Genetic variation in innate immunity and inflammation pathways associated with lung cancer risk†‡ §
For the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial, we thank Drs. Christine Berg and Philip Prorok, Division of Cancer Prevention, National Cancer Institute (NCI); the Screening Center investigators and staff of the PLCO Cancer Screening Trial; Mr. Tom Riley and staff, Information Management Services, Inc.; Ms. Barbara O'Brien and staff, Westat, Inc.; Mr. Tim Sheehy and staff, DNA Extraction and Staging Laboratory, Science Applications International Corporation (SAIC)-Frederick, Inc.; and Ms. Jackie King and staff, BioReliance, Inc. Most important, we acknowledge the study participants for their contributions to making this study possible.
For the NCI genome-wide association study, we thank the participants and collaborators of the Environment and Genetics in Lung Cancer Etiology (EAGLE), Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC), PLCO, and Cancer Prevention Study-II (CPS-II) Nutrition Cohort studies; the staff of the Core Genotyping Facility—specifically Amy Hutchinson, Aurelie Vogt, Kevin Jacobs, and Zhaoming Wang; the National Center for Biotechnology for assistance with data cleaning; Justin Paschall and Mike Feolo for data manipulation; and Adam Risch, Bill Wheeler, and Sihui Zhao of Information Management Services, Inc. for database support. *This article is US Government work and, as such, is in the public domain in the United States of America.
This article is US Government work and, as such, is in the public domain in the United States of America.
Pulmonary inflammation may contribute to lung cancer etiology. The authors conducted a broad evaluation of the association of single nucleotide polymorphisms (SNPs) in innate immunity and inflammation pathways with lung cancer risk and conducted comparisons with a lung cancer genome-wide association study (GWAS).
In total, 378 patients with lung cancer (cases) and a group of 450 controls from the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial were included. A proprietary oligonucleotide pool assay was used to genotype 1429 SNPs. Odds ratios and 95% confidence intervals were estimated for each SNP, and P values for trend (Ptrend) were calculated. For statistically significant SNPs (Ptrend < .05), the results were replicated with genotyped or imputed SNPs in the GWAS, and P values were adjusted for multiple testing.
In the PLCO analysis, a significant association was observed between lung cancer and 81 SNPs located in 44 genes (Ptrend < .05). Of these 81 SNPS, there was evidence for confirmation in the GWAS for 10 SNPs. However, after adjusting for multiple comparisons, the only SNP that retained a significant association with lung cancer in the replication phase was reference SNP rs4648127 (nuclear factor of kappa light polypeptide gene enhancer of B-cells 1 [NFKB1]) (multiple testing-adjusted Ptrend = .02). The cytosine-thymine (CT)/TT genotype of NFKB1 was associated with reduced odds of lung cancer in the PLCO study (odds ratio, 0.56; 95% confidence interval, 0.37-0.86) and the in the GWAS (odds ratio, 0.79; 95% confidence interval, 0.69-0.90).
A significant association was observed between a variant in the NFKB1 gene and the risk of lung cancer. The current findings add to evidence implicating inflammation and immunity in lung cancer etiology. Cancer 2012. Published 2012 by the American Cancer Society.
With 1.38 million deaths annually, lung cancer causes the largest number of cancer-related deaths worldwide.1 Cigarette smoking is the primary cause of lung cancer, increasing risk by 15-fold to 30-fold.2 In addition, pulmonary inflammation may contribute to lung cancer etiology through the production of reactive oxygen and nitrogen species, the proliferation of cells and increases in angiogenesis during tissue repair, and the up-regulation of antiapoptotic genes through the nuclear factor kappa B (NF-κB) pathway.3 Understanding the role of inflammation in lung cancer etiology may inform chemoprevention efforts and help identify high-risk individuals for screening.
Different strategies have been used to study the association between inflammation and lung cancer. Previous studies have demonstrated associations of lung cancer risk with inflammatory lung conditions, such as chronic obstructive pulmonary disease, pulmonary tuberculosis, and Chlamydia pneumonia.4-6 Furthermore, circulating levels of C-reactive protein (CRP), interleukin 6 (IL-6), and IL-8 and polymorphisms in immunity and inflammation-related genes (eg, IL-1 beta [IL1B]; IL-1 alpha [IL1A]; Fc fragment of immunoglobulin E, low-affinity II, receptor for CD23 [FCER2]; IL-10; tumor necrosis factor alpha [TNFα]; IL-8 receptor type A [IL8RA]; intercellular adhesion molecule 1 [ICAM1]; and IL-12 A [IL12A]) also have been associated with lung cancer risk.7-12 However, these genetic studies have been small, associations have not been replicated, and few studies have comprehensively evaluated polymorphisms in innate immunity and inflammation genes.
Information on genome-wide variation and lung cancer is available from genome-wide association studies (GWAS).13-16 None of the SNPs that have been associated significantly with lung cancer were identified previously in smaller genetic studies of inflammation-related genes. However, because of the stringent P values required for statistical significance in GWAS to prevent false-positive results, it is possible that moderate associations between polymorphisms in inflammation-related genes and lung cancer were missed.
We conducted a comprehensive evaluation of the association of 1429 SNPs across 211 genes in innate immunity and inflammation pathways with lung cancer risk in the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial. We independently confirmed our observations by conducting comparisons with the National Cancer Institute (NCI) lung cancer GWAS.13
MATERIALS AND METHODS
Study participants were enrolled in the PLCO Cancer Screening Trial, which included approximately 155,000 individuals ages 50 to 74 years.17 During the period from 1992 to 2001, participants were randomized to receive either cancer screening or routine health care. Our case-control study was nested within the screening arm of PLCO (N = 77,464). Lung cancer screening included a baseline chest x-ray and 2 (never smokers) or 3 (current/former smokers) annual chest x-rays. In addition, at baseline and annually for 6 years, blood samples were obtained and questionnaires were administered to collect demographic, behavioral, and dietary information.
Lung cancers were ascertained through self-report and confirmed by medical chart and death certificate review. Of the 898 lung cancers that occurred between baseline and December 31, 2004, 378 were included in our analysis. Cases were excluded if they were missing the baseline questionnaire (n = 17) or information on smoking behaviors (n = 11), if they had a history of cancer (n = 56), or if they had multiple cancers during follow-up (n = 88), did not consent to the use of their biologic specimens for genetic analyses (n = 217), or did not have available DNA (n = 131). Controls were matched to cases based on the following criteria: age at randomization (ages 55-59 years, 60-64 years, 65-69 years, or 70-74 years), sex, year of randomization (1993-1995, 1996-1997, 1998-1999, or 2000-2001), length of study follow-up (1-year intervals), and smoking behavior (status [current, former, or never], pack-years smoked at baseline [0-29, 30-39, 40-49, or ≥50 pack-years], and time since quitting [<15, ≥15 years]). Controls were matched to cases 1:1 for current and former smokers and 3:1 for never smokers.
DNA was extracted from buffy coats collected at baseline using a Qiagen kit (Qiagen Inc., Valencia, Calif). Genotyping for 1536 SNPs across 148 gene regions was performed using an Illumina GoldenGate oligonucleotide pool assay (Illumina, San Diego, Calif). Genes were selected onto the oligonucleotide pool assay platform based on their function in innate immunity and inflammation pathways (oxidative response, pattern recognition molecules and antimicrobials, integrins and adhesion molecules, complement, chemokines with their receptors and signaling molecules, and response genes and tissue factors). SNPs in these gene regions (defined as 20 kb upsteam to 10 kb past the polyA tail signal) from the International HapMap project were included using the Tagzilla algorithm.18 The following criteria were used for SNP inclusion: minor allele frequencies of >5% in the HapMap Caucasian samples, r2 values >0.8, and a greater weight for SNPs with a design score of 1.1. In addition, <5% of SNPs were included based on prior evidence from association studies. High-sensitivity CRP was measured using a chemiluminescent immunoassay (Diagnostic Products Corporation, Los Angeles, Calif).
Of the 1536 SNPs that were genotyped, 99 were excluded because of assay failure, and 8 were excluded because of deviations from Hardy-Weinberg equilibrium among controls (P < .001). The assay completion rate for the included SNPs ranged from 91.2% to 94.9%, and the concordance among 23 quality control samples was 87% to 95.7% for 29 SNPs and 100% for the remaining 1400 SNPs.
Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated for lung cancer by comparing heterozygous and homozygous variant genotypes with homozygous wild-type genotypes for each of 1429 SNPs. P values for trend (Ptrend) were calculated, treating the number of alleles as continuous variables. We conducted analyses stratified by lung cancer histology, smoking status, and family history; and we evaluated interactions between SNPs and smoking status, family history, and CRP levels. We also conducted analyses adjusted for age, race, sex, and smoking behavior (status, pack-years, and time since quitting) using unconditional logistic regression to retain the maximum number of cases and controls. Because adjustment did not alter the associations between SNPs and lung cancer, all results presented are unadjusted. In a sensitivity analysis that was restricted to white participants, the results also did not change. Therefore, we present results for all participants.
We identified SNPs that were associated significantly (Ptrend < .05) with lung cancer in our case-control study and then replicated our results in an independent sample, the NCI's lung cancer GWAS.13 The NCI GWAS included data on 5739 lung cancer cases and 5848 controls from the PLCO, the Environment and Genetics in Lung Cancer Etiology Study, the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study, and the Cancer Prevention Study II Nutrition Cohort.13 We estimated ORs with logistic regression in the NCI GWAS for 1) the same SNP estimated in the case-control study, if genotyped in the GWAS; or 2) an imputed value for the SNP estimated in the case-control study, if not genotyped in the GWAS. Imputed values were estimated from genotyped SNPs in the same gene region (20 kb upsteam to 10 kb past the polyA tail signal) as the SNP of interest using the IMPUTE statistical program19 with HapMap release 24 (National Center for Biotechnology Information, National Institutes of Health, Bethesda, Md) as the reference panel. We excluded PLCO participants from the GWAS analyses who were included in our case-control study (n = 355).
We performed permutations to adjust for multiple testing of 81 SNPs in the replication stage. Briefly, we permuted the case-control status in the PLCO data 1000 times, because genotype data were not available from the GWAS. For each permutation, we computed the 1-sided P value for each SNP with direction specified by the PLCO results and then computed the minimum P value. Multiple testing-adjusted P values were calculated as the proportion of permutations (n = 1000) in which the minimum P value was below the P value from the GWAS data. When the direction of the association differed in the PLCO and GWAS data sets, the multiple testing-adjusted P value was 1.0. The multiple testing-adjusted P values calculated through permutation were very similar to those estimated with a Bonferroni correction.
Baseline characteristics of cases (n = 378) and controls (n = 450) from the PLCO study are presented in Table 1. No significant differences were observed for the matching characteristics, with the exception of smoking status, because the study design oversampled never-smoking controls. Cases were significantly more likely than controls to have a history of bronchitis/emphysema at baseline (P = .007) and a family history of lung cancer (P = .01). The most common lung cancer histologic types included adenocarcinoma (n = 170), small cell carcinoma (n = 88), squamous cell carcinoma (n = 41), and large cell carcinoma (n = 20).
Table 1. Characteristics of 378 Lung Cancer Cases and 450 Controls in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial
|Total||450|| ||378|| || |
|Age at enrollment, y|| || || || || |
| 60-64||131||29.1||106||28|| |
| 65-69||147||32.7||128||33.9|| |
| 70-74||83||18.4||64||16.9|| |
|Sex|| || || || || |
| Men||270||60||247||65.3|| |
|Smoking statusb|| || || || || |
| Never smoker||99||22||30||7.9||—a|
| Former smoker, pack-years||217||48.2||217||57.4|| |
| 0-29 and quit for <15 y||21||4.7||23||6.1|| |
| 0-29 and quit for ≥15 y||49||10.7||44||11.6|| |
| 30-39, quit for <15 y||32||7.1||30||7.9|| |
| 30-39 and quit for ≥15 y||16||3.6||17||4.5|| |
| 40-49 and quit for <15 y||15||3.3||14||3.7|| |
| 40-49 and quit for ≥15 y||8||1.8||11||2.9|| |
| ≥50 and quit for <15 y||64||14.2||66||17.5|| |
| ≥50 and quit for ≥15 y||12||2.7||12||3.2|| |
| Current smoker, pack years||134||29.8||131||34.7|| |
| 0-29||32||7.1||26||6.9|| |
| 30-39||41||9.1||39||10.3|| |
| 40-49||9||2||11||2.9|| |
| ≥50||52||11.6||55||14.6|| |
|Race/ethnicity|| || || || || |
| Black||20||4.4||27||7.1|| |
| Hispanic||7||1.6||4||1.1|| |
| Other||13||2.9||9||2.4|| |
|Education|| || || || || |
| ≤High school graduate||157||34.9||148||39.2||.21|
| ≥High school graduate||293||65.1||230||60.8|| |
|Body mass index at enrollment, kg/m2|| || || || || |
| 18.5-24.9||142||31.6||121||32|| |
| 25.0-29.9||211||46.9||162||42.9|| |
| ≥30||89||19.8||88||23.3|| |
|Regular use of aspirin or ibuprofen|| || || || || |
| No||148||32.9||136||36|| |
|History of bronchitis/emphysema|| || || || || |
| No||394||87.6||299||79.1|| |
| Missing||9||2||17||4.5|| |
|History of heart disease|| || || || || |
| No||386||85.8||300||79.4|| |
| Missing||9||2||16||4.2|| |
|Family history of lung cancerc|| || || || || |
| No||387||86||288||76.2|| |
| Missing||15||3.3||24||6.4|| |
Of the 1429 SNPs that were included in our analysis, we observed significant associations between 81 SNPs located in 44 genes and lung cancer (Ptrend < .05). The results for all of the SNPs can be obtained by contacting the author. Of these 81 SNPS, there was evidence for confirmation in the GWAS data for 4 SNPs through direct replication and for 6 SNPs through imputation. Table 2 presents the ORs for the 10 confirmed SNPs in the PLCO and GWAS.
Table 2. Association of Selected Innate Immunity and Inflammation Single Nucleotide Polymorphisms With Lung Cancer in the PLCO Case-Control Study and the National Cancer Institute Genome-Wide Association Study
|NFKB1: rs4648127|| || || || || |
| CC||379||342||1.0||1.0b|| |
| CT||69||35||0.56 (0.36-0.87)||0.78 (0.68-0.90)|| |
| TT||2||1||0.55 (0.05-6.14)||0.83 (0.55-1.25)||.023c|
| CT+TT|| || ||0.56 (0.37-0.86)||0.79 (0.69-0.90)|| |
| Ptrenda|| || ||.0092||.0006|| |
|C2: rs497309|| || || || || |
| AA||374||293||1.0||1.0b|| |
| AC||71||76||1.37 (0.96-1.95)||1.26 (1.09-1.46)|| |
| CC||5||8||2.04 (0.66-6.31)||1.10 (0.86-1.41)||.18|
| AC+CC|| || ||1.41 (1.00-1.99)||1.21 (1.08-1.37)|| |
| Ptrenda|| || ||.040||.0058|| |
|TLR3: rs3775291|| || || || || |
| CC||229||172||1.0||1.0d|| |
| CT||192||165||1.14 (0.86-1.52)||1.05 (0.97-1.13)|| |
| TT||28||41||1.95 (1.16-3.28)||1.21 (1.06-1.38)||.32|
| CT+TT|| || ||1.25 (0.95-1.64)||1.07 (1.00-1.16)|| |
| Ptrenda|| || ||.024||.010|| |
|ALOX5: rs1487562|| || || || || |
| CC||321||237||1.0||1.0d|| |
| CT||119||122||1.39 (1.03-1.88)||1.11 (1.03-1.21)|| |
| TT||10||19||2.57 (1.18-5.64)||1.09 (0.91-1.30)||.44|
| CT+TT|| || ||1.48 (1.11-1.98)||1.11 (1.03-1.20)|| |
| Ptrenda|| || ||.0030||.02|| |
|C2: rs2734335|| || || || || |
| AA||149||96||1.0||1.0d|| |
| AG||192||180||1.46 (1.05-2.02)||1.02 (0.93-1.11)|| |
| GG||108||102||1.47 (1.01-2.13)||1.14 (1.03-1.27)||.46|
| AG+GG|| || ||1.46 (1.08-1.98)||1.06 (0.97-1.14)|| |
| Ptrenda|| || ||.038||.017|| |
|FCER1G: rs4489574|| || || || || |
| CC||229||169||1.0||1.0d|| |
| CT||169||152||1.22 (0.91-1.64)||1.14 (1.05-1.23)|| |
| TT||51||57||1.51 (0.99-2.32)||1.05 (0.92-1.19)||.70|
| CT+TT|| || ||1.29 (0.98-1.69)||1.12 (1.04-1.21)|| |
| Ptrenda|| || ||.041||.032|| |
|TLR4: rs10759932|| || || || || |
| TT||341||265||1.0||1.0b|| |
| CT||98||100||1.31 (0.95-1.81)||1.10 (1.01-1.21)|| |
| CC||9||12||1.72 (0.71-4.13)||1.08 (0.88-1.34)||.75|
| CT+CC|| || ||1.35 (0.99-1.84)||1.10 (1.01-1.20)|| |
| Ptrenda|| || ||.049||0.039|| |
|TLR10: rs7660429|| || || || || |
| CC||329||299||1.0||1.0b|| |
| CG||110||73||0.73 (0.52-1.02)||1.09 (0.99-1.19)|| |
| GG||11||6||0.60 (0.22-1.64)||1.14 (0.92-1.42)||1.0|
| CG+GG|| || ||0.72 (0.52-0.99)||1.10 (1.00-1.20)|| |
| Ptrenda|| || ||.044||0.036|| |
|ACAD11: rs11927882|| || || || || |
| GG||273||201||1.0||1.0b|| |
| AG||142||136||1.30 (0.97-1.75)||0.85 (0.72-0.99)|| |
| AA||35||40||1.55 (0.95-2.53)||0.71 (0.42-1.16)||1.0|
| AG+AA|| || ||1.35 (1.02-1.78)||0.83 (0.72-0.97)|| |
| Ptrenda|| || ||.027||.013|| |
|CCR6: rs9459883|| || || || || |
| GG||378||293||1.0||1.0b|| |
| CG||70||73||1.35 (0.94-1.93)||0.90 (0.80-1.02)|| |
| CC||2||12||7.74 (1.72-34.8)||0.72 (0.42-1.22)||1.0|
| CG+CC|| || ||1.52 (1.07-2.16)||0.89 (0.79-1.00)|| |
| Ptrenda|| || ||.003||.045|| |
After adjusting for multiple comparisons, only nuclear factor of kappa light polypeptide gene enhancer of B-cells 1 (NFKB1) (reference SNP no. rs4648127) retained a significant association with lung cancer in the replication phase (multiple testing-adjusted P = .02). The cytosine-thymine (CT)/TT genotype of NFKB1 (rs4648127) was associated with a 44% reduction in the odds (OR, 0.56; 95% CI, 0.37-0.86) of lung cancer in the PLCO case-control study and a 21% reduction in the odds (OR, 0.79; 95% CI, 0.69-0.90) of lung cancer when imputed in the GWAS. None of the 13 additional NFKB1 SNPs that were genotyped in the PLCO case-control study were associated significantly with lung cancer risk (all Ptrend < .05). It is noteworthy that rs4648127 was not correlated with the 12 other NFKB1 SNPs that were genotyped in PLCO and available in HapMap (r2 range, 0-0.49).
We carried out additional analyses in the PLCO case-control study. Among controls, NFKB1 (rs4648127) was not associated with family history of lung cancer, bronchitis/emphysema, or CRP (all P > .3). In addition, we observed no significant interaction between NFKB1 (rs4648127) and smoking status (Pinteraction = .19), CRP level (Pinteraction = .97), or family history (Pinteraction = .78). Across lung cancer histologies, we observed that the association between NFKB1 (rs4648127) and lung cancer was limited to adenocarcinoma, although we had limited power to detect associations with other histology groups.
A SNP in the NFKB1 gene (rs4648127) was associated with lung cancer in the screening arm of the PLCO Cancer Screening Trial, and a 44% reduction in lung cancer risk with at least 1 T allele was reported compared with the CC genotype. In addition, lung cancer risk was reduced by 21% when the same SNP was imputed with data from the NCI GWAS. Our observation of a significant association between the NFKB1 gene and lung cancer risk underscores the etiologic role of inflammation and immunity in lung carcinogenesis.
NF-κB, a transcription factor that is activated primarily by proinflammatory cytokines, plays an important role in development, immunity, tissue homeostasis, and inflammation and regulates gene expression, cell apoptosis, and proliferation.20 Although, to our knowledge, variation in the NFKB1 gene has not been associated previously with lung cancer risk, a polymorphism in the promoter region of NFKB1 has been associated with the severity of acute respiratory distress syndrome,21 suggesting biologic plausibility for the role of NFKB1 in pulmonary diseases. SNPs in NFKB1 have been associated with the risk of non-Hodgkin lymphoma, Hodgkin lymphoma, colon cancer, rectal cancer, meningioma, and cervical cancer.18, 22-25 Furthermore, haplotypes, including the same NFKB1 SNP that was identified in our study (rs4648127), were associated significantly with the risk of rectal cancer.25 It is noteworthy that NF-κB also can be activated by inducers other than cytokines, including viral and bacterial products, DNA damage, and hypoxia.26 Thus, the association between NFKB1and lung cancer also may reflect associations between other factors and lung cancer risk.
It is important to place results from candidate gene studies in the context of GWAS results. The evaluation of 515,922 SNPs in the NCI GWAS necessitated the use of very stringent P values to prevent false-positive associations. Therefore, false-negative associations may have arisen for SNPs with more moderate associations and weaker P values. Indeed, although 8 of 20 NFKB1 SNPs in the GWAS were significant at nominal levels (P < .05; range, .0004-.7),13 none of these met the GWAS P value threshold. In addition, there may have been poor coverage of gene regions associated with inflammation on the GWAS chip. Nonetheless, using a targeted approach focused on the immune and inflammation pathways, we observed support for an association between NFKB1 and lung cancer.
The main limitation of our study was the relatively small numbers of cases and controls, which limited our ability to detect weaker associations in a large number of SNPs or to conduct pathway-based analyses in the PLCO data set. However, we focused our analyses on genes that have functions related to inflammation and innate immunity, which support emerging, biologically plausible mechanisms in lung cancer development. Furthermore, we confirmed our results with the NCI GWAS study. We believe that this method has greater sensitivity than GWAS alone and can be used when there is a biologically motivated hypothesis.
In conclusion, our study supports the role of genetic variation in innate immunity in the development of lung cancer. Future studies should further examine the NFKB1 region to identify functional SNPs and corresponding protein levels that are associated with lung cancer risk. These findings add to the growing body of literature implicating inflammation and immunity in lung cancer etiology.
This study was supported by the Intramural Research Program of the National Cancer Institute. The Environment and Genetics in Lung Cancer Etiology (EAGLE), Prostate, Lung, Colon, Ovary Screening Trial (PLCO), and Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) studies were supported by the Intramural Research Program of the National Institutes of Health, National Cancer Institute (NCI), Division of Cancer Epidemiology and Genetics. PLCO also was supported by individual contracts from the NCI to the University of Colorado Denver (NO1-CN-25,514), Georgetown University (NO1-CN-25,522), the Pacific Health Research Institute (NO1-CN-25,515), the Henry Ford Health System (NO1-CN-25,512), the University of Minnesota, (NO1-CN-25,513), Washington University (NO1-CN-25,516), the University of Pittsburgh (NO1-CN-25,511), the University of Utah (NO1-CN-25,524), the Marshfield Clinic Research Foundation (NO1-CN-25,518), the University of Alabama at Birmingham (NO1-CN-75,022), Westat, Inc. (NO1-CN-25,476), and the University of California, Los Angeles (NO1-CN-25,404). ATBC also was supported by US. Public Health Service contracts (N01-CN-45,165, N01-RC-45,035, and N01-RC-37,004) from the NCI. The Cancer Prevention Study-II (CPS-II) Nutrition Cohort was supported by the American Cancer Society.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.