Inflammation, Genetic Polymorphisms in Proinflammatory Genes TNF-A, RANTES, and CCR5, and Risk of Pancreatic Adenocarcinoma

Adenocarcinoma of the exocrine pancreas is the fourth leading cause of cancer-related death in men and women in the U.S. (1). With the exception of cigarette smoking, few environmental risk factors for pancreatic cancer are known (2). Cytokines and other proinflammatory mediators have been implicated in inflammatory pancreatic diseases including pancreatitis and cancer (3). Inflammatory pancreatitis is believed to be a risk factor for pancreatic cancer (4).⁶

As a central mediator of inflammation and apoptosis, the cytokine tumor necrosis factor-α (TNF-α) may possess both protumor and antitumor activities (10). The chemokine, regulated upon activation, normally T cell–expressed, and presumably secreted (RANTES), and one of its receptors, CC chemokine receptor 5 (CCR5), are believed to play a role in antitumor immunity through immune cell recruitment (11). A number of polymorphisms have been identified in cytokine and chemokine genes. The polymorphisms examined in this study include TNF-A −308 (G/A) in the TNF-α promoter, RANTES −403 (G/A) in the RANTES promoter, and a 32 bp deletion in its receptor CCR5 (Δ32). TNF-A −308A is believed to exhibit increased TNF-A gene transcription (12) and has recently been shown to be genetically linked with transcription at the nearby lymphotoxin-α (LTA) locus (13). RANTES −403A has been associated with increased RANTES transcription and delayed HIV-1 disease progression (14), and the 32 bp deletion in CCR5 results in a frameshift at amino acid 185, protein truncation, and loss of expression on the cell surface (15). Some but not all studies have shown lower HIV-1 viral load and delayed progression to AIDS in CCR5-Δ32 heterozygotes and homozygotes (16, 17).

Because cytokines are released in response to various forms of cellular stress including inflammation and carcinogen exposure(18), we evaluated the main gene effects and potential interactions between genetic polymorphisms in cytokine/chemokine genes and indices of inflammation (history of pancreatitis, ulcer, or diabetes), elevated body mass index (BMI), saturated fat intake, and tobacco smoking in relation to the risk for pancreatic cancer.

A population-based case-control study of pancreatic cancer was conducted in six San Francisco Bay Area counties (Alameda, Contra Costa, Marin, San Francisco, San Mateo, and Santa Clara) between 1994 and 2005. Cases were identified using rapid case ascertainment by the Northern California Cancer Center (Union City, CA) with a goal to identify patients in the study area within 1 month of diagnosis. Eligible cases were newly diagnosed between 1995 and 1999 with adenocarcinoma of the exocrine pancreas and were between 21 and 85 years old, resided in one of the six Bay Area counties, were alive at the time of the first attempted contact, and could complete an interview in English. A total of 532 eligible cases completed the interview for a 67% response rate. Patient diagnoses were confirmed by participants' physicians and by the Surveillance, Epidemiology, and End Results abstracts that included histologic confirmation of disease.

Control participants were identified within the six San Francisco Bay Area counties using random-digit dial, and were frequency-matched to cases in an approximate 3:1 ratio by sex and age. Eligibility criteria were identical for case and control participants except for pancreatic cancer status. Control acquisition for those >65 years was supplemented by random selection from Health Care Finance Administration (now the Centers for Medicare and Medicaid Services) lists for the six Bay Area counties. A total of 1,701 eligible control participants completed the study interview for a 67% response rate. The main study included individuals with a previous cancer diagnosis (except for pancreatic cancer). Data from the main study showed that a similar proportion of case and control participants reported having had other cancers prior to their diagnosis or interview (14% and 13%, respectively). The analyses for this study are based on 308 cases and 964 controls who gave blood as part of the laboratory portion of the study. Detailed methods on case and control selection and the laboratory portion of the study have been published (19-23). The study interviewers obtained written informed consent from all participants and was approved by the University of California Institutional Review Board.

Exposure history including information on history of pancreatitis, stomach or duodenal ulcer, diabetes mellitus, smoking, diet, and anthropometric data were obtained at interview by self-report using structured questionnaires. No proxy interviews were conducted. Race/ethnicity was based on self-report and was broadly defined as Caucasian, Black/African-American, or Asian. Five cases and 15 controls did not fall into any of these three categories and were classified as “other race/ethnicity” for these analyses. Pancreatitis, ulcer, and diabetes were self-reported and considered positive when the respondent replied that the condition was physician-diagnosed. Participants were defined as smokers if they had smoked >100 cigarettes in their lifetime or had smoked pipes or cigars at least once per month for 6 months or more. Participants with passive smoke exposure at home as an adult (women, 32 cases and 95 controls; men, 5 cases and 21 controls) were removed from the referent group of never-smokers. A variable for smoking status was evaluated using the following three categories: (a) never active or passive, (b) former active or passive and/or cigar or pipe use, and (c) current active. BMI (weight in kilograms divided by height in meters squared) was based on self-reported usual adult height and weight and categorized based on quartiles of the distribution among the 964 controls (<21.15, 22.15-24.23, 24.24-26.59, and ≥26.60 kg/m²), and according to the WHO definition of normal weight (<25 kg/m²), overweight (25-29.9 kg/m²), and obesity (≥30 kg/m²; ref. 24). Daily saturated fat intake (g/d) was assessed using the Harvard food frequency questionnaire (intake prior to the previous year) and was adjusted for total energy intake using the residual method (25). Cut-points for saturated fat intake were based on quartiles of the distribution among the 964 controls who participated in the laboratory portion of the study.

Genomic DNA was extracted from whole blood using the QIAmp DNA Blood Mini kit (Qiagen Inc., Valencia, CA) according to the manufacturer's instructions. TNF-A −308G/A and RANTES −403G/A were genotyped using Masscode (formerly Qiagen Genomics, Inc., Bothell, WA; now BioServe Inc., Laurel, MD; ref. 26). For participants in whom the mass spectrometry method failed to yield a conclusive genotype, missing data were completed using PCR RFLP assays according to Wilson et al. (27) and Hajeer et al. (28). A random sample of the data (3%) for TNF-A −308 and RANTES−403 were repeated using both Masscode and PCR-RFLP and were found to agree for both genotyping methods. CCR5-Δ32 was genotyped according to a gel-based PCR method and primers published by Martinson et al. (29). DNA samples that yielded “no calls” after three attempts at genotyping were reported as missing.

Tests for Hardy-Weinberg equilibrium were conducted by comparing observed with expected genotype frequencies using a χ² test with 1 df. Expected genotype frequencies were estimated from allele frequencies. Odds ratios (OR; hereafter called risk) and 95% confidence intervals (CI) were estimated using unconditional logistic regression in SAS (v. 8.2; SAS Institute, Cary, NC). All statistical tests were two-sided. Potential confounders were included in multivariable models if their inclusion changed β variable estimates by >10%. Final multivariable models for cytokine/chemokine polymorphisms, inflammation, and pancreatic cancer included age at interview, sex, and smoking status. Potential confounders that did not change variable estimates by >10% were race or ethnicity, alcohol or coffee consumption, educational level, annual household income, family history of pancreatic cancer, history of diabetes, gallbladder disease, allergy, BMI, and vitamin B₁₂ deficiency. Gene-environment and gene-gene interactions were assessed using stratified models and by evaluating departures from additive effects. We evaluated departures from additive effects between two variables by coding a new variable with a common referent group based on a priori hypotheses (20). The magnitude of an interaction effect was determined by estimating age-adjusted interaction contrast ratios (ICR) with 95% confidence limits using PROC LOGISTIC in SAS (30). We calculated the ICR using the following formula:

where RR₁₁ is the risk ratio for exposure with a rare allele genotypes, RR₁₀ is the risk ratio for rare allele genotypes among the nonexposed, and RR₀₁ is the risk ratio for exposure with a common genotype. ICRs > 0 imply greater than additive effects (interaction), whereas ICRs = 0 imply additive effects (no interaction), and ICRs < 0 imply less than additive effects (negative interaction). For the present study, we considered ICRs of magnitude one or higher as indicative of an interaction. Confidence limits for ICRs that exclude the null value of zero can be considered statistically significant at α = 0.05.

All polymorphisms evaluated in this study were in Hardy-Weinberg equilibrium within each of the three control groups (Caucasian, African-American, and Asian; all P > 0.15; Table 1). The Δ32 deletion allele at the CCR5 locus was not detected in Asian participants, and was extremely rare in African-American participants (only two heterozygous individuals were detected). In African-American and Asian participants, allelic variation for the TNF-A (−308G/A) locus also was rare. In general, rare variants at RANTES and TNF-A loci were inversely associated with pancreatic cancer risk. However, all OR estimates overlapped unity, except at the RANTES locus in Asian participants (Table 1). Main gene OR estimates did not significantly differ when the data were stratified by median age at pancreatic cancer diagnosis or interview (<65.5 versus ≥65.5 years; data not shown).

Cases were about seven times more likely to have reported a history of pancreatitis than were controls (Table 2). ORs for chronic pancreatitis and acute pancreatitis were similar in magnitude. Having a history of ulcer (stomach or duodenal) was not associated with risk for pancreatic cancer in this study (Table 2). Reporting a history of non–insulin-dependent diabetes was only weakly associated with risk for pancreatic cancer (Table 2). Participants who reported that they were current tobacco smokers were at a 2.7-fold increased risk for pancreatic cancer (Table 2). Participants who were in the highest quartile of calorie-adjusted daily saturated fat intake (≥24.8 g/d) were at a nearly 2-fold increased risk for pancreatic cancer (Table 2). Participants in the upper third and fourth quartiles of BMI were at an increased risk for pancreatitic cancer (Table 2). With the exception of pancreatitis, adjusted ORs for all of the factors in Table 2 were similar in magnitude when the data were stratified by median age at pancreatic cancer diagnosis or interview (<65.5 versus ≥65.5 years; data not shown). The adjusted OR for pancreatitis in participants <65.5 years of age was 12 (95% CI, 4.2-35); the adjusted OR for pancreatitis in participants ≥65.5 years of age was 3.6 (95% CI, 1.3-10). ORs for the factors listed in Table 2 were similar in magnitude between the subset of the study with genomic DNA as described in this report (308 cases and 964 controls) and the entire San Francisco Bay Area pancreatic cancer study population (532 cases and 1,701 controls; data not shown; refs. 21, 31).⁶

ORs for the joint effect of self-reported history of pancreatitis and chemokine/cytokine gene polymorphisms suggested the possibility of modification of risk for pancreatic cancer, although some cell sizes were small (Table 3). Based on these joint ORs with common allele genotypes and no history of pancreatitis in the referent category, the risk of pancreatic cancer subsequent to pancreatitis was higher among individuals with rare allele genotypes in TNF-A (GA + AA) and RANTES (GA + AA), and among individuals with genotypes that included the common undeleted allele in CCR5 (wt/wt; Table 3). The age-adjusted ICR (95% CI) for pancreatitis and TNF-A was 13.6 (−13.8 to 41), 2.8 (−7.3 to 13) for pancreatitis and RANTES, and −6.9 (−17 to 2.9) for pancreatitis and CCR5. All 95% CIs for ICRs were wide and overlapped the null value of zero. Due to the small sample size, we were unable to stratify the data in Table 3 by median age at pancreatic cancer diagnosis or interview.

There were positive associations between having a history of pancreatitis and rare allele–containing genotypes at the TNF-A −308 locus (GA + AA) and the RANTES −403 locus (GA + AA; Table 4). The associations were significantly stronger when the analyses were restricted to pancreatic cancer cases only. Adjustment for smoking did not alter the results, so we only present ORs adjusted for age and sex. When the analyses were restricted to pancreatic cancer cases whose age at diagnosis was <65.5 years, the OR for TNF-A (GA + AA) and pancreatitis was 3.2 (95% CI, 1.1-9.5); in cases whose age at diagnosis was 65.5 years or greater, the OR was 3.1 (95% CI, 0.73-13). In pancreatic cancer cases whose age at diagnosis was <65.5 years, the OR for RANTES (GA + AA) and pancreatitis was 1.7 (95% CI, 0.62-4.8); in cases whose age at diagnosis was 65.5 years or greater, the OR was 4.5 (95% CI, 0.92-22).

There was no evidence for increased or decreased joint ORs for pancreatic cancer and having a positive history of stomach or duodenal ulcer and genotypes in RANTES, CCR5, or TNF-A (data not shown). There was no evidence for interactions between age at first ulcer (<39 versus ≥39 years) and genotypes in RANTES, CCR5, or TNF-A (data not shown). There was no evidence for interactions between having a self-reported history of diabetes and chemokine/cytokine polymorphisms in relation to pancreatic cancer risk (data not shown).

The results of joint ORs for smoking status and cytokine/chemokine genotypes in relation to pancreatic cancer risk suggested the possibility of interactions for current active smoking and the CCR5-Δ32 deletion allele (Table 5). The age-adjusted ICR (95% CI) for CCR5-Δ32 and smoking was 1.4 (−1.2 to 3.9); −0.9 (−2.2 to 0.47) for TNF-A and smoking; and −0.09 (−1.4 to 1.3) for RANTES and smoking. All 95% CIs for ICRs overlapped the null value of zero.

There was no evidence for interactions between cytokine/chemokine gene polymorphisms and quartiles of calorie-adjusted saturated fat intake (data not shown). Furthermore, joint ORs for BMI categories and chemokine/cytokine polymorphisms revealed no consistent evidence for interactions (data not shown).

Joint ORs for two-locus interactions (RANTES × CCR5, RANTES × TNF-A, and TNF-A × CCR5) revealed no evidence for interactions between these genes in relation to pancreatic cancer risk (data not shown). Stratification of these ORs by sex revealed no additional patterns in the data (data not shown).

In this population-based case-control study, we investigated the association between polymorphisms in the cytokine/chemokine genes TNF-A −308G/A, RANTES −403G/A, CCR5-Δ32, inflammation, and pancreatic adenocarcinoma. Our results suggest that the polymorphisms in TNF-A and RANTES are associated with having a history of pancreatitis, particularly as a possible early manifestation of pancreatic cancer. The same alleles (in TNF-A and RANTES) that were associated with pancreatitis were also associated with having a stronger risk of developing pancreatic cancer, suggesting that these genes are important determinants of pancreatic cancer in the presence of inflammation (pancreatitis). Whether pancreatitis is an independent risk factor or an early manifestation of pancreatic cancer has yet to be resolved. Recent data from our San Francisco Bay Area study show that an increasing number of years between the diagnosis of pancreatitis and pancreatic cancer is associated with a decreasing trend in ORs for pancreatitic cancer (trend P < 0.0001), suggesting that some pancreatitis is more likely to be an early manifestation of pancreatic cancer.⁶

TNF-A expression has been associated with the severity of acute pancreatitis (7) and may play a role in tissue remodeling following chronic pancreatitis (8). However, studies of the TNF-A −308 polymorphism and pancreatitis or pancreatic cancer generally have indicated a lack of association (32, 33). Our result may be different in that the association of TNF-A and RANTES with pancreatitis was found mainly among cases of pancreatic cancer. TNF-A −308A has been associated with increased TNF-A transcription, and TNF-α protein has been shown to inhibit the apoptosis of pancreatic cancer cells (34). Furthermore, TNF-A −308 has been genetically linked with transcription at the LTA locus located upstream from the TNF-A locus (13). LTA shares several biological and structural characteristics with TNF-A and is a potentially important mediator of the inflammatory response (35). Genetic variation at LTA has also been associated with the development of human cancers (36, 37).

Elevated levels of CCR5 and RANTES mRNA have been detected in human chronic pancreatitis tissue samples relative to normal pancreas tissues (6). The majority of the CCR5-positive cells were found to be macrophages (6). Thus, it is likely that cytokines/chemokines are predominantly expressed in the pancreas under conditions of inflammation. There is increasing evidence that chronic inflammatory conditions and persistent cell turnover could contribute to pancreatic cancer development (3, 38, 39). Our results showing a somewhat elevated OR for current active smoking and CCR5-32del genotypes in relation to pancreatic cancer risk suggest that intact CCR5 may offer pancreatic cells some protection from the damaging effects of tobacco smoking, but this result will need to be verified in other study populations.

Due to the limited number of individuals reporting a history of pancreatitis, we cannot rule out the possibility that our positive results were due to chance. However, risk estimates for pancreatic cancer in relation to pancreatitis are similar in the full data set and the subset who have DNA available. Other potential weaknesses in our study include the possible recruitment bias among pancreatic cancer cases associated with the rapidly fatal course for the disease. We cannot rule out the possibility that one or more genes under study are associated with more or less rapid case fatality. Despite this possibility, case refusal rates were quite low at 8% for the main study. As in all case-control studies, recall bias is possible, but less likely than for other common cancers because few risk factors are known.

The study of cytokines and chemokines in cancer development is extremely complex. For example, there are multiple families of chemokines with at least 50 different proteins in humans (40). Furthermore, multiple chemokines may bind the same receptor, or one receptor may react to multiple ligands. For example, RANTES ligand binds to CCR5, CCR1, and CCR3 (41), and CCR5 receptor can interact with RANTES, MIP-1α, and MIP-1β (15). Thus, this study represents the start of an evaluation of genetic variation in cytokines and chemokines in relation to inflammation and pancreatic cancer risk. Because invasion through the extracellular matrix is an important step in tumor spread and invasion, cytokines may interact in important ways with matrix metalloproteinases that also display genetic variation. Given the large number of possible interactions (42) and the evidence for cytokine and matrix metalloproteinase gene clusters on chromosomes 5q31 (43) and 11q21-q22 (44), respectively, evaluation of haplotype-related risk may provide additional insight to understanding pancreatic cancer development in humans. In conclusion, our results suggest that genetic variation in cytokine/chemokine genes in combination with proinflammatory conditions may influence the development of pancreatic cancer in humans.