Multiple sclerosis (MS) is a neuroinflammatory autoimmune disease with a multifactorial etiology characterized by inflammatory and neurodegenerative pathological phases.
Strong evidence supports the contribution of both genetic and environmental factors in the pathoetiology of MS, as demonstrated by increased disease concordance among monozygotic twins (approximately 25%) compared with dizygotic twins (approximately 5%).1 MS patients are at least twice as likely to be female, with the primary age at onset between 20 and 40 years of age.2 Variation within major histocompatibility-complex genes on chromosome 6p21 confers the greatest risk of MS, primarily within the human leukocyte antigen (HLA)-DRB1 locus (specifically, the HLA-DRB1*15:01 allele).3 Through International Consortium efforts, genome-wide association and replication studies have identified >100 MS susceptibility loci with modest effects (most odds ratios [ORs] <1.3); however, a large portion of the heritable component of MS (approximately 25–75%) remains unknown.4,5 Consideration of specific environmental factors important to development of MS may be necessary to further identify genetic risk factors.6 Exposure to tobacco smoke, limited sun exposure, vitamin D intake, and Epstein-Barr virus infection have all been strongly implicated as MS risk factors.7,8
Tobacco smoke history has consistently demonstrated an increased risk of MS. A comprehensive 2007 meta-analysis reported an elevated OR of 1.51 (95% confidence interval [CI] =1.22–1.87) for ever- versus never-smokers in retrospective studies and derived risk-ratio estimates of 1.24 (1.04–1.48) and 1.46 (1.28–1.67) for prospective studies.9 In 2011, results from a meta-analysis of 14 studies (representing data from 3052 cases and 457,619 controls) concurred with the initial estimates, demonstrating increased MS risk among smokers compared with nonsmokers (OR = 1.52 [95% CI = 1.39–1.66]).10 Although many studies have repeatedly demonstrated evidence for smoking as a risk factor for MS, only one has considered genetic predisposition, specifically HLA class I and II genotypes11; results await confirmation. Non-HLA genes have not been investigated.
Cigarettes and other tobacco products produce smoke composed of a highly complex aerosol of noxious gases and condensed tar particles that results from multiple thermolytic processes that occur with combustion of tobacco within the cigarette rod. Approximately, 4800 chemical substances are distributed between gas and particulate phases.12 Tobacco smoke contains an estimated concentration of 1017 oxidants per puff and many potent carcinogens or mutagens, including trace heavy metals and organic chemicals.13,14 Observational evidence emphasizes the importance of specific components of tobacco smoke. For example, a lack of association between Swedish snuff use and MS risk suggests that nicotine alone is not a risk factor for MS.15 There is also evidence supporting an increased risk of MS with passive tobacco smoke exposure in childhood and adulthood.16,17
Most xenobiotic tobacco smoke components are metabolized through one or more pathways that constitute the phase-1 and phase-2 enzymatic systems, whereby hydrophobic chemicals are converted into derivatives that can be readily excreted. Phase-1 reactions generally convert compounds to more hydrophilic metabolites through oxygenation, which bioactivates specific chemicals to become highly toxic intermediates with capacity for cellular damage.18,19 Phase-2 reactions detoxify these highly reactive species by attaching water-soluble functional groups, which diminishes their damaging potential and allows for subsequent excretion.18,19 Therefore, the efficacy of phase-2 enzymes is critical for minimizing cellular damage during the detoxification of exogenous chemicals. A large number of enzymes are involved in phase-1 and phase-2 detoxification; however, genetic variation within specific phase-2 enzymes, primarily N-acetyltransferases and glutathione S-transfereases, has consistently been shown to modify disease risk conferred by tobacco smoke in cancers and other autoimmune diseases.20–24 Given that not all smokers develop MS and only some people with MS were ever-smokers, we hypothesized that host genetic variation could contribute to metabolism efficiency of tobacco smoke constituents and thus to the risk of MS in smokers. We investigated variation within five phase-2 loci that encode the N-acetyltransferases (NAT1 and NAT2) and major glutathione S-transferases (GSTM1, GSTP1, and GSTT1). NAT1 and NAT2 enzyme phenotypes resulting from functionally relevant haplotypes (ie, fast vs. slow acetylators) were also studied.25
We investigated gene (G) by environment (E) interaction, in which a genetic variant might exacerbate the effect of the environmental risk factor but have no direct effect in unexposed persons. Specifically, variation within five candidate genes was examined for effect modification of MS risk conferred by exposure to tobacco smoke. Investigating G × E relation has proven challenging, owing to variation across studies in the environmental data collected and differences in allele frequencies across populations. We conducted a three-stage analysis using two population-based case-control datasets that consisted of a discovery population, replication population, and pooled analysis.
Discovery Study Population
The discovery study population consisted of non-Hispanic whites with MS and matched controls identified among members of Kaiser Permanente Medical Care Plan, Northern California Region (Kaiser study) using electronic medical records. This Kaiser Plan is an integrated health services delivery system with a membership of 3.2 million that comprises about 25% to 30% of the population of a 22-county service area in northern California. Comparisons with the general population have shown that the membership is generally representative, except that persons in impoverished neighborhoods are underrepresented.26 Briefly, MS cases were defined as having one or more outpatient MS diagnoses by a neurologist International Classification of Diseases (ICD9 code 340.xx; 95% had at least two MS diagnoses by a neurologist), self-identified white (non-Hispanic) race/ethnicity, age 18 through 69 years, and Kaiser membership at initial contact. We used medical records to validate diagnoses based on published diagnostic criteria.27 Controls were randomly selected from current Kaiser members who did not have a MS diagnosis or related conditions (optic neuritis, transverse myelitis, or demyelination disease; ICD9 codes: 340, 341.0, 341.1, 341.2, 341.20, 341.21, 341.22, 341.8, 341.9, 377.3, 377.30, 377.39, and 328.82) and were individually matched to cases on sex, birth date (± 1 year), race/ethnicity, and ZIP code of the case residence. Participants were mailed a consent form to be returned by mail and a biospecimen collection kit to be completed at a Kaiser clinic laboratory or returned by mail. Recruitment is ongoing; at the time of our data freeze, the study participation rate was 73% (of 2987 MS patients) for cases and 56% (of 2280 healthy persons) for controls. At the time of this study, not all cases had been matched.
The genotyped Kaiser dataset consisted of 1660 persons (1069 cases and 591 controls) with 594,345 autosomal SNPs derived using the Illumina Infinium 660K BeadChip (San Diego, CA) as previously described,4 subjected to stringent quality-control measures, and whole-genome imputation using IMPUTE2 (http://mathgen.stats.ox.ac.uk/impute/impute_v2.html). Data for 1052 persons were available from 1000 Genomes (eAppendix; http://links.lww.com/EDE/A777).28 The presence of relatedness and population substructure was determined using multidimensional scaling as implemented in PLINK v1.07 (http://pngu.mgh.harvard.edu/~purcell/plink/) (eAppendix; http://links.lww.com/EDE/A777).29 Genetic outliers and related persons were removed, resulting in 1016 MS cases and 583 controls with dense genetic data, of whom 1012 MS cases and 576 controls had detailed exposure data (see below). Imputed candidates of interest were subsequently genotyped in a subset of 134 persons (80 cases, 54 controls), using a custom Fluidigm SNPtype Assay (192.24GT Dynamic Array Integrated Fluidic Circuit [South San Francisco, CA]).
All variants within 5 kb of the five candidate genes (GSTM1, GSTP1, GSTT1, NAT1, and NAT2) were extracted from the MS case-control dataset. Both genotyped and imputed variants were extracted and then coded additively for the minor allele (0, 1, or 2 copies). A genotype for an imputed SNP was determined only if the probability (threshold calling) of that genotype was at least 90% and also had 90% imputation certainty (eAppendix; http://links.lww.com/EDE/A777). A total of 277 variants (39 genotyped and 238 imputed SNPs) within only three genes met quality-control criteria: minor allele frequency >0.10 (based on power analyses; eAppendix; http://links.lww.com/EDE/A777), Hardy Weinberg Equilibrium P > 0.0001, and missing genotypes <0.1. The overall genotype rate was 98.3%. In addition, a total of 13 functional haplotypes that correspond to specific NAT1 and NAT2 phenotypes that reflect acetylation capacity were generated using PLINK, where the most probable haplotype was assigned (eTable 1A; http://links.lww.com/EDE/A777). The phenotypes were constructed based on accepted definitions (http://n-acetyltransferase nomenclature.louisville.edu/). Furthermore, global NAT1 slow (combining NAT1*14, NAT1*15, and NAT1*17) and fast (combining NAT1*11, NAT1*30) acetylation status was assigned, and similarly for NAT2 using four SNPs (rs1801280 [NAT2*5], rs1799930 [NAT2*6], rs1799931 [NAT2*7], and rs1801279 [NAT2*14]) (eTable 1B; http://links.lww.com/EDE/A777).30 Persons who were homozygous for the major allele for all four SNPs were classified as rapid NAT2 acetylator phenotype, those heterozygous for any one of the SNPs were classified as intermediate NAT2 acetylator phenotype, and those homozygous for one or more SNPs or heterozygous for two or more SNPs were classified as slow NAT2 acetylator phenotype.
Study participants completed a standardized computer-assisted telephone interview (including questions on sociodemographic variables and environmental exposures) conducted by trained staff personnel. Cases answered questions about their illness, including timing of onset, diagnosis, course, and progression of MS. Questions were asked about exposures occurring during the year before the first year of self-reported symptoms (reference year). Matched controls, when available, were asked about the same exposures occurring in the year preceding the onset-year in their matched case (reference year). Participants were considered ever-tobacco-smokers if they reported having ever smoked at least one cigarette per day for one month or more. Participants were asked whether they were current smokers and at what age they first began to smoke daily. Former smokers were asked the age at which they first began smoking daily and when they last smoked. Those who started smoking before the age of 20 and were not current smokers were asked to recall their smoking status at age 20. A previous study has suggested that increased risk associated with tobacco smoking may persist only 5 years after cessation15; therefore, all smokers were asked whether they smoked in the appropriate reference year for the matched case and control pair. Three exposure measures were generated: (1) ever/never-smoker, (2) smoker at age 20 years, and (3) smoker in the reference year for the matched dataset only. Year of birth, sex, and education level (bachelor’s degree) were also collected (eAppendix; http://links.lww.com/EDE/A777).
Replication Study Population
The replication study population consisted of non-Hispanic whites from the Epidemiological Investigation of Multiple Sclerosis (Swedish study). This was a population-based nationwide case-control study of incident MS cases aged 16 to 70 years, recruited via hospital-based and privately run neurology units in Sweden. All university hospitals participated in the study; 40 study centers reported MS cases. All cases were examined at the unit of entry, and diagnoses fulfilling the McDonald criteria27 were confirmed by a neurologist. Controls were randomly selected from the national population register and matched to cases on sex, age (5-year age group), and residential area (county). At the time of this study, 1159 persons (632 cases and 527 controls) had completed their participation, although only 988 unmatched persons (494 cases and 494 controls) with nonmissing covariates and genetic information were available for this analysis.
Genotyping for the replication cohort was performed as part of a larger genome-wide analysis of MS susceptibility genes using Illumina Infinium 660K BeadChip.4 Imputation of NAT1 genotypes was done using MaCH v1.0.16 (http://www.sph.umich.edu/csg/abecasis/MACH/index.html) and Utah residents with Northern and Western European ancestry (HapMap), June 2010 as the reference panel, and standard settings.31 Only SNPs with an imputation probability >90% were extracted, resulting in 313 SNPs of which 225 passed quality control with genotyping rate = 1.
Data were collected by questionnaire on demographic variables, environmental exposures, and lifestyle behaviors. Information on current and previous smoking included duration of smoking and the average number of cigarettes/cigars/pipes smoked per day. Study participants were considered ever-smokers of tobacco if they had smoked or currently smoke on-or-off or regularly. Participants were asked to report the ages at which they had started and stopped smoking cigarettes, cigars, and pipes. Two exposures were examined: (1) ever/never-smoker and (2) smoker at age 20 years. Participants were also asked to report their ethnic origin (Swedish, Scandinavian, Finnish, and non-Scandinavian). All available participants provided their year of birth, sex, and education level (bachelor’s degree). To date, the response rate for the Swedish study is 90% (of 1974 contacted) for the case group and 70% (of 4816 contacted) for controls.
Unconditional logistic regression models were used to formally test for G × E interactions on a multiplicative scale. Models included main effects (G and E) and were adjusted for year of birth, sex, and education level. We conducted a three-stage analysis: (1) a discovery analysis using the Kaiser data adjusted for population ancestry; (2) a replication analysis using the Swedish data adjusted for ethnic origin; and (3) a meta-analysis to establish final measures of association. Using ever/never-smoker as the exposure of interest, we investigated 277 common variants (of which 8 correspond to specific NAT1 or NAT2 phenotypes; eTable 1A; http://links.lww.com/EDE/A777) and an additional 5 common NAT1 and NAT2 phenotypes in the discovery analysis by using PLINK v1.07.29 Given the extensive linkage disequilibrium (eFigure 1; http://links.lww.com/EDE/A777), we corrected for the number of independent tests, as recommended, based on the Solid Spine of linkage disequilibrium as implemented in Haploview v4.2.32,33 Using a minimum D′ value of 0.75, there were 11 independent regions (eFigure 1; http://links.lww.com/EDE/A777); therefore, our corrected significance threshold was P < 0.0045. G × E interactions were investigated in the replication analysis; data for 41 of 42 variants of interest were available. A replicating variant was subsequently investigated using a random-effects meta-analysis, with cohort of origin as the random effect, and year of birth, sex, and education level as fixed effects using xtlogit. We performed a test for heterogeneity (Cochran’s Q and I2) across the two studies for the genotype-stratified analyses was performed using metan as implemented in STATA v11.2 (StataCorp, College Station, TX).
We also investigated the risk of MS conferred by tobacco smoking in genotype-stratified analyses. These analyses were repeated using tobacco smoke exposure at age 20 years, in which analyses were restricted to MS cases with disease onset ≥20 years of age. Conditional logistic regression was conducted for a subset of the Kaiser study data (521 cases, 522 controls [one cases had two matched controls]), in which participants were matched on age, sex, and ZIP code. Conditional analyses investigated G × E interactions for the variant of interest using ever/never-tobacco-smoker, smoker at age 20 years, and smoking in the reference year.
To determine whether the replicating variant was the primary association signal within NAT1 from among the available genetic data, linkage disequilibrium analyses were conducted. All variants demonstrating r2 ≥ 0.5 with the replicating variants in either the Kaiser or Swedish datasets were identified and tested through meta-analysis. Haploreg v2 (2013.02.14) (http://www.broadinstitute.org/mammals/haploreg/haploreg.php) was used to assess the potential regulatory function of variants of interest.34 It includes an extensive library of SNPs (dbSNP 137), motif instances (ENCODE), enhancer annotations (Roadmap Epigenome Mapping Consortium), and expression quantitative trait loci (GTex eQTL browser).
The study protocol was approved by the Institutional Review Boards of Kaiser Permanente of Northern California and the University of California, Berkeley. The Swedish study was approved by the Ethical Review Board at Karolinska Institutet, and all participants provided written informed consent.
Overall, 2576 participants (1588 participants from the Kaiser study; 988 participants from the Swedish study) were studied (Table 1). On an average, study participants were in their fifties (Kaiser) and forties (Sweden); therefore, controls had aged past the highest risk period for MS (Table 1). For both crude and adjusted models, MS cases were more likely to have been ever-smokers (Kaiser study, adjusted OR = 1.27 [95% CI = 1.03–1.58]; Swedish study, 1.45 [1.12–1.88]) and smokers at age 20 years (Kaiser study, 1.51 [1.17–1.93]; Swedish study, 1.35 [1.04–1.74]) compared with controls (Table 2). When analyses were restricted to the subset of matched Kaiser cases and controls (521 MS cases, 522 controls), results were very similar (eTable 2; http://links.lww.com/EDE/A777). In addition, based on smoking status in the reference year (the year previous to onset of disease symptoms), smokers had a two-fold increased risk of developing MS compared with nonsmokers (1.97 [1.29–3.00]; eTable 2; http://links.lww.com/EDE/A777).
The discovery analysis investigated G × E interactions between ever/never-smoker status and genetic variation (277 SNPs, Table 3) within three phase-2 genes by using logistic regression models. We focused on interactions of 42 NAT1 SNPs with smoking status (Pcorrected < 0.05). No variant was marginally associated with MS, and there was no association of their main effects in the interaction models (data not shown). For the replication analysis, 41 of the 42 NAT1 variants were independently tested for G × E interactions in the Swedish dataset, using logistic regression. Results for one NAT1 variant (rs7388368A) were replicated (Table 4; Kaiser study, OR of interaction = 1.75 [95% CI = 1.19–2.56]; Swedish study, 1.62 [1.05–2.49]).
In the combined meta-analysis, ever-smokers had an increased risk for MS (Table 2; adjusted OR = 1.36 [95% CI = 1.16–1.60]), and evidence for an interaction with rs7388368A persisted (Table 4; OR for interaction = 1.65 [1.25–2.18]). Stratified analyses demonstrated smoking was a risk factor only for MS among persons who carried the rs7388368A minor allele (A/C or A/A genotype; Table 4). Tobacco smoking did not confer risk of MS among persons homozygous for the rs7388368C major/wildtype allele (C/C genotype: 893 cases and 643 controls). Among persons who were heterozygous for rs7388368A (A/C genotype: 508 cases and 358 controls), ever-smokers had a 60% increased risk of MS compared with nonsmokers (OR=1.60 [95% CI = 1.20–2.13]). Persons who were homozygous for rs7388368A (A/A genotype, 71 cases, 48 controls) had greater than five-fold risk for MS if they were ever-smokers compared with nonsmokers (5.17 [2.17–12.33]). Results were consistent in the Kaiser and Swedish study populations (Table 4), and there was no evidence for heterogeneity for the genotype-stratified meta-analyses (I2 = 0%; Cochran’s Q P > 0.5).
G × E interaction analyses were conducted for rs7388368A and smoker status at age 20 (Table 4). In the meta-analysis, there was evidence for G × E interaction (OR for interaction = 1.35 [95% CI = 1.01–1.81]). A similar relation between smoking status at age 20 and MS risk was observed when stratifying by rs7388368 genotype (C/C, OR = 1.25 [95% CI = 0.99–1.57]; A/C, 1.53 [1.13–2.08]; A/A, 3.43 [1.43–8.20]). There was no evidence for heterogeneity for the genotype-stratified meta-analyses (I2 = 0%; Cochran’s Q P > 0.5). When analyses were restricted to the subset of matched Kaiser cases and controls, results were very similar (eTable 3; http://links.lww.com/EDE/A777). When matched cases and controls were stratified by rs7388368A carrier status (C/C versus A/C and A/A), the three measures of tobacco smoke exposure were associated with MS among those with an A/C or A/A genotype only, despite very small sample sizes (<65 matched pairs; eTable 3; http://links.lww.com/EDE/A777).
To investigate whether rs7388368A might be the primary association within our data set, linkage disequilibrium between this associated SNP and other NAT1 SNPs was tested in each dataset. Nine NAT1 variants were in moderate linkage disequilibrium (r2 ≥ 0.5), of which two variants (rs4921877 and rs6586711) were in strong linkage disequilibrium (r2 > 0.65) with rs7388368A (eTable 4; http://links.lww.com/EDE/A777). Both rs4921877 and rs6586711 showed evidence for interaction with smoking status (eTable 4; http://links.lww.com/EDE/A777), consistent with that of rs7388368 results. To further characterize the association between NAT1 and tobacco smoke in MS, we analyzed 93 variants that had an uncorrected P < 0.05 in the discovery analysis, of which 81 SNPs were in the replication data, and we observed a total of 77 significant associations (P<0.05) in a combined meta-analysis of NAT1 variation. The most associated NAT1 SNPs were located in transcription factor-binding sites (Figure). To determine whether there were other independent G × E interactions, conditional analyses adjusting for the rs7388368–tobacco smoke interaction were conducted. The rs7388368–tobacco smoke interaction was present in all models (data not shown), and no other NAT1 variant demonstrated stronger evidence for an independent G × E signal (Figure).
G × E interactions with specific NAT1 and NAT2 phenotypes were also investigated in the discovery analysis using logistic regression models. A total of 13 common phenotypes (8 of which consisted of single SNPs) were assessed, including NAT1*3, NAT2*5, NAT2*5A, NAT2*5B, NAT2*5C, NAT2*5D, NAT2*6, NAT2*6A, NAT2*6B, NAT2*11A, NAT2*12A, NAT2*12C, and NAT2*13A; as well as global NAT1 and NAT2 slow and fast acetylator status. No evidence for interaction with smoking and MS risk was observed (eTables 1A and 1B; http://links.lww.com/EDE/A777).
MS is a complex autoimmune disease for which both genetic and environmental influences contribute to susceptibility. Despite progress on characterization of MS risk factors, particularly those with a genetic basis, more work is needed to further elucidate genetic and environmental triggers for disease onset. As in other complex diseases, genetic and environmental exposures are also likely to interact in a complex manner,35 and the timing of exposures may be important, particularly in the context of predisposing genetic backgrounds.36,37 We present strong evidence for G × E interaction in MS susceptibility based on a biologically driven hypothesis. The NAT1 rs73688368 genotype was shown to modify the effect of tobacco smoke exposure on disease risk.
A discovery, replication, and meta-analysis approach was used. We demonstrated that, among persons homozygous for rs7388368A, ever-tobacco-smokers were five times more likely to develop MS than never-smokers (Table 4). In addition, more precise measures of tobacco smoke exposure were used to demonstrate evidence for a temporal relation between exposure and disease onset. The first analysis examined smoking status at age 20 years. Among participants homozygous for rs7388368A, those who smoked at age 20 were three times more likely to develop MS than nonsmokers (Table 4). Using a matched case-control analysis nested within the larger Kaiser study, we tested for differences in tobacco smoke exposure in the reference year (the year before onset of first symptoms). Despite the much smaller sample size, similar results were observed: among carriers of rs7388368A, smokers had a 7-fold increased risk of MS compared with nonsmokers (eTable 3; http://links.lww.com/EDE/A777). In contrast, smokers had a 30% to 50% increased risk of MS compared with nonsmokers, overall, when NAT1 genotype was not taken into consideration.
The rs7388368 SNP and nearby variants in strong linkage disequilibrium (rs4921877 and rs6586711) reside within a dense transcription factor-binding region (Figure, eFigure 2; http://links.lww.com/EDE/A777). Variation at rs7388368 was predicted to affect four regulatory motifs (Foxl1, STAT, Pax-4, and Foxj1) where three transcription factors bind (USF1, USF2, NF-κB) by Haploreg (eFigure 2; http://links.lww.com/EDE/A777). Furthermore, rs7388368 resides within or adjacent to an insulator that modulates transcription patterns between enhancers and promoters, further suggesting a role in transcription (eFigure 2; http://links.lww.com/EDE/A777; Broad ChromHMM track, University of California, Santa Cruz genome browser [http://genome.ucsc.edu/cgi-bin/hgGateway]).38,39 These findings suggest that variation within NAT1 that affects gene expression may also modify the disease risk conferred by tobacco smoke. This is in contrast to cancer studies where specific NAT1 acetylation phenotypes have been shown to modify the association of tobacco smoke and cancer risk.20,22,23
This investigation had several strengths. We used two large MS research resources composed of clinically well-characterized cases and controls with both genetic and environmental exposure data. There are few datasets that allow G × E studies in MS. We performed a robust and well-powered multi-stage analysis based on detailed genetic information and a comprehensive G × E investigation of tobacco smoke exposure and four important phase-2 enzymes. Multiple measures of tobacco smoke exposure were assessed, and our findings were consistent among them. Although a prospective cohort design can be used to fully establish temporality and minimize potential recall bias, a case-control design for a less prevalent disease such as MS allows detection of modest G × E interactions with reasonable statistical power.40 In the current study based on case-control comparisons, the association between self-reported use of tobacco and MS was very similar to that of previous cohort and case-control study results.9,10 We confirmed the imputed genotypes (rs7388368; rs4921877) and Illumina 660K BeadChip genotypes (rs6586711) in 134 Kaiser subjects using a custom Fluidigm genotyping platform (100% concordant). We also accounted for multiple testing in the discovery analysis using Solid Spine of linkage disequilibrium to determine that there were 11 independent tests across the three genes (eFigure 1; http://links.lww.com/EDE/A777; corrected threshold P < 0.0045); other approaches, such as the spectral decomposition of matrices of pairwise linkage disequilibrium (http://gump.qlmr.edu.au/general/daleN/MatSpD/) suggest 39 effective tests (corrected threshold P < 0.0013).41 Therefore, interpretation of the discovery findings should be conservative. It is important to note that we used a replication dataset to confirm our initial results.
Whole-genome data were available for participants in the study; therefore, population outliers were removed before analysis, and stringent adjustment for population ancestry or country of origin was done to minimize potential confounding owing to population stratification. Within the discovery (Kaiser) dataset, results for G × E interaction and stratified analyses did not change when covariates for HLA-DRB1*15:01 status, history of infectious mononucleosis, latitude at birth, and latitude at age 10 variables were included in the models (data not shown). Therefore, adjustment for these covariates in replication analyses and final meta-analyses was not pursued. Study participants were drawn from two countries in which MS prevalence differs; Sweden has almost twice the prevalence of the United States. Furthermore, these study populations also differed with respect to their education level and smoking habits (Table 1). There was approximately 10% higher ever-tobacco smoking exposure in the Swedish study compared with the Kaiser study (Table 4) and approximately 20% higher tobacco smoking at age 20 (Table 4). However, despite these differences, similar G × E interactions were observed, and there was no evidence for heterogeneity in the meta-analyses.
Some limitations of this study should be acknowledged. First, the study populations are non-Hispanic whites; results may not be generalizable to other populations. Second, not all NAT1 and NAT2 haplotypes were individually investigated in the discovery analysis owing to low (<10%) frequency (eTable 1A; http://links.lww.com/EDE/A777). Further, recent results provide support for a complex interaction between HLA class I and II variation and tobacco smoke exposure.11 A larger study would be needed to fully characterize joint and independent contributions of HLA loci, the loci studied here, and interaction with tobacco smoke exposure (G × G × E). Our results await confirmation in larger, independent studies. Further research is also needed to determine whether NAT1 gene expression affects the metabolism of tobacco smoke constituents and subsequent risk of MS.
We thank Multiple sclerosis cases and controls for participating in the study. The International MS Genetics Consortium and the Wellcome Trust Case-Control Consortium 2 for their contributions to this study. The Epidemiological Investigation of Multiple Sclerosis has received grant support from the Swedish research council, Knut and Alice Wallenbergs foundation, the AFA, a Swedish Insurance company, foundation, the Swedish Council for Health, Working Life and Welfare (FAS) foundation and the Swedish Brain Foundation.
1. Hawkes CH, Macgregor AJ. Twin studies and the heritability of MS: a conclusion. Mult Scler. 2009;15:661–667
2. Oksenberg JR, Barcellos LF. Multiple sclerosis genetics: leaving no stone unturned. Genes Immun. 2005;6:375–387
3. Barcellos LF, Sawcer S, Ramsay PP, et al. Heterogeneity at the HLA-DRB1 locus and risk for multiple sclerosis. Hum Mol Genet. 2006;15:2813–2824
4. Sawcer S, Hellenthal G, Pirinen M, et al. Genetic risk and a primary role for cell-mediated immune mechanisms in multiple sclerosis. Nature. 2011;476:214–219
5. International Multiple Sclerosis Genetics Consortium (IMSGC). . Analysis of immune-related loci identifies 48 new susceptibility variants for multiple sclerosis. Nat Genet. 2013;45:1353–1360
6. Manolio TA, Collins FS, Cox NJ, et al. Finding the missing heritability of complex diseases. Nature. 2009;461:747–753
7. Ascherio A, Munger KL. Environmental risk factors for multiple sclerosis. Part I: the role of infection. Ann Neurol. 2007;61:288–299
8. Ascherio A, Munger KL. Environmental risk factors for multiple sclerosis. Part II: Noninfectious factors. Ann Neurol. 2007;61:504–513
9. Hawkes CH. Smoking is a risk factor for multiple sclerosis: a metanalysis. Mult Scler. 2007;13:610–615
10. Handel AE, Williamson AJ, Disanto G, Dobson R, Giovannoni G, Ramagopalan SV. Smoking and multiple sclerosis: an updated meta-analysis. PLoS One. 2011;6:e16149
11. Hedström AK, Sundqvist E, Bäärnhielm M, et al. Smoking and two human leukocyte antigen genes interact to increase the risk for multiple sclerosis. Brain. 2011;134(pt 3):653–664
12. Green CR, Rodgman A. The Tobacco Chemists’ Research Conference: a half century forum for advances in analytical methodology of tobacco and its products. Rec Adv Tob Sci. 1996;22:131–304
13. Pryor WA, Stone K. Oxidants in cigarette smoke. Radicals, hydrogen peroxide, peroxynitrate, and peroxynitrite. Ann N Y Acad Sci. 1993;686:12–27; discussion 27
14. Hoffmann D, Hoffmann I, El-Bayoumy K. The less harmful cigarette: a controversial issue. a tribute to Ernst L. Wynder. Chem Res Toxicol. 2001;14:767–790
15. Hedström AK, Bäärnhielm M, Olsson T, Alfredsson L. Tobacco smoking, but not Swedish snuff use, increases the risk of multiple sclerosis. Neurology. 2009;73:696–701
16. Mikaeloff Y, Caridade G, Tardieu M, Suissa SKIDSEP study group. . Parental smoking at home and the risk of childhood-onset multiple sclerosis in children. Brain. 2007;130:2589–2595
17. Hedström AK, Bäärnhielm M, Olsson T, Alfredsson L. Exposure to environmental tobacco smoke is associated with increased risk for multiple sclerosis. Mult Scler. 2011;17:788–793
18. Goldstein JA, Faletto MB. Advances in mechanisms of activation and deactivation of environmental chemicals. Environ Health Perspect. 1993;100:169–176
19. Shimada T. Xenobiotic-metabolizing enzymes involved in activation and detoxification of carcinogenic polycyclic aromatic hydrocarbons. Drug Metab Pharmacokinet. 2006;21:257–276
20. Sanderson S, Salanti G, Higgins J. Joint effects of the N-acetyltransferase 1 and 2 (NAT1 and NAT2) genes and smoking on bladder carcinogenesis: a literature-based systematic HuGE review and evidence synthesis. Am J Epidemiol. 2007;166:741–751
21. Keenan BT, Chibnik LB, Cui J, et al. Effect of interactions of glutathione S-transferase T1, M1, and P1 and HMOX1 gene promoter polymorphisms with heavy smoking on the risk of rheumatoid arthritis. Arthritis Rheum. 2010;62:3196–3210
22. Kilfoy BA, Zheng T, Lan Q, et al. Genetic variation in N-acetyltransferases 1 and 2, cigarette smoking, and risk of non-Hodgkin lymphoma. Cancer Causes Control. 2010;21:127–133
23. Cox DG, Dostal L, Hunter DJ, et al.Breast and Prostate Cancer Cohort Consortium. N-acetyltransferase 2 polymorphisms, tobacco smoking, and breast cancer risk in the breast and prostate cancer cohort consortium. Am J Epidemiol. 2011;174:1316–1322
24. Zhang ZJ, Hao K, Shi R, et al. Glutathione S-transferase M1 (GSTM1) and glutathione S-transferase T1 (GSTT1) null polymorphisms, smoking, and their interaction in oral cancer: a HuGE review and meta-analysis. Am J Epidemiol. 2011;173:847–857
25. Walker K, Ginsberg G, Hattis D, Johns DO, Guyton KZ, Sonawane B. Genetic polymorphism in N-Acetyltransferase (NAT): Population distribution of NAT1 and NAT2 activity. J Toxicol Environ Health B Crit Rev. 2009;12:440–472
26. Krieger N. Overcoming the absence of socioeconomic data in medical records: validation and application of a census-based methodology. Am J Public Health. 1992;82:703–710
27. Polman CH, Reingold SC, Edan G, et al. Diagnostic criteria for multiple sclerosis: 2005 revisions to the “McDonald Criteria”. Ann Neurol. 2005;58:840–846
28. Howie B, Fuchsberger C, Stephens M, Marchini J, Abecasis GR. Fast and accurate genotype imputation in genome-wide association studies through pre-phasing. Nat Genet. 2012;44:955–959
29. Purcell S, Neale B, Todd-Brown K, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet. 2007;81:559–575
30. Hein DW, Doll MA. Accuracy of various human NAT2 SNP genotyping panels to infer rapid, intermediate and slow acetylator phenotypes. Pharmacogenomics. 2012;13:31–41
31. Li Y, Willer CJ, Ding J, Scheet P, Abecasis GR. MaCH: using sequence and genotype data to estimate haplotypes and unobserved genotypes. Genet Epidemiol. 2010;34:816–834
32. Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics. 2005;21:263–265
33. Johnson RC, Nelson GW, Troyer JL, et al. Accounting for multiple comparisons in a genome-wide association study (GWAS). BMC Genomics. 2010;11:724
34. Ward LD, Kellis M. HaploReg: a resource for exploring chromatin states, conservation, and regulatory motif alterations within sets of genetically linked variants. Nucleic Acids Res. 2012;40:D930–D934
35. Thomas D. Gene–environment-wide association studies: emerging approaches. Nat Rev Genet. 2010;11:259–272
36. Goodin DS. The causal cascade to multiple sclerosis: a model for MS pathogenesis. PLoS One. 2009;4:e4565
37. Handel AE, Giovannoni G, Ebers GC, Ramagopalan SV. Environmental factors and their timing in adult-onset multiple sclerosis. Nat Rev Neurol. 2010;6:156–166
38. Ernst J, Kellis M. Discovery and characterization of chromatin states for systematic annotation of the human genome. Nat Biotechnol. 2010;28:817–825
39. Ernst J, Kheradpour P, Mikkelsen TS, et al. Mapping and analysis of chromatin state dynamics in nine human cell types. Nature. 2011;473:43–49
40. Thomas D. Gene–environment-wide association studies: emerging approaches. Nat Rev Genet. 2010;11:259–272
41. Li J, Ji L. Adjusting multiple testing in multilocus analyses using the eigenvalues of a correlation matrix. Heredity (Edinb). 2005;95:221–227
Supplemental Digital Content
© 2014 by Lippincott Williams & Wilkins, Inc