Drug resistance can occur at several levels and is the major cause of treatment failure in oncology. The ABC (ATP-binding cassette) transporters, beginning with the discovery of P-gylcoprotein (Pgp) almost 30 years ago, have been intensively studied as potential mediators of drug resistance. Although we understand that drug resistance is almost certainly multifactorial, investigators have attempted to link anticancer drug resistance to overexpression of ABC transporters and the consequent reduction in drug accumulation. A body of evidence implicated Pgp as being important in clinical outcome; however, critical studies aimed at proving the hypothesis using Pgp inhibitors in clinical trials have to date failed. Identification of the MRP (multidrug resistance protein)/ABCC subfamily expanded the possible mechanisms of reduced drug accumulation, and the discovery of ABCG2 added a new chapter in these investigations. Correlative studies examining ABCG2 and the ABCC subfamily members in clinical drug resistance have been less avidly pursued, while basic molecular studies of structure and function have proceeded briskly. Recently, studies have focused on how single nucleotide polymorphism in multidrug transporters might affect the pharmacokinetics and pharmacodynamics of anticancer agents. These studies suggest an important role for ABC transporters in pharmacology, independent of the ultimate determination of their role in multidrug resistance.
- ABC (ATP-binding cassette) transporter
- multidrug resistance
- multidrug resistance protein (MRP)
With approximately half of all cancer diagnoses leading to death in the modern era, there is a compelling need to address chemotherapy drug resistance. The role of specific drug resistance mechanisms has been revisited again with the development and success of agents identified for novel therapeutic targets. Drug resistance can occur at many levels including host drug metabolism, drug delivery, microenvironment and cellular mechanisms, and most of these mechanisms are poorly studied. Only cellular mechanisms of drug resistance have been examined in detail, illustrated in Figure 1, and even these studies have often been inconclusive. As one example, the existence of specific transport mechanisms for drug uptake was until recently completely unexamined, along with the obvious potential for chemotherapy resistance. Thus we are presented with the reality that, in 2005, the principal means by which cells escape chemotherapy are not known.
Having conceded that accounting for drug resistance completely cannot be accomplished, we can proceed to evaluate the evidence supporting a role for ABC (ATP-binding cassette) transporters in conferring drug resistance. With the cloning of the human genome, it became clear that there were 48 human ABC transporters , provoking consternation that any number of those could confer drug resistance if expressed in the right setting. However, further studies have suggested that many ABC transporters have a restricted substrate profile, and to date only 13 have been shown to transport chemotherapy drugs. Among these, only three have been pursued as potential mechanisms of resistance in the clinic. These three, MDR-1/Pgp (multidrug resistance 1/P-glycoprotein), MRP1 (multidrug resistance protein 1) and ABCG2, are as interesting for their role in normal physiology as for their role in drug resistance in clinical oncology .
The criteria upon which we should rely to define an ABC transporter as being important in drug resistance could be likened to the famous Koch's postulates, which were used to establish the role of bacteria in the generation of infectious disease: (1) the micro-organism must be found in all cases of the disease; (2) it must be isolated from the host and grown in pure culture; (3) it must reproduce the original disease when introduced into a susceptible host; and (4) it must be found in the experimental host so infected.
We can modify those criteria to something more appropriate for medical oncology, and apply the standards to each of the three transporters in turn: (1) is the transporter expressed in drug-resistant cancers?; (2) do levels increase after exposure to chemotherapy?; (3) is expression correlated with poor outcome?; and (4) is inhibition of the transporter associated with clinical benefit?
It should be noted, however, that detection of the ABC transporters in clinical samples requires reproducible, validated clinical assays. For Pgp, the sensitivity, specificity and reproducibility of the antibody-based assays has been a major problem. As a result, for many tumour types the questions cannot be answered with reliability.
Pgp, the first human ABC transporter discovered by Juliano and Ling [2a] as a 170 kDa protein overexpressed in drug-resistant cells, has been intensively studied both for its role in normal physiology and its potential role in clinical drug resistance. Among the list of Pgp substrates are anticancer agents that can be found in the up-front, or initial, therapy of almost all major cancer types, including doxorubicin, daunorubicin, paclitaxel, docetaxel, vincristine, vinblastine, vinorelbine and mitoxantrone. The possibility of increasing the intracellular accumulation of these agents offers the hope of increasing anticancer efficacy for a large number of tumour types.
One question that is frequently asked is the appropriate tumour type for testing Pgp-mediated drug resistance reversal. That question cannot at present be answered with certainty. We know that Pgp is widely expressed in multiple tumour types. Its near-ubiquitous expression in kidney cancer and adrenal cancer cannot explain the intrinsic refractoriness of those tumour types to both substrate and non-substrate drugs alike. It is likely to be more important in tumour types such as breast cancer, sarcoma and AML (acute myelogenous leukaemia), where a subset have detectable Pgp expression at diagnosis, and higher levels or more frequent detection in recurrent or relapsed disease (reviewed in ). Although not consistently confirmed in solid tumours, studies of paired samples have demonstrated increased Pgp expression following chemotherapy in a variety of tumour types, including breast, ovarian, bladder, central nervous system and cervical cancers. Increased levels suggest an important, or at least permissive, role in drug resistance. In a tumour type with less frequent expression, ovarian cancer, Pgp was recently reported to confer a survival disadvantage .
The disease for which Pgp has most consistently met our ‘Koch's postulates’ is AML. In this disease, expression has been documented in approx. 50% of clinical samples, with increasing levels in recurrent leukaemic cells. Consistent results have been obtained with multiple methodologies: functional assays by flow cytometry, expression by immunoflow cytometry and expression by RNA analysis [4–6]. Detection of expression has correlated with reduced response rate (or increased resistance rates) or decreased survival in numerous studies, with the precise endpoint varying from study to study. These assessments may need to be supported or revisited by recent reports that the C3435→T single nucleotide polymorphism (SNP) is associated with reduced levels of Pgp . An endogenous regulator of Pgp expression could have confounded previous correlations with patient outcome. Although some discrepancies were observed, most studies have suggested functional relevance of this polymorphism. For example, Jamroziak et al.  reported a correlation between improved clinical outcome and the C3435T polymorphism in patients with AML.
The clinical evaluation of Pgp inhibitors has provided more disappointment than success. Beginning with the first-generation Pgp inhibitors, such as verapamil, quinidine and cyclosporin A, and continuing with the second-generation inhibitors, such as PSC 833 and VX-710, clinical trials have sought to resolve the Pgp hypothesis by proving treatment benefits. Most clinical trials were conducted with Phase II designs that did not allow determination of the benefit of adding an inhibitor to the treatment regimen. Emblematic are the disappointing trials in breast cancer and ovarian cancer [9–11]. One major flaw in the trials with PSC 833 was the requirement for reduction of the dose of the anticancer agent due to a pharmacokinetic interaction most likely involving CYP3A (cytochrome P450 3A). In the setting of these clinical failures, two large randomized trials in AML deserve special mention. One, reported by List et al. , combined daunorubicin and cytosine arabinoside with or without cyclosporin A in patients with poor risk AML. A 2-year follow-up from this trial revealed a statistically significant improvement in relapse-free and overall survival. A second trial, reported by Baer et al. , combined PSC 833 with mitoxantrone, etoposide and cytosine arabinoside. This trial closed early due to greater toxicity in the treatment arm. However, among the subset of patients with detectable efflux positive leukaemia, there was a statistically significant improvement in complete remission rate, and a trend towards an improved disease-free survival. Had the patients been selected for enrolment in this trial on the basis of expression of Pgp, there may have been a detectable improvement in outcome.
New Pgp inhibitors, so-called third-generation agents, have been developed that are potent and non-toxic, and have minimized the pharmacokinetic interaction. One of these agents, tariquidar (XR9576), continues in clinical trials at the National Cancer Institute and at other sites. Imaging studies with 99mTc-sestamibi before and after tariquidar show clear inhibition of brisk renal and hepatic clearance, and in some patients increased tumour imaging . Functional assays in CD56+ cells by flow cytometry show inhibition of rhodamine efflux from the high Pgp-expressing CD56+ cells  for 48 hours after a single dose and beyond. Thus surrogate studies have confirmed the inhibition of Pgp-mediated efflux in patients. The real question is how to move beyond the confirmation of efflux inhibition in patients in a surrogate assay to documentation of the inhibition of Pgp in tumours. New imaging agents may aid in this undertaking and also provide a means of prospectively identifying patients in whom Pgp inhibition may be important.
The failure to convincingly prove the importance of Pgp in clinical oncology has hampered the clinical development of agents to inhibit other multidrug transporters. The multidrug-resistance-associated protein MRP1, cloned in 1992, has not been systematically studied in clinical samples. A low level of expression of MRP1 is ubiquitous, and higher levels of expression have been reported in only a handful of studies. These studies have suggested that a subset of patients exhibit high expression of MRP1 in lung, breast and ovarian cancer, chronic lymphocytic leukaemia and AML. The expression in lung cancer has been reported at 80–90%. 99mTc-sestamibi, which has the advantage or disadvantage of being a substrate for both Pgp and MRP, has been used in multiple small imaging studies in lung cancer. Thus it is interesting to note, although not with certainty attributable to MRP, the multiple reports demonstrating a correlation between response or survival from lung cancer and sestamibi retention . No studies have reported detection of increased MRP1 levels after treatment, and few studies have been able to correlate expression with clinical outcome. With the Pgp precedent, it is understandable that only a few clinical trials aimed at MRP1 inhibition have been undertaken. These trials have employed the dual Pgp and MRP inhibitor VX-710. In ovarian cancer, clinical benefit was observed in 31% of patients treated with VX-710 and paclitaxel . In sarcoma, 2 of 15 patients had a partial response following treatment with VX-710 and doxorubicin . With the Phase II design, it is difficult to discern whether VX-710 offered any improvement over that anticipated in each of the patient populations studied.
The third major ABC transporter currently understood to have a possible role in multidrug resistance is the BCRP (breast cancer resistance protein), or ABCG2 . Cell lines with ABCG2 overexpression are resistant to mitoxantrone, topotecan and irinotecan (and its active metabolite SN-38). Numerous recent reports have appeared, adding to the list of ABCG2 substrates, including methotrexate, flavopiridol, novel camptothecins, homocamptothecins, imatinib and gefitinib. In addition, conjugated forms of several compounds have also been described as substrates, including sulphated oestrogens, SN-38 glucuronides and methotrexate polyglutamates. The inclusion of these compounds as substrates suggests that ABCG2 may well be an organic anion transporter, like members of the ABCC subfamily of ABC transporters. Despite its characterization as a multidrug transporter, ABCG2 has been implicated in several normal physiological roles, based on distribution of expression in normal tissues. These roles include the maternal–fetal barrier, based on expression in the chorionic villi; the blood-brain barrier, based on high expression in endothelial cells in the brain; modulator of oral absorption; physiology of normal stem cells; transport of haem precursors, based on increased retention in mice with deleted abcg2; and xenobiotic protection.
Studies with both Pgp and MRP1 have demonstrated that amino acid variation, particularly within the transmembrane domains, could alter substrate specificity or transport efficiency. This has also been confirmed for ABCG2 . A mutation at amino acid position 482 in the third transmembrane domain of ABCG2 was acquired during selection for drug resistance in several human cell lines, including Arg482→Thr in MCF-7 AdVp cells (selected in doxorubicin) and Arg482→Gly in S1-M1–80 cells (selected in mitoxantrone). Mutation to serine or methionine also occurred in murine abcg2 following selection of cells in doxorubicin . These mutations resulted in a gain-of-function, adding resistance to anthracyclines and rhodamine, as well as increasing resistance to mitoxantrone. The altered substrate specificity has not yet been explained mechanistically. The importance of the observations concerning amino acid 482 was that altered amino acid sequences could alter substrate specificity, implying that important SNPs could be found.
Our laboratory and several others [22–26] have undertaken the sequence analysis of collections of DNA samples in order to identify SNPs. Figure 2 presents the results from our studies, as well as those published from other populations. Transfection studies with cell lines expressing variant ABCG2 proteins revealed that the Gln141→Lys SNP was likely to confer altered transport efficiency . These studies demonstrated a 2–5-fold reduction in IC50 values in these cells, despite selection of clones with comparable surface expression of ABCG2. Clinical studies followed, suggesting increased oral drug absorption in patients with the Gln141→Lys variant .
Confirmation that such an impact could result from an SNP will indicate that ABCG2 indeed plays an important role in oral drug absorption, and that its role in pharmacology will have to be carefully studied. An impact on drug distribution could also be linked with the SNP. In the case of the C3435→T polymorphism for Pgp, an improvement in survival in acute leukaemia linked to the polymorphism suggests an improvement in drug penetration or drug availability. For the Gln141→Lys SNP, increased drug availability could lead to increased toxicity, particularly for an agent like irinotecan with a narrow therapeutic window.
The clinical development of Pgp inhibitors has been limited by potency in the first-generation agents and drug interaction in the second-generation agent. However, it is also possible that effective Pgp inhibition results in increased bone marrow toxicity because of inhibition of efflux from haematopoietic precursors. With emerging data suggesting that stem cells express ABCG2, it is possible that inhibitors of ABCG2 will increase normal tissue toxicity, precluding the administration of standard doses of anticancer agents. Until a clinical trial is performed, the extent to which this will be a problem is unknown. Meanwhile, numerous compounds have been reported to inhibit ABCG2, including fumitremorgin C (FTC), tariquidar, flavopiridol, gefitnib and imatinib. One intriguing possibility is the use of a ‘multifunctional modulator’, a compound able to inhibit more than one of the transporters. Table 1 provides a list of reported multifunctional inhibitors [29–32].
A final approach in the development of strategies that circumvent drug resistance is to identify compounds that are not substrates for ABC transporters. With the difficulties inherent in proving that Pgp, MRP1 or ABCG2 are important in clinical drug resistance, and the corollary that inhibiting them is an important therapeutic strategy, effective non-substrate drugs become a real alternative. The epothilones are compounds that bind and stabilize tubulin in a manner comparable to that of the taxanes, but are not subject to Pgp-mediated resistance . A novel homocamptothecin, BN80915, is more effective than irinotecan in both ABCG2-expressing and non-expressing cells . Clinical trials with these agents in tumours that express the transporters will be needed to evaluate this strategy.
The recent approval of several novel targeted therapies offers a useful model for our thinking in the development of targeted ‘anti-transporter therapy’ or ‘drug resistance reversal’. Gefitinib, recently approved for lung cancer for its over-10% response rate in lung cancer [35–37], binds to the EGFR (epidermal growth factor receptor). Until recently, the factors that controlled whether or not a tumour responded to gefitinib were not understood. EGFR inhibition is accompanied by mitogen-activated protein kinase inhibition, regardless of whether a cell is sensitive to the agent, but it is mutation of the EGFR that determines sensitivity . Imatinib, effective in a significant percentage of patients with the intrinsically active bcr-abl kinase, also has activity in gastrointestinal stromal tumours, a disease with ckit gain-of-function mutations. Tumours without the ckit mutation are not sensitive. The anti-HER2/neu antibody Herceptin is only useful in patients with high-level (3+) expression of HER2. Deriving a paradigm, the development of ‘anti-transporter therapy’ should be approached by the identification of the subset of tumours in which ABC transporters are actively conferring resistance. This requires the development of new technologies, such as new imaging agents, that will allow the prospective functional evaluation of drug delivery in the clinical setting.
It can be strongly argued that ABC transporters are valid molecular targets, and that the ideal inhibitor for that target has not yet been identified. When considering the validity of ABC transporters as oncologic targets, several points should be considered. Upfront treatment for almost all malignancies includes at least one agent known to be a transporter substrate. Work with the National Cancer Institute drug screen suggested that Pgp substrates were numbered in the hundreds, if not thousands; and similar studies need to be carried out for ABCG2. Newly approved agents and agents in the chemotherapy drug development pipeline are substrates for multidrug transporters, including depsipeptide (FR901228), STI 571 (Gleevec), gefitinib, novel camptothecins and flavopiridol. Taken together, the evidence suggests that although the first- and second-generation inhibitor trials failed to confirm the importance of Pgp in clinical oncology, there is solid scientific rationale for developing the tools to study multidrug transporters as ‘novel molecular targets’.
↵1 The Cancer Research UK Lecture.
Transporters 2004: International Symposium on Membrane Transport and Transporters: Focused Meeting held at Selwyn College Cambridge, 2–5 September 2004. Edited by S.A. Baldwin (Leeds, U.K.) and P.M. Taylor (Dundee, U.K.).
Abbreviations: ABC transporter, ATP-binding cassette transporter; AML, acute myelogenous leukaemia; EGFR, epidermal growth factor receptor; MRP, multidrug resistance protein; Pgp, P-glycoprotein; SNP, single nucleotide polymorphism
- © 2005 The Biochemical Society