Vitamin D Receptor Gene Polymorphisms and Breast Cancer Risk

At least 1 in 10 women in the United Kingdom will develop breast cancer at some time in their lives, and there are more than 14,000 breast cancer deaths in the United Kingdom each year (1). Five to ten percent of breast cancer cases are caused by inherited germ-line mutations of genes such as BRCA1 and BRCA2 (2). Although the underlying causes for the development of the remaining sporadic cases are still poorly understood, it is likely that other genes exist that increase susceptibility to breast cancer. Because breast cancer is strongly influenced by the hormonal milieu, variation in genes that are responsive to hormonal activity is a possible candidate for increased risk. Henderson and Feigelson (3) have suggested that genes involved in steroid hormone metabolism and transport could act together to provide a high-risk profile for breast cancer. Variation in these genes might modify breast cancer risk through gene-gene or gene-environment interactions.

One potential candidate is the vitamin D receptor (VDR), a member of the steroid hormone family of nuclear receptors that are responsible for the transcriptional regulation of a number of hormone-responsive genes. VDR is expressed in normal mammary tissue, and more than 80% of breast tumor specimens are VDR positive (4). The natural ligand for VDR, 1,25-dihydroxyvitamin D₃ (1,25-D₃), as well as a number of novel synthetic vitamin D analogues inhibit proliferation and induce apoptosis in breast cancer cells in vitro (5, 6, 7, 8). Furthermore, in animal models of breast cancer, vitamin D analogues slow down tumor development and promote regression of established mammary tumors (reviewed in ref. 9).

Thus, the vitamin D pathway is a potential target for elucidating new breast cancer therapies. The gene encoding the VDR protein is known to display polymorphic variation. A polymorphic site in exon II at the 5′ end of the gene can be identified using the FokI restriction enzyme (10). This polymorphism is a T/C transition (ATG to ACG) at the first of two possible translation initiation sites and results in VDR proteins that differ in length by three amino acids. In individuals with the ACG sequence, initiation of translation occurs at the second ATG site and the three NH₂-terminal amino acids of the full-length VDR are missing. Alleles with an absent restriction site are designated F. In contrast, alleles with the ATG sequence at the first initiation site synthesize the full-length (427 amino acids) VDR protein and are designated f. The FokI f allele has been associated with increased breast cancer risk in African-American women (11). At the 3′ end of the VDR gene, three polymorphisms have been identified that do not lead to any change in either the transcribed mRNA or the translated protein. The first two sequences generate BsmI (12) and ApaI (13) restriction sites and are intronic, lying between exons 8 and 9. The third polymorphism lies in exon 9, generating a TaqI restriction site (14), and leads to a silent codon change (from ATT to ATC); both activities result in an isoleucine at codon 352. These three polymorphisms are linked to a further gene variation, a variable-length poly(adenylate) [poly(A)] sequence within the 3′-untranslated region (3′UTR). The poly(adenylate) sequence varies in length (15) and can be segregated into two groups; “long” (L), where A = _18–24, and “short” (S), where A = _13–17. Because of their close proximity on the VDR gene, there is strong linkage disequilibrium between the 3′ polymorphisms, BsmI, ApaI, and TaqI, and the variable-length poly(A) sequence so that, in Caucasian populations, two haplotypes are commonly observed: baTL (presence of BsmI, ApaI restriction sites, absence of TaqI site, long poly(A) sequence), and BAtS (15, 16). The baTL haplotype has been reported to be associated with increased risk of prostatic carcinoma (17, 18) and breast cancer (19, 20), whereas the BAtS haplotype is associated with increased risk of osteoporosis (16, 21).

In 2001, we carried out a pilot study in 422 Caucasian women to assess whether VDR polymorphisms are associated with breast cancer risk in a United Kingdom population (22). The pilot study was limited because there were very few volunteers with VDR genotype BB/SS, and so it was difficult to perform statistical analyses. In this report, we have extended the number of women in the study to 825, and determined the BsmI, FokI, and poly(A) genotypes of each individual. The aim is to assess more accurately whether specific genotypes are related to breast cancer risk in this Caucasian population, and to determine whether certain genotypes are associated with disease progression, or with pathological factors such as estrogen receptor status and tumor grade. We have also calculated association with breast cancer risk using a “cross-genotyping’” method similar to that described by Whitfield et al. (23), to investigate whether combining the 5′ FokI genotype with the 3′ BsmI/poly(A) genotype would reveal a more positive association with breast cancer risk than with each individually.

A retrospective case-control study was undertaken to investigate whether specific VDR gene polymorphisms are associated with breast cancer risk. The study was carried out exclusively on Caucasian subjects, because the VDR polymorphism distribution varies between different ethnic populations. St. George’s Hospital Medical School Ethics Committee approved the study, and written informed consent was obtained from participants in both control and cancer groups before interview and sampling.

Control women (n = 427) were recruited from the United Kingdom National Breast Screening Programme for South-West London. Women ages 50–65 years are invited for routine mammography screening every 3 years, and women under 50 or over 65 may attend by self-referral. Women recruited from the screening program all had a recent mammogram, confirming that there was no detectable breast cancer at the time of sampling, and all had no personal history of breast cancer. Mammography revealed that 147 women in the study had benign breast disease, the majority of these were benign calcifications, fibrocystic disease, fibroadenoma or other benign lump. These benign conditions are not conventionally associated with increased breast cancer risk. The age range of the control women was 36–80 years, with a median age of 54 years, at the time of sampling (Table 1). There were no age-related differences in VDR gene frequency between the oldest and youngest women. The majority of the control subjects (325 women, 76.1%) were postmenopausal and 229 women (53.6%) were current or past users of hormone replacement therapy (HRT; Table 1).

Of a possible 821 female breast cancer patients, 398 (48.5%) were recruited between the dates July 1, 1998, and October 31, 2002, through the Combined Breast Clinic at St. George’s Hospital. Not all patients were recruited because researchers attended the clinic only on an intermittent basis. Ninety-nine percent of the women who were asked to take part in the study consented. Cases were recruited without prior knowledge of diagnosis or tumor stage, so that as a result, some patients on the study had been diagnosed with their primary breast cancer before 1998 (prevalent cases), whereas some patients were newly diagnosed (incident cases). Details of primary tumor diagnosis, recurrences, and metastases were subsequently obtained from histopathology reports and medical records. Personal information such as number of children, age at menarche, HRT use, alcohol intake, and smoking status was obtained by asking the patient, but, when possible, this was also confirmed by medical records. The median age of the breast cancer volunteers was 56.0 (range, 26–89) years at diagnosis, and 60.5 (range, 27–90) years at the time of sampling (Table 1). Median time from diagnosis to sampling was 3.0 years (range, 0–26 years). A total of 260 women (65.3%) in this group were postmenopausal, and 113 (28.4%) were current or past users of HRT (Table 1). The majority of these women (99.1%) underwent surgery (wide local excision or mastectomy) with or without postoperative radiotherapy. In addition, 295 women were treated with hormonal therapy, 39 with chemotherapy, and 131 with a combination of both. The four women who did not undergo surgery presented with metastases. The characteristics of the tumors were validated from histopathological reports of biopsy and/or resection specimens (shown in Table 2). Tumor grade, estrogen receptor status, and lymph node involvement are presented for patients with invasive breast cancer. Thirty-four women reported a weak family history of breast cancer (defined as possessing one second-degree relative with breast cancer), and 79 women reported a strong family history (defined as having at least one first-degree relative, or two or more second-degree relatives with breast cancer; Table 1).

Polymorphism analysis for breast cancer risk was carried out using status at sampling, and all preceding data were collected at sampling. In addition, to carry out a preliminary investigation into the association of VDR polymorphisms with disease progression, the status of cancer patients was also reexamined at the end of the recruitment period (October 31, 2002), and these results are shown in Table 2. At this point, 40 women had local recurrence, (this number includes 11 women who also had metastases), 12 women had new primary breast tumors, and 58 had metastatic disease. Forty-three patients in the study were deceased at the end of recruitment, 38 from breast cancer. All of the women in the control group were still breast cancer free on October 31, 2002.

For all of the control women and for the 373 women in the cancer group, a 10-ml blood sample was collected into a lithium heparin tube, and whole blood was used for the extraction of DNA. In the remaining 26 cancer patients, DNA was extracted from a sample of their tumor tissue stored in liquid nitrogen. DNA extractions from blood and tissue were carried out using the GenElute mammalian genomic DNA kit (Sigma, Poole, United Kingdom). Genomic DNA was amplified by PCR using primers as described previously: BsmI (16), FokI (10), poly(A) (18). For BsmI and FokI genotyping, PCR product was digested with the appropriate restriction endonuclease (New England Biolabs Ltd, Hitching, United Kingdom), was separated on 1× 5 agarose gels, and was visualized by ethidium bromide staining. BsmI and FokI genotypes were defined by capital letters (B and F, respectively) in the absence of the restriction site and by small letters (b and f) where the restriction site was present. For the poly(A) analysis, a 425-bp PCR product was separated on a 6% PAGE-urea gel and was visualized by staining with a 0.1% solution of silver nitrate (Sigma). Under these conditions, the poly(A) region resolved into two distinct bands, long (L, A₁₈–A₂₄) and short (S, A₁₃–A₁₇).

The χ² test was used to assess any association between VDR polymorphisms and breast cancer risk, and to calculate whether allele frequencies deviated from expected Hardy-Weinberg equilibrium. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated to determine the risk of breast cancer associated with a given VDR genotype. Logistic regression was used to adjust the risk of breast cancer for age, menopausal status and HRT use. Odds ratios were adjusted primarily by using age of controls at sampling, and age of cases at diagnosis because age at diagnosis is related to breast cancer risk. To investigate whether results were biased by recall or time period effects, analysis was repeated and odds ratios were adjusted using age of controls and cases at sampling, because subjects were recruited on the basis of their status at time of sampling. A P value of ≤0.05 was considered as significant. All analysis was undertaken in Stata 7.0 (24).

There is a significant difference in the median age of the cases and controls at sampling (60.5 versus 54 years; P < 0.001; Table 1). However there is no difference in age of controls at sampling and age of cases at diagnosis, (54 versus 56 years; P = 0.72; Table 1). Significant differences were seen in menopausal status (premenopausal, 11.5% versus 30.9%) and HRT use (past or present use, 53.6% versus 28.4%) between the controls and cases (P < 0.001; Table 1); however, the reason that a higher proportion of cases are premenopausal is due to the broader age range of the cancer population. Nevertheless, among only 50–65 year olds, an unbiased sample excluding self-referrals, the distribution is very similar with 55% and 38% of controls and cases, respectively, having currently or previously used HRT. HRT use is lower in the cancer group because HRT use is usually ceased at diagnosis, and a smaller percentage of women will be prescribed HRT after having breast cancer than in the control population.

The frequencies of VDR polymorphisms found in our sample populations are shown in Table 3. There was no deviation from the expected Hardy-Weinberg frequency (P > 0.05) in either the control or cancer population. The 3′ VDR polymorphism BsmI and the variable-length poly(A) sequence are in linkage disequilibrium, such that, in this Caucasian population, bb genotype cosegregated with long poly(A) genotype, LL, in 97.7% of cases and controls (806 of 825). We found that both BsmI and poly(A) genotypes were significantly associated with breast cancer risk; the odds of developing breast cancer for a woman of genotype bb or LL were nearly twice that for a woman of genotype BB or SS, (bb versus BB genotype, OR = 1.75, 95% CI, 1.15–2.66, P < 0.01; LL versus SS, OR = 1.70, 95% CI, 1.11–2.60, P < 0.05; Table 3). Results were adjusted for age, HRT use, and menopausal status, which did not affect statistical significance of the BsmI OR, but slightly increased the significance of the poly(A) OR (Table 3). In addition, no significant difference was observed between genotype frequencies in control subjects with or without benign disease.

The 5′ VDR polymorphism, FokI, was not in linkage disequilibrium with the BsmI or poly(A) polymorphisms, as shown in other studies (15, 25), suggesting that they segregate independently. FokI genotype was not associated with breast cancer risk when analyzed in isolation (Table 3). However, some studies have shown that analysis of two VDR genotypes simultaneously can increase the statistical significance of any correlations found (reviewed in ref. 23). One example of this is in the study by Ferrari et al. (26), who used combined VDR genotype analysis in relation to bone mineral density and found stronger correlations in their study population. Therefore, we have investigated whether a combination of the FokI and poly(A) genotypes is associated with breast cancer risk. Fig. 1 shows the distribution of the combined genotypes, and illustrates that ffSS is the least common genotype in both cancer and control subjects, whereas the heterozygous FfLS is the most common in the control population, and FfLL is the most common in the cancer population. A χ² test including all genotypes showed that the distribution of FokI/poly(A) genotypes is significantly different in women with cancer compared with control women (P = 0.01; Fig. 1). Representing the genotype distributions graphically in this way suggested that the genotypes FFLL and FfLL are overrepresented in the cancer group (Fig. 1). Combining results from FokI and poly(A) analysis, 26% of controls and 38% of cases have genotypes FFLL or FfLL (Fig. 2). There was a highly significant association between FFLL/FfLL genotypes and breast cancer risk compared with all other genotypes (P < 0.001; Table 4). The odds of developing breast cancer for a woman of genotype FF/LL or Ff/LL were almost twice (OR = 1.82; CI, 1.35–2.45; Table 4) those of a woman of any other FokI/poly(A) combination, and this was slightly strengthened once the data were adjusted for age, HRT, and menopausal status (Table 4).

We divided the total cancer group (n = 398) into women with metastatic disease (n = 58) and women without metastatic disease (n = 340) to determine whether there was any correlation between combined VDR genotypes and cancer progression. Statistical analysis for disease progression was carried out using patient status at the end of the recruitment period (Table 2). This revealed that compared with controls, the adjusted OR for women with cancer but no metastases was 2.01 (95% CI, 1.43–2.84; P < 0.001) and this was increased to 3.35 (95% CI, 1.82–6.16; P < 0.001) in cancer patients with metastatic disease (Table 5). The difference in FFLL/FfLL genotypes between the cancer patients who have metastatic disease and those who do not is not statistically significant (P = 0.09). However, this metastatic group is relatively small and results suggest that the FFLL/FfLL genotypes are associated not only with the risk of breast cancer, but might also be associated with disease progression. It must also be remembered that patients in the cancer-without-metastases subgroup may well develop metastases subsequently.

BsmI, FokI and poly(A) genotypes were not associated with tumor grade, lymph node involvement, or estrogen receptor status in patients with invasive forms of breast cancer, either when assessed in isolation (data not shown) or when FokI and poly(A) were combined, (Table 6). Adjusting for age, HRT use, and menopausal status does not make any difference to these results (Table 6).

Twenty-five commonly used breast cancer cell lines were genotyped for VDR polymorphisms (Table 7); 56% were found to have the bb genotype (Fig. 2). This is a high proportion when compared with our volunteer control group (33% bb) and the total cancer patient group (43% bb; Fig. 2). In addition, analysis using the FokI/poly(A) cross-genotype method showed that 40% of the cell lines express the FFLL/FfLL combined genotype compared with 26% in the control volunteers, and 38% in the total cancer patient group, (Fig. 2). We have also observed that the highest proportion of bb and FFLL/FfLL genotypes in our case-control study (47% and 46.5%, respectively) occur in women with metastatic breast cancer (Fig. 2), although this is not statistically significant, possibly because of the relatively low number of patients with metastatic disease. It is notable then that most of the cell lines are derived from metastatic breast cancers (see Table 7).

We have investigated the association between VDR genotype and breast cancer risk in a United Kingdom Caucasian population. Results of this study show that, in accordance with the original pilot study (22), the 3′ polymorphisms BsmI and variable-length poly(A) are significantly associated with breast cancer risk, because the bb and LL variants are overrepresented in the cancer group.

Previous investigations of VDR polymorphisms and breast cancer risk by other groups have produced inconsistent results. Curran et al., (19) showed that the VDR polymorphisms ApaI and TaqI were significantly associated with breast cancer risk in a Caucasian population, such that haplotype aT increased risk. This corresponds with our findings because baTL alleles are in strong linkage disequilibrium in Caucasians (15). A similar association was reported in a study on a Japanese population, in which the bb genotype conferred an almost 4-fold increase in the risk of breast cancer (27). However, in contrast to our findings, four studies have reported no link between VDR polymorphisms and breast cancer risk in Caucasian subjects (20, 28, 29, 30). Three of these studies determined only the TaqI genotype, and it is not apparent whether the control women were assessed for the absence of breast cancer by mammography (20, 29, 30). Furthermore, the study by Lundin et al. (20) was carried out exclusively in premenopausal women (age range, 24–36 years), making it more likely that these cases were familial, and, therefore, cannot be directly compared with our results. Ruggiero et al., (28) found no difference between bb and BB frequencies in patients with primary breast cancer, although the number of subjects in each group was relatively small (n = 50 cases, 167 controls). Our study has looked at both prevalent and incident cases, and such a study design may lead to some survival bias in the prevalent cases; unfortunately, however, the number of incident cases (n = 184) would not be enough to analyze on its own.

The studies by Ruggiero et al. (28) and Lundin et al. (20) did however report that progression of breast cancer was associated with VDR polymorphisms; the bb genotype gave a four times higher risk of developing metastases compared with BB (28), whereas the TT genotype was associated with an increased risk of 1.8 for lymph node metastases (20). These results substantiate our preliminary finding that the “at risk” haplotype baTL is more prevalent in women who have developed metastatic disease, suggesting that, in Caucasian women, VDR polymorphisms are associated with progression of breast cancer. Routine VDR genotyping of women with breast cancer may, therefore, be a useful prognostic tool. Interestingly, although we suggest that the b allele is associated with increased breast cancer risk, other studies have reported a positive association with bone mineral density (16, 31), and this is concordant with an observed increased rate of breast cancer among women with high bone mass (32).

It has also been reported that low levels of circulating 1,25-D₃ are related to breast cancer risk (33) and development of bone metastases (34), which corresponds with the putative antitumorigenic role of 1,25-D₃. It is plausible that polymorphisms in the VDR gene alter the ability of 1,25-D₃ to interact with the VDR gene product so that, even if a woman is not deficient in 1,25-D₃, the ligand is not able to effectively induce transcription of genes via the VDR. The interaction of 1,25-D₃ with the VDR produces a conformational change in the VDR, permitting heterodimerization with the retinoid X receptor and leading to interaction with vitamin D response elements, and transcriptional regulation of vitamin D target genes. At least 26 genes are thought to be regulated in this way by 1,25-D₃ (35), and some of these are involved in cell cycle regulation, differentiation, and apoptosis, such as WAF1/CIP1 and c-myc. We hypothesize that polymorphisms in the VDR gene might lead to differential responsiveness of target cells to the action of 1,25-D₃. Specifically the baTL haplotype might modify the ability of 1,25-D₃ to regulate its target genes in a normal manner, by altering the expression or stability of the VDR, or by the ability of the VDR to interact with 1,25-D₃, retinoid X receptor, or vitamin D response elements.

The FokI polymorphism results in two differently sized proteins (F = 424 and f = 427 amino acids in length), and this may have an effect on VDR function. Although the bbFFLL genotype is associated with greater breast cancer risk and faster progression of breast cancer in this study, suggesting that F is less active than f, this is not reflected in other studies that have used different cell systems. For instance, functional studies have suggested that the F allele is the more active VDR haplotype, being more efficient in bringing about the effects of 1,25-D₃ than is f, in transfected HeLa cells (36) and peripheral blood mononuclear cells (37). In addition, by transfecting F-VDR and f-VDR into VDR-null monkey fibroblasts (COS-7) it has been shown that the F variant binds more efficiently with the transcription cofactor TFIIB, which elevates its transcriptional activity compared with f (38). By contrast, using transfected COS-7 cells, Gross et al., (39) found no difference in the ability of f or F to bind either 1,25-D_3, or DNA, or to transactivate the expression of vitamin D responsive genes. It should be noted that all of these studies (36, 37, 38, 39) investigated the FokI polymorphism in isolation. Whitfield et al., (23) also found that there was no difference in transcriptional activity when assessing VDR encoded by FokI variants in human fibroblast cell lines. However, differences were seen when FokI and poly(A) variants were assessed in combination; these workers designated an “allele score” (where FFLL = 4 and ffSS = 0) and found a positive association with transcriptional activity induced by 1,25-D₃ in which F/L was more active than f/S. Although we have found a significant association with breast cancer risk using cross-genotyping analysis of FokI and poly(A) variants, our results oppose those of Whitfield et al., (23), because we find that the FFLL genotype is more common in women with breast cancer and advanced disease. This would suggest that in breast cancer cells the F variant, when in combination with the long form L, of poly(A), is less active and, therefore, less able to exert the antitumorigenic effects of 1,25-D₃. To support this theory, one recent study has shown that the FF genotype is also more prevalent in aggressive prostate cancers than ff (40). It is, therefore, conceivable that F and L might be less active in breast (and prostate) cells if the functional effects of VDR polymorphisms were cell-type specific, although to our knowledge no functional experiments have been carried out on breast or breast cancer cells. However, because much of the literature does not support our hypothesis that F/L are less active VDR alleles, it is more likely that F and L may increase cancer risk in breast and prostate tissues via an unknown mechanism.

Because BsmI, TaqI, and ApaI polymorphisms do not lead to changes in the VDR protein amino acid sequence, it remains difficult to explain how these variations might influence VDR function. To date, most studies have shown no correlation between these 3′ VDR polymorphisms individually and the abundance of VDR message (41, 42), VDR protein expression and function (43), or mRNA stability (41, 44). However, Carling et al., (45) reported that Caucasian patients with parathyroid tumors are more often homozygous for the baT genotype than in controls and that VDR mRNA levels were lower in these patients compared with patients with a BBAAtt genotype.

It might be the case that in Caucasian subjects the BsmI/ApaI/TaqI cluster does not influence VDR function, but because of genetic linkage, it merely acts as a marker for the poly(A) sequence within the 3′ UTR. 3′ UTR sequences are often critical for determining transcript stability (reviewed in ref. 46), and the 3′ UTR regions of the steroid hormone receptor family are unusually long and highly conserved (47), suggesting that they play an important functional role. It is, therefore, possible that the area making up the internal poly(A) polymorphism in the VDR might be important for the regulation of mRNA stability. One possible mechanism could involve adenylate/uridylate-rich elements found to be present in the 3′ UTR. It is known that the 3′UTR of several genes (cited in ref. 47), including members of the steroid hormone receptor family, possess adenylate/uridylate-rich elements. Adenylate/uridylate-rich elements contain multiple copies of the motif AUUUA, which act as destabilizing elements by recruiting a set of adenylate/uridylate-rich element-binding proteins that facilitate mRNA degradation (reviewed in ref. 48). In a published sequence of the VDR gene (GenBank accession no. 004466) we have found three AUUUA motifs present in the 3′ UTR, near to the poly(A) sequence, leading us to speculate that the differences in the length of the poly(A) sequence might affect the binding of adenylate/uridylate-rich element-binding proteins. However, it has not yet been shown that adenylate/uridylate-rich elements are implicated in regulation of VDR mRNA stability, and there does not appear to be any literature regarding the potential importance of internal poly(A) sequences. Furthermore, Whitfield et al., (23) surmise that poly(A) binding proteins might bind differentially to the L and S forms of the poly(A) sequence within the 3′UTR, producing different functional consequences. Poly(A) binding proteins are commonly known to bind to the poly(A)+ tails of RNA and may be implicated in positively regulating translatability of mRNAs via a possible interaction with the 5′ end of genes (49). Interestingly, it has been shown that poly(A) binding proteins can bind to a region of adenylate residues 11–25 in length (50) and may, therefore, have the ability to bind to internal poly(A) tracts, such as the one in the VDR 3′ UTR.

In the future, we wish to establish in vitro models to further assess the functional significance of the polymorphisms; therefore, we assessed the VDR genotype of 25 commonly used breast cancer cells lines. Similar to the results from the patient study, the genotypes bb and FFLL/FfLL were overrepresented in the breast cancer cell lines providing support for the hypothesis that these genotypes increase the risk of breast cancer. We have found a range of VDR haplotypes in the breast cancer cell lines, which in future will enable us to investigate how VDR polymorphisms may influence the function of the VDR in breast cancer cells.

When analyzed in isolation, the 5′ FokI polymorphism (variants FF, Ff, ff) was not associated with breast cancer risk. However, we have found that by examining the 5′ and 3′ polymorphisms simultaneously, a statistically significant increase in risk of breast cancer occurs in which a subject possesses one or more F alleles together with the LL genotype. Many studies have investigated individual VDR polymorphisms in isolation when assessing the association with different disease states, but these results are often nonsignificant or conflicting (reviewed in ref. 51). However, a small number of studies have carried out cross-genotyping analysis on the VDR polymorphisms and have found that this can reveal a positive association with disease status. For example, Ingles et al., (11) observed that in African-American women, the LS and LL poly(A) variants had a 50% decreased risk of breast cancer compared with women of genotype SS, but that this protective effect was limited to women who have the FF FokI variant. Whitfield et al., (23) transfected 20 human fibroblast cell lines of varying VDR genotypes with a 1,25-D₃-responsive reporter-plasmid, and assessed their endogenous VDR activity. They found that results became statistically significant only when the FokI and poly(A) genotypes were examined simultaneously. It appears, from these studies and the present study, that it is not sufficient to analyze only one VDR polymorphism in relation to disease, because the 5′ and 3′ polymorphisms may be functionally linked either through their independent influences on VDR activity, or possibly by physical interaction.

In summary, this study has provided further support for a significant association between specific VDR gene polymorphisms and breast cancer risk. This finding may be useful in predicting the likelihood of women developing breast cancer, or whether a woman with breast cancer is likely to develop metastases. At diagnosis, patients at higher risk might then be targeted for more aggressive treatments. To this end, we will continue to monitor the 825 women in the study for any new breast tumor development, recurrence or metastatic spread, so that we can examine further whether the “at risk” genotypes are linked to risk and progression.

We would like to thank the women who have taken part in the study. We are grateful to Professor Nick Carter and Dr. Petros Syrris for help with the poly(A) genotyping, Alan McKay (Ludwig Institute, London, UK) for providing DNA from breast cancer cell lines, and all of the doctors and nurses in the breast cancer clinics at St. George’s for assisting with recruitment and sample collection. We are grateful for an equipment grant from the Breast Cancer Research Trust, United Kingdom.