Introduction

Doxorubicin is an anthracycline chemotherapy agent that is widely used for the curative, adjuvant and palliative treatment of various malignancies. In particular, the combination of doxorucibin and cyclophosphamide (AC regimen) is widely used in the adjuvant settings for the treatment of primary breast cancer. The concerted actions of various influx and efflux transporters that transfer doxorubicin across cellular membranes and the metabolizing enzymes responsible for its biotransformation have been postulated to have significant roles in determining intracellular doxorubicin levels and influencing its disposition (Figure 1). Its main metabolite, doxorubicinol, exhibits lower antitumor potency as compared with doxorubicin,1, 2 but has been shown to accumulate in cardiac tissues and is reportedly 10–30 times more potent than the parent drug at depressing cardiac contractility and causing doxorubicin-induced cardiotoxicity.2, 3 Variability in the pharmacokinetics of doxouribin and its metabolite doxorubicinol have therefore been postulated to contribute to the wide inter-individual variability in doxorubicin treatment outcomes and toxicity profiles.

Figure 1
figure 1

Pharmacology of doxorubicin.

PowerPoint slide

Full size image

Although several in-vitro studies have identified the prototypical ATP-binding cassette (ABC) transporter ABCB1 as the major candidate mediating doxorubicin resistance,4, 5, 6 doxorubicin transport from multidrug-resistant cells that do not overexpress ABCB1 suggests that efflux mechanism(s) distinct from ABCB1 may be involved in altering its disposition and mediating doxorubicin resistance.7, 8, 9 We have previously reported on the pharmacogenetic influence of transporters belonging to the ABC superfamily (ABCB1, ABCG2) and the solute carrier family (SLC22A16), drug-metabolizing enzymes such as carbonyl reductases 1 and 3 (CBR1 and CBR3, respectively), as well as regulatory nuclear receptors, such as the pregnane-X-receptor (PXR), in relation to doxorubicin pharmacokinetics in Asian breast cancer patients.10, 11, 12, 13 However, more recent studies have highlighted the potential importance of other transporters and proteins in modulating the disposition of doxorubicin, including ABCB5, ABCC5 and RLIP76.

ABCB5 is the third member of the human ABC transporter family.14, 15, 16 Similar to ABCB1, ABCB5 has been shown to mediate drug resistance in human cancer cells.17, 18 ABCB5-mediated efflux of doxorubicin has previously been identified as the mechanism underlying resistance to doxorubicin in ABCB5-expressing G3361 human melanoma cells.17 Specifically, melanoma cells expressing ABCB5 had significantly lower doxorubicin accumulation compared with those that did not express ABCB5.17 More recently, Yang et al.19 also found that ABCB5 was upregulated in doxorubicin-resistant breast MCF-7 clones. These findings suggest that ABCB5 may contribute to alterations in the disposition and intracellular levels of doxorubicin as well as doxorubicin resistance phenotype.

ABCC5 (Canalicular multispecific organic anion transporter C, MRP5/ABC33) belongs to the ABCC subfamily.20 Unlike the long type of ABCC proteins, it belongs to the short type of ABCC proteins that lacks of an N-terminal transmembrane domain.21 As an efflux transporter, ABCC5 has been shown to mediate the ATP-dependent transport of several anticancer drugs, including doxorubicin, and has been reported to confer resistance to a number of chemotherapy agents, including methotrexate, and the thymidilate synthase inhibitor raltitrexed. In a study by Yoshida et al.22, analysis of ABCC5 mRNA levels in doxorubicin-resistant human lung cancer cells SBC-3/ADM, AdR MCF-7 and K562/ADM showed markedly high expression of ABCC5 transcripts relative to their respective parental cell lines. Pratt et al.23 further demonstrated that ABCC5-transfected HEK cells had a twofold higher resistance to doxorubicin than non-transfected cells. Taken together, these findings suggest that ABCC5 expression and activity may contribute to variability in doxorubicin disposition observed in cancer patients.

RLIP76 (Ral-binding protein 1, 76 kD) is a multifunctional GTPase-activating transporter protein that is ubiquitously expressed.24 RLIP76 share some similarities with the ABC transporters. The sequence of its ATP-binding sites in the N-terminal (69GKKKGK74) and C-terminal domains (418GGIKDLSK425), for instance, are similar with the P-loop in the ABC transporters.25 However, transmembrane helices have not been identified in the RLIP76 sequence. Taking into account the overexpression of RLIP76 in some cancer cells, the ability of these cells to develop resistance to anticancer agents has highlighted the potential importance of RLIP76 as a target for cancer therapy in recent years. Indeed, RLIP76 has been demonstrated to be involved in the energy-dependent efflux of doxorubicin.26 RLIP76-overexpressing cells have also been shown to exhibit markedly increased efflux of doxorubicin and acquired resistance to doxorubicin-induced cytotoxicity.27, 28 Indeed, the inherently low intracellular accumulation of doxorubicin because of RLIP76-mediated efflux in non-small cell lung cancer has been attributed as the primary mechanism underlying doxorubicin resistance.29, 30 These findings suggest that transport by RLIP76 may prevent normal and malignant tissues from developing doxorubicin-induced toxicities. Although the RLIP76 gene is polymorphic, the impact of RLIP76 polymorphisms on doxorubicin disposition has not been evaluated in doxorubicin-treated cancer patients, including in the Asian population.

The present exploratory study examined the pharmacogenetics of additional efflux transporters, specifically ABCB5, ABCC5 and RLIP76 and their possible influence on the pharmacokinetics of doxorubicin in the same subset of Asian breast cancer patients. The pharmacogenetic profiles of the ABCB5, ABCC5 and RLIP76 genes were screened in three healthy Asian populations, namely the Chinese, Malays and Indians and the genotypic–phenotypic effects of identified polymorphisms in these genes were subsequently investigated in Asian breast cancer patients receiving the combination of doxorubicin and cyclophosphamide as adjuvant chemotherapy.

Materials and methods

Healthy subjects

The healthy population consisted of the three predominant ethnic groups in Singapore (Chinese, n=100; Malays, n=100; Indians, n=100), for whom ethnicities were confirmed against their National Registry Identification Cards. Participants provided both written as well as verbal informed consent to participate in the study. The study was approved by the ethics review committee of the National Cancer Center, Singapore.

Breast cancer patients

Patients with histologically confirmed invasive breast cancer and receiving adjuvant chemotherapy with doxorubicin and cyclophosphamide were recruited (n=62). Both verbal and written informed consent was obtained from all patients. The Patient Information and Consent forms, along with the study protocol were approved by the institutional ethics committee at the National Cancer Centre, Singapore. Patient selection criteria have been reported previously.12

Genotyping of ABCB5, ABCC5 and RLIP76 polymorphisms

Genomic DNA was extracted from the peripheral blood samples (5 ml) using the phenol–chloroform extraction method as described previously.12 Direct DNA sequencing was performed to screen for polymorphisms in the exon and exon–intron boundaries of the ABCB5 (Genbank accession no: AB353947) and RLIP76 (Genbank accession no: NM_006788) genes, as well as ABCC5 polymorphisms (5’-UTR: g.-1990G>A, g.-1821T>C, g.-1679T>A, g.-1205C>T, g.-793C>A, g.-889T>C; intron 5: i.374C>T (rs3749438); exon 8: c.1145A>G; exon 9: c.1185T>C (rs1132776); exon 12: c.1782T>C (rs939336); intron 1: i.1834C>T (rs4148557); intron 2: i.7980C>T (rs2292997); exon 25: c.3624C>T (rs3749442); exon 30: c.4896G>A (rs3749445), c.5557A>G (rs562) and 3’-UTR g.+6272A>G (rs1000002), g.+7161G>A (rs1533682)) as identified by Gwee et al.31

Pairwise correlation analysis between polymorphisms

Pairwise correlations between these polymorphisms and previously reported polymorphisms in ABCB1, ABCG2, SLC22A16, CBR1, CBR3 and PXR genes were quantified by |D'| and rho square (r2) values (Haploview, v4.2, Daly Lab, Broad Institute, Cambridge, MA, USA). The following polymorphisms were included: ABCB1 (c.1236C>T (rs1128503), c.3435C>T (rs1045642)), ABCG2 (c.421C>A (rs2231142)), SLC22A16 (c.146A>G (rs714368), c.755T>C (rs6907567), c.1226T>C (rs12210538), c.312T>C (rs6907567)), CBR1 (g.-48G>A, c.219G>C (rs25678), +967G>A, c.693G>A, c.627C>T), CBR3 (c.11G>A (rs8133052), c.255T>C, c.279C>T, c.606G>A, c.730G>A) and PXR (-31273A>G (rs9832958), -1570C>T (rs3814055), -566A>C (rs1523127), -298G>A (rs2276706), -23914T>G (rs3814056), -6994T>C (rs2472677), -601A>G (rs7643645), IVS2+55A>G (rs1464603), IVS2+78A>G (rs1464602), IVS2-1131C>T (rs2472680), IVS3+72T>G (rs3732356), IVS3+648T>C (rs2472681), IVS4+285G>A (rs3732357), IVS5+845C>A (rs2472682), IVS5-93A>G (rs6785049), IVS6-17C>T (rs2276707), 1448G>A, 1437G>A (rs3732358), 1792A>G (rs3732359), 1944T>C (rs3732360), 2180A>G (rs6438550), 2617A>C (rs3814057), 2654T>C (rs3814058)).10, 11, 12, 13

Doxorubicin administration and pharmacokinetic analysis

Intravenous doxorubicin was administered at 60 mg m–2 over 20 min with standard pre-medications, including intravenous dexamethasone 10 mg, diphenhydramine 50 mg, cimetidine 300 mg or ranitidine 50 mg. Intravenous cyclophosphamide was administered at a dose of 600 mg m–2 over 30 min. Each cycle lasted 3 weeks. Blood samples for pharmacokinetic analysis of doxorubicin were collected at 0, 5, 15 and 30 min, and at 1, 4, 8 and 24 h. Reverse-phase high-performance liquid chromatography with fluorescence detection was used for the quantification of plasma concentrations of doxorubicin and its major metabolite, doxorubicinol, as previously described.12

Pharmacokinetic parameters were determined using non-compartmental methods with a non-linear regression program on WinNonLin version 2.1 (Pharsight, Mountain View, CA, USA). Peak plasma concentrations (Cmax) were directly identified from the concentration-time curves of each individual. The trapezoidal rule was used to calculate the area under the plasma concentration-time curve (AUC) from time zero to the time (t) of the last detectable concentration (AUC0→t). AUC was extrapolated to infinity (AUC0-∞) by adding Ct/λz to AUC0→t, where Ct was the last detectable plasma concentration and λz is the elimination rate constant.

Statistical analyses

The chi-square (χ2) test was used to assess for the departure of genotype frequencies from Hardy–Weinberg equilibrium. The statistical differences between the pharmacokinetic parameters among the genotype groups were analyzed by the nonparametric Mann–Whitney U-test, with statistical significance set at P<0.05. The effects of demographics and clinical characteristics, such as ethnicity, age, body surface area (BSA), on the pharmacokinetic parameters of doxorubicin and doxorubicinol were evaluated using the univariate linear regression analysis. In addition, univariate linear regression analysis was performed to evaluate the associations between polymorphisms reported in this study as well as previously reported studies,10, 11, 12, 13 as listed above, with the pharmacokinetic parameters of doxorubicin. Covariates found to be significant (P<0.05) in the univariate analysis were included in the multivariate regression model and goodness-of-fit of the model was assessed using R2. All statistical analyses were performed using STATA v7.0 (College Station, TX, USA) and SPSS (Chicago, IL, USA).

Results

Patient demographics

As previously reported, the Asian breast cancer patients consisted of Chinese (N=46, 74%), followed by Malays (N=11, 18%) and Indians (N=5, 8%). The median age, height and BSA of the patients were 51 years (range: 29–73 years), 154 cm (range: 144–168 cm) and 1.52 m2 (range: 1.23–1.95 m2), respectively.

Doxorubicin pharmacokinetics

Plasma samples for pharmacokinetic analysis were available in 52 of the 62 recruited Asian breast cancer patients. The pharmacokinetics parameters of doxorubicin and doxorubicinol in these patients are summarized in Table 1. Wide interpatient variability in the pharmacokinetic parameters of doxorubicin and doxorubicinol was observed and has been described previously.12 Heights and weights were shown to be significant covariates affecting doxorubicin AUC0-∞ (P<0.015 and P<0.0001, respectively) and Cmax (P<0.028 and P<0.0001, respectively). Age and ethnicity were shown to be significant covariates affecting doxorubicin half-life (P<0.018). Markers of hepatic (aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase, bilirubin) and renal (serum creatinine) functions were not shown to significantly influence doxorubicin pharmacokinetic parameters.

Table 1 Summary of doxorubicin pharmacokinetic parameters in Asian breast cancer patients
Full size table

ABCB5, ABCC5 and RLIP76 genotype distributions

Three novel polymorphisms were identified by direct sequencing of the coding regions of the ABCB5 gene. All polymorphisms were nonsynonymous transitions: c.2T>C (exon 1; rs34603556), c.343A>G (exon 2; rs2301641) and c.1573G>A (exon 12; rs6461515) polymorphisms resulting in the p.M1T, p.K115E and p.E525K amino-acid changes, respectively. Table 2 summarizes the genotype and allele frequencies of the ABCB5 polymorphisms identified in the healthy subjects of three Asian ethnic groups and the breast cancer patients. All genotype frequencies conformed to Hardy–Weinberg equilibrium. The genotypic and allelic frequencies of these ABCB5 polymorphisms have been previously reported.32 The c.2T>C (rs34603556) polymorphism was approximately twofold higher in the healthy Indians (23%) compared with the Chinese (8%) and Malay (10%) populations. The allelic frequency of the c.2C variant was also significantly higher among the healthy Indian population (0.12) compared with the Chinese (0.05, P=0.007) and Malay (0.04, P=0.026) populations. The ABCB5 c.343A>G (rs2301641) and c.1573G>A (rs6461515) polymorphisms did not differ significantly in frequencies between the three healthy Asian ethnic groups. Among cancer patients, the c.343A>G (rs2301641) variant allele was significantly lower in frequency (0.05) compared with the healthy Chinese population (0.14), whereas the frequencies of ABCB5 c.2T>C (rs34603556) and c.1573G>A (rs6461515) were similar among the breast cancer patients as well as healthy subjects.

Table 2 Genotype and allele frequency of ABCB5 polymorphisms in healthy subjects and Asian breast cancer patients
Full size table

The genotype and allelic frequency of the ABCC5 polymorphisms among Asian breast cancer patients were in agreement with previously published reports in healthy Asian populations (Table 3).31 No polymorphisms were identified in the exon and exon–intron boundaries of the RLIP76 gene in all the three Asian ethnic groups.

Table 3 Genotype and allele frequencies of ABCC5 polymorphisms in Asian breast cancer patients (N=62)
Full size table

Pairwise correlations between polymorphisms

The ABCC5 and PXR genes are both located on chromosomes 3. The ABCB1 and ABCB5 genes are both located on chromosomes 7. The SLC22A16 and ABCG2 genes are located on chromosomes 6 and 4, respectively, whereas CBR1 and CBR3 genes are both located on chromosome 21. Although a classic linkage disequilibrium assessment could not be performed between the single-nucleotide polymorphisms in these genes as linkage disequilibrium analysis would necessitate the assessed single-nucleotide polymorphisms to be on the same chromosome, we performed pairwise correlation analysis between all the single-nucleotide polymorphisms that have been studied in these genes and the results are presented in Supplementary Table 1. Minimal correlations were observed between these single-nucleotide polymorphisms (pairwise r2<0.161), suggesting that most of them are independent from one another.

Genotypic–phenotypic correlates

Pairwise comparisons failed to show any significant associations between the ABCB5 c.2T>C (rs34603556), c.343A>G (rs2301641) and c.1573G>A (rs6461515) polymorphisms and the pharmacokinetic parameters of doxorubicin and doxorubicinol.

With regards to ABCC5 polymorphic variants, significant genotypic–phenotypic correlations were observed with ABCC5 g.+7161G>A (rs1533682) and g.-1679T>A polymorphisms (Table 4). Specifically, breast cancer patients homozygous for the variant g.+7161A (rs1533682) allele had significantly higher clearance of doxorubicin (CL/BSA (Lh−1m−2), median: 30.34; range: 25.41–33.60) when compared with patients who were homozygous for the g.+7161G allele (CL/BSA (Lh−1m−2), median: 22.46; range: 15.04–49.4, P=0.04). In addition, breast cancer patients homozygous for the g.-1679T allele had significantly higher exposure levels of doxorubicinol (AUC0-∞/dose/BSA (hm−5), median: 15.48; range: 6.18–67.17) as compared with patients who were heterozygous carriers of the polymorphism (AUC0-∞/dose/BSA (hm−5), median: 8.88; range: 3.68–21.71, P=0.0001).

Table 4 The influence of ABCC5 g.+7161G>A (rs1533682) and g.-1679T>A polymorphisms on doxorubicin pharmacokinetics in Asian breast cancer patients
Full size table

Multivariate analysis conducted with all the polymorphisms studied in the ABCB5 and ABCC5 genes, as well as ABCB1, ABCG2, SLC22A16, CBR1, CBR3 and PXR genes, revealed that only CBR1 +967G>A genotype status contributed significantly to the overall variability in doxorubicin clearance, CL/BSA (P=0.037) after adjustments for multiple covariates. The effects of other polymorphisms in the ABCB5, ABCC5, ABCB1, ABCG2, SLC22A16, CBR1, CBR3 and PXR genes were not significantly associated with the pharmacokinetics of doxorubicin.

Discussion

Screening the coding regions of the ABCB5 gene in 300 healthy subjects of distinct Asian ethnic groups revealed three novel nonsynonymous polymorphisms (c.2T>C (exon 1; rs34603556), c.343A>G (exon 2; rs2301641) and c.1573G>A (exon 12; rs6461515)). Although differences were noted in the allele and genotype frequencies of ABCB5 polymorphisms among the three Asian ethnic groups, genotype–phenotype analyses revealed a lack of significant associations between the identified ABCB5 polymorphisms and doxorubicin pharmacokinetics in the Asian breast cancer patients. The potential contribution of ABCB5 in mediating doxorubicin transport and chemoresistance was first demonstrated in human malignant melanoma by Frank et al.17 More recently, Kawanobe et al.18 also reported that HEK293 cells transfected with ABCB5 complementary DNA showed approximately 1.5-fold higher resistance to doxorubicin than the parental cells. However, the influence of ABCB5 polymorphisms on the disposition of doxorubicin and its major metabolite doxorubicinol has not been investigated to date. Our results indicate for the first time that polymorphisms in the coding regions of the ABCB5 gene do not significantly influence the disposition of doxorubicin and doxorubicinol.

It is important to note, however, that polymorphisms in both the intronic and promoter regions, which were not analyzed in this study, have been shown to influence ABCB5 gene expression and activity.33, 34 Therefore, identifying the presence of ABCB5 intronic and promoter variants in different populations may be important in elucidating their influence, if any, on the disposition of doxorubicin and doxorubicinol.

The sequencing of the ABCC5 gene and its 3’-flanking region in Japanese by Saito et al.35 identified 85 polymorphisms. Strong linkage disequilibrium between the 5’-UTR g.-1205C>T and 3’-UTR g.+7161G>A (rs1533682) polymorphisms has also been reported.31 These two polymorphisms are separated by >100 kb, suggesting that the ABCC5 gene is found within a region of strong linkage disequilibrium.31 The ABCC5 haplotype profiles were also observed to be highly similar between the Chinese and Malays, whereas those of Indian ethnicity had similar haplotype profiles with the Caucasians.31 Similar to ABCB5, ABCC5 has also been shown to mediate resistance to doxorubicin, particularly in non-small cell lung cancer.36, 37 However, no studies to date have investigated the influence of the identified ABCC5 polymorphisms on the pharmacokinetics of its substrates, including doxorubicin.

In this study, the ABCC5 3’-UTR g.+7161G>A (rs1533682) polymorphism was found to significantly influence the clearance of doxorubicin in our Asian breast cancer patients. Patients homozygous for the variant allele (AA) had approximately 35% significantly higher clearance of doxorubicin and 38% lower Cmax values compared with the reference group (Table 4). It is not known, however, if the higher clearance and lower exposure levels of doxorubicin observed in patients carrying the 3’-UTR g.+7161G>A (rs1533682) polymorphism are associated with altered pharmacodynamic outcomes. Given that its antitumor potency is mainly attributed to the parent drug doxorubicin, the significantly higher clearance of doxorubicin observed in patients with the homovariant genotype status (AA) at this locus may potentially translate to poorer outcomes in these patients as compared with patients carrying wild-type alleles. However, further investigations with a larger sample size are necessary to confirm this.

Recently, Zhu et al.38 demonstrated that miR-128 induced overexpression of ABCC5 in breast cancer cells and contributed to doxorubicin resistance. It is also interesting to note that the ABCC5 gene is highly expressed in myocardial tissues, including endothelial and smooth muscle cells where it has a crucial role in the efflux transport of the second messenger 3’,5’-cyclic GMP.39 The latter functions as a second messenger of nitric oxide, which regulates smooth muscle tone,40, 41 cardiac contractility42 and cardiomyocyte hypertrophy.43, 44 It is thus reasonable to hypothesize that these regulatory efflux functions of ABCC5 in cardiomyocytes coupled with the higher clearance of doxorubicin associated with the ABCC5 g.+7161G>A (rs1533682) polymorphism may serve as a protective mechanism against doxorubicin-induced cardiotoxicity, which merits further exploration.

Breast cancer patients homozygous for the ABCC5 g.-1679T allele were also found to have significantly higher exposure levels to doxorubicinol compared with patients who were heterozygous for the polymorphism. However, no significant influence on the exposure levels to doxorubicin was observed, and could be attributed to possible differences in the transport affinities of doxorubicin and doxorubicinol by the ABCC5 transporter protein. This finding is of potential clinical significance as doxorubicinol has been suspected to modulate doxorubicin-induced cardiotoxicity.2 Doxorubicin-treated patients homozygous for the ABCC5 g.-1679T reference allele may thus be at potentially higher risks of developing cardiotoxicity. However, this needs to be further investigated in a larger study that is adequately powered to detect differences in incidence of toxicities among the different genotype groups. The functional impacts of the g.+7161G>A (rs1533682) and g.-1679T>A polymorphisms in altering doxorubicin and doxorubicinol substrate specificity of ABCC5 have not been elucidated and probably need further characterization.

Mechanisms of doxorubicin transport and resistance mediated by RLIP76 have been investigated in several studies as Awasthi et al.26 demonstrated that an ATP-dependent uptake of doxorubicin into vesicles prepared from erythrocyte membrane, which do not express ABCB1 was completely inhibited by antibodies to RLIP76. The exceptionally broad substrate specificity (structurally unrelated anionic compounds, weakly cationic compounds (doxorubicin, dihydrodoxorubicin, daunomycin and vinblastine) and uncharged compounds (doxorubicinone, deoxydoxorubicinone and dihydrodoxorubicinone)) and its wide tissue distribution suggest that RLIP76 may function as an important efflux transporter for both xenobiotics and endobiotics.45, 46 In this study, no polymorphisms were identified, suggesting a lack of genetic variability in the RLIP76 gene in the Asian ethnic groups. However, it is conceivable that contributions from RLIP76 to variations in its activity or expression may arise from rare variants including those in the untranslated regions that were not detected in this study.

In conclusion, the present pharmacogenetic analysis of ABCB5, ABCC5 and RLIP76 transporter proteins in Asian breast cancer patients suggest that the ABCC5 g.+7161G>A (rs1533682) and ABCC5 g.-1679T>A polymorphisms may significantly influence the pharmacokinetics of doxorubicin and doxorubicinol, respectively. No significant influence of the newly identified ABCB5 polymorphisms on doxorubicin pharmacokinetic parameters was observed in this study. Taken together, the findings of this study as well as our previous findings on pharmacogenetic profiling across the doxorubicin pathway10, 11, 12, 13 further explain the importance of considering the polygenic impact of various functionally important candidate polymorphisms on the pharmacokinetics of doxorubicin and doxorubicinol.

Notably, the results of the multivariate analysis evaluating the relative contributions of all the polymorphisms studied to date in our Asian breast cancer patients revealed that only CBR1 +967G>A remained significantly associated with doxorubicin clearance. However, the study may be limited by sample size and may not be adequately powered for such multivariate approach. Further studies should be explored in a larger cohort of breast cancer patients belonging to other ethnic groups.