Effect of Insulin-Like Growth Factor Gene Polymorphisms Alone or In Interaction with Diabetes on the Risk of Pancreatic Cancer

Pancreatic cancer is the fourth leading cause of death from cancer and causes >33,000 deaths per year in the United States (1). The known and suspected risk factors for pancreatic cancer include cigarette smoking, type II diabetes, obesity, family history of pancreatic cancer, and diet (2). Germ-line mutations that are associated with several hereditary syndromes account for <10% of the pancreatic cancer burden. Few studies have examined the role of polymorphic variations of carcinogen metabolic genes (3, 4) and DNA repair genes (5-7), but the genetic susceptibility factors in the majority of sporadic pancreatic cancer cases are not well defined.

Accumulating evidence suggests that diet and related factors, such as physical activity and body size, may influence cancer risk through their effects on the serum concentration of insulin-like growth factor-I (IGF-I) and its binding proteins (8). IGF1 and IGF2 are structurally similar growth hormone-regulated polypeptides involved in both human development and the maintenance of normal function and homeostasis in most human cells. In addition to their classic role as endocrine hormones, IGFs regulate a wide range of biological functions, such as cell proliferation, differentiation, and apoptosis, through paracrine and autocrine mechanisms (9). Also, the IGF1 receptor-mediated initiation of signal transduction activates important intracellular signal pathways, including the Ras/Raf/mitogen-activated protein kinase and phosphoinositide 3-kinase pathways (10). IGF1 and IGF1 receptor are highly expressed in pancreatic cancer cells (11), and in pancreatic cancer cell lines, IGF1-mediated signaling transduction leads to increased proliferation and invasion, expression of angiogenesis mediators, and decreased apoptosis (12-14). IGF2 is overexpressed and its imprinting is disrupted in many tumors (15, 16). Furthermore, in vitro and in vivo studies have suggested that the overexpression of IGF2 results in a more malignant phenotype and increased tumor formation in nude mice (17).

Although strong experimental evidence suggests that IGFs play a role in carcinogenesis, the results of epidemiologic investigations are less persuasive. For example, polymorphic variants of the IGF1 gene and elevated serum levels of IGF1 protein have been associated with an increased risk of common cancers like prostate, colorectal, and breast cancers in some studies but not in others (18-24). Information on IGF2 polymorphisms and their association with cancer risk is scarce, and the results that have been published are similarly inconsistent (25, 26). In addition, no study we are aware of has yet examined the association between IGF gene polymorphisms and risk of pancreatic cancer. When cohort studies were conducted to examine the association between pancreatic cancer risk and prediagnostic plasma levels of IGF1 and IGF2, no association was observed (27, 28). However, in one of two case control studies we identified in our review of the literature, investigators found a slightly increased risk of pancreatic cancer associated with elevated plasma IGF1 in a limited number of samples (29, 30).

To fill in the gap between IGF and pancreatic cancer, we tested the hypothesis that polymorphic variants of the IGF1 and IGF2 genes alone or through an interaction with diabetes modify the risk of pancreatic cancer. To do so, we examined the associations between pancreatic cancer risk and selected six single nucleotide polymorphisms (SNP) of the IGF1 and IGF2 genes in 892 patients with pancreatic adenocarcinoma and in 793 healthy controls. Our results show for the first time that the IGF1 gene alone or in concert with diabetes to increase the risk of pancreatic cancer.

The study population and design of this hospital-based case control study have been described previously in detail (31). Briefly, consecutive patients with pathologically confirmed primary pancreatic ductal adenocarcinoma and controls were recruited at The University of Texas M. D. Anderson Cancer Center from 2000 to 2007. Controls were the healthy spouses, friends, or nonblood relatives of patients with various nongastrointestinal and non-smoking-related cancers. Controls were frequency matched to cases by age at enrollment (±5 years), sex, and race. Smoking and alcohol consumption history, medical history including diabetes, family history of cancer, and information on other risk factors were collected by personal interview. Body mass index (BMI; kg/m²) data were collected from study participants recruited in 2004 and later. All study subjects were U.S. residents and were able to communicate in English. Written informed consent was obtained from each study participant for the interviews and the collection of a blood sample. The study was approved by the Institutional Review Board of The University of Texas M. D. Anderson Cancer Center. A total of 1,318 cases were consented from the 1,635 cases approached with a response rate of 80.6% and 969 controls were recruited from 1,260 individuals approached with a response rate of 76.9%. Eighty-nine cases and 167 controls were excluded from the current study because of missing or incomplete risk factor information; 171 cases and 19 controls were excluded because of missing or inadequate DNA samples. In addition, 166 cases were excluded because of final diagnosis of disease other than pancreatic adenocarcinoma. Therefore, the total number of cases and controls involved in the current study is 892 and 783, respectively.

Peripheral blood mononuclear cells were collected from freshly drawn blood by Ficoll-Hypaque (Amersham Pharmacia Biotech) density gradient centrifugation and stored at −80°C. DNA was extracted with the use of a FlexiGene DNA kit (Qiagen) and a Maxwell16 automated system (Promega) and stored at 4°C for immediate use. DNA concentration was determined by using a NanoDrop ND1000 spectrophotometer (NanoDrop).

The IGF SNPs were selected from the National Cancer Institute SNP500Cancer database. Neither of the two genes has nonsynonymous SNPs listed in this database. We selected SNPs from the 3′-untranslated region (UTR) and the intron regions with minor allele frequency greater than 5%. Genotypes were determined using the TaqMan diallelic discrimination method (32). Details on the six SNPs examined in this study are given in Table 1. Probes and oligonucleotides were obtained from Applied Biosystems using the Assay-by-Design product. The reactions were prepared by using 2× TaqMan Universal Master Mix, 40× SNP Genotyping Assay Mix, DNase-free water, and 10 ng genomic DNA in a final volume of 5 μL/reaction. PCR amplification was completed using the ABI Prism 7900 HT sequence detector under the following conditions: 10 min at 95°C enzyme activation followed by 45 cycles at 92°C for 15 s and 60°C for 1 min (annealing/extension). About 5% of the samples were analyzed in duplicate and 100% consistency was achieved.

The distribution of categorical variables and genotype frequencies between cases and controls was tested by using the χ² test. The deviation of genotype distribution from the Hardy-Weinberg equilibrium was tested by using the goodness-of-fit χ² test to compare the observed genotype frequencies with the expected genotype frequencies with 1 df. Fisher's exact test was applied when any of the comparison groups had <5 subjects. Risk of pancreatic cancer was estimated as odds ratios (OR) and 95% confidence intervals (95% CI) calculated using unconditional logistic regression analysis.

The statistical models on epidemiologic factors were adjusted for gender, race (non-Hispanic White, Hispanic, African American, Asian), age (<50, 51-60, 61-70, >70 years), smoking status (nonsmoker, ≤20, >20 pack-years), diabetes (yes or no), and family history of cancer among first-degree relatives (yes or no), whenever appropriate. Genotype evaluations were restricted to non-Hispanic Whites because of the small sample size in the other groups and the known racial differences in genotype distribution. The association between genotype and risk of pancreatic cancer was adjusted for smoking and diabetes status and family history of cancer and, in some analyses, was further adjusted for duration of diabetes and BMI.

To detect possible interactions of specific genotypes and diabetes, ORs were estimated using unconditional logistic regression for the following groups, with nondiabetics with the wild-type genotype as the reference group: nondiabetics with the at-risk genotype (OR₁₀), diabetics with the wild-type genotype (OR₀₁), and diabetics with the variant genotype (OR₁₁). An OR₁₁ greater than the sum of OR₁₀ + OR₀₁ or greater than the product of OR₁₀ X OR₀₁ indicates a more than additive or more than multiplicative effect, respectively. To control for multiple testing, the multiplicity-adjusted P value was calculated using the Bonferroni method (33). For any statistically significant association we observed, we also estimated the false-positive report probability using the methods described by Wacholder et al. (34). We considered that a prior probability of 25% might be appropriate when there is biological plausibility and availability of previous epidemiologic evidence for such an association. The false-positive report probability value for noteworthiness was set as 0.2.

Haplotype was inferred from the genotyping data and the frequencies were compared between cases and controls using Haploview 4.1 software. All other statistical analyses were done using SPSS 14.0 and Stata 9.0 software. P values < 0.05 were considered statistically significant.

Table 2 shows the demographics and potential risk factors in patients and controls. The mean ± SD ages of cases and controls were 61.9 ± 10.2 and 61.0 ± 10.0 years, respectively (P = 0.85). Cases and controls were well matched by sex and age, but the study enrolled slightly fewer controls in all groups, except non-Hispanic Whites. As reported previously (31, 35), smoking status, diabetes status, BMI, and family history of cancer were all associated with a significantly increased risk of pancreatic cancer.

The genotype frequencies of IGF1 and IGF2 are presented in Table 3. The six SNPs were successfully amplified in 97.6% to 100% of the patients and controls. The distributions of these genotypes were all in agreement with Hardy-Weinberg equilibrium (P = 0.556-1.0). All three SNPs of each gene were in linkage disequilibrium with ∣D′∣ > 0.92. None of the IGF1 genotypes was significantly associated with pancreatic cancer. Four haplotypes of the IGF1 gene were inferred from the genotyping data. A haplotype containing the 16540 G, 1830 T, and -177 G alleles had a significantly lower frequency in cases (0.27) than in controls (0.41; P = 0.039). Although the number is small, a significant difference was observed between the IGF2 -233 C>T TT genotype distributions of cases and controls. Eleven controls and only 1 case had the homozygous mutant TT genotype, which translates into a 93% reduction of pancreatic cancer risk (P = 0.013). The multiplicity-adjusted P = 0.078. The false-positive report probability is 0.152 (P = 0.01), which indicates that the chance of this result being false-positive is small. Similarly, another IGF2 SNP, -69 TT, also showed a protective effect against pancreatic cancer, although this result was at borderline significant (P = 0.051). None of the three major haplotypes of the IGF2 gene was associated with pancreatic cancer. To avoid the reversal causation problem caused by pancreatic caner-associated diabetes, we performed risk analysis after excluding individuals with <2 years of diabetes duration. However, no significant change in the risk estimates was observed in these analyses compared with the results from the analyses of all study participants (Table 3).

To investigate the influence of genotype on the associations between both diabetes and BMI with risk of pancreatic cancer, we compared genotype distributions in subgroup analyses. Because the homozygous mutations occurred so infrequently in the study subgroups, they were combined with the heterozygous variants in this analysis. As shown in Table 4, a significant difference between the distributions of IGF1 -177 GC/CC genotype in cases and controls was observed among diabetics but not among nondiabetics. The GC/CC genotype frequencies in cases and controls were 19.7% and 18.8% among nondiabetics and 25.0% and 11% among diabetics (P = 0.013, χ² test). The IGF1 -177 GC/CC genotype frequencies tended to be lower among diabetic controls but higher among diabetic cases compared with the nondiabetic counterparts, but none of the differences was statistically significant (Table 4). The IGF1 -177 GC/CC genotype frequencies tended to be higher among both cases (24.1%) and controls (21.0%) with BMI > 25 kg/m² compared with cases (17.7%) and controls (16.1%) with BMI ≤ 25 kg/m², but the differences were not statistically significant. The other two IGF1 genotypes and all IGF2 genotypes were not distributed differently between cases and controls by BMI (data not shown).

Next, we used 2 × 4 tables and conducted logistic regression analyses to examine the interaction between the IGF1 genotype and diabetes and how it modifies the risk of pancreatic cancer. Using the nondiabetics with the wild-type genotype as the reference group, we observed a more than additive effect of the IGF1 -177 C allele and diabetes on the risk of pancreatic cancer. When compared with subgroups with neither the C allele nor diabetes, the adjusted OR (95% CI) for individuals with the C allele alone, diabetes alone, and both factors were 1.07 (0.81-1.42), 2.12 (1.53-2.93), and 5.69 (2.63-12.3), respectively. When the same analysis was done on data from individuals who had had diabetes for >2 years, the OR for individuals with diabetes alone dropped from 2.12 to 1.38, which indicates that the association of diabetes and pancreatic cancer was overestimated when all study subjects were included in the analysis. However, the OR remained at 5.50 for individuals with both diabetes and the IGF1 -177 GC/CC genotypes. Further adjusting for BMI reduced the magnitude of the ORs, but all ORs for individuals with diabetes and carrying the -177 C allele remained statistically significant (Table 5). Assuming the OR for individuals with both diabetes and the IGF variant allele was 3.5, the false-positive report probability for the observed ORs were <5%.

In this large case-control study, we found an IGF1 haplotype and the IGF2 Ex4 -233 C>T TT genotype was significantly associated with reduced risk of pancreatic cancer. We also observed a synergistic effect of the IGF1 Ex4 -177 G>C C allele and diabetes on the risk of pancreatic cancer. These observations support the hypothesis that polymorphic variants of the IGF genes modulate pancreatic carcinogenesis.

The relationship between diabetes and pancreatic cancer has been controversial (36). Although diabetes frequently occurs as a consequence of pancreatic cancer (37), increasing epidemiologic evidence supports the idea that long-standing diabetes has a role in causing pancreatic cancer (38, 39). Overt type II diabetes is usually characterized by hyperglycemia, insulin resistance, and hyperinsulinemia. Although insulin does not have direct carcinogenic effects on the pancreas, it may function in an integrated fashion with IGFs to promote tumor development (9). Specifically, it is known that >90% of the circulating IGF1 is bound to IGF-binding protein-3 and insulin affects IGF1 bioavailability by regulating the growth hormone receptor and reducing the level of IGF-binding proteins. Considering the role that insulin and IGF1 play in cell growth promotion, it is conceivable that individuals with a higher level of IGF1 activity as conferred either by genetic variation or by hyperinsulinemia would have much greater chance of tumor development.

Among the three IGF1 SNPs examined in this study, two are located in the 3′-UTR region of exon 4 of the gene. Although the 3′-UTR sequences do not translate into proteins, they may contain sequence motifs crucial for the regulation of transcription, mRNA stability, and cellular location of the mRNA or the binding of microRNA (40). Due to the small number of the homozygous mutant allele carriers in the current study, we could not directly explore the genotype and phenotype association. However, two previous studies have shown that the homozygous mutant allele of one of the two 3′ UTR SNPs (rs6220, IGF1 Ex4 +1830 C>T) was associated with elevated serum IGF1 level (41). Also, our finding that the SNP associated with an increased risk of pancreatic cancer among patients with diabetes (IGF1 Ex4 -177 G>C, rs5742714) is in linkage disequilibrium (D′ = 0.96) with SNP rs6220 indicates that the increased risk of pancreatic cancer associated with this SNP is most likely caused by the higher level of IGF1 activity conferred by the variant allele. The association of pancreatic cancer and IGF1 -177 GC/CC genotype seemed to be related to a lower frequency of diabetes in controls and a higher frequency of diabetes in cases among the genotype carriers. The mechanism underlying this differential association is not understood. It could be related to the different cellular functions that IGF1 plays in normal versus cancer cells (9, 10).

The current study has also observed a possible protective effect of the IGF2 gene 3′-UTR -233 C>T TT homozygous mutant against pancreatic cancer. However, the frequency of the variant genotype was extremely low (<2%) and the P value for the risk association was not statistically significant after adjusting for multiple test, so the possibility that this observation was made by chance alone cannot be excluded. Compared with levels of IGF1, high concentrations of IGF2 were present in circulation. IGF2 is involved in cell growth and development and is assumed to act in an autocrine manner, primarily via IGF1 receptor, where IGF2R functions as a “scavenger” receptor. IGF2 is overexpressed, and its imprinting is disrupted in many primary tumors and cancer cell lines (15, 16). Furthermore, in nude mice, the overexpression of IGF2 has been shown to promote more malignant tumor phenotypes and results in more efficient tumor formation (17). We presume that the reduced risk of pancreatic cancer associated with the IGF2 genotype was via a reduced growth-promoting activity that the polymorphic variant confers. The IGF2 Ex4 -233 C>T SNP is located within a CpG island of the 3′-UTR region; whether the C-to-T transition affected the methylation status as suggested in other studies (42-44) needs further investigation. Several studies have shown IGF2 polymorphisms to be significantly associated with BMI and birth weight (45-48). However, we failed to detect any association of IGF2 genotype and BMI among controls in this study. Because of the low frequency of the homozygous mutant variants we examined and incomplete BMI data, our study did not have sufficient power to examine the interaction of the IGF2 genotype with BMI on modifying the risk of pancreatic cancer.

Our study has several limitations. First, the study was conducted in a single tertiary referral hospital; results from this study population may not be generalized to the U.S. population. Second, the association of genotype and risk of pancreatic cancer could be biased if our study missed a lot of patients that were succumbed to this fatal disease rapidly and if the SNPs in question were associated with patient survival. Third, the study has a lower power to detect the main genotype effect in low-frequency homozygous mutants. For example, at the current study size, the power for detecting the main effect of the IGF2 -233 C>T SNP was only 60%. Although the IGF1 -177 G>C C allele frequency was >10%, the prevalence of diabetes was 10% among controls, and the synergistic effect of gene-diabetes was observed in only 44 cases and 8 controls, which yields less precision than we desired in the risk estimate. Fourth, the number of SNPs and genes included in this study were limited, the functional significance of the SNPs studied was unknown, and no phenotypic markers were included. Because it is known that IGF function can be significantly affected by the status of IGF receptors and IGF binding proteins, a comprehensive analysis of genes involved in the IGF axis and a systematic selection of SNPs of these genes would be required to fully elucidate the role of these genetic variations in pancreatic cancer. Illustrating genotype-phenotype associations by continuing this line of investigation may help reveal the mechanisms underlying the associations between genetic variations in IGF and the development of pancreatic cancer.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.