|Home | About | Journals | Submit | Contact Us | Français|
The mannose 6 phosphate/insulin-like growth factor 2 receptor (M6P/IGF2R) encodes a protein that plays a critical role in tumor suppression, in part by modulating bioavailability of a potent mitogen, insulin-like growth factor-2 (IGF2). We tested the hypothesis that the common non-synonymous genetic variants in M6P/IGF2R c.901C>G (Leu>Val) in exon 6 and c.5002G>A (Gly>Arg) in exon 34 are associated with risk of esophageal and gastric cancers. Study participants in this population-based study comprise 197 controls and 182 cases, including 105 with esophageal-gastric cardia adenocarcinoma (EGA), 57 with non-cardia gastric adenocarcinoma and 20 with esophageal squamous cell carcinoma (ES). Among white males, odds ratios (ORs) were elevated in relation to carrying at least one c.901C>G allele for EGA (OR= 1.9; 95%CI=1.0–3.6) and non-cardia gastric cancer (OR=2.5; 95%CI=1.2–5.5), but not ES. Exploratory subgroup analyses suggested that associations between EGA and this variant were stronger among irregular or non-users of non-steroidal anti-inflammatory drugs (NSAIDs) (OR=2.3; 95%CI=1.2–4.2) and cigarette smokers (OR=2.1; 95%CI=1.0–4.2). An association between carrying the c.5002G>A genotype and EGA was not evident. These findings suggest that non-synonymous polymorphisms in M6P/IGF2R may contribute to the risks of EGA and non-cardia adenocarcinomas. Larger studies are required to confirm these findings.
Esophageal cancer is the eighth most common cause of cancer-related mortality in the U.S., with a five-year survival rate of 14% 1. This cancer exhibits considerable gender, geographic, temporal and racial variation in incidence. Men comprise more than 80% of cases. Whereas squamous cell carcinoma of the esophagus (ES) remains the predominant histologic subtype globally, in the U.S. and other Western countries, esophageal adenocarcinoma (EA) has been the predominant subtype since the mid-1990s 2. The incidence of EA has increased >500 percent in the last three decades, an increase exceeding all other cancers 3. Whereas, ES is six-times more common in African American than white men, the incidence of EA is four times higher in white than African American men2.
Reasons for the sharp increase in incidence of EA and the closely linked gastric cardia adenocarcinoma (GCA) in Western countries and the preponderance among white men are unclear. However, the time trend parallels an increase in the prevalence of obesity in these countries 4, 5. Several lines of evidence suggest that most of these cancers arise in the setting of Barrett’s esophagus, a metaplastic condition in which the native squamous cell epithelium in the distal esophagus is replaced by specialized columnar cells in response to chronic gastroesophageal reflux (GERD). Indeed both GERD and obesity are strongly associated with esophageal and gastric cardia adenocarcinomas (EGA) in population-based case-control and cohort studies 6, 7. However, the prevalence of obesity/overweight or GERD does not vary substantially between African Americans and whites, or between men and women, and therefore fails to explain the racial or gender differences in EGA.
The Mannose 6 Phosphate/Insulin-like Growth Factor Receptor-2 (M6P/IGF2R); hereafter referred to as IGF2R, encodes a 300 kDa type 1 transmembrane glycoprotein that has been recently identified as a tumor suppressor 8. This glycoprotein has binding sites for M6P-bearing proteins and IGF2 8, the latter a potent mitogenic growth factor that when elevated in circulation, has been linked to increased risk of obesity, higher birth weight 9, and several adenocarcinomas. Extracellular binding of IGF2 to IGF2R results in internalization of IGF2 and its transport to the lysosomes where it is degraded, and this is the primary mechanism by which IGF2R modulates the bioavailability of IGF2.
IGF2R inactivation has been associated with risk of poorly differentiated breast10 and lung tumors 11. Recurrent loss of heterozygosity at the IGF2R locus, and mutations in the remaining allele appear to be early events in breast cancer, squamous cell lung cancer, hepatocellular carcinoma, non-Hodgkin lymphoma, ovarian cancer and renal cell carcinoma 10, 11. Chromosomal loss in a region that includes the IGF2R gene has also been linked with androgen insensitive prostate cancer 8. These findings point to the involvement of IGF2R as a tumor suppressor in the development of cancer, but the relation of genetic variants in IGF2R to EGA risk has not been evaluated.
More than 1,200 single nucleotide polymorphisms (SNPs) have been identified in IGF2R to date. Six are in the coding region and three of these (c.6206 A>G, Asn2020Ser; c.901C>G, Leu252Val and c.5002G>A Gly1619Arg) are non-synonymous12. Two of these, c.901C>G, Leu252Val and c.5002G>A Gly1619Arg, have a minor allele frequency that is more than 5% among Caucasians. Herein we report on analysis of the association between EGA risk and these two polymorphisms in IGF2R, utilizing resources from a multi-center population-based, case-control study of EGA across the United States.
Methods for the identification of study participants have been described previously 13. Briefly, this case-control study was conducted in the state of Connecticut, a 15-county area of New Jersey, and a 3-county area of the state of Washington. Eligible study participants were individuals newly diagnosed with EA, GCA, ES, and non-cardia gastric adenocarcinoma. In Connecticut, all individuals who spoke English and were aged 30 to 79 years diagnosed with primary invasive cancer of the esophagus or gastric cardia from February 1, 1993 through January 31, 1995 were invited to participate. Similar criteria were used to identify and enroll study participants in New Jersey, between April 1993 through November 1994, and in Washington, from March 1, 1993 through February 28, 1995. All EGA cases were eligible, whereas the ES and non-cardia gastric cancers were frequency-matched to the expected age distribution of cases with EGA, based on the state of identification, and five year age-group, for comparison. In New Jersey, further frequency-matching was also done on race (white versus other). Case eligibility was ascertained by two study pathologists using pathology reports and confirmed by hematoxylin and eosin stained slides. Disagreements were resolved by consensus.
For each study site, controls aged 30 to 64 years were identified using random digit dialing techniques. For those 65 years and older, Health Care Financing Administration rosters were used to identify population controls. Controls were frequency-matched to the expected distribution of cases by five-year age groups and sex, and additionally on race in New Jersey.
Trained interviewers administered structured questionnaires in person to 80% of eligible EGA cases, 74% ES cases and 74% of controls or their proxy respondents. The 60-minute questionnaire included assessment of the following: demographic characteristics; tobacco use, beverage consumption including alcohol; medical history, including weight and height in the year prior to the reference date (defined as date of diagnosis for cases and date of identification for controls); medication use including over-the-counter medications such as aspirin and other non-steroidal anti-inflammatory drugs (NSAIDS); occupational history, and food intake based on a semi-quantitative food-frequency questionnaire that included usual intake of fruit, vegetables and multivitamin supplements 3–5 years prior to diagnosis.
Up to 30mls of peripheral blood, from which DNA was extracted, was obtained from a sample of index subjects in Washington and nine of the 15 counties in New Jersey at the time of the in-person interview14. DNA samples were available for 197 controls and 162 cases (53 EA, 52 GCA and 57 other gastric adenocarcinomas, corresponding to 40% of controls, and 25%, 29%, and 23%, for EA, GCA and non-cardia gastric cases, respectively). We also included 20 ES cases for comparison. Although the proportion included was not large, individuals for whom DNA samples were available did not differ substantially from those for whom samples were not available with respect to age, gender, race, GERD and cigarette smoking (data not shown).
Lyophilized genomic DNA was re-suspended using PureGene system reagents (Gentra, Minneapolis, MN) and subsequently plated into two 384-well plates with 16 controls including four housekeeping genes. Primers for the c.5002 (rs629849) in exon 34 and c.901 (rs8191754) in exon 6 were designed by Applied Biosystem. For c.901 C>G, the forward primer was 5′-CTA AGG GTA CTG TGA TTA TCA CTC-3′ and the reverse primer was 5′-GAA AGT CAG GTC CTT GCT GGA G-3′. For the c.5002G>A, the forward primer was 5′-GAA ATT GAT GGT CCT GAC TTG CG-3′ and the reverse primer was 5′-GCA CTG GAG ATG CAC TTC TCC-3′. Genotyping was undertaken without the knowledge of case status of participants. Ten randomly selected replicate samples from each of the two 384-well plate were also included. Undetermined genotypes were excluded from analyses and constituted 4.2% and 5% of the total population for the c.901 C>G and c.5002 G>A variants, respectively.
We tested deviation from Hardy-Weinberg equilibrium (HWE) of IGF2R c.5002 and IGF2R c.901 genotypes among the controls for the overall and site-specific analyses using chi-square tests. Unconditional logistic regression models were used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the association between carrying the IGF2R c.5002 or the c.901 allele, and risk of esophageal or gastric cancer by histologic subtype. We compared carriers of at least one of the minor alleles (at risk genotype) to individuals homozygous for the common allele (referent genotype). Age at reference date, sex, race and study site, were included in all statistical models. Factors previously reported to be associated with these histologic subtypes were explored for effect modification using stratified analyses. These factors were then evaluated for confounding comparing logistic regression models with and without the variables of interest. Factors evaluated include household income, body mass index (BMI = kg/m2), number of years of education, current and former cigarette smoking, intake of alcohol in grams (wine, beer and liquor), use of aspirin and other NSAIDS, frequency of intake of fruit and vegetables and multivitamin supplement use. SAS version 9.1 (SAS, Cary, NC, USA) was used for all statistical analyses.
Table 1 summarizes the distribution of socio-demographic characteristics, anthropometric measures, dietary habits, and other risk factors for esophageal and gastric cancers among study participants with DNA available for these analyses. More than 95% of cases were white males born in the United States. To ensure that the study population for whom blood samples were available was not systematically different from the population in whom blood samples were not collected, we compared the strength of the association between each cancer subtype and factors previously associated with these cancers in the larger study, including SES as measured by educational level 13, BMI 15 and a history of cigarette smoking 13. Associations with BMI and educational levels in this sub-sample were similar to those found in the larger study at each cancer site, although we were unable to detect an association between cigarette smoking and adenocarcinomas of the esophagus and gastric cardia.
There was no evidence that genotype frequencies from the IGF2R c.901 deviated from Hardy-Weinberg equilibrium (p-values= 0.14), although the evidence for the c.5002 was less clear (p=0.05). Table 2 summarizes the prevalence of genotypes and adjusted ORs for the association between these genotypes and esophageal cancer risk, by histologic subtype. There was little evidence for an association between carrying the minor IGF2R c.5002 A-allele and risks for GCA (OR=1.21, 95%CI=0.58–2.54), non-cardia gastric cancer (OR=0.98, 95%CI=0.46–2.08), or EA (OR=0.60, 95%CI=0.25–1.44) or ES (OR=1.80, 95%CI=0.64–5.06). Restricting these analyses to white males did not substantially change these findings. We conducted additional stratified analyses in the combined category of EGA, to explore the association with IGF2R c.5002, since the origin of these junctional tumors is difficult to pinpoint, and also share risk factor profiles, as well as incidence trajectories. Although this combined case group allowed for increased statistical power, there was no apparent association when carrying this genotype was examined within categories of low or high BMI, past or current regular NSAID use, or high or low daily intake of fruits or vegetables.
There was no evidence of an association between IGF2R c.901 A>G and ES (OR=1.06, 95%CI=0.33–3.36), but a suggestive association was seen with non-cardia gastric adenocarcinomas (OR=1.80, 95%CI=0.92–3.57). Restricting analyses to white males generally strengthened these associations (OR=2.07 for GCA, 95%CI=0.96–4.50; OR=1.92 for EA, 95%CI=1.03–3.59; and OR=2.51; 95%CI=1.16–5.46 for non-cardia gastric cancer). Interestingly, the associations between this variant and risks of ES, GCA, EA and non-cardia gastric cancer were weaker for esophageal tumors and stronger and statistically significant for gastric cardia and gastric ones.
Exploratory stratified analyses also suggest that the associations between carrying the IGF2R c.901 G-allele and EGA varied with past exposures to NSAID use and cigarette smoking. Associations were stronger in those reporting no regular use of NSAIDs in the past or at the time of interview (OR=1.86, 95%CI=1.01–3.43), and these associations were strongest in white males (OR=2.28, 95% CI=1.18–4.42). These associations were not apparent in regular NSAID users. An association between this variant and EGA was also suggested in those reporting a history of cigarette smoking. However, interactions were not statistically significant, and adjusting for potential confounders did not substantially change these associations. Separate analyses for EA and GCA also did not alter these findings.
This is the first epidemiologic study to evaluate associations between non-synonymous variants of the IGF2R tumor suppressor gene and esophageal and gastric cancer. Several years ago, Killian 16 identified nine novel polymorphisms, six of them in the coding region of the IGF2R with three being non-synonymous. We analyzed the two polymorphic variants with minor allele frequencies more than 5%, with the working hypothesis that these variants altered this tumor suppressor. We found no evidence of an association between carrying the IGF2R c.5002 G>A minor allele and any subtype of esophageal or gastric cancer regardless of the subpopulation evaluated. However, an association was seen between carrying at least one IGF2R c.901C>G allele and EGA and with non-cardia gastric adenocarcinoma among white males with a history of cigarette smoking but infrequent use of NSAIDS.
The finding of no association between the IGF2R c.5002 A-allele and risk of esophageal or gastric cancers was unexpected since this polymorphism is located within extracellular domain 11 of the M6P/IGF2R where IGF2 binding occurs. Based on this location, carrying this variant was expected to alter IGF2 binding affinity 11, 12, thereby increasing the bioavailability of IGF2. Individuals carrying the c.5002 A-variant have been reported to be at higher risk for lung cancer 11 while no association was found in a recent study of osteosarcoma 17. Moreover, the racial distribution of this genetic variant parallels that of EA in that it is very low (<2%) in Africans where EA is rare; it is found in moderately low frequency in African Americans (13%) where these adenocarcinomas occur in slightly higher frequency, and its highest frequency is among Americans of European descent (36%) where the incidence of EA is also highest. However, a recent study suggests that the c.5002 A-allele does not significantly alter binding of IGF2 relative to the c.5002 G-allele nor does it alter protein trafficking or stability, although effects on IGF2R dimerization or the stability and binding of IGF2 to the soluble form of the IGF2R were not evaluted18. It is also possible that our inability to find associations may have been due to small sample sizes, the suboptimal call rate, or because the variant was not clearly in HWE (p-value=0.05).
The etiologic significance of finding an association between c.901 and esophageal and gastric cancer is presently unclear. The amino acid variant c.901Leu252Val is a conservative substitution with respect to hydrophobicity, but may lead to protein destabilization due to replacement of the isobutyl group with an isopropyl side chain 19. It is located in repeat domain 3 and is involved in binding M6P moieties on other proteins and this position has been strictly conserved throughout mammalian evolution 16. Because of this important function, altered affinity for M6P-bearing ligands may lead to changes in protein trafficking and secretion16, causing a shift in the protein milieu, within cells and tissues, including that of IGF2. The IGF2R is also responsible for binding and activating the latent form of TGFβ120, 21, a potent growth inhibitor. It is possible that this role may be impeded by the c.901 variant allele presumably enhancing cancer risk. Alternatively, the risk of developing esophageal and gastric cancer may involve decreased capacity for IGF2 binding due to steric hindrance mediated through binding of IGF2R to other lysosomal enzymes 22, thereby increasing bioavailability of this mitogenic growth factor. Indeed recent data from a cohort study suggest that esophageal and non-cardia adenocarcinoma risk may be determined, in part, in utero 23 and birth weight has been positively associated with carrying the c.901 variant in at least one allele 22. Nonetheless, the associations found are consistent with the tumor suppressive effect of IGF2R and with the association reported in relation to high birth weight 22 presumably reflecting increases in bioavailable IGF2 in-utero.
A limitation of this study was the small sample sizes and low p-value for the HWE test that may have contributed to our inability to detect associations between the c.5002 variant and the subtypes of esophageal and gastric cancers. However, this SNP is also not in HWE in whites in the HapMap and SNP500Cancer databases, confirmed by a recent report18. The small sample size also limited our ability to evaluate gene-environment effects on risk, even when we combined EA with GCA. Although other studies have also combined EA and GCA, tumors arising in these two locations may not always share the same etiologic mechanisms. An additional limitation often found in case-control studies of rapidly fatal diseases is the low response rate, which raises a possibility of selection (survival) bias as an alternative explanation for our findings. However, genotype frequencies in our control population were similar to publically available estimates (NCBI).hih.gov) of the prevalence of the c.901 variant in a white population. Hence, the inability to participate due to disease severity and death may not have differed by genotype. Furthermore, survival bias is an unlikely explanation since these SNPs are not known to be related to survival.
In summary, we found evidence of an association between esophageal, gastric cardia and non-cardia gastric adenocarcinomas and carrying the IGF2R c.901 G allele, particularly among white men reporting cigarette smoking and none or irregular use of NSAIDS. No association was found between these tumors and the IGF2R c.5002 genotype. Due to a limited sample size and low call rates for these genotypes, our findings should be considered preliminary.
This work was supported by K01 CA104517