Interindividual and interethnic variability of drug pharmacokinetics and pharmacodynamics may be contributed by commonly occurring genetic polymorphisms of drug-metabolizing enzymes and transporters. Polymorphisms of CYP2D6 in particular have been associated with effects on tamoxifen disposition and clinical efficacy, with interethnic differences in distribution of functional alleles that affect metabolizer phenotype. Other tamoxifen-related genetic variants of CYP3A4, CYP3A5, and sulfotransferase1A1 (SULT1A1) are also briefly reviewed here. Polymorphisms of CYP19A1 (aromatase gene) have been reported to correlate with clinical outcomes from aromatase inhibitors in small studies but require further confirmation. Many studies on chemotherapy are based on hypothesis-generating association studies and need to be validated through larger-scale cooperative group studies. For anthracyclines, polymorphisms in genes such as carbonyl reductase 3 (CBR3), ATP-binding cassette subfamily B, member 1 (ABCB1), glutathione-related transporter genes, and oxidative stress–related genes have been reported to correlate with clinical outcomes. The pharmacogenetics of taxanes has been extensively investigated, but associations of genetic polymorphisms in drug-metabolizing enzymes and transporters reported in earlier small studies have not been validated in a recent large clinical trial. Allelic variants associated with gemcitabine, capecitabine/5-fluorouracil, vinorelbine, and platinum disposition are reviewed. No pharmacogenetic studies have been published for targeted agents thus far, although several potential candidate genes warrant investigation. Future pharmacogenetic studies will need to focus on integration of multiple drug pathways to allow a more comprehensive analysis of genetic factors influencing drug efficacy and toxicity.
Breast cancer is a global health problem with an estimated age-standardized incidence rate of 37.4 per 100,000 and age-standardized mortality rate of 13.2 per 100,000 women (1). Increasing interest in personalized medicine has led to in-depth research into genetic pathways of drug metabolism and the role of biomarkers to optimize therapeutic decisions for each individual patient (2, 3). Although the current armamentarium of breast cancer therapies is improving rapidly, it is often hindered by drug resistance and treatment-related toxicities. In this regard, interethnic and interpatient variability is observed but not fully understood; the study of pharmacogenetics is making great strides in revealing possible mechanisms. More than a decade ago, amonafide, a site-specific intercalating agent and topoisomerase II inhibitor, showed activity in advanced breast cancer (4). However, genetic polymorphisms in the N-acetyltransferase-2 gene causing varying acetylation rates, interethnic differences in drug clearance, and unpredictable toxicities have hampered its clinical development (5, 6). Nonetheless, it is one of the early examples of interethnic differences playing an important role in pharmacogenetics. Drugs such as warfarin also exhibit interethnic differences in dose requirements with a possible genetic basis (7). Both tumor genomic factors (3) as well as heritable genetic factors may affect interindividual responses to drug therapy. This review aims to provide an overview of heritable genetic factors that might predict drug response and toxicities of major classes of breast cancer therapy (Table 1): hormonal therapies [tamoxifen and aromatase inhibitors (AI)] chemotherapeutic agents (anthracyclines, taxanes, etc.), and biological targeted therapy (trastuzumab, lapatinib, and bevacizumab), including possible interethnic differences.
Potentially relevant studies and review articles were obtained from a PubMed search spanning the period from 1980 to 2008. Search terms included the following combined subject headings: pharmacogenetics, pharmacogenomics, genetic polymorphisms, single nucleotide polymorphisms, breast cancer, tamoxifen, AIs, fulvestrant, anthracyclines, taxanes, paclitaxel, docetaxel, gemcitabine, capecitabine, vinorelbine, platinums, trastuzumab, lapatinib, and bevacizumab. The citation lists of all retrieved articles were examined to identify other potentially relevant articles.
Tamoxifen. Tamoxifen is used in treating estrogen receptor (ER)–positive early and advanced breast cancers, ductal carcinoma in situ, and as primary chemoprevention in high-risk women. Substantial interindividual variation exists in steady-state levels of tamoxifen and its metabolites following standard dosing (8). The majority of patients with metastatic disease and a significant proportion of patients receiving adjuvant tamoxifen eventually relapse, suggesting that the benefit is not uniform. Emerging data on tamoxifen pharmacogenetics have clinical relevance, as AIs have proven efficacy in both early and advanced postmenopausal breast cancers and offer a viable alternative.
Tamoxifen undergoes metabolism via the cytochrome P450–mediated pathway to several primary and secondary metabolites that show variable potencies toward the ER. N-desmethyl-tamoxifen, resulting from CYP3A4/5-mediated metabolism, is the major primary metabolite, accounting for ∼90% of primary tamoxifen oxidation, whereas 4-hydroxy-tamoxifen, mediated by CYP2D6 activity, is a minor metabolite. Both N-desmethyl-tamoxifen (via CYP2D6) and 4-hydroxy-tamoxifen (via CYP3A4/5) are secondarily metabolized to 4-hydroxy-N-desmethyl-tamoxifen (endoxifen). Both 4-hydroxy-tamoxifen and endoxifen are important active metabolites, exhibiting similar potency, although endoxifen plasma concentrations are up to 14-fold higher (9). Tamoxifen metabolites are inactivated via conjugation by sulfotransferases such as sulfotransferase1A1 (SULT1A1), or glucuronidation by the UDP-glucuronosyltransferases (UGT), including UGT1A8, UGT1A10, UGT2B7, UGT2B15, and UGT2B17. In addition, CYP1B1, CYP2B6, and CYP2C19 may be responsible for the isomerization of trans-4-hydroxy-tamoxifen to its weakly estrogenic cis-isomer form, a reaction that could be associated with drug-resistant phenotypes (Fig. 1; ref. 10). As many of the tamoxifen-metabolizing enzymes are polymorphic, genetic variations may account for interindividual or interethnic differences in tamoxifen-related outcomes.
CYP2D6 is the key enzyme catalyzing N-desmethyl-tamoxifen to endoxifen. More than 80 allelic variants have been described in CYP2D6, many of which are associated with increased, decreased, or absent enzyme activity. These variants result in phenotypes with a tetramodal distribution of metabolic activity: poor, intermediate, extensive (“normal”), and ultrarapid. The nonfunctional variants CYP2D6*3 (2637delA), CYP2D6*4 (1934G>A), CYP2D6*5 (gene deletion), and CYP2D6*6 (1795delT) constitute the majority of poor metabolizer phenotypes (11, 12), with reportedly lower endoxifen concentrations compared with wild-type. Five percent to 10% of Caucasians are poor metabolizers, with the CYP2D6*4 allele being the predominant allele (70-90%; ref. 13). In contrast, CYP2D6*4 is extremely rare in Asians or Black Africans. CYP2D6 alleles resulting in the intermediate metabolizer phenotype include CYP2D6*10 that occurs in 38% to 70% of Asians (14) and CYP2D6*17 that occurs in 20% to 34% of Africans (14), both of which are rare in Caucasians. In contrast, ultrarapid metabolizers carry gene duplications or multiduplications of functional alleles resulting in increased enzyme activity. These genotypes are rare in Caucasians and Asians but are common in Ethiopians and Saudi Arabians (Table 2; ref. 15).
In a study by Goetz et al. (16) on 190 postmenopausal breast cancer patients treated with adjuvant tamoxifen, CYP2D6*4 homozygotes (n = 13) had significantly worse relapse-free and disease-free survival (DFS) and lower incidence of hot flashes compared with women with either one or no CYP2D6*4 variant. This was confirmed by Schroth et al. (17) in a retrospective cohort of 486 patients who noted that tamoxifen-treated patients with the CYP2D6*4, CYP2D6*5, CYP2D6*10, and CYP2D6*41 alleles had significantly worse survival, whereas no association was found among the untreated. Three studies reported conflicting results (18–20), although confounding factors such as heterogeneity in patient populations may explain this difference. In Asia, the CYP2D6*10 allele is a major polymorphism resulting in the intermediate metabolizer phenotype. The frequency of this allele is approximately 37% to 70%, 50%, and 40%, respectively, in Chinese (21, 22), Koreans (23, 24), and Japanese (25, 26) compared with only 2% in European Caucasians (11). Two early Southeast Asian studies reported the CYP2D6*1/CYP2D6*10 to be the most common genotype in Malays and Chinese in Malaysia (27, 28). A Korean study in metastatic breast cancer patients found that CYP2D6*10 homozygotes had lower plasma levels of tamoxifen metabolites and a correspondingly shorter time to progression (TTP) compared with other CYP2D6 genotypes (13). In the adjuvant setting, CYP2D6*10 homozygotes have also been found to have significantly higher incidence of recurrence compared with wild-type (26, 29).
In addition to genetic variants, concomitant administration of CYP2D6 inhibitors may affect tamoxifen-related breast cancer outcomes by converting an extensive metabolizer to a phenotypic poor metabolizer and should be avoided. Potent CYP2D6 inhibitors include antidepressants such as fluoxetine and paroxetine, whereas moderate/weak inhibitors include cimetidine, amiodarone, ticlopidine, and haloperidol (30). These interactions were elegantly shown by Borges et al. (31) rand Goetz et al. (32), who reported that CYP2D6 extensive metabolizers who took potent CYP2D6 inhibitors had lower plasma endoxifen concentrations and poorer relapse-free survival, respectively, than those who did not.
A Food and Drug Administration–approved microarray-based pharmacogenetic CYP2D6 test (AmpliChip CYP450 Test) is currently available detecting 27 CYP2D6 variants, including nonfunctional variants (CYP2D6*3, CYP2D6*4, CYP2D6*5, and CYP2D6*6), deficient variants commonly found in Asians and Black Africans (CYP2D6*10 and CYP2D6*17), and ultrarapid variants (CYP2D6*2XN in Middle Easterns and Ethiopians). A label change has been proposed to discuss the option of CYP2D6 testing before tamoxifen use, although Food and Drug Administration has yet to reach a consensus to recommend testing. There is currently no established treatment algorithm based on CYP2D6 genotype, and it is unclear whether women who are intermediate or poor CYP2D6 metabolizers should avoid tamoxifen altogether or receive a higher dose. Measuring plasma levels of tamoxifen metabolites has not been shown to be informative; the studies that correlated CYP2D6 variants with tamoxifen outcome did not establish the predictive importance of these measurements; this is likely because CYP2D6 influences the exposure to several active metabolites that collectively account for tamoxifen pharmacodynamics (PD). Hartman and Helft (33) have suggested that it may be clinically rational to carry out CYP2D6 testing on postmenopausal women with early breast cancer, in comparison with premenopausal women, in view of the availability of AIs for the former group but a lack of an evidence-based alternative for the latter, pending the Suppression of Ovarian Function Trial (SOFT). Goetz et al. (30) concurred that CYP2D6 genotyping may be considered in postmenopausal women with alternative treatment options and should include at least the common alleles encoding for the poor/intermediate metabolizer phenotypes to identify patients least likely to benefit from tamoxifen. In addition, the identification of ultrarapid metabolizers may help to select patients who may benefit from longer tamoxifen use (up to 4-5 years) before switching to AI (30).
Other genetic variants responsible for tamoxifen metabolism have been studied. A CYP3A4 promoter variant, CYP3A4*1B (−392A>G) has been associated with cancer phenotype and has an allele frequency of 35% to 67% in African-Americans, 2% to 9% in Caucasians, and 0% in Taiwanese and Chinese (34, 35). No studies to date have linked CYP3A4*1B with altered tamoxifen metabolism, although it has been reported to confer a 3-fold increased risk of endometrial cancer for tamoxifen-treated women (36). A CYP3A5 genetic polymorphism, CYP3A5*3 (6986A>G), results in severely decreased CYP3A5 activity, although several studies have failed to show its association with tamoxifen metabolism (37, 38) or breast cancer outcomes (16). However, in a recent study, postmenopausal patients treated with adjuvant tamoxifen, who were homozygous for the CYP3A5*3C variant, displayed significantly improved recurrence-free survival (RFS; ref. 20). These conflicting findings underscore the complexity of tamoxifen metabolism and resistance mechanisms and suggest that the evaluation of variants representing different interacting mechanisms may be more informative than evaluating single variants.
The enzymes responsible for elimination and inactivation of tamoxifen and its metabolites through conjugation with either a sulfate or a glucuronide may also have important genetic variation. A polymorphism of the SULT1A1 gene, SULT1A1*2 variant (638G>A, Arg213His) encodes for a protein with reduced activity. However, two studies exploring its association with tamoxifen pharmacokinetics (PK) were negative (20, 37), whereas a third study reported conflicting results of greater risk of death in tamoxifen-treated patients (n = 160; ref. 39). In vitro studies suggested a UGT1A4 variant (Leu48Val) to show increased glucuronidation activity against tamoxifen and its metabolites, although the clinical significance is still unexplored (40). A nonsynonymous polymorphism in UGT2B15 (UGT2B15*2; 253G>T) in a putative substrate binding domain has been assessed (18); adjuvant tamoxifen-treated breast cancer patients possessing SULT1A1*2/*2 and either UGT2B15*1/*2 or UGT2B15*2/*2 had significantly reduced 5-year survival (18). This novel finding requires further confirmation. In addition to drug-metabolizing enzymes, a recent study reports that ER genotypes ESR-XbaI and ESR2-02 may be associated with tamoxifen-induced lipid changes and may further contribute to interindividual variability to tamoxifen benefits (41).
Aromatase inhibitors. AIs have proven efficacy in both advanced and early postmenopausal hormone receptor–positive breast cancers. Pharmacogenetic associations pertaining to their efficacy and toxicities have not been well described, although efforts are ongoing to elucidate these mechanisms. The target of AIs is the cytochrome P450 enzyme aromatase, which is encoded by the CYP19A1 or aromatase gene. Ma et al. (42) detected 88 CYP19 polymorphisms resulting in 44 haplotypes in 60 patients from each of four ethnic groups, Caucasian Americans, African-Americans, Han Chinese Americans, and Mexican Americans, and found wide interethnic variation in genotype distribution. Functional characterization showed the Cys264, Thr364, and double variant Arg39Cys264 allozymes to show decreased activity compared with wild-type; the Arg39Cys264 allozyme also displayed a significant increase in inhibitor constant for letrozole, suggesting relative resistance. Interestingly, the Arg39 variant is present in 6.7% Han Chinese Americans but is very rare in the other three ethnic groups, whereas the Cys264 variants occurred at higher frequencies in Han Chinese Americans and African-Americans (11.7% and 22.5%) than Caucasian Americans or Mexican Americans (2.5% and 5%). These findings suggest that patients with an aromatase enzyme with lower activity may derive less benefit from an AI and that interethnic differences in AI responses could exist, although confirmation in clinical studies is required. More recently, a small Spanish study (n = 67) described the association of a common CYP19A1 3′-untranslated region variant (rs4646) with higher complete response rate and improved TTP in postmenopausal, hormone receptor–positive, metastatic breast cancers treated with letrozole (43). Although the mechanism by which the variant affected letrozole activity was not directly addressed, possible mechanisms postulated included increased enzyme activity through mRNA stabilization or enhanced transcription. Another Spanish study (n = 94) evaluated 13 CYP19A1 polymorphisms in women receiving neoadjuvant letrozole and derived a predictive model comprising eight most informative polymorphisms, including rs4646, which could predict letrozole response at 4 months with 86% accuracy (44). These promising data warrant validation before clinical application.
Other CYP enzymes may also affect the PK of AIs. For example, anastrozole inhibits CYP1A2, CYP2C9, and CYP3A4; letrozole is metabolized by CYP2A6 and CYP2C19; and exemestane is metabolized by CYP3A4. Genetic polymorphisms in these enzymes could affect their efficacy or resistance.
To determine the clinical significance of genetic influence on AIs, several large studies are being carried out under the auspices of the Pharmacogenetics Research Network (45), including a study on single nucleotide polymorphisms and intragene haplotypes in adjuvant anastrazole PK and estrogen PD pathways (45). Further research into the genetic polymorphisms affecting specific subtypes of AIs is needed before they can be used to guide clinical decisions and provide further insight into therapy selection for the individual patient.
Fulvestrant. Fulvestrant is a pure antiestrogen that antagonizes the hormone-dependent activation of ERs. It is glucuronidated by UGT1A1, UGT1A3, UGT1A4, and UGT1A8 (46). There have been no studies to date about the clinical outcomes of UGT polymorphisms and fulvestrant and there is room for more research in this area.
Anthracyclines. Anthracyclines (doxorubicin and epirubicin) have wide interindividual variation in PK and PD. A prospective study had suggested interethnic variations, noting Chinese breast cancer patients to experience more profound neutropenia from adjuvant doxorubicin/cyclophosphamide than Caucasians (47). Anthracyclines have complex disposition pathways involving various metabolizing enzymes and transporters that may contribute to interindividual variability. Doxorubicin and epirubicin undergo phase I reduction reactions to doxorubicinol and epirubicinol by carbonyl reductases (CBR1 and CBR3) and aldoketoreductases (AKR1A1 and AKR1C2); efflux out of cells is mediated by transporters including ABCB1, ABCC1, ABCC2, and ABCG2, whereas inactivation is by the cytochrome P450 enzymes (CYP3A4 and CYP3A5). Epirubicin also undergoes phase II reactions by conjugation, mainly by UGT2B7. The orphan nuclear receptors regulate transcription of multiple cytochrome P450 enzymes and transporters and may further contribute to the variability.
A recent prospective study on CBR1 and CBR3 in Asian breast cancer patients treated with single-agent doxorubicin (n = 101) revealed two common CBR3 variants to affect doxorubicin PK and/or PD (48). The CBR3 11G>A variant was associated with lower conversion of doxorubicin to doxorubicinol, greater tumor reduction, and hematologic toxicities, whereas the CBR3 730G>A variant was associated with increased conversion of doxorubicin to doxorubicinol but did not correlate with hematologic toxicity. The CBR3 11A variant was more common in Chinese than Caucasians (57% versus 36%) and may contribute to the greater doxorubicin-induced myelosuppression observed in Chinese, although this hypothesis requires validation in larger cohorts. Comprehensive analysis of three orphan nuclear receptors (PXR, CAR, and HNF4α) and CYP3A5*3C in the same cohort of patients revealed no significant correlation between these genotypes and doxorubicin PK or PD (49).
Among the ATP-binding cassette transporters, the ABCB1 gene, which encodes P-glycoprotein implicated in drug resistance, is one of the most studied. Common ABCB1 variants (1236C>T, 2677G>T/A, and 3435C>T) that show linkage and seem to impair ABCB1 substrate transport have been extensively studied with respect to their influence on drug disposition. Kafka et al. (50) reported significant correlation between the 3435T variant and better clinical response to anthracyclines with or without taxanes in locally advanced breast cancer patients (n = 68). Another retrospective study in Asian breast cancer patients (n = 62) receiving adjuvant doxorubicin-based chemotherapy reported the ABCB1 c.1236-2677-3435 CC-GG-CC haplotype to be associated with significantly lower doxorubicin exposure than the CT-GT-CT and TT-TT-TT haplotypes, although effects on doxorubicin toxicities were not reported (51). A variant (c.421C>A) in another transporter, ABCG2, did not correlate with doxorubicin PK in the same study.
In addition to genes that encode for metabolizing enzymes or transporters, genetic variants that may affect anthracycline-mediated tumor cell kill have been investigated. The glutathione S-transferases (GST) catalyze the reduction of secondary organic oxidation products produced by chemotherapy that contribute to further cellular damage, and individuals lacking or deficient in these enzymes may have better treatment outcome. In a retrospective study on 251 breast cancer women who predominantly received doxorubicin-based combination chemotherapy, women with null GSTM1 or GSTT1 genotypes had reduced mortality (52). Ambrosone et al. (53) studied variants in genes related to oxidative stress, including manganese superoxide dismutase (MnSOD), catalase (CAT), and myeloperoxidase (MPO), in radiotherapy- and/or chemotherapy-treated (including doxorubicin combinations) breast cancer survivors, and found those with genotypes associated with higher levels of reactive oxygen species [MnSOD C (Ala16) and MPO-463G variants] to have better survival. Concordant findings were reported in a preliminary Southwest Oncology Group study, which described association of the MPO-463G allele with better DFS in breast cancer women given adjuvant CAF [cyclophosphamide, doxorubicin, 5-fluorouracil (5-FU)] or CMF (cyclophosphamide, methotrexate, 5-FU) but not in untreated women (54).
Although anthracyclines have been used for decades, the mechanisms underlying their interindividual variability remain unclear. Although several prospective candidate genes and polymorphisms have been studied, most reports are retrospective, were confounded by the concurrent administration of other chemotherapeutic agents, and did not incorporate drug PK or report direct drug PD effects such as tumor responses or toxicities, making it difficult to derive clear correlations between genotype and anthracycline effects. Larger prospective studies incorporating both PK and PD end points, preferably in response to single-agent anthracyclines, will be desired to discover and validate prospective genetic predictors that may be applied clinically.
Taxanes. Together with anthracyclines, taxanes are the most active cytotoxic agents in breast cancer treatment. Both paclitaxel and docetaxel are good substrates of P-glycoprotein located at the biliary canalicular membrane and bind to tubulin residues to stabilize microtubules (55). Docetaxel is predominantly metabolized by CYP3A4/5, whereas paclitaxel is metabolized by CYP2C8/CYP3A4 to inactive hydroxylated metabolites (55). The hepatic influx protein transporter OATP1B3 has been shown to mediate transport of paclitaxel and docetaxel into the hepatocytes at the basolateral membrane (56). Given these potential candidate genes that have well-described polymorphisms, much effort has been put into correlating genetic variants with taxane PK and PD. This is particularly pertinent given that both taxanes have highly variable PK and unpredictable toxicities.
Pharmacogenetics will only be useful clinically if there is validated and mechanistic evidence for the influence of allelic variants on one or more specific PD end points. Taxanes present challenges in validation because of their schedule dependence, necessitating evaluation of genetic variants on PD at different regimens and doses. Furthermore, paclitaxel shows saturable distribution kinetics, and exposure time above threshold concentrations is better correlated with PD than area under the curve or maximum concentrations, complicating interpretation of any genotype-phenotype relationships.
As noted for anthracyclines, ABCB1 polymorphisms may affect the activity of P-glycoprotein, thereby potentially perturbing the disposition of taxanes (57–59). Although some studies have found ABCB1 genotype and haplotypes to correlate with taxane PK and toxicities (60–62), there have been many others that have not shown any correlation (63–65). For example, in Japanese patients with ovarian cancer, variant alleles at ABCB1 129T>C, 1236T>C, and 2677G>A/T were associated with lower paclitaxel area under the curve (60). For docetaxel, C1236T homozygotes had significantly lower clearance in a study of 92 patients (61), and the 3435TT genotype was associated with greater neutropenia in 58 patients (66). A previous report found that patients with the 2677TT-3435TT diplotype had significantly worse nadir neutropenia secondary to docetaxel treatment (67). On the other hand, several larger studies have not shown any effect of ABCB1 polymorphisms on taxane PK or PD (64, 65). No associations have been found between the genotype of membrane transporters MRP2, breast cancer resistance protein (BCRP), and OATP1B3 with paclitaxel or docetaxel PK or PD (68, 69).
Functional CYP2C8 polymorphisms include CYP2C8*2 (Phe269Ile) and CYP2C8*3 (Lys139Arg), which have impaired intrinsic clearance for paclitaxel in vitro; however, no clinically significant associations have been found with paclitaxel clearance in vivo (70). Possibly, CYP metabolism of xenobiotics has evolved considerable redundant pathways, often including CYP3A4, which could be activated to compensate for dysfunction of the main metabolic pathway. CYP3A4 and CYP3A5 polymorphisms have also been studied in correlation with taxane disposition. CYP3A4*1B is a promoter polymorphism frequently found in African-Americans (45%) and Caucasians (2% to 9.6%) but rare in Southeast Asians and has been associated with increased CYP3A4 transcriptional activity (71). CYP3A5*3C is an intron 3 splicing variant that results in protein truncation. In Caucasian breast cancer patients receiving paclitaxel, doxorubicin, and cyclophosphamide, ABCB1 (1236C>T, 2677G>T/A), ABCG2 (421C>A), CYP1B1*3, CYP3A4*1B, CYP3A5*3C, CYP2C8*3, and CYP2C8*4 genotypes were not correlated with paclitaxel clearance, although homozygotes of CYP1B1*3 had significantly longer progression-free survival (PFS) (64). Alone, CYP3A5*3C was not associated with docetaxel clearance in other studies (63). In a recent study in Caucasian patients, subjects harboring simultaneous CYP3A4*1B and CYP3A5*1A alleles had 64% higher docetaxel clearance than those without (68). However, given the rarity of CYP3A4*1B in Asians, the finding is likely ethnic specific.
PXR, CAR, and HNF4α are orphan nuclear receptors that regulate transcription of multiple cytochrome P450 enzymes and transporters, including ABCB1 and BCRP (ABCG2), and are activated by ligand binding (49, 72, 73). Incorporating common polymorphisms of PXR, CAR, and HNF4α and CYP3A5*3C with other covariables in a nonlinear mixed effect model (NONMEM) for docetaxel clearance showed no significant contribution from these polymorphisms in explaining variability of clearance (74).
In summary, the contribution of pharmacogenetics to individualizing taxane therapy in breast cancer has been limited. To emphasize this issue, a large study, the Scottish Ovarian Cancer Study, showed that toxicities of paclitaxel or docetaxel and carboplatin did not correlate with known polymorphisms involved in the taxane disposition pathway (65). In this study carried out without PK, toxicities of the taxanes (paclitaxel/docetaxel) and platinums were studied in association with 27 polymorphisms of 16 genes in 914 patients, allowing application of a test set and a validation set, which gives more robust statistical conclusions than single-arm studies involving small patient numbers that could at best be considered exploratory. Possible reasons for discrepancy apart from statistical chance may include less patient or physician bias, different schedules or combinations, or ethnic makeup of patients. Yet, other questions remain unresolved; for example, interethnic differences in docetaxel PK and toxicities have been described, with Chinese and Indians experiencing more frequent hematologic toxicity compared with Malays (49). However, no explanations for this have thus far been elucidated, warranting more studies.
Other Chemotherapy Agents
Chemotherapy agents, such as gemcitabine, capecitabine, vinorelbine, and platinums, are used widely, alone or in combination with other drugs, beyond first-line therapy in metastatic breast cancer. The pharmacogenetic effects of these drugs, however, are likely to be confounded by combination regimens and prior treatments resulting in relative tumor resistance and greater toxicities.
Gemcitabine. Gemcitabine is transported into cells by nucleoside transporters (SLC28 and SLC29; 75, 76), phosphorylated by deoxycytidine kinase (DCK) to its active monophosphate form, and subsequently by nucleotide kinases to gemcitabine triphosphate, and finally incorporated into DNA, inhibiting its repair and synthesis. Gemcitabine and gemcitabine monophosphate are inactivated by cytidine deaminase (CDA) and deoxycytidylate deaminase, respectively (77, 78). Gemcitabine also inhibits thymidylate synthase (TS) and ribonucleotide reductases (RRM1 and RRM2) involved in DNA repair.
An RRM1 haplotype 2455G-2464A correlated with less neutropenia, granulocyte colony-stimulating factor requirements, and poorer PFS and OS in a prospective study of 74 Korean breast cancer patients treated with single-agent gemcitabine (79). In a restrospective study in advanced breast cancers treated with gemcitabine/carboplatin (n = 41), two RRM1 promoter region alleles (−37C and −524T) predicted for better tumor response, PFS and OS (80). A prospective study (n = 256) in Japan reported patients with the CDA 208G>A (Ala70Thr) variant to have decreased gemcitabine clearance and greater hematologic toxicities following gemcitabine coadministered with fluorouracil or platinums (81); the variant is more common in Africans than in Japanese or Europeans (13% versus 4.3% versus 0%; refs. 82, 83). A synonymous variant CDA 435C>T (Thr145Thr) was associated with lower response rates and shorter TTP in Asian lung cancer patients receiving carboplatin/gemcitabine (84). In the same study, SLC28A1 1561G>A was related to increased hematologic toxicity (84) and showed interethnic differences in frequencies (Caucasians, 73%; Chinese, 12%; Malays, 30%; Indians, 35%) and warrants further evaluation (84). Ethnic differences have been observed in DCK with Asians showing a higher frequency of promoter variants −C360G/−C201T than Caucasians, which may predispose to greater gemcitabine toxicities (85).
Capecitabine and 5-FU. Capecitabine is an oral prodrug of 5-FU. TS, the target of 5-FU, methylenetetrahydrofolate reductase (MTHFR), a key regulatory enzyme in folate metabolism, and dihydropyrimidine dehydrogenase (DPD), the catabolizing enzyme of 5-FU, are potential targets that could contribute to interindividual variability to 5-FU–based therapy. In the TS enhancer region, varying copies of 28-bp tandem repeated sequences (TSER*2, TSER*3, TSER*4, and TSER*9) have been described that increase TS expression and portend a worse response (86). MTHFR polymorphisms, 677C>T and 1298A>C, are associated with low enzyme activity but did not correlate with capecitabine toxicities (87). Complete or partial DPD deficiency resulting in severe 5-FU toxicities occurs in approximately 0.1% and 3% to 5% of individuals, and at least 20 functional DYPD variants have been described, with the DYPD*2A splice site variant being the most common. Unfortunately, the rarity and heterogeneity of DYPD mutations has precluded the routine use of DPD testing (88).
Vinorelbine. Vinorelbine is a Vinca alkaloid that prevents microtubule assembly, thereby inhibiting DNA replication. Studies relating to vinorelbine pharmacogenetics are limited. Metabolism is principally by CYP3A4 enzymes (89), whereas resistance has been said to be mediated by ABCB1 (90). However, in a prospective study involving 41 Caucasian cancer patients, including 10 breast cancers, no association was found between vinorelbine clearance and ABCB1 1236T>C, 2677G>T/A, and 3435C>T (91).
Platinums. Platinums cause interstrand DNA cross-linking leading to cessation of DNA synthesis and apoptosis, and toxicity due to platinums is modulated by proteins involved in DNA repair (ERCC1, ERCC2, and XRCC1), DNA detoxification (GSTP1 and MPO), and transport (SLC31A1, ABCC2, and ABCG2; ref. 65). However, there are no genetic polymorphisms that are ready for clinical application. In the large Scottish randomized trial of carboplatin and a taxane in ovarian cancer patients, genetic polymorphisms of genes relevant to the platinum pathway (ABCC2, ABCG2, ERCC1, ERCC2, GSTP1, MPO, and XRCC1) were not significantly associated with clinical outcomes or toxicity (65).
Heritable genetic factors may increase breast cancer risk; susceptibility loci include high-penetrance genes that are rare (e.g., BRCA1/2) and low-penetrance genes that are common. Understanding of these susceptibility loci has led to the elucidation of the biological mechanisms of breast cancer (92). For example, ∼80% of BRCA1-related breast cancer are triple negative and basal-like by microarray, a phenotype for which no target exists for tailored therapy (93). However, BRCA1/2 mutations have been reported to confer differing tumor chemosensitivities. For example, tumors with BRCA1 mutations may be associated with taxane resistance (94), and those with BRCA1/2 mutations are unable to carry out DNA repair by homologous recombination, rendering them susceptible to platinum agents and poly(ADP-ribose) polymerase inhibitors (94–97). Perhaps one of the most recent illuminating findings in the arena of platinum agents and breast cancer is the discovery of secondary intragenic mutations that restore the open reading frame of BRCA2, resulting in cells that are now competent in homologous recombination, leading to resistance to platinums and poly(ADP-ribose) polymerase inhibitors (98, 99). These novel findings warrant further evaluation and confirmation from clinical trials.
Targeted therapy such as anti-HER2 (trastuzumab and lapatinib) and antiangiogenesis agents [bevacizumab directed against vascular endothelial growth factor (VEGF)] are now approved treatments for advanced breast cancers and have entered clinical trials for early-stage breast cancer. No pharmacogenetic studies have been published for these agents, although several potential candidate genes warrant evaluation. No polymorphisms have been found to affect trastuzumab metabolism, efficacy, or toxicity thus far (100). However, binding of trastuzumab to its target at the extracellular domain of ERBB2 may be affected by differential amino acid expression, and functional polymorphism in this region may influence individual drug responses. CYP2C19, CYP3A4/5, P-glycoprotein, and BCRP are involved in the metabolism and transport of lapatinib, and polymorphisms in genes encoding these proteins may affect lapatinib disposition (101). Several polymorphisms have been described in genes that code for proteins or receptors involved in the VEGF pathway, including VEGF (102–105), VEGF-R2 (KDR; ref. 106), and HIF1α (107–109), which regulates transcription of VEGF. Some of these polymorphisms have functional activity in vitro, affecting VEGF production, VEGF-R2 expression, or downstream effects, and may potentially influence individual responses to anti-VEGF therapy. More research is eagerly awaited in this field for better selection of patients to improve cost-effectiveness in this area of treatment where the expense is often prohibitive.
In addition to single nucleotide polymorphisms, copy number variations may alter the expression levels and possibly the activity of a given gene. For example, copy number variations have been reported in PXR, CBR1, and ESR1 (110–112) and may well influence the response to commonly used chemotherapeutic agents or hormonal therapy in breast cancer, warranting investigation. Furthermore, as drugs are inevitably involved in complex metabolic and transport pathways, individual genetic polymorphisms are likely to be less predictive than a panel of polymorphisms. The availability of high-throughput gene chips that allow simultaneous analysis of multiple prospective polymorphisms and copy number variations would facilitate research in these fields. The integration of multiple drug pathways into future trials would also be critical to allow more comprehensive analysis of genetic factors that could influence drug efficacy and toxicity.
It seems promising that pharmacogenetics may, in the future, be used in conjunction with tumor genomics and clinical disease/tumor characteristics to better tailor our treatment of each individual (Fig. 2). However, despite the slew of exciting data generated over the years, there has been a lack of validation studies. Clinical application of pretreatment pharmacogenetic testing to determine drug response and toxicity is still limited in oncology, with CYP2D6 testing representing one of the first examples that may become more widely clinically applicable. Although some studies had suggested interesting interethnic differences in drug disposition and/or genetic distribution that may have correlations, such as CYP2D6 and tamoxifen, CYP19A1 and AIs, CBR3 and anthracyclines, CYP3A4 and tamoxifen and taxanes, and CDA and SLC28 and gemcitabine (Table 3), more work is required to compare drug efficacy and toxicity across different ethnic groups, where host-genotype interactions may be significant. These gaps emphasize the importance of prospective collection of germ-line DNA from early-phase clinical trials as well as for the concept of global consortia to accelerate the momentum in this area of research. With rapid biotechnology advances, personalized treatment based on genotype may hopefully be available in the future.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
- Received July 9, 2008.
- Revision received September 23, 2008.
- Accepted October 10, 2008.