We investigated the FGFR2-breast cancer association in greater detail, examining the association of four intronic SNPs within strata defined by clinical and pathological characteristics, and the modification of the association by other breast cancer risk factors and exogenous exposures including smoking, alcohol, and obesity.
is a member of the receptor tyrosine kinase family, involved in mammary gland proliferation and development (23
). It has been shown that FGFR2
can transform normal human mammary epithelial cells and is over-expressed in up to 15% of breast tumors (25
). Several transcription binding sites, including a putative ER-binding site, have been indentified in intron 2 of FGFR2
), and SNPs in this region have been reported to be associated more with ER-positive than ER-negative breast tumors in several studies (5
). We did not find a difference in the association by ER status. We did find significant positive associations of FGFR2
variants for PR-positive cases, consistent with previous reports in Dutch and Irish studies (4
). PR can be directly activated in a ligand-independent fashion through the MAPK
pathway which can be stimulated by FGFR2
). Additionally, the FGF2/FGFR2
axis may have a role in PR activation, inducing hormone-independent mammary tumor growth most likely through MAPK
pathway activation (28
In this study, although we did not replicate the predominant association with ER-positive tumors, we did identify significant association of FGFR2
variants among women with risk factors related to endogenous estrogen exposure such as lower number of births. In this population, we did not see an association with lymph node positive tumors as reported previously (12
), although it should be noted that those previous findings were not consistent with the overall trend toward association with risk among those with favorable tumor characteristics (5
). Similar to what others have found (6
), we found that there was a significant association of FGFR2
and breast cancer among those with a history of breast cancer in a first degree family member.
One of our most interesting results is that the association of FGFR2
SNPs and breast cancer was limited to current and former smokers, with no significant association in never smokers. Smoking has been associated with breast cancer risk in several studies, but to date there is inconclusive epidemiological evidence for a strong significant association (29
), and modification by unknown susceptibility alleles may be partly to blame. There is evidence suggesting that smoking may modify the association between genetic factors and breast cancer risk, e.g. smokers that are BRCA1/2
mutation carriers have a significant increase in risk (31
). A recent meta-analysis and pooled analysis found that variants in the N-acetyltransferase 2
) modify the association between cigarette smoking and breast cancer risk; increased risk is associated with smoking among those with the NAT2
-slow acetylator genotype (33
). In our study, we found effect modification of the association between FGFR2
SNPs and breast cancer risk by smoking, with a more than two-fold increase in breast cancer risk for smokers with the variant genotype. Interestingly, there was no indication that the amount of cigarettes smoked had any additional effect per analyses of pack years. One limitation is the amount of never smokers in our study, which may have limited our power to detect associations in this group. Besides a direct carcinogenic effect, tobacco constituents induce and promote cell proliferation, invasion and epithelial-mesenchymal transformation in several types of cells, including normal mammary epithelial cells and breast cancer cells, e.g., by activating the MAPK
pathway either independently or most likely through growth factor receptors (35
). It is possible that smoking activates the MAPK
pathway either directly or by increasing the FGFR2
activity, which in turn could be already over-expressed in carriers of the variant genotype. Smoking could potentially act synergistically with, or directly on, the FGFR2/MAPK
pathway to induce and promote the carcinogenic process in the breast. If this were the case, the effect might be independent of dose.
Alcohol consumption is another well established risk factor for breast cancer (37
). While the mechanism of the observed association is unknown, the established impact of alcohol on estrogen levels may explain this phenomena, at least in part (37
). In our study, both lifetime consumption and the intensity of alcohol drinking did not modify the association of the investigated variants and breast cancer.
There is consistent evidence that higher BMI and central adiposity is associated with increased postmenopausal breast cancer risk possibly through hormone-mediated mechanisms (39
). When investigating adiposity expressed by BMI in interaction with the FGFR2
variants and breast cancer, we found no significant interaction in the entire sample or in pre- or post-menopausal women separately. The gene-only results for FGFR2
and breast cancer in our study were similar for pre- and postmenopausal women, which may suggest that the association of adiposity and breast cancer is independent of the FGFR2
-breast cancer pathway.
To our knowledge, this is the first study to examine interactions between exogenous exposures such as alcohol consumption and smoking and FGFR2
genotypes in a Caucasian population. However, the interest in gene-environment interaction in the context of FGFR2
-breast cancer association is emerging; one study in the Japanese population has investigated the interaction with smoking and drinking with null results, and another study nested within the Women's Health Initiative investigated the interaction with a low-fat dietary pattern (15
). In general, our findings are consistent to the study in the Japanese population, except for the association with smoking, suggesting that the effect of FGFR2
variants on breast cancer risk is similar in different racial groups as also reported by two recent meta-analyses on the subject (42
). The discrepancies regarding the association with smoking might be due to the differences in smoking pattern among these two populations or the fact that the Japanese study had a much smaller sample size compared to ours, and therefore very small numbers in different strata.
Taken together, these data support the prevailing hypothesis that FGFR2
variants play a role in breast cancer etiology, particularly for women with a smoking history. The previously reported association with hormone receptor positive tumors found by previous studies could simply be an artifact of the increased proportion of receptor-positive cases in all data sets investigated so far, as suggested recently (14
A hormone-mediated mechanism would likely not be consistent with our finding that smokers who carry FGFR2
variant alleles are at higher risk of developing breast cancer, in that smoking is thought to have an anti-estrogenic effect (44
). However, several epidemiology studies have not found significant differences in endogenous estrogen levels between smokers and non-smokers (45
) and the observed modification by smoking may be related to other effects of smoking on the tissue environment. On the other hand, we observed a consistent association of FGFR2
SNPs with increased breast cancer risk within women with low parity, consistent with previous reports in other populations (8
), however we did not find increased risk with parity in the presence of FGFR2
minor alleles as compared to parity in the presence of common FGFR2
genotype. These reproductive risk factors are thought to be related to the endogenous hormone exposure, and thus the hormone mediated mechanism underlying the association of FGFR2
variants and breast cancer risk cannot be ruled out, suggesting that tobacco and hormonal exposures may be independent factors. Interestingly, a recent study of hormone therapy within the Women's Health Initiative trial showed an interaction of one FGFR2
SNP and breast cancer risk in case-only analyses (17
). Their findings suggest that in women who carry two variants of rs3750817, estrogen alone is protective with regard to breast cancer risk. Clearly, large well-powered replication studies are needed to support this hypothesis. Together, our findings of a smoking interaction and these purported protective effects of estrogen in the presence of FGFR2
minor alleles suggests that estrogen plays a complex role in FGFR2
-mediated breast cancer risk.
Consistent with the general epidemiological data on breast cancer, the cases in our study were significantly different when compared to controls in terms of age, parity, family history and the presence of previous benign breast disease. Although there seem to be an indication of a slight increase in risk for carriers of one or two copies of the minor allele in combination with these factors in our study, the only multiplicative term of interaction that reached significance was parity as discussed above. There are several other known genetic risk factors for breast cancer such as high penetrance mutations (e.g. BRCA1
). If the case sample contains a significant number of BRCA
carriers or breast cancer with familial aggregation which could contribute to the unbalanced number of first-degree relatives with breast cancer between cases and controls, it may have an effect on the risk estimates in our study, as recently suggested (47
). However, we could not investigate this in our study due to the lack of data regarding high penetrance mutations and/or familial aggregation.
The strengths of our study include the population-based design and well characterized clinical and pathological features of cases, as well as extensive data collected on demographics and other exposures such as lifetime smoking and drinking. We believe the wealth of information particularly in regard to breast cancer tumors, lifetime exposure to tobacco smoke and alcohol intake in this study presented a unique opportunity to both test hypotheses regarding the association of genetic variants with tumor phenotypes and generate new hypotheses via investigation of gene-environment interactions.
We performed multiple statistical tests herein and therefore cannot rule out the possibility of our findings being due to chance. In many cases the strata were small, thus limiting power to detect such effects. It is important to note that the significant interaction of FGFR2 and smoking does not remain significant with adjustment for multiple testing. There is a great deal of further statistical and biological confirmation of this putative interaction needed as well as a need to test whether these results are generalizable to women of other race/ethnic groups.
Little is known about the functional role of these FGFR2 SNPs and the biologic mechanisms by which they contribute to breast cancer etiology. The interaction of FGFR2 variants with smoking may provide some insight into the etiology and the biological mechanisms underlying breast carcinogenesis, hence generating hypotheses for further studies that could have implications in prevention, detection and treatment.