Polymorphisms in genes involved in the estrogen pathway and mammographic density

There is substantial evidence that steroid hormones such as estrogens play an important role in the etiology of breast cancer. In particular, estrogens are believed to contribute to tumor growth by promoting the proliferation of cells with existing mutations or perhaps by increasing the opportunity for mutations [1]. Therefore, several enzymes and receptors involved in the estrogen pathway have been suggested to play a role in the development of breast cancer [1, 2]. Among them, 17β-hydroxysteroid dehydrogenase 1 (HSD17B1) is the enzyme responsible for the conversion of estrone to estradiol [3] which is the most potent estrogen. Cytochrome P450 1B1 (CYP1B1) catalyses the conversion of estrone and estradiol to potentially carcinogenic catechol estrogen 4-hydroxyestrogen (4-OH) [4‐6]. The catechol-O-methyltransferase (COMT) enzyme is principally responsible for both the inactivation and detoxification of catecholestrogens, which can cause oxidative damage [4, 7]. Both estrogen receptors alpha and beta (ERα, ERβ) are members of the nuclear receptor family of ligand-inducible transcription factors, but they are believed to have different transcriptional activation properties [8]. The interaction of estrogen and estrogen receptors stimulates cell proliferation which is believed to be one of the mechanisms responsible for the carcinogenicity of estrogens in the human breast [1]. Specific polymorphisms (SNPs) in these estrogen-related genes could directly or indirectly lead to variations in their activities and therefore, may have an effect on breast cancer risk. For instance, certain SNPs located in ERα, ERβ, HSD17B1, COMT and CYP1B1 genes have been associated with breast cancer risk among premenopausal or postmenopausal women [9‐12] or in some strata of other estrogen-related factors such as parity, hormone replacement therapy (HRT) use, age at menarche or body mass index (BMI) [9, 13‐15].

Those SNPs located in ERα, ERβ, HSD17B1, COMT and CYP1B1 genes may also be associated to mammographic density (MD), which is a strong and independent risk factor for breast cancer [16]. It has been repeatedly found that women with 75% or more of MD have a four to six fold greater risk of developing breast cancer compared to women with no measurable dense breast tissue [16, 17]. Elevated MD represents a higher proportion of fibroglandular cells and therefore reflects a higher proliferative activity within this tissue [18]. The link between MD and breast cancer is not well understood but it could be related to estrogen. For instance, in addition to the proliferative effect on fibroglandular cells of the breast by endogenous or exogenous estrogens [1, 19], several breast cancer risk factors related to estrogens such as age, parity, menopausal status or HRT use, are also associated with MD [16, 17]. Because MD is a highly heritable factor [20], a number of studies examined the association of SNPs located in estrogen-related genes with MD [21‐33]. In these studies, although no associations were found among the entire population, some associations were observed in certain strata of estrogen-related factors such as menopausal status or the use of HRT [23, 28, 29, 31, 32]. Up to now, few studies were conducted among premenopausal women [28‐33] and no data is available for stratified analysis based on estrogen-related factors, even though heritability of MD has been suggested to be higher in this population [34]. The purpose of this study was to evaluate the relation between 11 polymorphisms in five genes involved in the estrogen pathway (ERα, ERβ, HSD17B1, COMT and CYP1B1) and premenopausal MD, and to examine the modifying effect of four estrogen-related factors (parity, hormonal derivatives use, age at menarche and BMI) on these associations.

Characteristics of the study population are summarized in Table 2. Of the 741 premenopausal women, 739 reported to be Caucasian (99.7%). The mean age at the time of mammogram was 46.8 years (standard deviation (SD) 4.6 years). The mean percent and absolute density were 42.3% (SD 24.3%) and 47.1 cm² (SD 28.7 cm²) respectively.

Relation between polymorphisms in genes involved in the estrogen pathway and MD are presented in Table 3. Women carrying increased number of copies of the rare allele of CYP1B1 rs1056836 SNP had decreased adjusted means of percent (P _trend = 0.09) and absolute (P _trend = 0.08) density. However, no SNP in ERα, ERβ, HSD17B1, COMT and CYP1B1 genes was significantly associated to percent or absolute density (all P _trend ≥ 0.05). Moreover, we observed no significant association of any haplotypes within ERα or HSD17B1 genes with percent or absolute density (data not shown).

Table 4 presents the effect modification of parity, hormonal derivatives used, age at menarche and BMI found statistically significant on the relation between polymorphisms in estrogen-related genes and MD. The results from these analyses regardless of statistical significance are detailed in Additional files 1, 2, 3 and 4. Among nulliparous women, an increase of rare alleles of CYP1B1 rs1056836 SNP was negatively associated with AD (P _trend = 0.004), while no association was observed among parous women (P _trend = 0.62; P _i = 0.02). Effect modification of parity on the association between COMT rs4680 SNP and AD was also observed (P _i < 0.05), but no association reached statistical significance among nulliparous (P _trend = 0.15) or parous (P _trend = 0.12) women. We found that an increase in the number of rare alleles of the HSD17B1 SNP (rs598126 and rs2010750) was associated with a decrease in AD among women who ever used hormonal derivatives (P _trend = 0.04 and P _trend = 0.08, respectively), but with an increase in AD among women who never used them (P _trend = 0.06; P _i = 0.02 and P _trend = 0.04; P _i = 0.01, respectively). Moreover, among women who never used hormonal derivatives, an increase in the number of rare alleles of the COMT rs4680 SNP was found to be negatively associated with AD (P _trend = 0.02; P _i = 0.03). These associations and effect modifications remain similar when the analysis is limited to women who ever or never used oral contraceptive (data not shown). Effect modification of age at menarche on the association of COMT rs4680 SNP with AD was also observed (P _i = 0.02), revealing a negative association between an increase of rare alleles of this genotype and AD limited to women with an age at menarche above the median value (P _trend = 0.03). Among women with a BMI above the median value, an increase in the number of rare alleles of the HSD17B1 rs598126 SNP was associated with a decrease in AD (P _trend = 0.01), while no association was observed among those with a BMI below or equal to the median value (P _trend = 0.53; P _i = 0.02). When percent density was used as the dependent variable, similar trends were observed but they were less statistically significant. No significant effect modification of estrogen-related factors on the association of haplotypes within ERα or HSD17B1 genes with percent or absolute MD was observed (data not shown).

In the present study, we evaluated the association of 11 polymorphisms in five genes involved in the estrogen pathway with MD, and found no overall association in premenopausal women. However, some SNPs located in the CYP1B1, COMT or HSD17B1 genes were found to be associated with MD among nulliparous women, non-hormonal users, those with late menarche or among women with high BMI. These results highlight the need to consider the modifying effect of estrogen-related factors when evaluating associations of SNPs involved in the estrogen pathway with MD.

Although no overall association was found between three SNPs (rs598126, rs2010750 and rs676387 in strong linkage disequilibrium) in the HSD17B1 gene and MD, we found some associations within strata of BMI and hormonal derivatives used. We observed a decrease in MD with an increasing number of the rare allele of rs598126 SNP among women with higher BMI or past hormonal derivatives users. In contrast, among women who never used hormonal derivatives, MD increased by about 20 cm² or 14% for women who are homozygote for the rare allele (TT for rs598126 or AA for rs2010750) compared to women homozygote for the common allele (CC or GG respectively), and this association was borderline or statistically significant even with a restricted number of women in this stratum. To our knowledge, these SNPs have not been investigated in relation with MD. However, Feigelson et al. who reported no evidence of an association between HSD17B1 rs598126 or rs2010750 SNP and overall risk of breast cancer, found that, among tumors that were negative for the estrogen receptors (ER), additional copy of rare alleles was associated with an increased risk of breast cancer [41]. It has been suggested that ER-negative tumors have less estrogenic activity demonstrated by lower levels of estrogen as compared to ER-positive tumors [46]. Since women who never used hormonal derivatives are believed to have less estrogenic activity than women using them [47], results from Feigelson and colleagues on breast cancer risk seem consistent with ours on MD. Little is known about the biology of SNPs in HSD17B1 gene, therefore these associations need to be confirmed in further studies.

The evaluation of the relation between CYP1B1 rs1056836 SNP and MD did not show a significant association in our overall population of premenopausal women which is consistent with other studies conducted in premenopausal [28, 32, 33] as well as in postmenopausal [21, 22, 28] women. However, we observed that, among nulliparous women, carriers of increasing number of G (Val) allele of CYP1B1 gene have lower MD. Since nulliparous women have been related to higher estrogenic activity then parous women [48], our results suggest that an increasing number of the Val allele is related to lower MD among women presenting higher estrogenic activity. Although this is the first study to find such association among nulliparous women, Lord et al. observed a similar one among another group of women presenting higher estrogenic activity; postmenopausal hormone (PMH) users [23]. Using data from clinical trials, they reported that among women that used combined estrogen and progesterone for the previous year, the increase in MD was lower for women carrying Val/Val genotype than those carrying Leu/Leu genotype [23]. However, no interaction between CYP1B1 rs1056836 SNP and PMH status (current users vs. never or past use) was reported in the cross-sectional study conducted by Haiman et al. [28]. The fact that the activity of the CYP1B1 enzyme is regulated by estrogen may explain why associations of CYP1B1 rs1056836 SNP with MD are observed among women presenting higher estrogenic activity. However the biological effect of this polymorphism is still unclear. Data from laboratory studies suggest that the Leu allele might have an increased action in the activation of mammary procarcinogens [49] known to stimulate cell proliferation [50, 51]. Conversely, the substitution of leucine to valine in position 432 of the CYP1B1 enzyme has been related to a higher production of 4-OH, increasing the 4-OH/2-OH ratio which could lead to higher mitogenic stimulation of breast cells [49].

Among studies that examined the relation of SNP rs4680 in the COMT gene with MD, most of them [22, 23, 28‐31], although not all [31, 32], found no association among premenopausal [28‐30] or postmenopausal [22, 23, 28, 30, 31] women. However, two studies conducted in multiethnic premenopausal women suggested that MD decreased with each additional copy of the A (Met) allele [31, 32]. Our findings support a similar association but only among women that had lower estrogenic activity like among those who never used hormonal derivatives or had a late menarche [47, 52]. To our knowledge no other study evaluated the possible modifying effect of age at menarche, but three examined the modifying effect of PMH use on the relation between COMT rs4680 SNP and MD [23, 28, 29]. In one study, Haiman and colleagues reported that an increase in the Met allele was associated with a decrease in MD among never and past PMH users, while no association was observed among current users [28]. In another study conducted by the same group, an increase of A (Met) alleles was associated to an increase in MD for current PMH users, and no association was found for never and past users [29]. However, data from a clinical trial showed no association within strata of different PMH uses (estrogen or combined estrogen and progesterone) [23]. Estrogen is known to down-regulate the activity of the COMT enzyme [44] which inactivates by O-methylation the catechol metabolite of estrogen 4-OH. The variant allele of the gene coding for COMT, the Met allele, has been associated to lower enzymatic activity [53, 54]. Therefore, the Met allele should be associated with an increased proliferative activity, because of less inactivation of 4-OH in presence of estrogen. It would be biologically reasonable to presume that additional copies of the Met allele would be associated with an increase in MD, particularly among women with increased estrogenic activity. Up to now, observations from our group and others suggest an increase in MD with additional copies of the Met allele among women with higher estrogenic activity (hormonal derivatives users) or a decrease in MD with additional copies of the Met allele among women with lower estrogenic activity (late age at menarche or those who never hormonal derivatives) [28, 29].

We evaluated the relation of SNPs in ERα or ERβ gene with MD among premenopausal women only, and our results showed no significant association or effect modification by estrogen-related factors. Up to now, the association between SNPs in ER genes and MD has been inconsistent [22, 24, 26, 27, 33]. Crandall et al. found no association between SNP rs9340799 (ERα) or rs1256049 (ERβ) and MD among premenopausal women but the genotype homozygote for the rare allele of SNP rs2234693 (ERα) was related to a higher MD [33]. Among a population composed of premenopausal and postmenopausal women, MD was not related to ERα rs2234693 SNP but it had a negative association with the rare allele of rs9340799 SNP [26]. Among postmenopausal women, no association was found for ERα rs2234693 [22] or rs2077647 [24] SNP and MD, but the rare allele of rs9340799 SNP was related to a higher MD [24].

This study has some limitations. We cannot exclude that our findings may be due to chance because we evaluated several associations. Type I errors or false-positive results are therefore possible. Population stratification can be a concern in this type of study [55, 56]. However, it might not be a problem here because our population was mostly composed (99.7%) of Caucasian women and of over 87.7% French-Canadians [39]. Non-differential misclassification bias from genotyping measurements is possible although it is unlikely in the absence of Hardy-Weinberg disequilibrium and complete concordance for repeated samples. Non-differential misclassification of MD is also possible. However, if present, it should be relatively small and most likely random because all of the mammograms were performed at the same clinic and MD was assessed blindly by one observer whose reliability of reading was shown to be high.

This study proposes that some SNPs located in the CYP1B1, COMT or HSD17B1 genes are associated with MD among nulliparous women, non-hormonal users, those with late menarche or among women with high BMI. Our findings suggest that modifying effects of estrogen-related factors should be considered when evaluating associations of polymorphisms in estrogen-related genes with premenopausal mammographic density.