Functional Polymorphisms in UDP-Glucuronosyl Transferases and Recurrence in Tamoxifen-Treated Breast Cancer Survivors

Tamoxifen is a selective estrogen receptor (ER) modulator that binds the ER and inhibits breast cancer growth (1). Members of the cytochrome P450 enzyme family (e.g., CYP2D6) catalyze phase I metabolism of tamoxifen to its most active metabolites, 4-hydroxy-tamoxifen (4HT) and 4-hydroxy-N-desmethyl tamoxifen (endoxifen; ref. 2). These metabolites bind the ER with approximately 100-fold higher affinity than tamoxifen itself and are ostensibly responsible for tamoxifen's protection against breast cancer recurrence (3). Most tamoxifen phase II reactions are catalyzed by UDP-glucuronosyl transferases (UGT; refs. 4, 5). UGTs catalyze the addition of glucuronide moieties to 4HT and endoxifen, which negates their anti-estrogenic properties (6) and promotes their excretion (7). Three UGTs are encoded by polymorphic genes. UGT2B15 has a variant allele (*2/rs1902023) that confers an approximately 2-fold increased rate of catalysis on the translated enzyme, compared with wild type (8). UGT2B7 has a variant allele (*2/rs7439366) that encodes an enzyme with approximately 2- and 5-fold lower glucuronidation activity toward 4HT and endoxifen, respectively (9, 10). UGT1A8 harbors 2 variant alleles: one (*2/rs1042597) produces an enzyme with catalytic activity similar to the wild type, and the other (*3/rs17863762) eradicates glucuronidation activity completely (10). Therefore, the UGT2B15*2, UGT2B7*2, and UGT1A8*3 polymorphisms (Table 1) have the potential to alter substantially the elimination rates of 4HT and endoxifen, either prolonging or decreasing their circulating half-lives, and potentially modifying tamoxifen effectiveness.

A number of studies have measured the association between functional polymorphisms in genes encoding tamoxifen phase I enzymes (e.g., CYP2D6) and breast cancer recurrence in tamoxifen-treated women, including a recent study from our group (11). The sum of the evidence from these studies points to a null association (12).

Two earlier studies measured the association between the UGT2B15*2 polymorphism and recurrence among tamoxifen-treated breast cancer patients (13, 14). Nowell and colleagues (13) and Wegman and colleagues (14) both observed a higher recurrence rate among carriers of the *2 variant. However, neither association was measured precisely enough to provide clear evidence refuting a null relationship.

The objective of this study was to measure with high precision the associations between functional polymorphisms in UGT2B15, UGT2B7, and UGT1A8 and breast cancer recurrence, while accounting for phase I metabolic status.

This study was approved by the Regional Committee on Biomedical Research Ethics of Aarhus County, Denmark, and by the Boston University Medical Campus Institutional Review Board.

We studied the association between UGT polymorphisms and breast cancer recurrence by using a population-based case–control design; the methods have been detailed elsewhere (11). Briefly, the source population consisted of women ages 35 to 69 years, residing on Denmark's Jutland Peninsula, who were diagnosed with Unio Internationale Contra Cancrum (UICC) stage I, II, or III primary breast cancer between 1985 and 2001 and who were reported to the Danish Breast Cancer Cooperative Group (DBCG) registry (15). Two groups were formed from this study population: (a) women whose primary tumors expressed ER and who were treated with tamoxifen for at least 1 year (ER⁺/TAM⁺), and (b) women whose primary tumors did not express ER, who were not treated with tamoxifen, and who survived for at least 1 year after diagnosis (ER⁻/TAM⁻). We included the ER⁻/TAM⁻ group to estimate the direct effect, if any, of the UGT variants on breast cancer recurrence risk. Cases were women who experienced a recurrence after at least 1 year of tamoxifen treatment (if in the ER⁺/TAM⁺ group) or after surviving at least 1 year beyond diagnosis (if in the ER⁻/TAM⁻ group). Eligible controls were women who were not diagnosed with a recurrence after the same amount of follow-up time as each matched case.

Cases of breast cancer recurrence were identified from the DBCG registry. Risk sets were specified for each case, matched on ER/TAM group membership, menopausal status at diagnosis, date of breast cancer surgery (±12 months), county of residence, and UICC tumor stage. One control was sampled at random and without replacement from the risk set of each case. Follow-up ended with the first of diagnosis of recurrent breast cancer, death from any cause, emigration from Denmark, or September 1, 2006.

Procurement procedures for archived primary tumors, DNA extraction and amplification methods, and details about central confirmation of ER expression are described elsewhere (11). Commercially available TaqMan kits (Applied Biosystems) were used to genotype the UGT2B15*2 and UGT1A8*3 alleles, as detailed in Table 1. Genotyping of CYP2D6*4 also was done with a commercially available TaqMan kit (kit C-27102431-D0; ref. 11). A high proportion of the nucleotides flanking the UGT2B7*2 single-nucleotide polymorphism (SNP) are guanine or cytosine residues, which precludes development of an oligonucleotide probe for this locus. Using HaploView software (Broad Institute, Cambridge, Massachusetts), we downloaded public data from the International HapMap Project (version 2, release 22), and searched the 5 kb region flanking the target SNP (rs7439366) for candidate proxy SNPs with high estimated linkage disequilibrium (LD) parameters. We identified SNP rs7434332, estimated to be in perfect LD with UGT2B7*2 [R² = 1; D′ = 1 (95% CI, 0.95–1)]. On the basis of these LD parameters, detecting a variant allele for the proxy SNP is essentially equivalent to detecting the UGT2B7*2 allele. A custom TaqMan kit for this proxy SNP was developed through contract with Applied Biosystems, Inc.

Genotypes were classified as 2 normal (wild-type) alleles, 1 variant allele, or 2 variant alleles by the auto-call feature of the analytic software (MXPro QPCR version 4.1; Stratagene). Because the UGT1A8*3 variant is rare, we additionally classified UGT1A8 genotype as either 2 normal alleles or at least 1 variant allele. Impaired CYP2D6 function was defined as presence of 1 or 2 copies of the *4 variant allele (rs3892097). Normal CYP2D6 function was defined as presence of 2 normal alleles.

Recurrence was classified according to DBCG convention as any type of breast cancer or metastasis diagnosed after the initial course of therapy. In addition to the matching factors, covariates included age at diagnosis (continuous), histologic grade (low, medium, or high), type of breast cancer surgery (mastectomy or breast conserving), and receipt of adjuvant systemic chemotherapy or radiotherapy.

Within ER/TAM groups we tabulated the frequency and proportion of cases and controls according to UGT genotypes and tumor, treatment, and demographic characteristics. We tested whether genotype frequencies among controls were consistent with those predicted under Hardy–Weinberg equilibrium by calculating χ² test statistics.

We used logistic regression models to estimate the OR (approximating the rate ratio because of the control sampling strategy) associating UGT polymorphisms with breast cancer recurrence among women taking tamoxifen (ER⁺/TAM⁺). The analyses were repeated in the group of women who did not take tamoxifen (ER⁻/TAM⁻) to evaluate non-tamoxifen–mediated associations between UGT genotypes and breast cancer recurrence. Logistic regression models were specified 2 ways for each of the UGTs: (a) conditioned on the matching factors and (b) adjusting for tumor stage, menopausal status, receipt of adjuvant systemic chemotherapy, receipt of adjuvant radiotherapy, type of surgery, time to recurrence, and histologic grade. Conditional and adjusted models were also specified, in which all UGT variants were included as independent variables. We also applied these models to the subset of cases and controls whose ER expression was concordant in their original and repeated assays, within strata of genetically normal and impaired CYP2D6 function.

All statistical tests were 2-sided (α = 0.05). Analyses were conducted by SAS version 9.2 (SAS Institute).

In the ER⁺/TAM⁺ group, we identified 541 breast cancer recurrence cases, to whom we matched 541 breast cancer controls. In the ER⁻/TAM⁻ group, we enrolled 300 breast cancer recurrence cases, to whom we matched 300 breast cancer controls. Table 2 shows the distribution of the cases and controls according to UGT genotypes, CYP2D6 genotype, and key tumor treatment and demographic characteristics. Compared with controls, women with a recurrence had tumors of higher histologic grade, but were otherwise similar. A large proportion of cases and controls in the ER⁺/TAM⁺ group were postmenopausal and diagnosed with stage II or III tumors, which is expected on the basis of the DBCG criteria used to assign women to tamoxifen treatment during the 1985 to 2001 period (16). Subjects were initially assigned to 1, 2, or 5 years of tamoxifen treatment, based on clinical protocols at the time of diagnosis. Medical record review for a subset of cases and controls found that many women initially assigned to shorter tamoxifen protocols actually received longer courses as the recommended duration increased to 5 years (11).

Genotypes of UGT2B15*2, UGT2B7*2, and UGT1A8*3 were successfully assayed for 94% or more of all cases and controls. Among controls in the ER⁺/TAM⁺ and ER⁻/TAM⁻ groups, observed genotype frequencies were consistent with those expected under Hardy–Weinberg equilibrium, and observed minor allele frequencies were similar to benchmark values reported in Caucasian samples (data not shown).

Table 3 summarizes the associations between the variant UGTs and breast cancer recurrence. In both the ER⁺/TAM⁺ and ER⁻/TAM⁻ groups, none of the UGT SNPs was associated with breast cancer recurrence in either conditional or adjusted models [in the ER⁺/TAM⁺ group, for UGT2B15*2: adjusted OR for 2 variant alleles compared with 2 normal alleles = 1.0 (95% CI, 0.70–1.5); for UGT2B7*2: adjusted OR for 2 variant alleles compared with 2 normal alleles = 0.96 (95% CI, 0.65–1.4); for UGT1A8*3: adjusted OR for 1 or 2 variant alleles compared with 2 normal alleles = 0.95 (0.49–1.9)]. These estimates did not differ substantially when we included all UGT SNPs in the same conditional or adjusted logistic regression models (data not shown), nor when we repeated analyses within strata of CYP2D6*4 genotype and included only women with concordant ER expression between diagnosis and re-assay (Table 3).

We observed no association between genetically modified UGT activity and the rate of breast cancer recurrence in a large population-based case–control study of tamoxifen-treated and untreated breast cancer patients. There are several plausible explanations for our null findings.

First, it is possible that glucuronide moieties are cleaved by enteric bacterial glucuronidase activity during enterohepatic circulation of the 4HT and endoxifen glucuronides. Reabsorption of restored active metabolites may negate any fluctuation in their plasma concentrations attributable to genetic variation in the UGTs.

Second, any effect of the UGT variants on excretion of active tamoxifen metabolites may be counteracted by simultaneous effects on estrogen metabolism. Sex steroids, including estradiol, are also deactivated and eliminated via glucuronidation (17, 18). The UGT2B15*2 polymorphism, for example, would be expected to speed clearance of 4HT, endoxifen, and estradiol simultaneously. So it may be that a harmful impact of lower tamoxifen metabolite concentrations is averted by a corresponding reduction in estradiol levels, at least in this population of predominantly postmenopausal ER⁺ women.

Another possibility is that tamoxifen and its metabolites circulate in such vast molar excess to ER binding sites on residual breast cancer cells that modest changes in their concentrations due to metabolic enzyme variants do not prevent their saturation of estradiol binding sites and inhibition of tumor growth (19).

To detect the UGT2B7*2 variant allele, we used a proxy SNP estimated to be in perfect LD with the target SNP [R² = 1; D′ = 1 (95% CI, 0.95–1)]. Because these LD parameters were estimated with uncertainty in an independent sample of Caucasians, UGT2B7*2 genotype may have been misclassified in our study population. However, given the high and precise LD estimate, it is unlikely that the error is sufficient to have masked a truly non-null association.

Residual confounding is an unlikely source of bias in our measurements, primarily because no factor affecting recurrence is expected to be causally related to genotype, and also because adjustment for known recurrence risk factors did little to change the estimated ORs.

In summary, we observed no association between functional polymorphisms in key UGT enzymes involved in phase II metabolism of tamoxifen, whether or not we took CYP2D6*4 genotype into account. These results do not support the notion that these biomarkers of metabolic enzyme activity can be used to improve the prediction of clinical response to adjuvant tamoxifen therapy.

No potential conflicts of interest were disclosed.

This study was supported by U.S. National Cancer Institute at the NIH [R01 CA118708 (T.L. Lash) and T32 CA09001-35 (T.P. Ahern)]; Danish Cancer Society [DP06117 (S. Hamilton-Dutoit)]; Karen Elise Jensen Foundation (H.T. Sørensen); Congressionally Directed Medical Research Programs [BC073012 (T.P. Ahern)].