Antioxidant genes, diabetes and dietary antioxidants in association with risk of pancreatic cancer

Abstract

To test the hypothesis that polymorphic variants of antioxidant genes modify the risk of pancreatic cancer, we examined seven single-nucleotide polymorphisms (SNPs) of genes coding for superoxide dismutase (SOD) 2, glutathione S-transferase alpha 4 (GSTA4), catalase and glutathione peroxidase in 575 patients with pancreatic adenocarcinoma and 648 healthy controls in a case–control study. Information on risk factors was collected by personal interview and dietary information was collected by a self-administered food frequency questionnaire. Genotypes were determined using the Taqman method. Adjusted odds ratio (AOR) and 95% confidence interval (CI) were estimated by unconditional logistic regression. No significant main effect of genotype was observed. A borderline significant interaction between diabetes and SOD2 Ex2+24T>C CT/TT genotype was observed (P_interaction = 0.051); the AORs (95% CI) were 0.98 (0.73–1.32) for non-diabetics carrying the CT/TT genotype, 1.73 (0.94–3.18) for diabetics carrying the CC genotype and 3.49 (2.22–5.49) for diabetics carrying the CT/TT genotype compared with non-diabetics carrying the CC genotype. Moreover, the SOD2 −1221G>A AA genotype carriers had a significantly increased risk for pancreatic cancer among those with a low dietary vitamin E intake but decreased risk among those with a high vitamin E intake (P_interaction = 0.002). There was a non-significant interaction between diabetes and GSTA4 Ex5−64G>A genotypes (P_interaction = 0.078). No significant interaction between genotype with cigarette smoking or vitamin C intake was observed. These data suggest that genetic variations in antioxidant defenses modify the risk of pancreatic cancer in diabetics or individuals with a low dietary vitamin E intake.

Introduction

Pancreatic cancer is the fourth leading cause of cancer-related death in the USA. Cigarette smoking, obesity and diabetes are the major modifiable risk factors for this disease (1). Increased oxidative stress and proinflammatory response may be the common mechanisms by which cigarette smoking, obesity and diabetes contribute to pancreatic cancer development (2–5). Antioxidant system includes enzymatic and non-enzymatic mechanisms that work together against oxidative stress (6). Dietary antioxidants (e.g. vitamin C, vitamin E and selenium) provide the primary non-enzymatic defense mechanisms. Serum levels of antioxidants have been positively associated with fruit and vegetable intake (7). An inverse association of dietary intake of fruits, vegetables and vitamin C with the risk of pancreatic cancer has been observed in previous studies (8,9).

The enzymatic antioxidant system mainly includes superoxide dismutase (SOD), catalase (CAT), glutathione peroxidase (GPX) and other relevant enzymes and proteins. Genetic variations in antioxidant pathways have been shown to modify the risk of cancers of the breast, prostate and lung (10–12), especially in presence of low intake of antioxidants (11). To date, there is only one study with limited sample size that has reported the association of antioxidant genes polymorphisms with pancreatic cancer (13). To fill this gap, we examined the single-nucleotide polymorphisms (SNPs) of several common antioxidant genes, SOD2, CAT, GPX and GSTA4, alone or in interaction with other risk factors in pancreatic cancer development.

SOD2 codes for mitochondrial manganese SOD (Mn-SOD), which catalyzes the dismutation of superoxides into water and hydrogen peroxide (6). CAT catalyzes the decomposition of hydrogen peroxide into water, thereby preventing cell damage from highly reactive oxygen free radicals (14). GPXs are selenoproteins that reduce organic peroxides and hydrogen peroxide through the coupled oxidation of glutathione. GSTA4 is a member of the glutathione S-transferase family with a substrate affinity to lipid peroxidation products. Expression of the SOD2, CAT and GPX genes has been detected in the human pancreas (15,16). Overexpression of SOD2 is known to suppress the cell growth or reverse the malignant phenotype of pancreatic cancer (17).

We hypothesized that polymorphic variations in these antioxidant genes modify the risk of pancreatic cancer, especially when oxidative stress is increased and non-enzymatic antioxidants are deficient.

Materials and methods

Study population

The study population is a subset of those enrolled in a case–control study of pancreatic cancer at The University of Texas M. D. Anderson Cancer Center from January 2000 to May 2007. The subjects with DNA sample, complete risk factors as well as dietary information were included in this study. Details of the study design and methods have been described previously (18). Briefly, case patients were diagnosed with a pathologically confirmed pancreatic ductal adenocarcinoma. Control subjects were cancer-free healthy individuals recruited from non-blood relatives (mainly spouses) and friends of patients with cancers other than gastrointestinal cancer or lung cancer. The response rate to recruitment was 83.9% for patients and 83.6% for control subjects. Patients and control subjects were frequency matched by age (±5 years), sex and race. Each participant provided written informed consent for an interview and blood sample. The Institutional Review Board of M. D. Anderson approved this study.

Data collection

Using a structured questionnaire in in-person interviews, trained personnel collected information on demographics, cigarette smoking, alcohol use, diabetes, medical history and family history of cancer. A self-administered food frequency questionnaire (FFQ) that was previously developed and validated by the Department of Nutrition at the Harvard School of Public Health (19) was used to collect dietary information about the period 1 year prior to the patients’ cancer diagnoses or control subjects’ recruitment to the study. The FFQ captured intake information ∼84 food items and contained 19 questions related to dietary habits and use of supplements. The FFQ return rate was 76.2% for control subjects and 59.3% for cancer patients. Those who did not return the FFQ tended to be minorities (both patients and control subjects), male control subjects who were <50 or >70 years old and patients with poor performance status or late-stage disease. Each participant in the current study had complete epidemiological and dietary information and a DNA sample, which represent 56% of the total study population. The dietary variables examined in the current study included total daily intake of vitamins C and E or supplements.

DNA extraction and genotyping

Peripheral blood mononuclear cells were collected from freshly drawn blood by Ficoll–Hypaque (Amersham Pharmacia Biotech, Piscataway, NJ) density gradient centrifugation and stored at −80°C. DNA was extracted with a FlexiGene DNA kit (QIAGEN, Valencia, CA) and a Maxwell 16 automated system (Promega, Madison, WI) and stored at 4°C for immediate use. The seven examined SNPs—rs1802061 (GSTA4 Ex5−64G>A, Q117Q), rs182623 (GSTA4 −2816T>A), rs316141 (GSTA4 IVS6+1214T>C), rs4880 (SOD2 Ex2+24T>C, V16A), rs2758346 (SOD2 −1221G>A), rs1001179 (CAT −369 C>T, aka −262 C>T) and rs1050450 (GPX1, EX1−226 C>T, P200L)—were genotyped using the TaqMan method and an ABI 7900HT Fast Real-time PCR System (ABI, Foster City, CA). These SNPs were selected based on their functional significance and previously reported associations with the risk of cancers (20–23). Each of the SNPs had a minor allele frequency of >10% among Caucasians. About 5% of the samples were genotyped in duplicate and 100% consistency was achieved.

Statistical analysis

For statistical analyses, we used SPSS 15.0 (SPSS, Chicago, IL) and Stata 9.0 (Stata, College Station, TX) software. All the tests were two tailed, and P < 0.05 was considered statistically significant. A chi-square test was used to examine the differences in risk factor and genotype distributions between the patients and control subjects, as well as the deviation of the genotype distribution from Hardy–Weinberg equilibrium (HWE) in control subjects. Genotype-specific risks were estimated as adjusted odds ratios (AORs) and 95% confidence intervals (CIs) by unconditional logistic regression models; adjustments for age (continuous), race (non-Hispanic white, Hispanic and black), sex, smoking (never or ever), alcohol (≤60 g/day, >60 g/day), diabetes (yes or no) and family history of cancer among first-degree relatives (yes or no). Body mass index was not included in the model because this information was not collected from individuals recruited before 2004. Individuals who were homozygous for the common alleles were used as the referent group. Carriers of heterozygous and homozygous mutations were combined in the analyses of interactions with exposure variables. Dietary intake of antioxidant nutrients were calorie adjusted using the residual method described by Willett et al. (24). After adjustment, we divided the dietary data into low versus high using the median of control value as the boundary arbitrarily. Because diabetes could be a manifestation of pancreatic cancer, individuals with recent diabetes onset (i.e. within 2 years before the patients’ cancer diagnoses and before control subjects’ recruitment to the study) were excluded from the analyses of interaction between genes and diabetes.

To detect possible interactions between genotype and smoking, genotype and diabetes or genotype and antioxidant intake, individuals without the risk factor (e.g. non-diabetic) and without the at-risk genotype were used as the reference group, and AOR for non-diabetics with the at-risk genotype (OR₁₀), diabetics with the referent genotype (OR₀₁) and diabetics with the at-risk genotype (OR₁₁) were estimated using unconditional logistic regression. Interaction contrast ratios (ICRs), calculated as OR₁₁ − OR₁₀ − OR₀₁ + 1, were used to measure the interaction effect of genotype and risk factor (25). ICR > 0 indicates a more than additive effect (positive interaction), ICR = 0 indicates a simple additive effect (no interaction) and ICR < 0 indicates a less than additive effect (negative interaction). ICR > 0 and 95% CI excluding 0 was considered as statistically significant. Similarly, OR₁₁ > OR₁₀ × OR₀₁ − 1 indicates a more than multiplicative effect (25). The cross-product term of the genotype and the variable of interest was generated with logistic regression models. The significance of the interaction term (P_interaction) was tested using a likelihood ratio test, with the full model containing the interaction term, the main effect of the genotype and the exposure variable and the reduced model lacking the interaction term. The Bonferroni correction was applied to control false discover rate related to multiple comparisons.

Results

Demographics and risk factors

The demographics and risk factors for pancreatic cancer in this study population are summarized in Table I. We found no significant difference in demographic distribution between patients and control subjects. Cigarette smoking, heavy alcohol consumption, history of diabetes and family history of cancer were significantly associated with increased risk of pancreatic cancer, as previously reported (18). When the median intake value of controls was used as the cut point, a higher intake of total vitamin C (both dietary and supplemental) or dietary E (including food fortification) was associated with significantly reduced risk of pancreatic cancer. The median daily intakes of total vitamin C and vitamin E (dietary and supplement) were 163.10 and 6.97 mg for patients and 181.41 and 7.21 mg for control subjects, respectively.

Table I.

Distribution of selected variables among patient and control subjects

Variables	Case (N = 575) n (%)	Control (N = 648) n (%)	P (χ2)	OR (95% CI)a
Sex
Female	229 (39.8)	246 (38.0)	0.445
Male	346 (60.2)	402 (62.0)
Race
White	523 (91.0)	600 (92.6)	0.238
Hispanic	29 (5.0)	33 (5.1)
Black	23 (4.0)	15 (2.3)
Age
<50	68 (11.8)	88 (13.6)	0.448
51–60	151 (26.3)	183 (28.2)
61–70	226 (39.3)	251 (38.7)
>70	130 (22.6)	126 (19.4)
Education
Bachelor's degree or less	457 (79.5)	522 (80.6)	0.638	1.00
Higher than bachelor's degree	118 (20.5)	126 (19.4)		1.20 (0.89–1.62)
History of diabetes
No	437 (76.0)	582 (89.8)	0.000	1.00
Yes	138 (24.0)	66 (10.2)		2.78 (2.00–3.86)
Cigarette smoking
Never	245 (42.6)	323 (49.9)	0.011	1.00
Ever	330 (57.4)	325 (50.2)		1.33 (1.05–1.70)
Alcohol (g/day)
≤60	509 (88.5)	603 (93.1)	0.006	1.00
>60	66 (11.5)	45 (6.9)		1.65 (1.09–2.50)
Cancer history in first-degree relativesb
No	198 (34.6)	291 (45.2)	0.000	1.00
Yes	375 (65.5)	353 (54.8)		1.57 (1.24–2.00)
Total vitamin C (mg/day)
≥169.77	245 (42.6)	288 (49.9)	0.013	1.00
<169.77	330 (57.4)	289 (50.1)		1.37 (1.07–1.75)
Dietary vitamin C (mg/day)
≥110.39	308 (53.6)	288 (49.9)	0.215	1.00
<110.39	267 (46.4)	289 (50.1)		0.88 (0.69–1.12)
Total vitamin E (mg/day)
≥19.01	278 (48.4)	287 (49.8)	0.616	1.00
<19.01	297 (51.7)	289 (50.2)		0.91 (0.71–1.16)
Dietary vitamin Ec (mg/day)
≥7.07	246 (42.8)	289 (50.0)	0.013	1.00
<7.07	329 (57.2)	288 (49.9)		1.45 (1.14–1.86)

^aOdds ratios was adjusted for sex, age, race, education, cigarette smoking, alcohol, history of diabetes and cancer history in first-degree relatives.

^bInformation was missing from two cases and four controls because of adopted family.

^cInclude vitamin E from food fortification.

Genotype distribution

Except GSTA4 Ex5−64G>A (P = 0.027) and GSTA4 −2816T>A (P = 0.006), genotype distributions were in HWE in controls. GSTA4 Ex5−64G>A failed to follow HWE because the extremely low frequency of the homozygous mutant. GSTA4 −2816T>A followed HWE among cases but did not among controls; perhaps due to some unidentified experimental errors. Genotype distributions were comparable between patients and control subjects for all the SNPs examined (Table II).

Table II.

Distribution of genotype frequencies

Genotype	Case n (%)	Control n (%)	P (χ2)	OR (95% CI)a
GSTA4 Ex5−64G>A
GG	460 (83.3)	470 (83.0)	0.895	1.00
AG	89 (16.1)	96 (17.0)		0.97 (0.70–1.35)
AA	3 (0.5)	0 (0.0)		—
GSTA4 −2816T>A
AA	284 (50.6)	338 (54.6)	0.270	1.00
AT	225 (40.1)	220 (35.5)		1.16 (0.76–1.76)
TT	52 (9.3)	61 (9.9)		1.02 (0.68–1.52)
GSTA4 IVS6+1214T>C
CC	180 (33.1)	207 (34.4)	0.894	1.00
CT	272 (50.0)	294 (48.8)		1.00 (0.76–1.31)
TT	92 (16.9)	101 (16.8)		1.01 (0.71–1.45)
SOD2 Ex2+24T>C
TT	143 (25.6)	167 (26.2)	0.738	1.00
CT	278 (49.8)	309 (48.4)		1.12 (0.84–1.50)
CC	137 (24.6)	162 (25.4)		1.07 (0.77–1.50)
SOD2 −1221G>A
AA	120 (22.5)	131 (22.1)	0.710	1.00
GA	268 (50.2)	294 (49.6)		1.08 (0.79–1.48)
GG	146 (27.3)	168 (28.3)		1.00 (0.71–1.42)
CAT −329T>C
CC	349 (63.3)	366 (60.8)	0.599	1.00
CT	174 (31.6)	207 (34.4)		0.90 (0.69–1.17)
TT	28 (5.1)	29 (4.8)		1.03 (0.59–1.80)
GPX1 EX1−226C>T
CC	263 (47.6)	316 (51.3)	0.347	1.00
CT	240 (43.5)	242 (39.3)		1.15 (0.90–1.46)
TT	49 (8.9)	58 (9.4)		2.49 (0.47–13.10)

^aOdds ratio was adjusted for sex, age, race, education, cigarette smoking, alcohol, history of diabetes and cancer history in first-degree relatives.

Interaction between genotypes and risk factors

Next, we examined the possible interactions of genotypes with cigarette smoking or genotypes with diabetes. We did not detect any interaction of genotypes with smoking (data not shown). We observed a significant additive interaction of diabetes with the SOD2 Ex2+24T>C genotype with an ICR of 1.78 (95% CI: 0.16–3.39). Furthermore, this interaction was borderline significant on a multiplicative scale (P_interaction = 0.051) (Table III). When we excluded individuals with diabetes onset ≤2 years before their cancer diagnoses or recruitment to the study from the analyses, the interaction between the SOD2 Ex2+24T>C and diabetes diminished. The ICR was 1.22 (−0.20 to 2.64) and P_interaction = 0.111. Meanwhile, a non-significant interaction between GSTA4 Ex5−64G>A and diabetes was observed (P_interaction = 0.078). We did not detect a significant interaction of diabetes with CAT or GPX genotypes (Table III).

Table III.

Interaction between genotype and diabetes

Genotype	Diabetes	Case/control (n)	OR (95% CI)a	Pb	OR (95% CI)c	Pd
GSTA4 Ex5−64G>A
GG	No	348/419	1.00	0.469	1.00	0.071
AG/AA	No	71/88	0.93 (0.65–1.32)		0.93 (0.65–1.32)
GG	Yes	112/51	2.52 (1.74–3.66)		1.57 (0.98–2.49)
AG/AA	Yes	21/8	3.33 (1.43–7.74)		4.33 (1.37–13.71)
ICRe			0.88 (−2.03 to 3.78)		2.83 (−2.17 to 7.84)
GSTA4 −2816T>A
AA	No	221/305	1.00	0.431	1.00	0.170
AT/TT	No	204/253	1.06 (0.82–1.38)		1.06 (0.82–1.38)
AA	Yes	63/33	2.55 (1.60–4.07)		1.51 (0.85–2.66)
AT/TT	Yes	73/28	3.54 (2.19–5.73)		2.91 (1.57–5.39)
ICR			0.93 (−1.00 to 2.86)		1.34 (−0.57 to 3.24)
GSTA4 IVS6+1214T>C
CC/CT	No	341/450	1.00	0.308	1.00	0.302
TT	No	70/95	0.95 (0.67–1.35)		0.95 (0.67–1.34)
CC/CT	Yes	111/51	2.84 (1.96–4.12)		1.99 (1.25–3.18)
TT	Yes	22/6	4.58 (1.81–11.60)		3.39 (1.05–10.96)
ICR			1.78 (−2.54 to 6.11)		1.45 (−2.61 to 5.51)
SOD2 Ex2+24T>C
CC	No	106/138	1.00	0.051	1.00	0.099
CT/TT	No	316/436	0.98 (0.73–1.32)		0.98 (0.73–1.32)
CC	Yes	31/24	1.73 (0.94–3.18)		1.22 (0.59–2.50)
CT/TT	Yes	105/40	3.49 (2.22–5.49)		2.42 (1.40–4.17)
ICR			1.78 (0.16–3.39)		1.22 (−0.20 to 2.64)
SOD2 −1221G>A
GG	No	111/150	1.00	0.542	1.00	0.469
GA/AA	No	288/383	1.02 (0.76–1.36)		1.02 (0.76–1.37)
GG	Yes	35/18	2.55 (1.36–4.80)		1.59 (0.73–3.50)
GA/AA	Yes	100/42	3.27 (2.09–5.13)		2.35 (1.37–4.03)
ICR			0.70 (−1.20 to 2.61)		0.74 (−0.89 to 2.37)
CAT −329T>C
CT/TT	No	156/212	1.00	0.505	1.00	0.638
CC	No	259/330	1.06 (0.81–1.38)		1.05 (0.81–1.38)
CT/TT	Yes	46/24	2.55 (1.48–4.39)		1.90 (0.97–3.72)
CC	Yes	90/36	3.40 (2.16–5.35)		2.22 (1.28–3.87)
ICR			0.80 (−1.07 to 2.66)		0.27 (−1.39 to 1.94)
GPX1 EX1−226C>T
CT/TT	No	224/267	1.00	0.219	1.00	0.625
CC	No	192/288	0.81 (0.63–1.05)		0.81 (0.63–1.05)
CT/TT	Yes	65/33	2.41 (1.53–3.81)		1.79 (1.03–3.12)
CC	Yes	71/28	2.98 (1.81–4.89)		1.82 (0.97–3.43)
ICR			0.75 (−0.95 to 2.46)		0.22 (−1.23 to 1.68)

^aOdds ratio was adjusted for sex, age, race, education, cigarette smoking, alcohol and cancer history in first-degree relatives.

^bLikelihood ratio test for genotype and diabetes without excluding individuals with diabetes duration ≤2 years.

^cIndividuals with diabetes duration ≤2 years were excluded from the logistic regression model.

^dLikelihood ratio test for genotype and diabetes excluding individuals with diabetes duration ≤2 years.

^eInteraction contrast ratio calculated as OR₁₁ − OR₁₀ − OR₀₁ + 1.

Interaction of genotypes and dietary antioxidants

Next, we examined the genotype associations with the risk of pancreatic cancer among individuals with different levels of vitamin C and vitamin E intake. We observed a significant interaction of SOD2 −1221G>A with dietary vitamin E intake (P_interaction = 0.002, Table IV). This association remained statistically significant after Bonferroni correction (P < 0.007). We also observed a borderline significant interaction of dietary vitamin E intake with GSTA4 IVS6+1214T>C (P = 0.069) and SOD2 Ex2+24T>C (P = 0.080) genotype (Table IV). In subgroup analysis (Table V), we observed that the TT genotype of GSTA4 IVS6+1214T>C was associated with decreased risk of pancreatic cancer among individuals with a lower intake of dietary vitamin C or vitamin E but increased risk among those with a higher intake of those dietary antioxidants (P_interaction = 0.062 for the interaction with vitamin C). Furthermore, the AG and AA genotype in SOD2 −1221G>A was associated with a significantly increased risk of pancreatic cancer among individuals with a low dietary vitamin E intake with AOR (95% CI) of 1.86 (1.21–2.86) and 1.98 (1.17–3.36), respectively. In contrast, the AA genotype was associated with a significantly decreased risk among those with a high dietary vitamin E intake (AOR: 0.57, 95% CI: 0.35–0.95). No other genotypes were significantly associated with risk of pancreatic cancer in this analysis.

Table IV.

Interactions between genotype and dietary Vitamin E intake^a

Genotype	Vitamin Eb	Case (n)	Control (n)	OR (95% CI)c	Pd
GSTA4 Ex5−64G>A
GG	≥7.07	195	235	1.00	0.715
AG/AA	≥7.07	39	47	1.02 (0.63–1.64)
GG	<7.07	265	235	1.39 (1.07–1.82)
AG/AA	<7.07	53	49	1.26 (0.81–1.96)
GSTA4 −2816T>A
AA	≥7.07	117	161	1.00	0.164
AT/TT	≥7.07	125	121	1.38 (0.97–1.96)
AA	<7.07	172	150	1.59 (1.14–2.21)
AT/TT	<7.07	148	130	1.55 (1.10–2.19)
GSTA4 IVS6+1214T>C
CC/CT	≥7.07	199	238	1.00	0.069
TT	≥7.07	38	37	1.33 (0.81–2.20)
CC/CT	<7.07	253	213	1.48 (1.13–1.94)
TT	<7.07	54	58	1.07 (0.70–1.65)
SOD2 Ex2+24T>C
CC	≥7.07	56	82	1.00	0.080
CT/TT	≥7.07	182	202	1.36 (0.91–2.04)
CC	<7.07	81	61	1.97 (1.21–3.21)
CT/TT	<7.07	239	225	1.63 (1.10–2.42)
SOD2 −1221G>A
GG	≥7.07	73	64	1.00	0.002
GA/AA	≥7.07	158	202	0.70 (0.47–1.05)
GG	<7.07	73	93	0.71 (0.45–1.13)
GA/AA	<7.07	230	179	1.18 (0.79–1.77)
CAT −329T>C
CT/TT	≥7.07	89	108	1.00	0.954
CC	≥7.07	146	167	1.05 (0.73–1.51)
CT/TT	<7.07	113	103	1.37 (0.93–2.04)
CC	<7.07	203	167	1.46 (1.02–2.09)
GPX1 EX1−226C>T
CT/TT	≥7.07	135	138	1.00	0.160
CC	≥7.07	103	143	0.74 (0.52–1.05)
CT/TT	<7.07	168	151	1.14 (0.82–1.60)
CC	<7.07	149	129	1.19 (0.85–1.68)

^aVitamin E intake without supplements but including food fortification.

^bMedian value of control was used as the cut points.

^cOdds ratio was adjusted for sex, age, race, education, cigarette smoking, alcohol, history of diabetes and cancer history in first-degree relatives.

^dLikelihood ratio test for genotype and dietary vitamin E intake.

Table V.

Association of genotype and risk of pancreatic cancer in subgroups of vitamin C and E intake

	Case/control (n)	OR (95% CI)a	Case/control (n)	OR (95% CI)a	Case/control (n)	OR (95% CI)a	Case/control (n)	OR (95% CI)a
Genotype	Total vitamin C				Dietary vitamin C
	≤169.77 (mg/day)		>169.77 (mg/day)		≤110.39 (mg/day)		>110.39 (mg/day)
GSTA4 Ex5−64G>Ab
GG	242/220	1.00	218/250	1.00	224/249	1.00	236/221	1.00
AG/AA	47/47	0.86 (0.54–1.37)	45/49	1.07 (0.68–1.69)	45/45	1.09 (0.68–1.75)	47/51	0.91 (0.58–1.42)
GSTA4 −2816T>A
TT	32/35	1.00	23/29	1.00	29/36	1.00	26/28	1.00
AT	107/89	1.26 (0.70–2.26)	111/98	1.35 (0.72–2.56)	104/103	1.19 (0.66–2.16)	114/84	1.35 (0.72–2.53)
AA	155/140	1.22 (0.70–2.15)	134/171	0.99 (0.53–1.83)	138/151	1.23 (0.69–2.17)	151/160	0.95 (0.52–1.73)
GSTA4 IVS6+1214T>C
CC	85/88	1.00	95/99	1.00	79/87	1.00	101/100	1.00
CT	147/128	1.11 (0.74–1.67)	125/136	0.92 (0.63–1.35)	136/136	1.00 (0.66–1.51)	136/128	1.03 (0.70–1.50)
TT	51/40	1.34 (0.78–2.31)	41/55	0.75 (0.45–1.25)	44/58	0.75 (0.44–1.28)	48/37	1.33 (0.78–2.25)
SOD2 Ex2+24T>C
CC	71/67	1.00	66/76	1.00	60/74	1.00	77/69	1.00
CT	149/130	1.24 (0.80–1.91)	129/147	1.01 (0.67–1.54)	142/145	1.30 (0.83–2.02)	136/132	0.99 (0.65–1.50)
TT	73/73	1.07 (0.65–1.76)	70/77	1.04 (0.64–1.67)	69/77	1.24 (0.75–2.07)	74/73	0.92 (0.57–1.47)
SOD2 −1221G>A
GG	71/78	1.00	75/79	1.00	67/81	1.00	79/76	1.00
AG	144/120	1.25 (0.79–1.99)	124/143	0.99 (0.63–1.55)	134/134	1.40 (0.88–2.24)	134/129	0.87 (0.55–1.36)
AA	64/55	0.86 (0.51–1.44)	56/63	1.07 (0.65–1.76)	53/65	1.11 (0.66–1.88)	67/53	0.84 (0.51–1.38)
CAT −329T>C
CC	180/159	1.00	169/175	1.00	167/173	1.00	182/161	1.00
CT	91/89	0.93 (0.64–1.37)	83/96	0.93 (0.64–1.35)	81/97	0.93 (0.63–1.36)	93/88	0.92 (0.63–1.34)
TT	18/9	1.75 (0.74–4.15)	10/17	0.68 (0.29–1.57)	18/12	1.63 (0.73–3.61)	10/14	0.61 (0.25–1.46)
GPX1 EX1−226C>Tc
CC	142/130	1.00	110/142	1.00	115/145	1.00	137/127	1.00
CT/TT	146/134	0.98 (0.69–1.40)	157/155	1.28 (0.91–1.81)	151/144	1.27 (0.89–1.81)	152/145	0.99 (0.70–1.40)
Genotype	Total vitamin E				Dietary vitamin E
	≤19.01 (mg/day)		>19.01 (mg/day)		≤7.07 (mg/day)		>7.07 (mg/day)
GSTA4 Ex5−64G>Ab
GG	232/228	1.00	228/242	1.00	249/217	1.00	211/253	1.00
AG/AA	48/45	1.06 (0.66–1.69)	44/51	0.90 (0.57–1.42)	49/46	0.84 (0.53–1.34)	43/50	1.09 (0.69–1.73)
GSTA4 −2816T>A
TT	32/35	1.00	23/29	1.00	31/31	1.00	24/33	1.00
AT	106/84	1.40 (0.78–2.51)	112/103	1.18 (0.62–2.23)	106/92	1.05 (0.57–1.95)	112/95	1.42 (0.77–2.62)
AA	150/152	1.09 (0.63–1.91)	139/159	1.07 (0.57–1.99)	164/136	1.24 (0.69–2.23)	125/175	0.91 (0.50–1.65)
GSTA4 IVS6+1214T>C
CC	89/95	1.00	91/92	1.00	95/83	1.00	85/104	1.00
CT	140/120	1.20 (0.80–1.78)	132/144	0.86 (0.58–1.27)	141/112	1.08 (0.71–1.62)	131/152	0.94 (0.64–1.38)
TT	47/47	1.02 (0.60–1.73)	45/48	0.92 (0.55–1.55)	52/55	0.74 (0.45–1.25)	40/40	1.28 (0.74–2.20)
SOD2 Ex2+24T>C
CC	72/69	1.00	65/74	1.00	76/58	1.00	61/85	1.00
CT	133/134	1.12 (0.73–1.73)	145/143	1.11 (0.73–1.70)	154/126	0.99 (0.64–1.53)	124/151	1.21 (0.80–1.85)
TT	80/74	1.12 (0.69–1.82)	63/76	0.98 (0.60–1.61)	70/81	0.67 (0.41–1.10)	73/69	1.57 (0.97–2.54)
SOD2 −1221G>A
GG	78/77	1.00	68/80	1.00	65/87	1.00	81/70	1.00
AG	133/125	1.17 (0.77–1.79)	135/138	1.13 (0.75–1.72)	151/119	1.86 (1.21–2.86)	117/144	0.71 (0.47–1.08)
AA	60/56	0.98 (0.59–1.63)	60/62	1.09 (0.66–1.79)	68/46	1.98 (1.17–3.36)	52/72	0.57 (0.35–0.95)
CAT −329T>C
CC	180/162	1.00	169/172	1.00	189/157	1.00	160/177	1.00
CT	85/88	0.88 (0.60–1.30)	89/97	0.97 (0.67–1.40)	89/82	0.92 (0.62–1.36)	85/103	0.93 (0.65–1.35)
TT	15/10	1.32 (0.56–3.12)	13/16	0.86 (0.39–1.91)	18/10	1.95 (0.83–4.55)	10/16	0.65 (0.28–1.52)
GPX1 EX1−226C>Tc
CC	137/139	1.00	115/133	1.00	142/122	1.00	110/150	1.00
CT/TT	146/131	1.14 (0.81–1.63)	157/158	1.10 (0.78–1.56)	155/137	0.91 (0.64–1.30)	148/152	1.39 (0.98–1.97)

^aOdds ratio was adjusted for sex, age, race, education, cigarette smoking, alcohol, history of diabetes and cancer history in first-degree relatives.

^bThree cases with AA genotype were combined with those with AT genotype.

^cFive cases and two controls had TT genotype.

Discussion

The major findings of this study indicate that specific antioxidant gene SOD2 and GSTA4 polymorphisms were associated with an increased risk of pancreatic cancer among diabetics or individuals with low dietary vitamin E intake. These findings support the hypotheses that genetic variations in antioxidant defense modify the risk of pancreatic cancer among individuals with increased oxidative stress or a low intake of dietary antioxidants.

SOD2 (Mn-SOD) is perhaps the most extensively studied antioxidant gene in the cancer field and the results are inconsistent. The functional SNP SOD2 Val¹⁶Ala (Ex2+24T>C), which confers higher than normal antioxidant activity (26,27), has been associated with a reduced risk of lung cancer (12), bladder cancer (28) and pancreatic cancer (13) but an increased risk of breast cancer (10) and prostate cancer (11). Null associations have been reported in other studies of breast cancer and bladder cancer (29–31). The current study failed to confirm the previously reported association of SOD2 Val¹⁶Ala genotype and increased risk of pancreatic cancer that was observed in a previous study of 122 cases and 331 controls (13). This discrepancy may have been resulted from a selection bias induced by small sample size because the allele frequency was much lower in the control group of the previous study compared with the reported frequency in SNP500 and HapMap databases and in the current study. Nevertheless, the current study showed that diabetics carrying the Val¹⁶ allele (the low activity allele) had a 3.49-fold increased risk of pancreatic cancer compared with non-diabetics carrying the Ala homozygote. These observations are consistent with previous reports that the Val allele was associated with an increased risk of pancreatic cancer (13) and type II diabetes (32). In addition to its antioxidant activity, SOD2 has been suggested as a tumor suppressor gene in pancreatic cancer because of its growth-regulating activity (33). Like many other solid tumors, pancreatic tumors have been demonstrated to have low levels of antioxidant enzymes compared with normal human pancreas (15). In pancreatic carcinoma cell lines, SOD2 levels were associated with cell doubling time and SOD2 overexpression inhibited the growth and reversed the oncogenic phenotypes of pancreatic cancer cells (16,33). Our observation that the low activity allele of the SOD2 Ex2+24T>C genotype was associated with a higher risk for pancreatic cancer is consistent with the gene's tumor suppressor role.

The SOD2 −1221G>A sequence variant is located in the regulatory region of the gene. Although no functional study has been conducted on this SNP, it could affect the gene's transcription factor binding. We found the A allele of the SOD2 −1221G>A genotype was associated with a higher risk of pancreatic cancer among individuals with a low dietary vitamin E intake but a lower risk among those with a higher vitamin E intake (P_interaction = 0.002). SOD2 catalyzes the conversion of superoxide anion to oxygen and hydrogen peroxide (H₂O₂). H₂O₂ is an important intermediate product in reactive oxygen species metabolism. It can be either reduced to H₂O by CAT or GPx, or it can go through Fenton reaction to generate more reactive hydroxyl free radicals that initiate the lipid peroxidation chain reactions (34). Vitamin E is one of the most efficient antioxidants to break this chain reaction (35). Thus, under vitamin E sufficient conditions, a higher level of SOD2 would protect cells from the superoxide anion-caused damages without the risk of increasing hydroxyl radicals. In contrast, under vitamin E deficient conditions, a higher level of SOD2 may lead to a higher level of H₂O₂ and subsequently a higher level of hydroxyl free radical and lipid peroxidation reactions. Further investigation is required to confirm this observation and to illustrate the functional significance of this SNP under different levels of antioxidant nutrients.

GST gene polymorphisms have been studied extensively in association with the risk of human cancers, including pancreatic cancer. For example, the GSTT1 null genotype in smokers and GSTP1 SNPs in younger individuals have been associated with an increased risk of pancreatic cancer (36,37). GSTA4 genotypes have previously been investigated in liver cancer (38) and lung cancer (39) but not in pancreatic cancer. The current study found that GSTA4 Ex5−64G>A (Q117Q) genotypes did not modify the risk of pancreatic cancer in general but did do so in diabetics. Although the functional significance of the SNPs examined is not well known, their association with an increased risk of pancreatic cancer is biologically plausible. GSTA4 has a substrate affinity to lipid peroxidation products (40,41); thus, it plays a major role in protecting cell membranes from lipid radicals. Vitamin E, α-tocopherol especially, has been claimed the most important lipid-soluble antioxidant, and it is known to protect cell membranes from oxidation by reacting with lipid radicals produced in the lipid peroxidation chain reaction (35). If the variant alleles of the GSTA4 gene reduced the enzyme's activity, the combined enzymatic antioxidant deficiency and increased oxidative stress in diabetes would render the cells subject to excessive lipid peroxidation-induced damages. Furthermore, a differential effect of the GSTA4 IVS6+1214T>C genotype by dietary vitamin C and E intake was observed. The variant genotype unexpectedly showed a protective effect among individuals with a lower intake of dietary antioxidants but a detrimental effect among those with a higher intake of dietary antioxidants. We do not have a biological explanation for this intriguing observation. Since these associations were statistically non-significant, they could be spurious associations related to small sample size. However, the consistent observations in genotype interaction with both vitamin C and vitamin E intake suggest that further investigation on the functional significance of this SNP and the gene expression regulation at different levels of antioxidant is needed.

Accumulating evidence supports the role of long-term type II diabetes in pancreatic cancer development. Diabetes mellitus is a metabolic disorder characterized by hyperglycemia, which often leads to abnormally high levels of superoxide anion radicals and high oxidative stress (2,42). The inflammatory response and insulin resistance associated with diabetes may also contribute to pancreatic cancer development. Previous studies have shown that antioxidants such as vitamin C and vitamin E can improve insulin resistance, reduce the risk of diabetes in overweight individuals and reduce the risk of heart attack in diabetics (43,44). Consistent with these observations, the current study demonstrated a protective effect of total vitamin C intake and dietary vitamin E against pancreatic cancer, which suggests that diabetes and pancreatic cancer may share a common mechanism, e.g. increased oxidative stress.

Our study is a hospital-based case–control study with inherent limitations. For example, the low return rate of the FFQ and the patients’ referral pattern led to selection bias, i.e. underrepresentation of minorities, older people and patients with most advanced diseases. However, there was no significant difference in the race and age distribution between cases and controls in this study. Furthermore, we did not find any association between the nutrient intake level and disease stage (data not shown). Thus, it is unlikely that the risk estimates in this study were significantly affected by these factors. Some patients may have changed their dietary pattern 1 year before their cancer diagnosis because of early symptoms, so the dietary information collected was subject to information bias and recall bias. Some of the observations of genotype association with risk of pancreatic cancer may be derived by chance alone. Because of the limited sample size, we could not examine the association of genotype with extremely low dietary antioxidant intake. The fact that diabetes is easily diagnosed among patients than among control subjects may lead to overestimation of diabetes risk on pancreatic cancer. Nevertheless, our study has generated some very interesting data that are consistent with previous reports that genetic variations in antioxidant genes could reduce the risk of pancreatic cancer in the presence of diabetes or low antioxidant intake.

Funding

National Institutes of Health (RO1 grant CA98380); National Institute of Environmental Health Sciences Center (grant P30 ES07784).

Glossary

Abbreviations

AOR

adjusted odds ratio

CAT

catalase

CI

confidence interval

FFQ

food frequency questionnaire

GPX

glutathione peroxidase

GSTA4

glutathione S-transferase alpha 4

HWE

Hardy–Weinberg equilibrium

ICR

interaction contrast ratio

SNP

single-nucleotide polymorphism

SOD

superoxide dismutase