Associations between Androgen and Vitamin D Receptor Microsatellites and Postmenopausal Breast Cancer Free

The role of androgen stimulation in breast carcinogenesis has been disputed. Although in vitro experiments indicate that androgens inhibit breast cell growth (1, 2), higher circulating androgen levels in breast cancer cases compared with controls (3) point to a possible adverse influence of androgenic stimulation in breast cancer tissue. Androgens act through the androgen receptor (AR), which is genetically polymorphic. There is a (microsatellite) trinucleotide repeat polymorphism (CAG_n) in exon 1 of the gene. This polymorphism affects the transactivation capacity of the receptor; the longer the repeat the less efficient the transactivation (4-6). In line with this, the short CAG_n has been associated with an increased risk of prostate cancer (7-10), and the long CAG_n with male infertility (11-13). Investigations of the relation between the AR polymorphism and breast cancer (Table 1) have been conflicting, variably reporting that short repeats are associated with a decreased risk (14-18) or are not associated with risk (19-21) or are associated with decreased breast cancer survival (16, 22, 23).

1α,25-Dihydroxyvitamin D3 (vitamin D) is another steroid hormone involved in cell growth and differentiation (24, 25). Vitamin D acts via a specific nuclear receptor, the vitamin D receptor (VDR). There are several strongly linked polymorphisms in the 3′ untranslated region of VDR that are of unclear functional significance (26-28) but that nevertheless have been associated with risk of prostate cancer (29-31) and osteoporosis (28, 32–34). Some investigators have also shown associations between variants in this region of the gene and breast cancer risk (35–40), whereas others (20, 41–44) have shown no association (Table 2).

We have chosen to study the AR CAG microsatellite and a polyadenylic acid [poly(A)] microsatellite in the 3′ untranslated region of VDR in relation to breast cancer in a large case-control study in a genetically homogenous population.

As described previously in detail (45), this nationwide population-based case-control study encompassed all incident cases of primary invasive breast cancer among women 50 to 74 years of age resident in Sweden between October 1993 and March 1995. Cases of breast cancer in situ were not included. Breast cancer patients were identified at diagnosis through the six Swedish regional cancer registries, to which reporting of all malignant tumors is mandatory. All Swedish residents are assigned a unique national registration number. This number is recorded in all registries, including the Total Population Register. It is possible for researchers, provided that the appropriate permissions are granted, to approach the authority in charge of the Total Population Register (currently the Tax Authority) and ask for national registration number and addresses of people that fulfill certain criteria specified by the researcher. Control women were randomly selected from the general population according to the expected age frequency distribution (in 5-year age groups) of cases.

Cases were asked to participate in the study by their respective physicians. When patients consented, they received a mailed questionnaire asking for detailed information about intake of menopausal hormones and oral contraceptives, weight, height, reproductive history, medical history, and other lifestyle factors. Controls were contacted directly with the questionnaire. Eighty-four percent (n = 3,345) of eligible cases and 82% (n = 3,454) of the controls ultimately participated in the parent study. Among the participating controls, 455 who failed to return the mailed questionnaire were interviewed by phone. Results from the parent study are available in previous publications (45-48).

We randomly selected 1,500 women with invasive breast cancer and 1,500 controls (frequency matched by age) among postmenopausal participants without any previous malignancy (except in situ cervix carcinoma or nonmelanoma skin cancer) in the parent study. To increase statistical power in subgroup analyses, we additionally selected all remaining eligible cases and controls who had taken menopausal hormone treatment (either preparations with medium potency estrogen only, mainly estradiol and conjugated estrogens, or medium potency estrogen in combination with progestin) for at least 4 years (191 cases and 108 controls) and all women with self-reported diabetes mellitus (110 cases and 104 controls). In total, 1,801 cases and 1,712 controls were selected. In addition, 345 controls from the parent study selected for a parallel endometrial cancer study (49) who fulfilled the inclusion criteria could be added to our sample of breast cancer–free controls. The present study was approved by the Institutional Review Boards at Karolinska Institutet and in the six other Swedish regions and was done in compliance with the Helsinki Declaration.

We contacted all selected living women by mail and those who gave informed consent received a blood sampling kit by mail. Whole blood samples were drawn at a primary health care facility close to the woman's home. Breast cancer cases who declined to donate a blood sample were asked to authorize our use of archived paraffin-embedded tissue taken at breast cancer surgery. We also attempted to retrieve archived tissue samples from all deceased breast cancer cases. We obtained blood samples from 1,322 (73% of selected cases) and archived tissue samples for 247 (14% of selected cases; total participation rate among gases is 87% of all selected) breast cancer patients. Among the chosen control women, 1,524 (74%) contributed blood samples. Reasons for nonparticipation included lack of interest in or skepticism about genetic research and, in some instances, advanced disease or death. We thus obtained final population-based participation rates of 73% and 61% in cases and controls, respectively.

We isolated DNA from 3 mL whole blood using Wizard Genomic DNA Purification kit (Promega) according to the manufacturer's instructions. From nonmalignant cells in paraffin-embedded tissue, we extracted DNA using a standard phenol/chloroform/isoamylalcohol protocol (50). Slides from each block were scrutinized by a pathologist. Areas found to contain malignant cells were marked on the slides and removed from the 50-μm cuts used for DNA extraction.

We amplified fragments corresponding to the CAG_n in the AR gene and the A_n in the VDR gene by PCR using the following primers: 5′-AGAGGCCGCGAGCGCAGCACCTC-3′ (AR, forward), 5′-GCTGTGAAGGTTGCTGTTCCTCAT-3′ (AR, reverse), 5′-GTGTAGTGAAAAGGACACCGGA-3′ (VDR, forward), and 5′-GACAGAGGAGGGCGTGACTC-3′ (VDR, reverse). A “touch-down” PCR was used, in which both reactions were simultaneously run in an ABI Prism 877 Integrated Thermal Cycler robot (PE Applied Biosystems). We used AmpliTaq Gold kits and standard reagents (Applied Biosystems). The amplification profile consisted of denaturation at 95° for 10 min followed by 36 cycles of denaturation at 96° for 30 s, annealing at 59° to 57° for 40 s, elongation at 72° for 60 s, and final extension at 72°. The annealing temperature was 59° in the first 3 cycles, 58° in the following 12 cycles, and 57° in the last 21 cycles. We set up separate PCRs for samples that could not be amplified in the touch-down reaction. These reactions were done on a GeneAmp PCR System 9700 (Perkin-Elmer Co.) programmed for denaturation at 96° for 10 min followed by 36 cycles of denaturation at 96° for 30 s, annealing at 55° or 56° for 40 s, elongation at 72° for 60 s, and final extension at 72° for 7 min. The amplification products were read on a Genescan run ABI 377 DNA gel-slab electrophoresis sequencer (Perkin-Elmer) with a TAMRA-labeled internal length standard (Genescan-500 TAMRA, Applied Biosystems). We used Genotyper software to determine the genotypes (Genotyper version 2.0, Perkin-Elmer).

We were able to successfully genotype 1,542 breast cancer cases for the AR microsatellite and 1,511 cases for the VDR microsatellite and 1,260 controls for both polymorphisms. For both the AR CAG_n and VDR A_n, the exact number of repeats for a range of fragment lengths was determined by direct DNA sequencing of the fragments (data not shown). In the following statistical analysis, we included 251 additional genotyped controls from the same source population that were genotyped for the parallel endometrial cancer study (see above). Thus, the total number of controls included was 1,511.

We determined whether AR and VDR genotype frequencies were in Hardy-Weinberg equilibrium using the web version of the Genepop software (51). Based on a priori decisions, we dichotomized the AR CAG_n at the median repeat length among controls (22 repeats) and VDR A_n between the two peaks in the bimodal distribution of repeat lengths among controls (18 repeats). In secondary analyses, we also used cutoffs determined after visually examining the distributions of allele lengths among cases and controls.

We calculated odds ratios (OR) and 95% confidence intervals (95% CI) from conditional logistic regression models using maximum likelihood methods. We conditioned on the variables used for selection (i.e., age in 5-year categories, use of menopausal hormones for 4 years or more, and self-reported diabetes mellitus). For detailed descriptions of how the above variables were defined, please see ref. 47. All covariates were introduced into the logistic regression model to detect confounding, indicated by changes in risk estimates, or other associations potentially affecting the primary association between genotypes and breast cancer. We investigated interactions between AR or VDR genotype and duration of menopausal hormone use, body mass index, parity, diabetes mellitus, and family history of breast cancer by doing separate analyses over strata of these exposures. Formal tests for interaction were done by comparing models containing interaction terms with models containing only main effects using likelihood ratio tests.

We did all analyses using SAS system PHREG procedure (release 8.02, SAS Institute, Inc.).

Selected characteristics of the breast cancer cases and controls (Table 3) largely conformed to known epidemiologic breast cancer case-control differences. On average, cases had fewer children, were older at first birth, heavier, and more often had a family history of breast cancer. We found 30 alleles of the AR microsatellite (range, 6-43 CAG_n) that were approximately normally distributed (Fig. 1). In the VDR microsatellite locus, we found 21 alleles that were bimodally distributed (range, 7-34; Fig. 2). The genotype frequencies at the two loci were not in Hardy-Weinberg equilibrium, neither when dichotomized nor when all alleles were considered (P < 0.001 for both AR and VDR microsatellites). When we considered the alleles as long or short among controls, there were no associations between AR genotype and other known or suspected breast cancer risk factors (data not shown). However, the VDR SS genotype was more common among women who had their first child above the age of 30 years (P = 0.02) and the VDR LL genotype was more common among women who had a lean body build at ages 7 and 18 years (P = 0.07 and 0.05, respectively; data not shown).

There was no difference in the mean repeat length between cases and controls for the AR (P = 0.22) or the VDR microsatellite (P = 0.96) and no association between mean repeat length and age at breast cancer diagnosis (data not shown). When modeled by logistic regression, neither AR nor VDR genotypes (with alleles divided into long or short) were significantly associated with breast cancer risk overall (Table 4). Similarly, there was no association when breast cancers were subclassified into ductal and lobular types (Table 4). When we considered only short alleles, only long alleles, or the sum of alleles as continuous variables, we also found no significant association. The ORs for each unit increase in length for short, long, and sum of alleles for AR were 0.99 (95% CI, 0.96-1.03), 0.98 (95% CI, 0.96-1.01), and 0.99 (95% CI, 0.98-1.01), respectively. For the VDR, these estimates were 1.00 (95% CI, 0.98-1.02), 1.00 (95% CI, 0.97-1.02), and 1.00 (95% CI, 0.99-1.01), respectively. There was no indication of interaction between AR and VDR genotypes in effects on breast cancer risk (P_interaction = 0.50). In subgroup analyses (Tables 5 and 6), we found only one statistically significant interaction, namely one between VDR genotype and parity. Women with two short alleles had a reduced risk for breast cancer, irrespective of parity, compared with nulliparous women with two long alleles.

In exploratory analyses, we defined new cutoffs that were intended to maximize the contrasts between cases and controls. With these cutoffs, carrying two short (<20) AR CAG repeats or two short (<21) VDR poly(A) was associated with increased breast cancer risks overall, ORs of 1.63 (95% CI, 1.13-2.35) and 1.25 (95% CI, 1.03-1.51), respectively. These overall associations were also present in subgroup analyses; no indications of interaction emerged using these cutoffs and the VDR-parity interaction was weakened (data not shown).

There was no association between AR genotype and histologic type, tumor size, or clinical stage at diagnosis (data not shown). AR genotype and estrogen receptor status in the tumor were not associated among the 65% cases for whom receptor information was available. VDR genotype was associated with stage at diagnosis (P = 0.05). Homozygosity for the short allele was overrepresented among stage I tumors and homozygosity for the long allele overrepresented in stage IV. VDR genotype was not associated with any other clinical characteristic (data not shown).

We show that AR and VDR microsatellites do not have any substantial influence on the risk of postmenopausal breast cancer among Swedish women. Shorter alleles at one or the other locus might entail a slightly increased risk, but these findings were data driven and should be interpreted with considerable caution.

Our study was population based and large and it was done in a genetically homogenous population. The latter limits the potential for confounding by population stratification. There are no convincing reasons to believe that differential participation associated to genotype would operate to cause selection bias. We had extensive information about other breast cancer risk factors, which enabled us to evaluate effect modification or confounding. The genotyping methods that we used are well established. The laboratory personnel were blinded to case-control status and could thus not have scored genotypes systematically with regard to case-control status.

The genotype frequencies of the AR and VDR microsatellites, although similar to previously published reports, were not in Hardy-Weinberg equilibrium. The reason for this deviation is unclear. The Swedish population is not particularly inbred, and there is no reason to believe that there has been recent mutations at the AR and VDR microsatellite loci nor significant genetic admixture. It is very unlikely that our study participation was dependent on genotype. For quality control reasons, we repeated the genotyping analyses of ∼1% of our samples with identical results, which indicates that our allele calling procedure was reproducible. It is possible, however, that preferential amplification of the shorter of two alleles occasionally occurred in heterozygotes. In samples of very low DNA yield, one of the peaks of a heterozygote may have been below the detection limit of our assays. Factors such as these would have caused misclassification of some heterozygotes as homozygotes and could explain the deviation from Hardy-Weinberg equilibrium. Such misclassification would weaken observed associations, especially for recessive or codominant penetrance. This scenario has to be mentioned as a possible reason for our finding of nonassociation despite previous reports (14-16, 35-40) of significant associations.

The role of androgens in the development of breast cancer is complex. Whereas in vitro experiments show that androgen stimulation inhibits the stimulatory effect of estrogens in breast epithelium (52), epidemiologic studies indicate that high circulating levels of androgens confer an increased breast cancer risk (3), possibly because androgens are precursors for estrogens. Recently, there has been an increased interest in the feasibility of administering androgens to women with menopausal symptoms, the underlying idea being that the side effect profile, including cancer risks, of androgen therapy may be more beneficial than that of estrogen or estrogen-progestin therapy (53). Because we do not see an effect of androgen signaling, our results suggest that the androgen strategy might indeed be safer, at least in relation to breast cancer risk.

Several studies have established that long CAG_n causes reduced AR transactivation (4, 6). In line with this, there is also fairly consistent evidence for an increased risk of prostate cancer with short CAG_n, (consistent with enhanced androgen signaling) and an increased risk for male infertility with long repeats (consistent with attenuated androgen signaling). Because AR is located on the X chromosome, men only have one copy of the AR gene. Women, on the other hand, have two copies of the gene and one of them is inactivated, most likely in a random fashion, at least with regard to CAG length. Unmeasured X inactivation status may be one explanation for the divergent results about the influence of AR CAG_n on diseases and conditions in women. In a recent investigation, short CAG_n was associated with an earlier age at menarche (54), whereas Westberg et al. (55) found that short repeats were associated with higher androgen levels in women.

Several previous studies (Table 1) have shown long CAG_n to be associated with increased sporadic breast cancer risk among women (14-18), with breast cancer risk among men (56) and increased risk among BRCA1 (57) or BRCA1/2 (58) mutation carriers. Other studies, now supported by us, have reported no association (20, 21, 59), no modifying effect on BRCA1/2 mutation penetrance (60), and no association with age at breast cancer presentation (61, 62). However, there is one report that breast cancers in women with shorter CAG are of higher grade and confer a shorter survival (22).

Vitamin D is involved in cell growth and differentiation (24, 25) and seems to have antiproliferative effects (63). Although VDR is expressed in normal as well as malignant breast tissue (64), the functional significance of genetic variants of the receptor is unresolved (26, 27). Previous studies of the receptor gene variants and breast cancer risk have been conflicting (Table 2), some showing association (35-40, 65) and others showing no association (20, 41-44). The ambiguous state of knowledge in conjunction with our present results point that VDR polymorphism has no major overall influence on the risk for breast cancer. The interaction between VDR genotype and parity present in our data has not been described previously. Bearing in mind the number of comparisons in this study, it probably represents a chance finding. On the other hand, both parity and vitamin D are assumed to influence the breast in a prodifferentiating manner and thus there is a basis for further hypotheses about their interaction.

Our data do not support major roles for AR or VDR polymorphism as breast cancer risk factors.

We thank all women who participated in this study; Anna Christensson, Boel Bissmarck, and Anders Holmberg for providing invaluable people skills and logistic expertise to the project; Hans-Olov Adami for contributing scientific input during all stages of the study; and primary health care centers and pathology departments all over Sweden for cooperating unselfishly in the collection of biological samples.