journals | Submit | Contact Us | Français|
Thorgeir E. Thorgeirsson firstname.lastname@example.org
Daniel F. Gudbjartsson si.edoced@gfd
Ida Surakka email@example.com
Jacqueline M. Vink firstname.lastname@example.org
Najaf Amin email@example.com
Frank Geller firstname.lastname@example.org
Patrick Sulem email@example.com
Thorunn Rafnar firstname.lastname@example.org
Tõnu Esko email@example.com
Stefan Walter firstname.lastname@example.org
Christian Gieger email@example.com
Rajesh Rawal firstname.lastname@example.org
Massimo Mangino email@example.com
Inga Prokopenko ku.ca.xo.llew@agni
Reedik Mägi ku.ca.xo.llew@kideer
Kaisu Keskitalo firstname.lastname@example.org
Iris H. Gudjonsdottir si.edoced@hsiri
Solveig Gretarsdottir si.edoced@gievlos
Hreinn Stefansson si.edoced@nnierh
John R. Thompson ku.ca.el@jrt
Yurii S. Aulchenko email@example.com
Mari Nelis ee.tu@leniram
Katja K. Aben firstname.lastname@example.org
Martin den Heijer email@example.com
Asger Dirksen kd.tenldad@ida
Haseem Ashraf kd.tenldad@meesah
Nicole Soranzo ku.ca.regnas@6sn
Ana M Valdes firstname.lastname@example.org
Claire Steves email@example.com
Andre G Uitterlinden firstname.lastname@example.org
Albert Hofman email@example.com
Anke Tönjes firstname.lastname@example.org
Peter Kovacs email@example.com
Jouke Jan Hottenga firstname.lastname@example.org
Gonneke Willemsen email@example.com
Nicole Vogelzangs firstname.lastname@example.org
Angela Döring ed.nehcneum-ztlohmleh@gnireod
Norbert Dahmen ed.zniam-inu@nemhadn
Barbara Nitz email@example.com
Michele L. Pergadia ude.ltsuw.yrtaihcysp@pelehcim
Berta Saez rf.oohay@1002zeas
Veronica De Diego moc.liamtoh@ogeidedorev
Victoria Lezcano moc.liamtoh@acsoibonaczelv
Maria de Garcia-Prats moc.liamtoh@starpaicragalol
Samuli Ripatti firstname.lastname@example.org
Markus Perola email@example.com
Johannes Kettunen ku.ca.regnas@5kj
Anna-Liisa Hartikainen firstname.lastname@example.org
Anneli Pouta email@example.com
Jaana Laitinen firstname.lastname@example.org
Matti Isohanni email@example.com
Shen Huey-Yi firstname.lastname@example.org
Maxine Allen email@example.com
Maria Krestyaninova ku.ca.ibe@kairam
Alistair S Hall ku.ca.sdeel@hsasvc
Gregory T. Jones firstname.lastname@example.org
Andre M. van Rij email@example.com
Thomas Mueller ta.bal-sb@relleuM.samohT
Benjamin Dieplinger ta.bal-sb@regnilpeiD.nimajneB
Meinhard Haltmayer ta.bal-sb@reyamtlaH.drahnieM
Steinn Jonsson si.ilatipsdnal@jnniets
Stefan E. Matthiasson si.gninkeal@mes
Hogni Oskarson si.sumuh@ingoh
Thorarinn Tyrfingsson si.aas@nniraroht
Lambertus A. Kiemeney ln.ncmu.hbe@yenemeiK.B
Jose I. Mayordomo moc.liamtoh@omodroyamesoj
Jes S Lindholt kd.MR.grobiV@tlohdniL.S.seJ
Jesper Holst Pedersen firstname.lastname@example.org
Wilbur A. Franklin email@example.com
Holly Wolf ude.revnedcu@floW.ylloH
Grant W. Montgomery ua.ude.rmiq@Mtnarg
Andrew C. Heath ude.ltsuw@htaeha
Nicholas G. Martin ua.ude.rmiq@Mkcin
Pamela A.F. Madden ude.ltsuw.yrtaihcysp@pneddam
Ina Giegling firstname.lastname@example.org
Dan Rujescu email@example.com
Marjo-Riitta Järvelin firstname.lastname@example.org
Veikko Salomaa email@example.com
Michael Stumvoll firstname.lastname@example.org
Tim D Spector email@example.com
H-Erich Wichmann ed.nehcneum-ztlohmleh@nnamhciw
Andres Metspalu ee.cbe@serdna
Nilesh J Samani ku.ca.el@sjn
Brenda W. Penninx firstname.lastname@example.org
Ben A. Oostra email@example.com
Dorret I. Boomsma firstname.lastname@example.org
Henning Tiemeier email@example.com
Cornelia M. van Duijn firstname.lastname@example.org
Jaakko Kaprio email@example.com
Jeffrey R. Gulcher firstname.lastname@example.org
Mark I McCarthy email@example.com
Leena Peltonen ku.ca.regnas@aneel
Unnur Thorsteinsdottir firstname.lastname@example.org
Kari Stefansson si.edoced@snafetsk
Smoking is a risk factor for most of the diseases leading in mortality1. We conducted genome-wide association (GWA) meta-analyses of smoking data within the ENGAGE consortium to search for common alleles associating with the number of cigarettes smoked per day (CPD) in smokers (N=31,266) and smoking initiation (N=46,481). We tested selected SNPs in a second stage (N=45,691 smokers), and assessed some in a third sample (N=9,040). Variants in three genomic regions associated with CPD (P< 5·10−8), including previously identified SNPs at 15q25 represented by rs1051730-A (0.80 CPD,P=2.4·10−69), and SNPs at 19q13 and 8p11, represented by rs4105144-C (0.39 CPD, P=2.2·10−12) and rs6474412-T (0.29 CPD,P= 1.4·10−8), respectively. Among the genes at the two novel loci, are genes encoding nicotine-metabolizing enzymes (CYP2A6 and CYP2B6), and nicotinic acetylcholine receptor subunits (CHRNB3 and CHRNA6) highlighted in previous studies of nicotine dependence2-3. Nominal associations with lung cancer were observed at both 8p11 (rs6474412-T,OR=1.09,P=0.04) and 19q13 (rs4105144-C,OR=1.12,P=0.0006).
Smoking behavior and nicotine dependence (ND) are under genetic influence4. While environmental influences play a strong role in the initiation of smoking5, the heritability of smoking persistence, smoking quantity and ND is high in most twin studies5,6. Sequence variants within a cluster of nicotinic acetylcholine receptor (nAChR) genes on chromosome 15q25 have recently been shown to associate with the number of cigarettes smoked per day (CPD)7-8, ND3,7, and smoking-related diseases such as lung cancer (LC)7,9-10, peripheral arterial disease (PAD)7, and chronic obstructive pulmonary disease (COPD)11
To search for additional common variants associating with smoking behavior we performed meta-analyses of GWA studies, mainly using samples of European ancestry from the ENGAGE consortium (See URL section), focusing on two smoking phenotypes: CPD and smoking initiation. The smoking initiation analysis was performed with a total of 30,431 ever-smokers and 16,050 never-smokers, using data from 12 GWA studies, Corogene, deCODE, EGPUT, ERF, NFBC, KORA, NTR-NESDA, Rotterdam, SUSOD, TwinUK and WTCCC-CAD. For CPD we combined data from the same studies and the NL-BLC with a total of 31,266 subjects. Information on the meta-analysis studies for CPD and smoking initiation is provided in Supplementary Table 1, the Supplementary Note, and in the Methods online. After genomic control correction of each component study, we combined association data for ~2,500,000 imputed and genotyped autosomal SNPs with a fixed-effects additive meta-analysis using the inverse-variance method for CPD and smoking initiation. QQ-plots for CPD, excluding markers in the 15q25 region, displayed only modest inflation of the χ2-test statistic (λGC=1.02) (Supplementary Figure 1A). In addition to the 15q25 locus, SNPs at two loci, 19q13 and 7p14, were genome-wide significanct (GWS) for CPD (P<5·10−8) in the meta-analysis data. The QQ-plot for smoking initiation displayed weak inflation of the χ2-test statistic (λGC=1.03) and no GWS associations (Supplementary Figure 1B).
We selected 15 regions for smoking initiation totaling 277 SNPs, and 14 regions totaling 443 SNPs for CPD, for in silico replication in samples from the Tobacco and Genetics (TAG) and the Oxford-Glaxo Smith Kline (Ox-GSK) consortia (see accompanying papers12-13) (Supplementary Table 2). For CPD we included a region on chromosome 8p11 based on large number of SNPs exhibiting suggestive associations with CPD, strong candidacy of region genes (encoding nAChR subunits α6 and β3 (CHRNA6 and CHRNB3)), and prior suggestive evidence for association between SNPs within this region and ND3-4.
In addition to the 15q25 locus, three novel loci, 7p14, 8p11 and 19q13, were GWS for CPD after combining the results from the ENGAGE meta-analysis set with those of TAG and OX-GSK (Table 1, Fig 1, and Supplementary Table 2). No GWS associations for the selected smoking initiation regions were observed in the combined analysis of the meta-analysis and the in silico data (Supplementary Table 2).
For further confirmation of the CPD association signals at the 7p14, 8p11 and 19q13 loci, selected markers from these regions were genotyped in additional samples (n=9,040) from Iceland, Australia, Denmark, Germany and Spain (Table 1). Markers at 8p11 and the 19q13 loci had effects in the same directions, but not the marker on 7p14 (Table 1). After combining these data with ENGAGE results and the in silico replication the 8p11 and the 19q13 loci remained GWS but not the 7p14 locus (Table 1). The CPD association results for the SNPs in Table 1 for each study are presented in Supplementary Table 3 and Supplementary Figure 2.
Nominally significant heterogeneity in the strength of association with CPD was observed for rs1051730 at 15q25 (P=0.035, I2=32%) and rs215614 at 7p14 (P=0.018, I2=34%). There is little doubt that the 15q25 association with CPD is real. Therefore its heterogeneity must be due to differences between the study population such as different CPD information ascertainment, different types of cigarettes being used, different phenotypic and demographic ascertainment strategies or different genetic structures. The heterogeneity observed at 7p14 could be caused by a false positive finding or, if it is indeed a true positive, some combination of the “winner’s curse” and the same sort of differences driving the heterogeneity at 15q25.
Not surprisingly, the strongest associations observed with CPD in the combined analysis were with SNPs within the previously identified region on chromosome 15q25 (rs1051730-A, P= 2.4 10−69, effect size = 0.80 CPD) (Table 1, Figure 1). We searched for additional association signals in the 15q25 region, not accounted for by rs1051730, by performing linear regression using the rs1051730 allele count as a covariate in a subset of the ENGAGE samples (N=23,089). The residual signals were mostly tagged by two SNPs in relatively low LD, rs2869046-T (P=4.8·10−5, Effect=0.5 CPD) and rs2036534-T (P=9.1·10−5, Effect=0.5 CPD) (r2=0.080 and D’=0.65 in the HapMap CEU samples) (Supplementary Figure 3). These two SNPs are also in fairly weak LD with rs1051730 (r2=0.12, D’=0.49 and r2=0.18, D’=1.05 in the HapMap CEU samples for rs2869046 and rs2036534, respectively). These data suggest that, either, the three variants, rs2869046-T, rs2036534-T and rs1051730-A, represent independent association signals, or that a variant(s) captured by a combination of these variants remains to be identified. A SNP, rs578776, in LD with rs2036534 (r2=0.74, D’=0.95 in the HapMap CEU samples) was previously reported to associate with ND to an extent that could not be accounted for by rs105173014. As with rs2036534, rs578776 is in weak LD with rs1051730 (r2=0.21, D’=1.0 in the HapMap CEU samples). However, rs578776 is not correlated with rs2869046 (r2=0.0.038, D’=0.41 in the HapMap CEU samples) and thus does not explain the signal described here for rs2869046 at the 15q25 locus (Supplementary Figure 3).
The SNPs on chromosome 8p11 that reached genome–wide significance, rs6474412 (effect=0.29 CPD, P=1.4 10−8) and rs13280604 (effect=0.31 CPD, P=1.310−8), are in perfect LD in the HapMap CEU samples with a variant (rs13277254-T) that was previously highlighted as suggestively associated in a GWA study of ND using controls who had been exposed to smoking but had not developed ND (rs13277254-T, OR=1.19, P=6·10−5)2-3. The other SNPs in the region showing association with CPD (Figure 1) are all in strong LD with these SNPs. Rs6474412 is located about 2.1 kb from the 5′ end of the β3 nAChR subunit gene (CHRNB3), and belongs to a group of highly correlated SNPs, that includes two SNPs in exons of CHRNB3: a synonymous SNP (rs4593), and a non-synonymous SNP in CHRNB3 (rs4952)3. Although the CHRNB3 gene is implicated by the location of the associating SNPs, these markers could be tagging variation elsewhere within the LD block that also contains the α6 nAChR subunit gene (CHRNA6) (Figure 1).
Nine different nicotinic cholinergic receptor subunits (α2-α7, β2-β4) are expressed in the human brain, and they combine with each other in diverse patterns to form various types of functional pentameric receptors distinguished by subunit composition and sensitivity to nicotine15. Rodent studies have implicated CHRNA6 and CHRNB3 receptor subunits in nicotine-induced dopamine release16. Neither CHRNA6 nor CHRNB3 are expressed in lung tissue17.
The CPD associated markers on chromosome 19q13 are located in a region harboring CYP2A6, which encodes CYP2A6, an enzyme that plays a major role in the oxidation of nicotine in human liver microsomes, as well as several other genes and pseudogenes belonging to the CYP gene family (Figure 1). A number of sequence variants in or near CYP2A6 that reduce CYP2A6’s enzymatic activity have been identified18. For some of these variants, effects on smoking behavior have been suggested18. In the present study, the most significant association in the region was observed with rs4105144. This SNP is in LD with CYP2A6*2 (rs1801272) (r2=0.13 and D’=1.0 in the HapMap CEU samples) and the CYP2A6*2 reduced function allele is only found on the background of rs4105144-C which associates with reduced smoking quantity. Although the effect of rs4105144 (0.39±0.06) is smaller than that of rs1801272 (0.68±0.18)(Table 1), its association is more significant (lower P value) because of higher minor allele frequency. This suggests that rs4105144-C may be tagging many reduced function variants. The second most significant association in the region was with rs7937, in the untranslated 3′ end of the RAB4B gene, which is in LD with rs4105144 (r2=0.32, D’=0.82 in the HapMap CEU samples). The third most significant association in the region was with rs7260329 that is almost independent of rs4105144 (r2=0.0064, D’=0.091 in the HapMap CEU samples). Rs7260329 is an intronic SNP in CYP2B6, but its product converts nicotine to cotinine with about 10% of the catalytic activity of the CYP2A6 enzyme, and also metabolizes several drugs of abuse, and buproprion, an atypical antidepressant also used as a smoking cessation aid18. The CYP2B6 levels in the human brain are higher than those of CYP2A6 and are altered in smokers and alcoholics18-19.
One of the ENGAGE studies (GenMetS) had information on immune-reactive serum cotinine levels for a set of samples (N=485)20, that reflects the catalytic activity of the CYP2A and CYP2B gene products18. One of the SNPs associating with CPD at 19q13 (rs7937-G) showed nominally significant association with cotinine levels (effect=1.16,P=0.0031), while the markers at 8p11 and 7p14 did not associate with cotinine levels in this sample (P>0.26). However, two SNPs at 19q13 showed a stronger association with cotinine levels than rs7937, rs2233152-A and rs2287692-A (effect=1.92, P=0.00021), but did not associate strongly with CPD in the ENGAGE samples (P=0.013) and most of the markers showing the strongest association with CPD did not associate with cotinine levels.
We next assessed the SNPs from the novel regions associating with CPD for association with ND, defined as a score of four or higher on the Fagerstrom Test for Nicotine Dependence (FTND), or endorsement of at least three of the seven Diagnostic and Statistical Manual of Mental Disorders 4th edition (DSM-IV) criteria. Allele frequencies for 1,979 Icelandic (deCODE) and 835 Dutch (NTR-NESDA) ND cases were compared to 36,202 Icelandic and 611 Dutch population controls. SNPs on chromosome 8p11, and chromosome 7p14 associated nominally with ND, but none of the SNPs on chromosome 19q13 (Supplementary Table 4).
As we had previously found association of the 15q25 region with LC and PAD7, we directly genotyped selected markers from the 7p14, 8p11 and 19q13 regions for association with LC (2,019 cases and 40,509 controls) and PAD (2,855 cases, 40,424 controls) in samples of European ancestry. The LC data were also combined with summary-level data from the publicly available GWA dataset on LC (2,518 case and 1,921 controls) from the International Agency for Research on Cancer (IARC)10 (Table 2). Nominally significant associations with LC were observed for rs6474412-T on 8p11 (OR =1.12, 95%CI: 1.05-1.20, P=0.00060), rs215614-G on 7p14 (OR =1.07, 95%CI: 1.02-1.13, P=0.011), and rs7260329-G and rs4105144-C on 19q13 (OR =1.06, 95%CI: 1.00-1.12, P=0.041 and (OR =1.09, 95%CI: 1.00-1.18, P=0.040) (Table 2). As for the effect on CPD (Table 1) the effects of these variants on LC is substantially weaker than that of the 15q25 variants (OR =1.31, P=1.5 10−8)7,9-10(Table 2), warranting further analysis in additional sample sets. No significant associations with PAD were observed for the markers tested (Supplementary Table 5). The potential effect of rs7260329-G and rs4105144-C on LC, is interesting in light of the fact that CYP2A6 gene product activates procarcinogenic nitrosamines18.
The 13 regions that were selected from the ENGAGE meta-analysis of CPD but were not GWS, nominate a number of interesting functional candidate genes, including genes encoding for γ-aminobutiric acid receptor subunits (GABRA1 and GABRG2), and PDE1C, CDH13, and A2BP1, that were all highlighted in a GWA study of ND and smoking cessation21-22, and some of these genes may play a role in smoking behavior.
In conclusion, we have discovered sequence variants associating with smoking behavior within regions harboring nAChR genes (CHRNB3-CHRNA6, 8p11) and nicotine-metabolizing enzymes (CYP2A6 -CYP2B6, 19q13). The 8p11 association is reminiscent of that with chromosome 15q25; both regions contain nAChR genes, and the key variants associate with ND and LC, bringing up the question of whether the risk for LC is through the effect on smoking behavior, or involves increased vulnerability to the harmful effects of smoking as well7,23-27. However, the dissection of the causal pathway will be even more difficult in case of these new variants, as their effects on both CPD and LC are smaller, and further studies are warranted.
We thank all the participants in the genetic studies whose contributions made this work possible. This work was supported in part by NIH (R01-DA017932) and the European Commission’s Sixth Framework Program, Integrated Project GENADDICT (LSHM-CT-2004-005166). The ENGAGE smoking consortium was formed through a component of the Integrated Project ENGAGE, supported by the European Commission’s Seventh Framework Program, grant agreement HEALTH-F4-2007- 201413. ENGAGE projects have benefited from the SIMBioMS platform28 that has greatly facilitated data exchange and annotation. Additional acknowledgements are listed in the Supplementary Note online.
COMPETING INTERESTS STATEMENT Authors whose affiliations are listed as deCODE genetics are employees of deCODE genetics, a biotechnology company.
URLs. ENGAGE Consortium (www.euengage.org).