Variation in the FGFR2 Gene and the Effect of a Low-Fat Dietary Pattern on Invasive Breast Cancer Free

The Women's Health Initiative (WHI) randomized controlled trial included evaluation of the health risks and benefits of four prevention interventions in a partial factorial design (1, 2). A total of 68,132 postmenopausal women were enrolled during 1993 to 1998 at 40 U.S. clinical centers.

The dietary modification (DM) trial component tested whether a low-fat dietary pattern could reduce the incidence of breast and colorectal cancer. It enrolled 48,835 women, with random assignment of 40% to the intervention group and 60% to a usual-diet comparison group, and proceeded to its planned termination after an average of 8.1 years of intervention and follow-up. The low-fat dietary pattern intervention goals included a reduction in dietary percent of energy from fat from baseline levels (estimated by food frequency questionnaire to be above 32% as an eligibility criterion) to 20%, an increase to five of daily vegetables and fruit servings, and an increase to six in daily grain servings. The assessed change in percent of energy from fat was only ∼70% of that assumed in the trial design (1), and the lower invasive breast cancer incidence in the intervention versus the comparison group was ∼70% of that hypothesized in trial design [hazard ratio, 0.91; associated 95% confidence interval (CI), 0.83-1.01; 1,727 incident cases during follow-up; ref. 2].

However, the overall trial comparison may be diluted by the inclusion of many women having a relatively low-fat diet at baseline, who may have made little or no dietary change, if assigned to the intervention group. In fact, there was a significant (P = 0.04) interaction between baseline percent of energy from fat, as assessed using 4-day food records, and intervention versus comparison group hazard ratio, with women in the top quartile (≥36.8%) of percent of energy from fat having a hazard ratio of 0.78 (95% CI, 0.64-0.95; ref. 2).

Findings from recent breast cancer genome-wide association studies provide an opportunity to seek genetic factors that may relate to dietary intervention effects on breast cancer. The strongest breast cancer association to emerge from these studies (3, 4) involves single nucleotide polymorphisms (SNPs) in intron 2 of the FGFR2 gene. For example, SNP rs2981582, with a minor allele frequency of ∼40%, has been found to convey a per-minor allele increase in breast cancer risk of ∼30% (3, 4). FGFR2 splice variants have been shown to transform human mammary epithelial cells (5), and blocking FGFR2 signaling inhibits breast cancer cell proliferation (6). Intron 2 includes highly conserved regions and is dense in transcription factor binding sites, including sites thought to be relevant to the estrogen receptor (7).

We tested the FGFR2 SNP just mentioned, along with seven others in intron 2 of this gene, among women who developed invasive breast cancer during the follow-up period of the WHI clinical trial. Each case was matched to a corresponding control without a breast cancer diagnosis during the follow-up period. As recently reported, each of the eight SNPs was significantly related to breast cancer risk in this WHI cohort, with seven of the eight having P < 1.0 × 10⁻⁷ (8). Moreover, the genotype of one of the eight SNPs, rs3750817, was associated with the breast cancer odds ratio (OR) for both postmenopausal estrogen plus progestin therapy (P = 0.03) and estrogen-alone therapy (P = 0.05), whereas the genotype of a second SNP, rs2981582, may also be associated (P = 0.05) with estrogen-alone therapy. Here, we report corresponding analyses for the dependence of ORs for the DM intervention on these SNPs.

WHI clinical trial enrollees were postmenopausal, in the age range 50 to 79 y, without a history of invasive breast cancer, and with no suggestion of breast cancer on baseline mammogram and clinical breast exam. Clinical outcomes were self-reported at 6-mo intervals. Breast cancers were confirmed (9) by review of pathology reports by local physician adjudicators, followed by adjudication at the Clinical Coordinating Center that included coding of estrogen receptor status and progesterone receptor status (positive or negative per local pathology report), histology, and extent of disease using the National Cancer Institute's Surveillance, Epidemiology and End Results system. The trial was approved by human subject review committees at each participating institution, and each study participant provided written informed consent.

All 1,727 invasive breast cancer cases developing between randomization (1993-1998) and the end of the intervention phase for the DM trial component (March 31, 2005) were considered for inclusion in the present analyses. Of these, 1,676 (97.0%) had adequate quantity and quality of DNA for SNP genotyping.

The eight FGFR2 SNPs were included in a larger project involving 9,039 SNPs selected from previous genome-wide association studies (3, 4), including a WHI-Perlegen Sciences collaboration (10). Genotyping and quality control methods at Perlegen, where genotyping took place, have been described previously (11). The average call rate for SNPs meeting quality assurance criteria was 99.8% in the overall project, and the average concordance rate for duplicate samples (157 pairs in data set) was also 99.8%.

Six of the eight FGFR2 SNPs selected (rs2981582, rs1219648, rs2912774, rs2981579, rs11200014, and rs2420946) were from a single linkage disequilibrium (LD) region where much of the interest from genome-wide breast cancer association studies has focused. These SNPs have a minimum squared pairwise correlation (r²) of 0.83 among Caucasian women. A seventh SNP, rs17102287, is in the same genomic region, but is in much lower LD with the other six SNPs (maximum r² of 0.32 among Caucasian women). The eighth SNP, rs3750817, is a little distant from the other seven and has a maximum pairwise r² of 0.47 with the other SNPs among Caucasian women, but falls within a conserved region of the gene (3). For example, in the WHI clinical trial cohort studied here, SNP rs2981582 with a minor allele (A) frequency of 0.38 had an estimated per minor allele breast cancer OR of 1.27 (P_trend = 9.3 × 10⁻⁸), similar to that for the other SNPs in the same LD region, whereas SNP rs3750817 with a minor allele (T) frequency of 0.39, and negatively correlated with the other SNPs, had an estimated per minor allele OR of 0.78 (P_trend = 8.2 × 10⁻⁸; ref. 8).

Tests for SNP interactions with the DM intervention OR were carried out using case-only data analyses (8, 12-14). This approach, which here can be expected to yield essentially unbiased and highly efficient OR estimates, involves logistic regression of intervention assignment (0, comparison; 1, intervention) on indicator variables for 0, 1, or 2 minor SNP alleles, with constant term given by log [q/(1−q)], where q is the fraction (0.40) of the trial cohort assigned to active intervention in the DM trial. Likelihood ratio tests with two degrees of freedom were used to test for intervention OR variation with SNP genotype. The treatment indicator variable was randomly permuted and the likelihood ratio statistic was recalculated to generate significance levels that are adjusted for the fact that eight FGFR2 SNPs were tested for OR interaction. The multiple testing–adjusted significance level was calculated as the fraction of 1,000 permutation samples for which the likelihood ratio statistic was as large or larger than that observed.

The characteristics of the women participating in the DM trial have been published (2, 15). Women were postmenopausal at enrollment, with an average age of 63 years, and ∼20% were of minority race/ethnicity. About two-thirds were overweight or obese. All women were without a prior breast cancer diagnosis at enrollment.

Table 1 shows DM intervention OR estimates (95% CI) according to the number of minor SNP alleles, for all randomized women, and separately for those having baseline percent of energy from fat in the upper quartile. Significance levels for tests of independence of ORs with SNP genotype are also given. OR estimates for the entire randomized cohort may be slightly lower among women homozygous for the major allele for the first six highly correlated SNPs and slightly lower for women having one or more minor alleles of rs3750817, although these variations are far from significant (P ≥ 0.18). These patterns are clearer, and mostly significant (P < 0.05), however, upon restricting the analysis to women having baseline percent of energy from fat in the upper quartile. Relatively lower ORs among women having one or more T alleles for rs3750817 is particularly evident (P = 0.005). To compare the evidence for interaction with the DM intervention effect among the FGFR2 SNPs studied, we carried out case-only analyses for women having baseline percent of energy from fat in the upper quartile in which the intervention effect was allowed to depend simultaneously on the number of minor alleles of rs3750817 and the original tagging SNP in the first LD region, rs2981582. A dependence of the intervention effect on rs3750817 genotype could be detected (P = 0.03) when the intervention effect was allowed to depend on rs2981582 genotype, but this was not the case for rs2981582 (P = 0.25) when the intervention effect was allowed to depend on rs3750817 genotype.

With adjustment for the fact that eight FGFR2 SNPs were tested for interaction, the significance level for interaction with rs3750817 increases to 0.03. This multiple testing–adjusted test of interaction remained significant (P = 0.05) when the analysis was restricted to Caucasian women (361 breast cancer cases), whereas the corresponding adjusted P values were not close to significant (P ≥ 0.17) for the other seven FGFR2 SNPs.

Table 2 shows a further breakdown of ORs for women having baseline percent of energy from fat in the upper quartile, according to the estrogen and progesterone receptor status of the breast tumor, both for rs2981582 (as a representative SNP from the first LD set) and for rs3750817. The patterns just described are not very apparent within tumor receptor status subtypes for rs2981582 but are pronounced for tumors that are estrogen receptor–positive (P = 0.008) or progesterone receptor–positive (P = 0.003) for rs3750817. With multiple testing adjustment, the P values for rs3750817 increase to 0.05 for estrogen receptor–positive tumors and to 0.02 for progesterone receptor–positive tumors. We tested whether the ORs shown in Table 2 for estrogen receptor–positive tumors differed from those for estrogen receptor negative by a simple multiplicative factor and did not find evidence to the contrary (P = 0.74), presumably due to the relatively few estrogen receptor–negative tumors. The corresponding test comparing progesterone receptor–positive and progesterone receptor–negative tumor ORs was also not significant (P = 0.34).

The low-fat dietary pattern intervention implemented in the WHI provided suggestive evidence of a reduction in invasive breast cancer risk overall, and stronger evidence among women whose prerandomization diet was relatively high in fat and whose dietary change was larger.

Among women in the upper quartile of percent of energy from fat, the dietary intervention OR varied with the genotype of SNPs in the FGFR2 gene. This interaction was particularly evident (P_nominal = 0.005) in relation to SNP rs3750817, where evidence for a breast cancer risk reduction was confined to women having one or more minor (T) alleles. These patterns were apparent for estrogen receptor–positive and for progesterone receptor–positive tumors and persisted upon allowing for the testing of eight FGFR2 SNPs. The randomized assignment of women to the low-fat dietary pattern intervention prevents population stratification from biasing dietary intervention ORs for the overall study population. However, stratification could influence OR variations with SNP genotype, if both genotype and dietary intervention ORs varied among population strata. In fact, the International HapMap data do indicate a lower frequency of the T allele of rs3750817 among persons of African, compared with European, ancestry. However, when analyses were restricted to the 84.3% of breast cancer cases among Caucasian women, the dietary intervention OR pattern was unchanged, and the multiple testing–adjusted P value of 0.05 (versus 0.03 for the combined ethnicities) is as expected with the reduced sample size if DM ORs are unchanged. These findings tend to strengthen the evidence for a DM intervention benefit among women having a high fat content in their customary diets because one would not expect evidence for interaction with rs3750817, or other SNPs, to arise if the DM intervention had no effect on breast cancer incidence.

It is interesting that rather similar breast cancer OR patterns as a function of rs3750817 also arose for the randomized placebo-controlled hormone therapy interventions in the WHI clinical trial (8). The elevated breast cancer risk with estrogen plus progestin arose from women having one or more major alleles of this SNP, whereas a suggested risk reduction with estrogen alone arose from women homozygous for the minor allele. As previously noted (8), there is a high degree of sequence homology around rs3750817, and the region has a density of transcription factor binding sites, including the prediction (8) that the T allele introduces a YY1 transcription factor binding site into the human sequence. YY1 is involved in breast cancer cell migration (16), is involved in ERBB2 oncogene expression (17), and represses TRAIL-induced apoptosis (18).

The ability to assess the biological plausibility of an interaction of a DM intervention effect interaction with rs3750817 genotype is limited by the incomplete knowledge of the regulation of FGFR2 protein expression. Meyer et al. (19) studied protein expression in relation to several SNPs in the vicinity of those in the first LD region studied here and concluded that SNPs associated with risk correlated with FGFR2 expression itself, rather than functioning through receptor-ligand interactions. They also identified two transcription factors whose binding affinity was altered by the SNPs studied. Corresponding studies have not been reported for rs3750817, or other SNPs in close proximity to rs3750817.

Reduced ORs among women having the T allele of rs3750817 could suggest a sensitivity of risk to variations in the circulating hormonal milieu. The low-fat dietary pattern intervention led to some reduction in plasma estradiol and increase in sex hormone–binding globulin (2), whereas estrogen plus progestin and estrogen alone cause major increases in both plasma estradiol and sex hormone–binding globulin, (20).

Similarly, a high body mass index (BMI) is an established risk factor for postmenopausal breast cancer. We observe in the case-control data for this project a significant (P = 0.02) interaction between BMI and the rs3750817 genotype in relation to breast cancer risk, with comparatively stronger evidence for BMI association with risk among women having one or two minor SNP alleles. However, the ORs for rs3750817 shown in Table 1 do not differ significantly (P = 0.60) between obese (BMI, ≥30) and nonobese women.

The strengths of this study are the randomized controlled trial design of the DM trial, which implies precise orthogonality between genotype and dietary intervention, justifying the highly efficient case-only data analysis method used. Other strengths include prediagnostic blood specimens, collected and stored according to a standardized protocol, and quality controlled SNP genotyping.

Study limitations include incomplete knowledge of FGFR2 variants involved in breast cancer risk determination and in related dietary influences. In addition, replication of the interactions reported here will be useful. Readily available WHI observational data are not well suited to this purpose because 4-day food records have not undergone nutrient analysis for most of the controls in the present genotyping case-control studies, and food frequency questionnaire data, in contrast to food record data, provide little evidence of association between dietary fat and breast cancer (21). A replication study effort that will obtain FGFR2 genotypes for study subjects included in the earlier 4-day food record analyses (21) is in the planning stage.

In summary, SNP rs3750817 in a conserved region of the FGFR2 gene is strongly associated with breast cancer risk and is reported here to interact with the putative effect of a low-fat dietary pattern on breast cancer incidence. Women having one or more minor (T) alleles of this SNP may benefit from reduction from a high-fat to a lower fat dietary pattern.

No potential conflicts of interest were disclosed.

Decisions about study design, data collection and analysis, interpretation of the results, the preparation of the manuscript, or the decision to submit the manuscript for publication resided with committees composed of WHI investigators that included NHLBI representatives.

Program Office: Elizabeth Nabel, Jacques Rossouw, Shari Ludlam, Joan McGowan, Leslie Ford, and Nancy Geller (National Heart, Lung, and Blood Institute, Bethesda, Maryland).

Clinical Coordinating Center: Ross Prentice, Garnet Anderson, Andrea LaCroix, Charles L. Kooperberg, Ruth E. Patterson, Anne McTiernan (Fred Hutchinson Cancer Research Center, Seattle, WA); Evan Stein (Medical Research Labs, Highland Heights, KY); Steven Cummings (University of California at San Francisco, San Francisco, CA).

Clinical Centers: Sylvia Wassertheil-Smoller (Albert Einstein College of Medicine, Bronx, NY); Aleksandar Rajkovic (Baylor College of Medicine, Houston, TX); JoAnn E. Manson (Brigham and Women's Hospital, Harvard Medical School, Boston, MA); Charles B. Eaton (Brown University, Providence, RI); Lawrence Phillips (Emory University, Atlanta, GA); Shirley Beresford (Fred Hutchinson Cancer Research Center, Seattle, WA); Lisa Martin (George Washington University Medical Center, Washington, DC); Rowan Chlebowski (Los Angeles Biomedical Research Institute at Harbor-University of California at Los Angeles Medical Center, Torrance, CA); Yvonne Michael (Kaiser Permanente Center for Health Research, Portland, OR); Bette Caan (Kaiser Permanente Division of Research, Oakland, CA); Jane Morley Kotchen (Medical College of Wisconsin, Milwaukee, WI); Barbara V. Howard (MedStar Research Institute/Howard University, Washington, DC); Linda Van Horn (Northwestern University, Chicago/Evanston, IL); Henry Black (Rush Medical Center, Chicago, IL); Marcia L. Stefanick (Stanford Prevention Research Center, Stanford, CA); Dorothy Lane (State University of New York at Stony Brook, Stony Brook, NY); Rebecca Jackson (The Ohio State University, Columbus, OH); Cora E. Lewis (University of Alabama at Birmingham, Birmingham, AL); Cynthia A Thomson (University of Arizona, Tucson/Phoenix, AZ); Jean Wactawski-Wende (University at Buffalo, Buffalo, NY); John Robbins (University of California at Davis, Sacramento, CA); F. Allan Hubbell (University of California at Irvine, CA); Lauren Nathan (University of California at Los Angeles, Los Angeles, CA); Robert D. Langer (University of California at San Diego, La Jolla/Chula Vista, CA); Margery Gass (University of Cincinnati, Cincinnati, OH); Marian Limacher (University of Florida, Gainesville/Jacksonville, FL); J. David Curb (University of Hawaii, Honolulu, HI); Robert Wallace (University of Iowa, Iowa City/Davenport, IA); Judith Ockene (University of Massachusetts/Fallon Clinic, Worcester, MA); Norman Lasser (University of Medicine and Dentistry of New Jersey, Newark, NJ); Mary Jo O'Sullivan (University of Miami, Miami, FL); Karen Margolis (University of Minnesota, Minneapolis, MN); Robert Brunner (University of Nevada, Reno, NV); Gerardo Heiss (University of North Carolina, Chapel Hill, NC); Lewis Kuller (University of Pittsburgh, Pittsburgh, PA); Karen C. Johnson (University of Tennessee Health Science Center, Memphis, TN); Robert Brzyski (University of Texas Health Science Center, San Antonio, TX); Gloria E. Sarto (University of Wisconsin, Madison, WI); Mara Vitolins (Wake Forest University School of Medicine, Winston-Salem, NC); Michael Simon (Wayne State University School of Medicine/Hutzel Hospital, Detroit, MI).

WHI Memory Study: Sally Shumaker (Wake Forest University School of Medicine, Winston-Salem, NC).

Grant Support: National Heart, Lung, and Blood Institute, NIH, U. S. Department of Health and Human Services (contracts HHSN268200764314C, N01WH22110, 24152, 32100-2, 32105-6, 32108-9, 32111-13, 32115, 32118-19, 32122, 42107-26, 42129-32, and 44221). Clinical Trials Registration: ClinicalTrials.gov identifier: NCT00000611. R.L. Prentice was partially supported by grant CA53996 from the National Cancer Institute.

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.