Introduction

Tamoxifen, a selective estrogen receptor (ER) modulator, has been widely used for the treatment and prevention of recurrence for patients with hormone receptor (ER or progesterone receptor)-positive breast cancers. As >70% of breast cancers are hormone receptor-positive, thousands of breast cancer patients worldwide initiate to take endocrine treatment including tamoxifen each year. In pre- and postmenopausal patients with primary breast cancer, 5 years of adjuvant tamoxifen significantly reduced recurrence rate as well as cancer-specific mortality for 15 years after their primary diagnosis.1 However, approximately one-third of patients treated with adjuvant tamoxifen experience a recurrent disease,1, 2 implicating possible individual differences in responsiveness to tamoxifen.

Tamoxifen is metabolized to more active metabolites or inactive forms by phase I and phrase II enzymes, including cytochrome P450s (CYPs), sulfotransferases (SULTs) and UDP-glucuronosyltransferases (UGTs). The polymorphisms in these drug-metabolizing enzymes are considered to affect individual differences in plasma concentrations of active tamoxifen metabolites and clinical outcome in breast cancer patients treated with tamoxifen. Among these enzymes, CYP2D6 has been most extensively investigated owing to its significant role in production of active metabolites, endoxifen and 4-hydroxytamoxifen.

This review summarizes current reports on the relationships of genetic polymorphisms and other biomarkers to individual differences in clinical outcome of breast cancer patients with tamoxifen treatment. In addition, we investigate reasons or causes of discordant results for the association between CYP2D6 genetic variations and clinical outcome, and would like to highlight various problematic issues in pharmacogenomics studies.

Tamoxifen metabolism

Tamoxifen is extensively metabolized by phase I and phase II enzymes in the human liver (Figure 1).3, 4 Tamoxifen itself has low affinity to the ER as only 1.8% of the affinity of 17β-estradiol.3 The major metabolite N-desmethyltamoxifen is formed by N-demethylation, which is catalyzed mainly by CYP3A4 and CYP3A5, with small contribution by CYP2D6, CYP1A2, CYP2C9 and CYP2C19.5, 6, 7, 8, 9 N-desmethyltamoxifen shows weak affinity to the ER similar to tamoxifen.3, 4 However, 4-hydroxytamoxifen, which is formed by 4-hydroxylation of tamoxifen, has 100-fold higher affinity to the ER and 30- to 100-fold greater potency in suppressing estrogen-dependent breast cancer cell proliferation than tamoxifen.3, 10, 11, 12 This conversion is catalyzed by CYP2D6, CYP2B6, CYP2C9, CYP2C19 and CYP3A4.5, 13, 14, 15 Endoxifen (4-hydroxy-N-desmethyltamoxifen) has a potency equivalent to 4-hydroxytamoxifen,10, 16, 17 and its plasma concentration level exceed that of 4-hydroxytamoxifen by several folds, suggesting endoxifen to be a principal active metabolite.9, 10, 11 Endoxifen formation from N-desmethyltamoxifen is predominantly catalyzed by CYP2D6.18 Several additional metabolites, such as N,N-didesmethyltamoxifen, 4′-hydroxy-N-desmethyltamoxifen and α-hydroxytamoxifen were reported, but no other highly active metabolite has been described so far.4

Figure 1
figure 1

Metabolic pathways of tamoxifen in human. Major metabolic pathways are highlighted with bold arrows.

Tamoxifen and these metabolites are further metabolized by phase II enzymes, such as SULTs and UGTs. SULT1A1 is considered to be the primary SULT responsible for the sulfation of 4-hydroxytamoxifen and endoxifen.19, 20 UGT1A8, UGT1A10, UGT2B7, UGT2B15 and UGT1A4 are involved in the O-glucuronidation of 4-hydroxytamoxifen and endoxifen.21, 22, 23 Tamoxifen and 4-hydroxytamoxifen are glucuronidated by UGT1A4 to the corresponding N+-glucuronides.24, 25 The genetic variations of these drug-metabolizing enzymes are possible to affect tamoxifen metabolism.

Genetic polymorphisms of CYP2D6

CYP2D6 is one of the most important CYP isoforms owing to its central role in the metabolism of a number of clinically important drugs.26 The CYP2D6 gene is located on chromosome 22q13.1, containing two neighboring pseudogenes, CYP2D7 and CYP2D8. This locus is extremely polymorphic with over 80 allelic variants, a subset of which should affect the gene product and result in wide interindividual and ethnic differences in CYP2D6 activity.27 Commonly, four CYP2D6 phenotypes are defined on the basis of their in vivo metabolic capacities: poor metabolizer (PM), intermediate metabolizer (IM), extensive metabolizer (EM) and ultra-rapid metabolizer (UM).28, 29 It has been reported that the PM phenotype, which is caused by carrying two null alleles, is present in 5–10% of Caucasians.30 The CYP2D6*3, CYP2D6*4, CYP2D6*5 and CYP2D6*6 are major null alleles that are related to the PM phenotype and account for nearly 95% of the PMs in Caucasians.31 Among them, CYP2D6*4 shows the highest frequency as 17.5–23.0%.27 CYP2D6*5, which is found at a frequency of ∼5%, lacks an entire CYP2D6 gene. In contrast, <1% of Asians show the PM phenotype,32 and most Asians are categorized as IMs because of the high frequency of a CYP2D6*10 allele.33, 34 The CYP2D6*14, CYP2D6*18, CYP2D6*21, CYP2D6*36 and CYP2D6*44 were null alleles found in Asian populations, although their frequencies are very low.35, 36, 37, 38 The frequencies of UMs, who carry a duplicated/multiplied wild-type CYP2D6 gene(s), are 10–15% in Caucasian, whereas UMs are uncommon in Asians. As described here, because the CYP2D6 gene locus is complex, genotyping of CYP2D6 variants, especially CYP2D6*5, is technically not so easy. Although the accuracy of genotyping partly depends on the quality of DNAs and the platforms of genotyping, wrong genotyping results sometimes cause incorrect interpretation of the research outcome, and result in both false-positive conclusions and false-negative conclusions.

CYP2D6 genotype and clinical outcome of tamoxifen therapy

In recent years, we have seen an explosion of interest in the clinical relevance of CYP2D6 genotype on outcome of breast cancer patients who are treated with tamoxifen. Prospective cohort studies of adjuvant tamoxifen treatment have revealed a wide interindividual variation in the steady state plasma concentrations of active metabolites, endoxifen and 4-hydroxytamoxifen during tamoxifen treatment in patients carrying CYP2D6 genetic variants.8, 9, 11 The patients homozygous for null alleles (categorized as PM) showed nearly one-fourth of endoxifen concentration in plasma, compared with those carrying two normal alleles (categorized as EM).8, 9 The patients carrying two alleles that encode a low-function enzyme, including CYP2D6*10 and CYP2D6*41 (categorized as intermediate metabolizer), had nearly 50% of plasma endoxifen concentration compared with the controls.4, 39, 40, 41 These patients with low endoxifen concentration were suspected to have a poorer clinical outcome.

As shown in Table 1, a number of studies have reported the association between the CYP2D6 genotype and clinical outcome of breast cancer patients receiving the tamoxifen therapy. One of the first studies reported by Goetz et al.42, 43 demonstrated that homozygous carriers of CYP2D6*4 allele had a shorter relapse-free survival (RFS) and disease-free survival than the patients for heterozygous or homozygous for the wild-type allele (hazard ratio (HR), 1.85; P=0.18 for RFS: HR, 1.86; P=0.089 for disease-free survival). Following these reports, Schroth et al.44 published retrospective analysis of 1,325 breast cancer patients with adjuvant tamoxifen monotherapy, and observed that PMs revealed a significantly higher risk of recurrence than EMs with HR of 2.12 for a time to recurrence (P=0.003). These associations were supported by several research groups.45, 46, 47, 48, 49, 50 In Asians, we reported the significant effects of CYP2D6 genotype (especially CYP2D6*10) on RFS in Japanese patients receiving adjuvant tamoxifen monotherapy (HR, 9.52; P=0.000036).40, 51 The worse clinical outcome of tamoxifen therapy in the patients carrying CYP2D6*10 was confirmed in Chinese, Korean, Thai and Malaysian populations.52, 53, 54, 55 However, several discordant results have been also reported.56, 57, 58, 59, 60, 61 More recently, two retrospective analyses of large prospective trails, the ATAC (Alimidex, Tamoxifen, Alone or in Combination) trial and the Breast International Group (BIG) 1-98 trial, were reported.62, 63 In the ATAC analysis, there was no significant association between any of CYP2D6 phenotypic groups and recurrence rates in 588 patients treated with tamoxifen (HR, 1.22; P=0.44; PM relative to EM).62 Similarly, in the BIG 1-98 analysis, no significant difference was found among different CYP2D6 metabolizer groups and cancer-free survival in 973 breast cancer patients (HR, 0.58; P=0.35; EM vs PM).63 As discussed in previous reports, there may be several confounding factors or critical errors in the experimental designs to explain these discrepancies.

Table 1 Summary of studies evaluating association of CYP2D6 genotype with response to adjuvant tamoxifen therapy

One of the most important issues in the pharmacogenomics study is the quality of genotype data. This should be influenced by (i) the accuracy of genotyping methods, (ii) coverage of genotyped alleles and (iii) DNA source. In both of the ATAC and BIG 1-98 studies,62, 63 the authors mentioned the high reproducibility of genotyping methods because of the concordance of genotyping results in duplicate determinations. However, this does not fully guarantee the accuracy of their genotype results. Their genotype results were highly deviated from Hardy–Weinberg equilibrium (χ2 P=10−92 for CYP2D6*4) probably because they used the low-quality genomic DNA extracted from formalin-fixed paraffin-embedded tumor tissues.64, 65, 66, 67 Therefore, they excluded CYP2D6*5 from the analyses, and performed 60-cycle PCR to detect 1846 G>A (CYP2D6*4), which is likely to lead to the misgenotyping results. The importance of wide coverage of CYP2D6 alleles was clearly demonstrated by Schroth et al.68 In the report, the increase of genotyping coverage was shown to increase HR for RFS as well as enhance the statistical power. In our samples, we also detected a lower HR of 5.83 without CYP2D6*5 genotyping data than that of 9.52 (wt/wt vs V/V, N=282; unpublished data). In addition, nearly 30% frequency of loss of heterozygosity at the chromosome 22q, where the CYP2D6 gene is located, in breast cancer cells definitely causes misclassification of patients and leads to misinterpretation of the results if one uses DNAs isolated from tumor tissues (particularly caner-cell rich samples).69

The second critical issue is selection of study participants. To evaluate the effects of CYP2D6 genotype on tamoxifen efficacy, it is scientifically certain that the patients treated only with tamoxifen should be selected. As shown in Table 1, most of studies showing the ‘null’ association included the patients who were treated with a combination of tamoxifen and chemotherapy. We reported significant effects of CYP2D6 genotypes on shorter RFS when we analyzed patients treated with the tamoxifen monotherapy (HR, 9.52; P=0.0032; N=282), but not when we analyzed those with the combination chemotherapy (HR, 0.64; P=0.44; N=167).70, 71 In a combined population (total 449 patents, including 37.2% of those with the combination therapy), HR dropped to 2.45 (95% confidence interval, 1.30–4.54) for wt/wt vs V/V (unpublished data).

These lines of evidence clearly tell us the importance of complete CYP2D6 genotyping using germline DNAs isolated from very carefully selected samples with tamoxifen monotherapy. All of ‘null’ association studies lacked one or multiple elements of these essential factors, as shown in Table 1. Therefore, large prospective studies satisfying these conditions are needed to make a definite conclusion for the value of CYP2D6 genotyping in tamoxifen therapy.

The patients carrying decreased- or impaired-function CYP2D6 alleles consistently showed lower plasma endoxifen concentrations than those having the homozygous normal genotype.4, 8, 9, 11, 39, 40, 41 Plasma endoxifen levels were suggested to associate with clinical outcome of tamoxifen-treated patients.72 Therefore, several research groups recently conducted CYP2D6 genotype-based dose-adjustment studies.73, 74, 75 Irvin et al.74 demonstrated that endoxifen levels were significantly increased when the dose was increased from 20–40 mg in intermediate metabolizer and PM patients; however, endoxifen levels in PM patients were still significantly lower than the normal individuals. We also investigated the effects of the increase of tamoxifen dose from 20 to 30 mg or 40 mg in the patients heterozygous or homozygous for variant alleles, respectively, and demonstrated that endoxifen concentrations were significantly increased to a similar level of the CYP2D6-normal patients who took 20 mg of tamoxifen (Figure 2).75 In these studies, the incidence of adverse events was not affected by the dose adjustment. Although further verification is required especially for PM patients, these results suggest that increased tamoxifen dose is an effective way to maintain the effective endoxifen concentration for the patients carrying decreased function or null alleles of CYP2D6.

Figure 2
figure 2

Steady state plasma concentration of endoxifen before and after dose escalation of tamoxifen in breast cancer patients. The horizontal line indicates the median concentration, the box covers the 25th–75th percentiles, and the maximum length of each whisker is 1.5 × the interquartile range, dots outside the whiskers are outliers. Data from Kiyotani et al.75

Polymorphisms in other genes and clinical outcome of tamoxifen therapy

Other CYPs, including CYP2C9, CYP2C19, CYP3A4 and CYP3A5, UGTs and SULTs are also involved in the metabolism of tamoxifen. Among them, CYP3A5*3 is well investigated in association with tamoxifen metabolism or clinical outcome of tamoxifen therapy; however, no significant association was observed.4, 9, 42, 45, 76, 77 For CYP2C19, a significant association with clinical outcome of tamoxifen treatment was found in carriers of CYP2C19*17,45 but not in the carriers of CYP2C19*2 or CYP2C19*3.45, 59 However, the results have also been contradictive and not conclusive.78, 79 Several investigations on genetic variations in the SULT1A1 gene, including single-nucleotide polymorphisms (SNPs) and copy number variations, found no clear association with tamoxifen efficacy56, 58, 79 and tamoxifen metabolism.9, 57 Further analysis would be required by consideration of ‘allele copy number’ of SULT1A1, as demonstrated in the case of CYP2D6.80, 81, 82

There are several reports investigating the involvement of drug transporters in disposition of tamoxifen and its active metabolites, endoxifen and 4-hydroxytamoxifen. ABCB1 (P-glycoprotein, multidrug resistance protein 1) is an ATP-dependent, efflux transporter with broad substrate specificity widely appreciated for its role in mediating cellular resistance to many anticancer agents.83 ABCB1 is reported to be involved in the transport of active tamoxifen metabolites.84, 85 Several ABCB1 polymorphisms have been reported, including 2667 G>A/T and 3435C>T; however, no SNPs were significantly associated with clinical outcome of tamoxifen therapy.40, 54 ABCC2 (multidrug resistance-associated protein 2) has an important role in the biliary excretion of glucuronides or sulfates of drugs, including tamoxifen and its metabolites.17 We found an intronic SNP of ABCC2 (rs3740065), which is in strong linkage disequilibrium (r2=0.89) with −1774 G/delG, to be significantly associated with clinical outcome of patients with tamoxifen therapy through the screening using haplotype-tagging SNPs.40, 86 An in vitro study reporting that ABCC2 was expressed at higher levels in tamoxifen-resistant breast cancer cells suggests the possibility that active metabolites of tamoxifen are transported by ABCC2 from breast cancer cells.87

We also identified a novel locus, containing C10orf11, associated with RFS in the breast cancer patients treated with tamoxifen alone by the genome-wide association study encompassing a total of 462 Japanese patients (HR, 4.53; P=6.28 × 10−8).88 At present, however, no report is available regarding the function of the C10orf11 protein. Large-scale replication study and further functional analysis are required to verify these associations, and to clarify their biological significance and mechanisms that have effects on the clinical outcome of patients receiving tamoxifen therapy.

Other factors affecting clinical outcome of tamoxifen therapy

As well as the genetic polymorphisms modifying the tamoxifen pharmacokinetics, characteristic of cancers, including gene expression profiles or genomic alterations, are also one of important determinants of individual response to tamoxifen. Many molecules have been identified to be involved in the tamoxifen resistance.89, 90 Several microarray analyses revealed the gene signatures to predict the outcome of adjuvant tamoxifen therapy, such as breast cancer intrinsic subtype,91, 92 21-gene signature (used as OncotypeDX)93 and HOXB13/IL17BR expression ratio.94, 95 Goetz et al.96 reported that combination of CYP2D6 genotype and HOXB13/IL17BR was significantly associated with disease-free survival (log-rank P=0.004) and overall survival (log-rank P=0.009). More recently, Ellis et al.97 clarified the elevated frequency of somatic mutations and genome-structure changes in aromatase inhibitor-resistant tumors by whole-genome sequencing. Therefore, prediction of individual response to tamoxifen using cancer characteristics seems to be effective, and may affect the association results of genetic markers.

Conclusion

Although a large number of investigations on tamoxifen pharmacogenomics have been performed, the association results between CYP2D6 genotype and clinical outcome are still controversial. However, accumulation of the evidence clarifies some of the causes of these controversial results, particularly some scientific issues in the false-negative results, and implies the importance of the quality of genotyping as well as sample selections in the tamoxifen pharmacogenomics study. The important issues learned from the tamoxifen and breast cancer studies are commonly applicable in pharmacogenomics studies. As we are aiming to establish the personalized medicine system in which we select a right patient and provide an appropriate dose of a right drug, the pharmacogenomics study also requires the accurate genotyping using a sufficient number of appropriate patients in order to obtain truly positive results and avoid false-positive and false-negative results. Finally, genotype-guided dose-adjustment based on the CYP2D6 genotypes will be a good example for the personalized medicine. To reduce the medical care cost without losing the quality of medical care, it is very important to use the drugs, which are available at lower cost, on the basis of individual genetic information. As several novel associated SNPs/loci have been identified, integration of genotypes of CYP2D6 and other genes as well as tumor characteristics should be the future approach to predict clinical efficacy of tamoxifen and provide better quality of lives to breast cancer patients.