IGF-I Genetic Variation and Breast Cancer: the Multiethnic Cohort

The insulin-like growth factor-I (IGF-I) signaling pathway regulates both cellular proliferation and apoptosis, thereby making it a compelling candidate gene for cancer pathogenesis. Epidemiologic studies indicate that high levels of circulating IGF-I are associated with elevated risk of breast cancer in premenopausal women (1, 2). Few studies have investigated the role of genetic variation in IGF-I in relation to breast cancer and have focused solely on the polymorphic dinucleotide repeat (CA) in the promoter region (3-7). Evidence of an effect of this polymorphism on breast cancer risk remains inconsistent (8). To date, no studies have thoroughly characterized common variation in IGF-I and examined this in relation to breast cancer risk. We characterized genetic variation across the IGF-I locus using a combination of direct (resequencing) and indirect (haplotype-based approach) methods and then did a large case-control analysis to assess association between this inherited variation and sporadic breast cancer risk in a multiethnic cohort.

Detailed information about the Multiethnic Cohort–nested breast cancer case-control study has been reported previously (9). A total of 1,615 incident breast cancer cases (21% African American, 7% Native Hawaiian, 26% Japanese, 21% Latina, and 25% White) and 1,962 controls (22% African American, 15% Native Hawaiian, 21% Japanese, 20% Latina, and 22% White) were included in the present study. Controls were women without breast cancer before the cohort entry and without a diagnosis up to April 2002. Controls were frequency matched to cases by age and ethnicity.

The details of exon resequencing, linkage disequilibrium characterization and tagging single nucleotide polymorphism (tSNP) selection have been described (10). Briefly, IGF-I exons were sequenced in 95 advanced cases of breast cancer (n = 19 per racial-ethnic group). No missense SNP was identified during this sequencing effort. To characterize the linkage disequilibrium patterns, 64 SNPs (1 SNP/2.4 kb) spanning 156 kb were genotyped in a multiethnic panel of 349 unrelated women with no history of cancer. Haplotype blocks (regions of strong linkage disequilibrium) were defined using the methods of Gabriel et al. (11). Twenty-nine tSNPs were selected to predict the common haplotypes with high probability (average R_h² = 0.90) and 35 unmeasured SNPs (i.e., SNPs not genotyped in case-control samples). The method of predicting unmeasured SNPs has been described elsewhere (10).

All SNPs were genotyped by Taqman assay (Applied Biosystems, Foster City, CA). Taqman primers, probes, and conditions for genotyping assays are available upon request. The average genotyping success rate was 97%. All genotyping was done with laboratory personnel blinded to case-control status of the samples, which included quality control samples for validation. Concordance for quality control samples was 99.7%.

We used the χ² test to assess departures of the genotype distribution from Hardy-Weinberg equilibrium among controls in each ethnic group. All tSNPs conformed to Hardy-Weinberg equilibrium at P < 0.01 level. Methods for haplotype estimation and case-control analysis were previously described (9). A likelihood ratio test was done to globally test for haplotype effect in each block. Unconditional logistic regression models were used to estimate odds ratios (OR) and 95% confidence intervals for haplotype-specific and genotype-specific risks, adjusting for age and ethnicity. We also did multivariate analyses with established breast cancer risk factors included in the models (9). All Ps are two sided. SAS version 8.2 (SAS Institute, Cary, NC) was used for all analyses.

We tested both haplotypes and single tSNPs to maximize the likelihood that we have captured all of the unmeasured variation at the locus. Four regions of strong linkage disequilibrium (haplotype blocks), ranging from 11 to 60 kb, were identified, and 5 to 11 common haplotypes (i.e., frequency ≥ 5%) were observed within each block. The haplotypes for each ethnic group accounted for 77% to 100% of the chromosomes in the population (10).

There were no significant associations between inherited variation at IGF-I and breast cancer risk. The haplotype frequencies between cases and controls were similar in each block (Ps of the global test for association with breast cancer > 0.46; Table 1). This observation provides little support for the existence of a common disease allele at the IGF-I locus strongly contributing to breast cancer risk. Within each block, no individual haplotype was observed to be associated with breast cancer risk. We also assessed association between single tSNPs (and combinations of tSNPs) and breast cancer risk. Using the 29 tSNPs and genotype data obtained from the multiethnic panel, we estimated for each individual the alleles of the 35 unmeasured SNPs (10). We were able to predict 33 of 35 SNPs (average R_s² = 0.91); SNP 18 (hCV2801091) and SNP 63 (rs2971578) could not be predicted with R_s² > 0.7. SNP 18 was then genotyped in case-control samples, whereas SNP 63 failed assay design. Figure 1 shows the association between 63 SNPs and breast cancer risk, and no convincing association was observed. Two nominally significant negative associations were observed: SNP 1 (rs855228; P_trend = 0.03) and SNP 4 (rs7965399; P_trend = 0.04). We performed permutation test to evaluate how often these nominally significant associations would be expected by chance (10). The permutation P was 0.64 for a nominal P = 0.03, which indicates a similar result would be seen by chance 64% of the time. Analysis of cases with advanced disease only (n = 421) or adjustment for established breast cancer risk factors and menopausal status did not alter the results.

To date, our study is the largest study to evaluate the genetic contribution of IGF-I to breast cancer risk. The main strength of this study is the thorough and systematic approach for studying common variation in IGF-I: direct sequencing of exons for putative functional SNPs and genotyping a dense set of variants for the noncoding regions. We found no evidence of an association between inherited variation at IGF-I and breast cancer risk. Our power calculation indicates that our sample of 3,577 individuals provides 80% power (assuming a log additive model, a two-sided test and using α = 0.001 to account for multiple testing) to detect an odds ratios of 1.5 for SNP with 5% allele frequency or for haplotypes predicted with R_h² ≥ 0.9 and frequency as low as 5%.

This is the first study that used SNPs in assessing association between IGF-I genetic variation and breast cancer risk. Five published studies of IGF-I and breast cancer have focused only on the (CA)_n repeat polymorphism (3-7). The (CA)₁₉ allele, the most common allele in various populations, has been associated with increased risk in two studies (3, 6), whereas one study reported an inverse association (7), and two other studies (including ours) observed no association (4, 5). In our previous study, we found that the (CA)₁₉ allele was not associated with breast cancer among postmenopausal women (odds ratios, 1.2; 95% confidence interval, 0.8-1.8; ref. 5). This study was based on 400 cases and 400 controls from four racial-ethnic groups: African American, Japanese, Latino, and White from the Multiethnic Cohort.

Our results suggest that common germ line variation in IGF-I is not a major contributor to breast cancer risk in this multiethnic cohort. Whether other genes in IGF pathway (e.g., IGF-2 and IGFBPs) have a more important role or whether the combined effects of several modest associations within this pathway have greater cumulative effect on breast cancer susceptibility remains to be determined. Future work will include evaluation of whether IGF-2, IGF2R, IGFBP-1, and IGFBP-3 genetic variations are associated with breast cancer risk.

We thank Noel Burtt, Loreall Pooler, David Wong, and Stephanie Riley for their laboratory assistance; and Elena Giorgi and Hank Huang for their technical support; and the participants of the Multiethnic Cohort Study for their participation and commitment to the study.