Physiological Reviews

Transmembrane Transport of Endo- and Xenobiotics by Mammalian ATP-Binding Cassette Multidrug Resistance Proteins

Roger G. Deeley, Christopher Westlake, Susan P. C. Cole


Multidrug Resistance Proteins (MRPs), together with the cystic fibrosis conductance regulator (CFTR/ABCC7) and the sulfonylurea receptors (SUR1/ABCC8 and SUR2/ABCC9) comprise the 13 members of the human “C” branch of the ATP binding cassette (ABC) superfamily. All C branch proteins share conserved structural features in their nucleotide binding domains (NBDs) that distinguish them from other ABC proteins. The MRPs can be further divided into two subfamilies “long” (MRP1, -2, -3, -6, and -7) and “short” (MRP4, -5, -8, -9, and -10). The short MRPs have a typical ABC transporter structure with two polytropic membrane spanning domains (MSDs) and two NBDs, while the long MRPs have an additional NH2-terminal MSD. In vitro, the MRPs can collectively confer resistance to natural product drugs and their conjugated metabolites, platinum compounds, folate antimetabolites, nucleoside and nucleotide analogs, arsenical and antimonial oxyanions, peptide-based agents, and, under certain circumstances, alkylating agents. The MRPs are also primary active transporters of other structurally diverse compounds, including glutathione, glucuronide, and sulfate conjugates of a large number of xeno- and endobiotics. In vivo, several MRPs are major contributors to the distribution and elimination of a wide range of both anticancer and non-anticancer drugs and metabolites. In this review, we describe what is known of the structure of the MRPs and the mechanisms by which they recognize and transport their diverse substrates. We also summarize knowledge of their possible physiological functions and evidence that they may be involved in the clinical drug resistance of various forms of cancer.


Development of multidrug resistance (MDR) occurs during treatment of many forms of malignant and infectious disease and is often the ultimate cause of treatment failure. Although MDR may develop in response to a specific drug or drug combination, it often encompasses agents to which individuals have not been previously exposed, and which may or may not share targets and mechanisms of action with those that elicited development of resistance. In the case of cancer, these agents include conventional cytotoxic natural product type drugs (e.g., the Vinca alkaloids, including vinblastine and vincristine, and anthracyclines, such as doxorubicin and daunorubicin), alkylating agents (e.g., melphalan), platinum-containing compounds (e.g., cisplatin and carboplatin), antimetabolites (e.g., methotrexate), and nucleoside/nucleotide analogs (e.g., gemcitabine and cytosine arabinoside). However, it is becoming increasingly apparent that resistance can also limit the efficacy of newer, so-called targeted therapeutic agents, such as imatinib mesylate (Gleevec) (133, 264) and various anti-human immunodeficiency virus (HIV) drugs (167, 235, 263, 265). Not surprisingly, no single mechanism has been identified that can account for resistance to the entire spectrum of anticancer drugs in common use, and it is widely accepted that clinical resistance is multifactorial. However, the discovery more than 30 years ago of P-glycoprotein (P-gp/MDR1) demonstrated that it was possible for a single protein to confer resistance to a relatively large number of structurally diverse drugs with different mechanisms of action (219). The cross-resistance spectrum conferred by P-gp/MDR1 typically encompasses a broad range of natural product type drugs (151, 502), but it does not include other clinically important chemotherapeutic agents, such as platinum compounds, nucleoside analogs, or alkylating agents.

P-gp/MDR1 is a member of the ATP binding cassette (ABC) superfamily of transmembrane transporters, and it functions as a direct active transporter of the drugs to which it confers resistance (10). Thus increased drug efflux is a defining characteristic of the MDR phenotype conferred by this transporter protein. In humans, the ABC superfamily includes 49 genes that have been assigned to a “family tree” with 7 branches, designated A through G (94). Although the vast majority of the proteins are energy-dependent transporters, the superfamily also contains examples of one channel gated by ATP binding and hydrolysis, i.e., the cystic fibrosis transmembrane conductance regulator (CFTR/ABCC7) (417) and ATP-dependent potassium channel regulators, such as the sulfonylurea receptors SUR1/ABCC8 (3, 203) and SUR2/ABCC9 (204). The range of exogenous and endogenous substrates transported by mammalian ABC proteins is vast, and defects in a number of these genes are the cause of inherited disorders, with cystic fibrosis being the most common and extensively studied (258, 390, 417).

The core functional unit of the ABC transporters consists of two polytropic membrane spanning domains (MSDs) [each of which typically contains 6 transmembrane (TM) helices but may range from 5 to 10] and two nucleotide binding domains (NBDs) (171). All four of these domains can be encoded in a single polypeptide in the order NH2-MSD-NBD-MSD-NBD-COOH (Fig. 1). Alternatively, the functional transporter may be a homo- or heterodimer of polypeptides, each of which contributes an MSD and an NBD. In some cases, the more typical NH2-MSD-NBD-COOH order may be reversed, as exemplified by breast cancer resistance protein (BCRP/ABCG2) (108). The NBDs of ABC proteins all contain Walker A and B motifs first identified in F1 ATPases that are essential for ATP binding and hydrolysis (522). In addition, they also contain a motif known as the “C” signature that is characteristic of the ABC ATPases and that has the core sequence LSGGQ (170). Although the C signature is clearly intimately involved in ATP hydrolysis, its role in nucleotide binding has not been established, and in general, it appears not to be essential (19, 253, 291, 397, 448, 490, 496).

FIG. 1.

Topology of short and long ABCC proteins. The figure illustrates a simple, probable topology of the “long” and “short” multidrug resistance proteins (MRPs). In the case of MRP1, the model is supported by considerable experimental data (see sect. ivA). The NH2 terminus of MRP2 has also been shown to be extracellular. However, the topologies predicted by different computer-assisted algorithms differ, and in the case of MRP2 MSD2, several favor a topology involving 4 rather than 6 transmembrane helices.

Over the last two decades, it has become evident that P-gp/MDR1 is far from being the only human ABC transporter that, in vitro at least, can confer resistance to clinically important chemotherapeutic agents (32, 61, 71, 72, 85, 108, 154, 192, 252, 268, 537). The number of additional transporters has now reached 10, and with the exception of one protein (BCRP/ABCG2), they all belong to the multidrug resistance protein (MRP/ABCC) family (94). In vitro, the MRP-related proteins can collectively confer resistance to natural product drugs and their conjugated metabolites, as well as platinum-containing compounds, folate antimetabolites, nucleoside and nucleotide analogs, arsenical and antimonial oxyanions, peptide-based agents and, in concert with changes in conjugating or biosynthetic enzymes, alkylating agents (Fig. 2) (96, 161, 256, 287). Aside from their potential role in drug resistance, several MRPs together with P-gp/MDR1 (10, 114) and BCRP/ABCG2 (109) have been recognized as major components of the distribution and elimination pathways for a wide range of both anti-cancer and non-anti-cancer drugs and metabolites (161, 256). In addition, because of their presence in a number of tissue/blood barriers protecting sanctuary sites in the body, they also markedly influence the disposition of drugs and other chemicals (80, 113, 270, 287, 360, 404, 536, 571). In this review, we focus primarily on the structure of the MRPs and their mechanisms of substrate recognition and transport. We also summarize the state of knowledge of their normal physiological functions and the evidence that they may or may not play a role in the clinical drug resistance of various forms of cancer.

FIG. 2.

Subcellular localization and substrate specificity of the MRPs. A cartoon is shown of two polarized cells, one expressing long MRPs (left) and the other, short MRPs (right). The subcellular location(s) of each protein on the apical (upper) or basolateral membranes is shown. The presence of an individual protein in both locations indicates cell type specific differences in subcellular distribution. In the case of MRPs 7, 8, and 9, the question mark indicates that their subcellular localization is not known. Some of the major classes of substrates for each protein are indicated, as well as specific examples of substrates chosen to illustrate the overlap in substrate profiles among the MRPs. Additional details are provided in sections vvii.



The first member of the MRP transporter family was cloned in 1992 from the drug-selected human lung cancer cell line H69AR (72, 342). This cell line is a multidrug resistant derivative of the small cell lung cancer cell line H69 that was selected by repeated exposure to the anthracycline doxorubicin. Although selected in a single drug, H69AR cells display cross-resistance to a wide range of structurally unrelated natural product cytotoxic drugs that have a variety of different molecular targets. The multidrug resistance phenotype of the H69AR cells was somewhat similar to that previously, exclusively associated with the overexpression of P-gp/MDR1 (151, 219, 502). However, despite extensive investigation, no evidence of increased expression of P-gp/MDR1 was detectable in the H69AR cells (70, 342), nor was its resistance phenotype readily reversible by agents that reverse P-gp/MDR1-mediated resistance (68, 71).

The primary cause of the multidrug resistance phenotype of H69AR cells was ultimately identified by screening for mRNAs that were stably overexpressed following drug selection (72). Several overlapping cDNA clones were obtained that corresponded to an mRNA with an open reading frame encoding a protein of 1,531 amino acids (72). Cytogenetic analyses indicated that the gene encoding the overexpressed mRNA was located at chromosome 16p13.1 and was amplified ∼100-fold in H69AR and HT1080/DR4 cells (72, 470). Analysis of the predicted sequence of the protein, initially named multidrug resistance-associated protein (MRP) and subsequently multidrug resistance protein 1 (MRP1), suggested the presence of numerous TM helices and two cytoplasmic domains containing motifs associated with the ABC superfamily of transmembrane transporters. Thus MRP1 (ABCC1) appeared to belong to the same protein superfamily as P-gp/MDR1 and, as a consequence, was considered a possible cause of the multidrug resistance characteristics of H69AR cells. However, the computer-predicted topology of the MSDs of MRP1 was unusual for an ABC transporter. Furthermore, the two putative NBDs, which in many ABC proteins are identical or very similar, were relatively divergent, even with respect to highly conserved elements such as the Walker B motif and the C motif or signature sequence. The NBDs of MRP1 were in fact much more similar to those of CFTR which, as indicated above, functions as a chloride channel rather than a drug pump (72).

The lack of structural identity, only ∼19% overall, between MRP1 and P-gp/MDR1 was unexpected given the similarity between the phenotypes of H69AR cells and cells overexpressing P-gp/MDR1 (72). Nevertheless, gene transfer experiments established that MRP1 was capable of conferring resistance to several major families of natural product drugs, including anthracyclines, Vinca alkaloids, and epipodophyllotoxins (73, 146). In transfected cells, overexpression of MRP1 also resulted in lower intracellular levels of drug and increased rates of efflux, indicating that, like P-gp/MDR1, MRP1 functioned as a drug efflux pump (73, 553). However, these early studies with MRP1 transfected cells revealed some significant differences between the drug resistance profiles conferred by the two proteins. In addition to conferring resistance to many natural product drugs included in the P-gp/MDR1 drug resistance profile, MRP1 and its subsequently cloned murine ortholog conferred resistance to heavy metal oxyanions such as sodium arsenite, sodium arsenate, and antimony potassium tartrate (73, 482). Although mammalian P-gp/MDR1 does not confer resistance to these compounds, a previously identified ABC transporter in Leischmania tarentoliae (ltpgpA), originally thought to be a P-gp/MDR1 homolog, had been shown to increase resistance to arsenicals (379). Sequence comparisons indicate that ltpgpA is a homolog of MRP1 rather than P-gp/MDR1 (72). With respect to clinically important cancer drugs, P-gp/MDR1 and MRP1 also differ in their ability to confer resistance to taxanes. Although these compounds are very good P-gp/MDR1 substrates (340), MRP1 confers only low levels of taxane resistance (73, 553).


Discovery of the second member of the MRP family, MRP2 (ABCC2), occurred ∼4 years after the discovery of MRP1 (39, 381, 494). By this time, considerable information concerning the substrate profile and tissue distribution of MRP1 had accumulated, and it was known that the protein was capable of transporting a wide range of conjugated organic anions. The substrate profile of MRP1 was very similar to that of a functionally characterized bile canalicular transporter, the canalicular multispecific organic anion transporter (cMOAT), which was defective in certain strains of mutant rats and sheep and in individuals suffering from a mild, inherited form of conjugated hyperbilirubinemia known as Dubin-Johnson syndrome (110, 430, 476). The similarity in substrate profiles was such that MRP1 was considered a possible candidate for the canalicular transporter known as cMOAT. However, the protein responsible was subsequently shown to be an MRP1 homolog (224, 230, 382). MRP2 was first cloned in 1996 from normal rat liver using cDNA probes corresponding to predicted highly conserved regions of human MRP1 (49, 381).

C. MRPs 3–6 (ABCC3–6)

Among all MRP family members, MRP2 displays the highest similarity to MRP1 with respect to substrate specificity (60, 85, 119, 187, 384). However, from an evolutionary standpoint, MRP1’s closest relative is MRP3, whose partial sequence was first reported in 1997, together with MRP4 and MRP5 (251). These three MRPs were identified by mining of EST databases. Discovery of an additional member of the family was facilitated by sequencing of chromosome 16, which revealed an MRP-related gene, MRP6, the 3′-end of which is located within 9 kb of the 3′-end of MRP1 (251). Despite the likely origin of MRP1 and MRP6 by gene duplication and the possibility for gene homogenization, as is thought to have occurred with some mammalian isoforms of P-gp, MRP6/ABCC6 is less similar to MRP1 than either MRP2 or MRP3. One of the most intriguing discoveries involving MRP6 is the striking association between mutations in the gene and the rare connective tissue disorder pseudoxanthoma elasticum (PXE) (36, 41, 416). Why defects in MRP6 result in the disease is presently unknown.

D. MRPs 7–9 (ABCC10–12) and MRP10 (ABCC13)

In the last 2–3 years, three, possibly four, additional MRP-related genes have been identified (34, 191, 493), bringing the total number of potential family members to 10. MRPs 1–8 have all been shown to encode functional ATP-dependent transporters and their combined substrate profiles encompass a wide array of endo- and xenobiotics and/or their conjugated metabolites. Whether MRP9 encodes a functional protein is not yet known, but there has been a report of high levels of a truncated MRP9 mRNA in breast cancer (35), and in the mouse, Mrp9 mRNA has been detected at high levels in seminiferous tubules (465). The most recently identified “MRP,” MRP10? (ABCC13) (546), represents an interesting example of a gene that has degenerated to varying extents among mammals (12). In all species examined, with the possible exception of the rhesus monkey, the ABCC13 gene appears incapable of encoding a functional transporter (see below).


The ”C“ branch is one of the largest of the 7 branches of ABC superfamily, and in humans, it is composed of 13 proteins. In addition to 10 MRPs, the C branch proteins include CFTR and the sulfonylurea receptors SUR1 and SUR2A/B mentioned previously. Thus MRPs 1–6 and MRPs 7–10 have been designated ABCC1–6 and ABCC10–13, respectively, while CFTR, SUR1, and SUR2A/2B are designated ABCC7, ABCC8, and ABCC9, respectively (94).

A. Long and Short ABCC Proteins

Based on the predicted topology of their membrane-spanning regions, the ABCC proteins (with the exception of MRP9) fall into two groups. CFTR (ABC7), MRPs 4, 5, 8, and ABCC13 have a “typical” ABC transporter structure with two MSDs (MSD1 and MSD2), each of which is predicted to contains six TM helices (12, 18, 34, 177, 191, 227, 493, 500), while the remaining MRPs and the SURs (ABCC8 and ABCC9) have an additional NH2-terminal region comprised of ∼200 amino acids (Fig. 1). The NH2-terminal extensions of the long ABCC proteins are relatively poorly conserved. However, they are all hydrophobic and predicted to contain from four to six transmembrane helices (MSD0). The longer, five-domain MRPs are, for the most part, relatively closely related to MRP1 (Fig. 3A) (amino acid identity: 45, 56, and 44% for MRPs 2, 3, and 6, respectively). In contrast, those lacking an NH2-terminal extension share almost equivalent amino acid identity with CFTR (34, 250, 493). The exceptions to this generalization are MRP7 and ABCC13. MRP7 contains an NH2-terminal extension, but its overall identity with MRP1 is only slightly higher (31%) than its identity with the SURs (28 and 26% with SUR1 and SUR2, respectively) (191). ABCC13, on the other hand, lacks an extended NH2-terminal region but has a higher amino acid identity with the core region of MRP1 than MRP6 (546) (Fig. 3, A and C). As indicated above, the ABCC13 gene appears to have undergone a process of gradual pseudogenization in mammals, and presently, the only species found to contain an ABCC13 gene potentially capable of encoding a full-length protein is the rhesus macaque (12). Thus it appears likely that one of the earliest events in this degenerative process may have been deletion of exons encoding the NH2-terminal region of a progenitor relatively closely related to MRP1.

FIG. 3.

Dendrograms of the structural similarities of the full-length ABCC proteins, their nucleotide binding domains (NBDs) and membrane spanning domains (MSDs) of the long ABCC proteins. Dendrograms are based on ClustalW alignments and were generated using Tree View (65). All three panels are drawn to the same scale. A: dendrogram of the human ABCC proteins based on their entire sequence. B: dendrogram of the individual NBDs of the human ABCC proteins together with E. coli HlyB and the NBDs of human MDR1/ABCB1. C: dendrogram of MSD0s of the “long” ABCC proteins.

Comparison of the amino acid sequences of the long and short ABCC proteins reveals that the NH2-termini of CFTR and the shorter MRPs align with a relatively conserved sequence that appears to define the COOH-terminal boundary of the additional regions present in MRPs 1, 2, 3, 6, and 7 and the SURs. In the MRP1 gene, this sequence is found at the 5′-end of exon 6, strongly suggesting that the first five exons were acquired as the result of a gene fusion event (147). Because putative orthologs of MRP1 exist in plants and yeast, this event must be quite ancient. To date, six ABCC-like proteins from Streptomyces cerevisiae have been identified that have NH2-terminal extensions: Ycf1, Bpt1, Ybt1, YKR103w/YKR104w, YHL035c, and Yor1p (86, 216, 292, 331, 375, 389, 528). Like the human proteins, these domains are poorly conserved, but with the exception of Yor1p, they are all predicted to contain several TM helices. The function of MSD0 in the “long” MRPs remains poorly defined (see sect. ivB). However, in SUR1, the comparable domain is involved in an interaction with the potassium channel Kir6.2 required for the trafficking of SUR1 to the plasma membrane and for the gating function of the KATP channel (16, 54, 378).

B. ABCC Proteins Are Characterized by an Atypical NH2-Proximal NBD

Regardless of whether the MRPs have two or three MSDs, they all share certain highly conserved features in their NBDs, particularly in NBD1, that are hallmarks of the entire family. These features are also conserved in CFTR and to a lesser extent the SURs and provide the most compelling evidence for a common ancestor of all C branch members (Figs. 3B and 4) (94, 95). Like other ABC proteins, the NBDs of the C branch proteins contain Walker A and B motifs, as well as a copy of the C sequence that is the signature of ABC NBDs (170). However, in the ABCC proteins, these elements deviate in some cases from the form commonly found in other ABC proteins (72). For example, the amino acid following the Walker B motif in most ABC proteins is glutamate, and this residue is critical for cleavage of the β-γ phosphodiester bond of ATP (345, 472, 515). Although this residue is present at the appropriate location in NBD2 of the ABCC proteins, it is not present in NBD1 (72). In MRP1 and other ABCC proteins, the residue is an aspartate, with the exception of CFTR which has a Ser residue at this location (Fig. 4). Despite the conservative nature of the substitution in NBD1 of MRP1, the lack of glutamate has a profound effect on the ATP binding and hydrolysis characteristics of the NBD and as a consequence, on the catalytic cycle of MRP1 and the other ABCC proteins (386) (see sect. ix). Similarly, the ABC signature sequence in NBD2 of the ABCC proteins differs from that typically found in other members of the superfamily (72), but the functional consequences of this deviation are less well defined (387, 490).

FIG. 4.

Sequence alignment of the NBDs of ABCC proteins and selected more distantly related ABC transporters. Multiple sequence alignment of the NBDs of MRP1 with those of the protein’s presumed yeast ortholog, YCF1, human cystic fibrosis transmembrane conductance regulator (CFTR), the human antigen processing protein TAP1, and the bacterial transporters HlyB, MsBA, and MJ0796. The alignment was performed using ClustalW (65). The figure was colored according to the ClustalW color scheme and formatted using Jalview (67).

In addition to the variations in conserved motifs described above, alignment of the sequences of the ABCC NBDs reveals additional structural features that distinguish these proteins from most other ABC transporters (Fig. 4). The NBDs of most prokaryotic ABC transporters are identical (171). In many eukaryotic ABC proteins, the NBDs are structurally very similar (59), and in P-gp/MDR1, the positions of the NBDs in P-gp/MDR1 can be exchanged with little or no effect on function (26, 199). In contrast, the NH2-proximal NBDs of ABCC proteins are quite divergent from their COOH-proximal NBDs, which have a more typical ABC NBD structure (72). The most obvious distinguishing feature of the NH2-proximal NBDs of the MRPs and CFTR is a decrease in the spacing between the Walker A and ABC signature motifs relative to their COOH-terminal NBDs and the NBDs of proteins such as P-gp. The decrease in spacing is attributable to an apparent deletion at the same location that has eliminated 13 amino acids, the only exception being MRP7/ABCC10 which lacks just 10 amino acids (191). As might be expected from these structural differences, the two NBDs play distinct functional roles in the catalytic cycle of the ABCC proteins (54, 136, 194, 195, 386) (see sect. ix).


A. Experimental Evidence for 3 MSDs, 17 Transmembrane Helices, and Extracellular NH2 Termini

Some, but not all, computer-assisted hydropathy algorithms predict that the NH2-terminal extensions of the MRP1-like MRPs and the SURs contain five TM helices and that the NH2 termini of the proteins are extracellular (18, 74, 177, 398, 500). Thus these proteins are predicted to have 3 MSDs (MSD0, MSD1, and MSD2) and a total of 17 TM helices arranged in a 5 (MSD0) +6 (MSD1) +6 (MSD2) configuration. Considerable experimental evidence supports the predicted topology of MRP1 (Fig. 5), as well as the extracellular locations of the NH2 termini of both MRP1 and MRP2 (18, 177, 178, 180, 227, 228, 234, 243, 245, 350). The first evidence that the NH2 terminus of MRP1 was extracellular came from mutational studies that revealed the presence of two functional N-linked glycosylation sites very close to the start of the protein at Asn19 and Asn23 (177). The fact that a third potential site at Asn71 was not used and presumed to be intracellular provided additional support for the existence of five TM helices in MSD0. The extracellular location of the NH2 termini of MRP1 and MRP2 was subsequently confirmed by epitope insertion studies and the use of an antibody to the first 25 amino acids of rat MRP2, respectively (176, 178, 180, 243). The presence of a third N-linked glycosylation site at Asn1006 in MRP1 confirms that the loop connecting predicted TM12 and TM13 is extracellular (177). Subsequent epitope mapping studies with the MRP1 monoclonal antibodies (MAbs) MRPr1, QCRL-1, MRPm5, and MRPm6, determined that amino acids residues 238–247, 918–924, 1063–1072, and 1511–1520 were intracellular (176, 178, 180, 243). The MRPr1 epitope is located in CL3, the intracellular loop connecting TM helices 5 and 6, while MAb QCRL-1 detects a region beginning at 124 residues downstream of the NBD1 Walker B element in the linker region preceding MSD2. The epitope for MAb MRPm5 is localized to CL6, which connects TMs 13 and 14, while MAb MRPm6 recognizes the COOH-terminal region.

FIG. 5.

Predicted topology of MRP1 showing the locations of mutations shown to affect overall activity, substrate specificity, and protein processing. Molecular dynamic modeling of MRP1 MSD1 and MSD2 was used to predict the positions of helical elements (shown as spirals of amino acids) and their locations relative to a lipid bilayer of 40 Å (52). Helical elements in MSD0 were predicted using HMMTOP. Mutations shown to affect the overall activity of MRP1 are colored yellow; those affecting only certain substrates are shown in green. The conserved Walker A (red), signature C (orange), and Walker B (yellow) are shown in colored boxes. Glycosylation sites at Asn-19, -23, and -1006 are indicated (Y).

Another approach used to determine the topology of MRP1 has involved expressing mutant proteins containing inserted hemagglutinin (HA) epitopes in predicted extracellular and cytosolic regions (227, 228). Immunocytochemical detection of HA peptide inserts in intact and permeabilized cells confirmed the orientation of the NH2 terminus CL3, the region linking NBD1 to MSD2 and the position of the extracellular loop connecting TM12–13 (Fig. 5). In addition, HA tags at position 163 and 574 were detected outside the cell, while the region containing amino acid 653 was located on the cytosolic face of the membrane. Together, these results support the five and six TM arrangement of MSD0 and MSD1, respectively.

Some topology algorithms predict that MSD2, most notably in MRP2, contains four rather than six TM helices (179, 247). No experimental data supporting one topology or the other are presently available for MRP2, but epitope insertion studies of this domain in MRP1 favor the presence of six rather than four TM helices (228). However, the evidence rests on detection of an epitope inserted at amino acid 1222, which only became detectable extracellularly when two copies of the HA epitope were inserted in tandem. Whether this region can adopt more than one topology, as demonstrated recently for certain TM helices of voltage-dependent ion channels (169), remains an intriguing possibility.

B. What Are the Functions of MSD0 and the Cytoplasmic MSD0-MSD1 Linker?

As the defining structural characteristic of the longer MRPs and the SURs, the function of MSD0 in these proteins has been the subject of considerable interest and speculation. As mentioned previously, only MSD0 of SUR1 has been shown to fulfill a well-defined functional role. Interaction between MSD0 of SUR1 and the Kir6.2 potassium channel masks an endoplasmic reticulum (ER) retention signal and, in so doing, allows the protein complex to exit the ER (16, 54). This mechanism serves to ensure that the stoichiometry of SUR1 and Kir6.2 in the plasma membrane is 1:1. Once in the plasma membrane, MSD0-mediated interactions between the two proteins are required for nucleotide-dependent gating of the channel. The NH2-terminal MSDs of MRP2 and the yeast MRP1 ortholog Ycf1 have been shown to be required for apical membrane and vacuolar localization, respectively. At present, no additional function has been ascribed to them, and it is unclear whether they are required simply for correct folding of the protein or contain specific targeting signals (120, 330). Until very recently, no function had been ascribed to MSD0 of MRP1.

Initial studies suggested that MSD0 was necessary for the function of MRP1 (135). This suggestion stemmed from the demonstration that removal of part of MSD0, or the whole MSD0 and various segments of CL3, resulted in the loss of leukotriene C4 (LTC4) transport, a high-affinity physiological substrate of MRP1 (see sect. viA). Furthermore, coexpression of MRP1 lacking MSD0 and most of CL3 with a polypeptide corresponding to the missing region restored activity indicating that the NH2-terminal fragment containing MSD0 formed stable interactions with the core of the protein. However, the loss of activity was subsequently attributed to partial deletion of CL3, rather than the elimination of MSD0 (20). Thus MRP1 lacking the first 203 amino acids MRP1204–1531, which removes MSD0 but leaves most of CL3 intact, traffics to the plasma membrane and transports at least some substrates with little or no loss in efficiency (20, 394, 409, 529). Removal of as few as six additional amino acids (MRP1210–1531) from CL3 causes retention in the ER of mammalian cells. When expressed in insect cells, the truncated protein remains able to traffic to the plasma membrane, but its LTC4 transport activity is severely diminished (529). The regions of CL3 necessary for trafficking and activity of MRP1 have now been defined as lying between Cys208 and Lys270. Although the primary sequence of this region of CL3 is not particularly conserved among ABCC proteins, its predicted secondary structure is, even among more distantly related orthologs and homologs (Fig. 6). When this 60-amino acid region of MRP1 is expressed alone, it associates strongly with membranes and can at least partially rescue trafficking of an NH2-terminally truncated protein lacking MSD0 and most of CL3 (21, 529).

FIG. 6.

Alignment of the NH2-terminal and cytoplasmic loop 3 (CL3) regions of human ABCC proteins and probable MRP1 orthologs from Drosophila, yeast, and nematode. The alignment was generated using ClustalW, and two secondary-structure algorithms were used to predict α-helical elements. Regions shown in red were predicted by PSIPRED, and those in bold face by PhD (217). The CFTR NH2-terminal α-helix boxed in green was determined by NMR (81). Disease-causing CFTR mutations are indicated with a blue circle (source: CFTR mutation database,

Although there is no discernible primary structure homology, the NH2-terminal region of the recently crystallized bacterial vitamin B12 transporter subunit BtuC contains a relatively long α-helix (residues 2–32) that appears to be laterally associated with, or embedded in, the inner leaflet of the lipid bilayer (297). The essential segment of CL3 in MRP1 contains two predicted amphipathic α-helices that have counterparts in all ABCC proteins, even those that do not contain the third NH2-terminal MSD (147, 529). In CFTR, 19 disease-associated mutations have been identified in this region of the protein, underlining its functional importance (CFTR mutation database, One of these mutations involves a highly conserved Pro residue corresponding to Pro209 in MRP1 that defines the NH2-terminal boundary of the essential region of CL3. Nevertheless, mutation of Pro209 either alone or in combination with two nearby Pro residues had little or no effect on expression of MRP1, or its transport activity (210). However, as described below, the mutational studies were carried out in full-length MRP1, and it is possible that the presence of MSD0 may attenuate the effects of these mutations on the trafficking and activity of the protein (531).

A probable explanation for the apparent lack of a functional requirement for MSD0 of MRP1 has been provided by the recent demonstration that the protein contains at least partially redundant trafficking signals (531). Thus MRP1 lacking either MSD0 or its COOH-terminal 30 amino acids remains competent to traffic to the basolateral membrane (530, 531). However, in the absence of both regions, the protein cannot exit the ER. Furthermore, mutations of conserved dileucine and phenylalanine residues in the COOH-terminal regions of other ABCC proteins, such as SUR1, that cause ER retention (462), do not impede trafficking of full-length MRP1 (530), but do result in ER retention in the absence of MSD0 (531). Coexpression of MSD0 with these mutated proteins rescues trafficking, indicating that the NH2-terminal domain must interact with the core of the protein in the ER. Detailed comparative analyses of the subcellular distribution and internalization of full-length and MSD0-less MRP1 have also revealed that MSD0 promotes retention in, or recycling to, the plasma membrane (531). Thus ∼50% of MSD0-less MRP1 can be detected in recycling endosomes compared with ∼10% of the full-length protein. In general, the observations are compatible with evolution of the MRP1-like ABCC proteins by fusion of a four-domain ancestor with another integral membrane protein capable of trafficking independently to the plasma membrane (147), followed by the subsequent loss to varying extents of duplicated and redundant processing and trafficking signals.



The tissue distribution and subcellular targeting of the MRPs is quite variable. MRP1 is widely expressed, with high levels reported in the lung, testis, kidney, skeletal and cardiac muscles, and the placenta (72, 129, 479). Notably, MRP1 is barely detectable in adult human liver, but in proliferating hepatocytes and liver cancer cell lines, such as HepG2, expression is considerably higher (421). Although MRP1 expression is widespread in the body, it is nevertheless found primarily in specific cell types, including bronchial epithelial cells and hyperplastic type II pneumocytes in the lung; proliferating Paneth cells in the colon; Leydig and Sertoli cells in the testis; placental syncytiotrophoblasts and epithelial cells of the endoplacental yolk sac; and mast cells, eosinophils, helper T cells, and erythrocytes in the circulatory system (15, 46, 287, 304, 352, 380, 388, 480, 481, 498, 535, 539). MRP1 is also expressed in a cell specific manner in the brain, as well as the blood-brain barrier and the choroid plexus of the blood cerebrospinal fluid barrier (99, 339, 536).

MRP1 typically localizes predominantly to the plasma membrane and traffics selectively to the basolateral component in polarized cells (116, 175, 287, 421, 539). This contrasts with the apical membrane localization of other efflux pumps such as P-gp/MDR1 (495), BCRP (318), and MRP2 (224). However, in some cell types, such as placental syncytiotrophoblasts and brain microvessel endothelial cells, MRP1 localizes to apical membranes (479, 571). What determines this cell-type specific targeting to different membrane compartments is not known. The proportion of MRP1 targeted to cell surface membranes in cultured cells also varies. For example, 80–90% of MRP1 localizes to the plasma membrane in transfected HeLa and HEK293 cells (8, 146, 175, 531) compared with only ∼50% in multidrug resistant H69AR cells (8, 72). Significant intracellular accumulation of MRP1, possibly in the Golgi, has also been reported in the drug-selected human small cell lung carcinoma cell line GLC4-Adr (72, 511). In GLC4-Adr and H69AR cells, intracellular MRP1 colocalized with vesicles that accumulated fluorescent substrates (96, 511). Similarly, exposure of cells transiently transfected with MRP1 fused to green fluorescent protein (GFP) to the fluorescent anthracycline doxorubicin resulted in accumulation of drug in vesicles bounded by membranes containing the fusion protein (401). Consequently, MRP1 is believed to be functional on the plasma membrane and in intracellular compartments.


MRP2 has a more restricted tissue distribution than MRP1. The protein is expressed in the liver, kidney, small intestine, colon, gallbladder, placenta, and lung (131, 224, 247, 287, 445). MRP2 is the only MRP that is consistently found in apical membranes. Thus, in the liver, it is present in canalicular membranes (224) and on apical membranes in renal proximal tubules, placental syncytiotrophoblasts, and intestinal epithelium (118, 131, 229, 384, 445). Its highest expression in the gut is in the villi of the proximal jejunum (347). Why MRP2 traffics exclusively to the apical membrane is poorly understood, and there are conflicting reports in the literature. Efficient trafficking of MRP2 clearly requires the presence of MSD0, but it has not been established that this region of the protein contains functional apical targeting signals, or whether it is required for correct folding of the remainder of the protein (120). For example, MSD0 of MRP2 when fused to the core region of MRP1 does not alter the basolateral localization of the protein (531). Unlike MRP1, MRP2 contains a PDZ-domain located at its COOH terminus (237). It has been proposed that this domain may interact with scaffolding proteins such as radixin that could link MRP2 to the F-actin cytoskeleton in a manner analogous to that proposed for CFTR (234, 244, 348). In support of this suggestion, the hepatocanalicular localization of MRP2 is disrupted in radixin knock-out mice (234). However, it has been reported that the PDZ motif of MRP2 can be deleted without loss of apical localization (358). Other regions of the protein may also contain signals that are involved in the apical targeting of MRP2 (164, 249, 358), but attempts to locate them by making various MRP1/MRP2 hybrid proteins have generated inconsistent results.


MRP3 is expressed in the adrenal gland, pancreas, gut, gall bladder, and placenta, with lower levels being found in liver, kidney, and prostate (30, 236, 246, 250, 447, 479). In the liver, MRP3 is present in basolateral membranes of hepatocytes close to bile ducts, as well as in cholangiocytes lining the ducts themselves (252, 474). The levels of MRP3 increase under any conditions that result in cholestasis (107, 181, 246) while the presence of MRP2 in the canalicular membrane markedly decreases (497). Thus there appears to be a reciprocal relationship between the two transporters that is presumed to protect the liver from accumulation of potentially toxic bile constituents (107, 474, 497). In the intestine, MRP3 is present in enterocytes in the ileum and the colon, and in the kidney it is found in the distal tubules (428).

D. MRPs 4 and 5

MRP4 mRNA is expressed at low to moderate levels in ovary, testis, adrenals, lung, and intestine and at somewhat higher levels in the prostate (267). Like MRP1, expression of MRP4 in normal liver is very low. Also as observed with MRP1, in some cell types MRP4 is targeted to apical rather than basolateral membranes (43). For example, apical localization of MRP4 occurs in the kidney proximal tubule and endothelial cells of the brain capillaries (270, 508), but the protein is found basolaterally in prostate tubuloacinar cells, choroid plexus, and HepG2 cells (268, 270, 419). MRP5 is more widely expressed than MRP4, with the highest levels of MRP5 mRNA being detected in skeletal muscle and cardiac and cardiovascular myocytes (30, 43, 93, 186, 250, 337, 571). In the brain, MRP5 colocalizes with MRP1 and MRP4 on the luminal (apical) side of capillary endothelial cells and is also present in astrocytes and pyramidal neurons (186, 360). Despite the apical location of MRP5 in microcapillary endothelial cells, the protein is found on the basolateral membrane of polarized epithelial cells (537, 571).


The tissue distribution of MRP6 is of particular interest because of the association between mutations in this protein and the degenerative connective tissue disease PXE (36, 277, 416). Individuals with this disease develop calcified elastic fibers and abnormal collagen fibers in elastic tissues of the skin, retina, and arteries. The mechanism by which mutations of MRP6 cause this degeneration is not known. Earlier reports indicated that expression of MRP6 may be restricted to kidney and liver, prompting the suggestion that the protein might be involved in elimination of a metabolite responsible for the degeneration observed in affected tissues (31, 251, 504). This hypothesis remains a possibility, but more recent in situ hybridization and immunohistochemical studies have detected murine Mrp6 mRNA and protein in epithelial cells from many tissues (27). In addition to parenchymal cells in the liver and the proximal tubules in the kidney, relatively high levels of the protein and/or mRNA were found in keratinocytes of the skin, tracheal and bronchial epithelium, intestinal mucosa and corneal epithelium, as well as endothelial and smooth muscle cells of the cardiovascular system.

F. MRPs 7–9 and ABCC13

MRP7 transcripts have been detected in many tissues by RT-PCR, but expression levels appear to be relatively low (19, 223). In the mouse, the highest levels of mRNA were found in heart, skeletal muscle, and kidney (223). These studies also revealed two splice variants of Mrp7 mRNA, designated Mrp7A and Mrp7B, the latter of which contains two, short additional 5′-exons capable of encoding 41 amino acids. The relative levels of the two variants differ among tissues, but the possible functional implications of the observation are not known. RT-PCR indicates that human MRP8 mRNA is widely expressed, with the exception of kidney, spleen, and colon (34, 493, 545). Finally, there is some uncertainty surrounding the tissue distribution of MRP9, although transcripts have been consistently detected in the testis and in breast (35, 93, 493). However, even the longest MRP9 transcripts detected (∼4.5 kb) do not encode a full-length protein, as predicted from the exon structure of the gene (35, 493). In contrast, a full-length transcript of the orthologous mouse gene has been identified, and expression appears to be restricted to the testis (465). Furthermore, the protein encoded by murine Mrp9 mRNA displays high homology with the predicted open reading frame of a full-length transcript of the human gene. Relatively high levels of expression of the apparently degenerate human ABCC13 have been detected in colon, bone marrow, salivary gland, and fetal liver, but the major transcript is only ∼1 kb (12, 212). In the macaque, where the gene may be functionally intact, a mRNA of ∼5.0 kb can be found with high expression in colon and small intestine (12). Currently, the tissue and subcellular distributions of ABCC10–13 proteins are not known.


Defining the substrate specificities of the MRPs has been, and remains, an area of considerable activity. Many studies have focused on xenobiotic substrates because of the potential role of the MRPs in clinical drug resistance and in protection against a wide range of environmental toxicants (42, 78, 96, 161, 247, 287). Others have sought to identify potential endogenous substrates to gain insight into possible physiological functions of the proteins. Identification of exogenous substrates of the MRPs has frequently been based on an assessment of their ability to confer resistance to candidate cytotoxic drugs and xenobiotics. In some cases, such studies have also incorporated cellular accumulation and efflux assays of radiolabeled or fluorescent xeno- and endobiotics. Identification of potential physiological substrates has relied heavily on the use of inside-out membrane vesicles from transfected cells to directly measure ATP-dependent transport of candidate compounds, or their ability to inhibit transport of an established substrate (22, 183, 213, 273, 298, 299, 300, 301). For several of the MRPs, information obtained from in vitro substrate specificity studies has been augmented by examining the consequences of knocking-out the gene in mice or in cell lines (311, 312, 405, 534). In the case of MRP2, much of the work defining substrate specificity has been done by examining biliary transport defects in MRP2-deficient GY/TR or EHBR rats (140).

A. MRP1 and MRP2

The MDR phenotype of the H69AR cells from which MRP1 was cloned was well characterized before identification of the protein (68, 69, 70, 72, 342). Consequently, some of MRP1’s potential drug substrates were readily predictable and confirmed using transfected cells (73, 146, 482, 553). These included primarily natural product cytotoxic agents, such as anthracyclines, epipodophyllotoxins, and Vinca alkaloids, as well as certain heavy metal oxyanions. MRP1-transfected cells also displayed decreased drug accumulation and increased drug efflux, strongly suggesting that the drugs were directly transported by MRP1 (73, 553, 554). However, initial attempts to confirm that MRP1 was a primary, active transporter of these compounds using vesicle transport assays were unsuccessful. The first MRP1 substrate to be identified by in vitro transport studies was the proinflammatory mediator LTC4, which remains the most well-characterized, confirmed physiological substrate of the protein (272, 303, 349, 534). The discovery that LTC4 was a high-affinity substrate arose from investigation of the possibility that MRP1 might be a broad specificity transporter known as the MOAT that had been identified functionally in hepatocanalicular membranes and membranes from mast cells (273, 444). Although nothing was known of the primary structure of MOAT at the time, a great deal was known about its substrate specificity. As described above, naturally occurring MOAT mutations in rats and sheep with biliary transport defects similar to human Dubin-Johnson syndrome (110, 430, 476) enabled extensive profiling of MOAT substrates by comparative analysis of the bile constituents of mutant and wild-type animals. The profiles indicated that MOAT transported a diverse array of conjugated organic anions, as well as possibly free glutathione (GSH) and heavy metal GSH complexes. Fortunately, the substrate specificities of MRP1 and MOAT, now known to be MRP2, overlap extensively. Thus, although the original premise was a case of mistaken identity, it proved to be of tremendous assistance in identifying substrates for MRP1.

In vitro studies using membrane vesicles have confirmed that MRP1 and MRP2 can transport a vast array of organic conjugates (97, 231, 287). The existence of transporters capable of effluxing GSH conjugates, collectively termed GS-X pumps, was postulated (205) well before the MRPs were identified. However, the substrate specificities of MRP1 and MRP2 are broader and encompass not only GSH conjugates, but also many glucuronide and sulfate conjugates of both xeno- and endobiotics (Fig. 2). Consequently, MRP1 and MRP2 are important contributors to cellular extrusion and elimination of the relatively hydrophilic products of phase II conjugation reactions that are frequently involved in detoxification of hydrophobic xenobiotics, such as the potent fungal carcinogen aflatoxin B1 (300). Both MRP1 and MRP2 transport conjugated leukotrienes such as LTC4 and its metabolites LTD4 and LTE4 (85, 272, 299). Steroid and bile salt conjugates are also among the physiological substrates of both proteins (66, 85, 212, 214, 298, 395, 558). Thus the jaundice observed in Dubin-Johnson patients is attributable to a loss of MRP2-mediated canalicular efflux of bilirubin glucuronides (222, 224, 230, 382), and possibly the inability to transport cholestatic estrogen conjugates such as E217βG, another well-characterized substrate of both proteins (85, 213, 298).

B. Transport and the Role of GSH

The lack of transport of unmodified drugs, combined with the demonstrated ability of MRP1 to transport organic anion conjugates, suggested that the protein might confer resistance by effluxing drug conjugates (213, 299). MRP1 is indeed capable of transporting the glucuronide conjugate of etoposide and a GSH conjugate of doxorubicin (213, 393, 433). However, while the transport of drug or xenobiotic conjugates may contribute the resistance profiles conferred by MRP1 and MRP2 in some situations, it is now recognized that other mechanisms are of more general relevance. The transport of a number of unmodified drugs to which both MRP1 and MRP2 confer resistance, such as vincristine and doxorubicin, is dependent on or stimulated by GSH (85, 299, 301, 413). The first indication of a requirement for GSH came from the demonstration that GSH depletion decreased drug efflux by cells overexpressing MRP1 (451, 510, 554). Vesicle transport assays confirmed that GSH enhanced the ability of unmodified drugs to inhibit transport of the known MRP1 substrates, E217βG and LTC4, and that it was possible to detect ATP-dependent transport of vincristine when GSH was included in the transport assays (299, 301, 302). Similarly, GSH stimulates transport of unmodified drugs by MRP2, including etoposide and vinblastine (119, 507). Although it was initially presumed that GSH stimulation was limited to unmodified substrates, this is not the case and GSH stimulated transport of physiologically relevant sulfated steroids, such as estrone-3-sulfate and DHEAS, as well as of the tobacco smoke carcinogen, NNAL-O-glucuronide, and etoposide glucuronide by MRP1 has now been described (281, 395, 433, 558). Interestingly, MRP2 is also able to transport NNAL-O-glucuronide, but transport is not stimulated by GSH and indeed is inhibited by GSH (281). This probably reflects the fact that NNAL-O-glucuronide establishes different interactions with each protein despite their considerable structural and functional conservation (see sect. ix). On the other hand, none of the three known nicotine glucuronide metabolites appears to be a substrate of MRP1 or MRP2 (288). MRP2 also does not transport estrone-3-sulfate even in the presence of GSH (440) but does transport other sulfate conjugates such as acetaminophen sulfate (556).

The mechanism by which GSH stimulates transport is complex and not fully understood. GSH itself is a relatively poor substrate for both MRP1 and MRP2 with a Km of >1 mM (280, 301, 384, 434). In contrast, GSSG is transported by MRP1 with a much higher Vmax and a Km of ∼100 μM (274). However, in the presence of some GSH-dependent substrates or modulators of MRP1, the Km for GSH decreases to approximately the same range as GSSG (284, 299, 301). Reciprocal stimulation of transport of GSH and second substrate (e.g., vincristine and aflatoxin B1) has also been observed with both MRP1 and MRP2, indicative of positive cooperativity between binding of the two substrates (119, 300, 301, 326). The estimated stoichiometry of transport of GSH and a drug such as vincristine is compatible with the possibility that transport of both substrates is coupled, but this has not been demonstrated formally. In other cases, such as estrone-3-sulfate and NNAL-glucuronide, no stimulation of GSH transport can be detected (281, 395). In addition, a number of examples exist of compounds that strongly stimulate GSH transport, also by decreasing the Km for GSH by ∼10-fold, without themselves being transported. These include drugs such as verapamil and bioflavonoids, such as apigenin (282, 284, 302).

The GSH-dependent stimulation of transport of organic substrates by MRP1 is not dependent on the reducing potential of the peptide, since short-chain S-alkyl tripeptides, such as S-methyl-GSH, and non-sulfur-containing tripeptides such as ophthalmic acid are also effective (281, 283, 301, 395). However, the effectiveness of S-alkyl-GSH analogs tends to decrease with increasing alkyl chain length, possibly because the larger alkyl derivatives begin to compete for transport of the second substrate. GSH derivatives in which the cysteine residue has been replaced by a number of different amino acids are also able to stimulate transport (283). The effectiveness of these derivatives increases with the hydrophobicity of the amino acid side chain, suggesting that the GSH cysteine may be in a relatively hydrophobic region of the protein’s binding pocket. The transport of heavy metal oxyanions to which MRP1 and MRP2 confer resistance may also be GSH dependent, and it has been suggested that it occurs via a cotransport mechanism analogous to that proposed for organic substrates (435). However, the major forms of arsenic identified by analysis of bile from normal Wistar rats are arsenic triglutathione and methyl arsenic diglutathione (221, 383). These conjugates are not found in bile from mutant TR Wistar rats lacking MRP2, strongly suggesting that they are substrates of the protein (149). Very recent in vitro transport studies of MRP1 have confirmed that arsenic triglutathione is a high-affinity substrate (Km <1 μM) and that it can be formed enzymatically by GSTP1–1 (286). Whether other heavy metals are transported as conjugates or via some type of cotransport mechanism with GSH is not yet known.

C. Physiological Roles of MRP1 and MRP2

The substrate diversity and widespread expression of MRP1 suggest numerous possible physiological functions, many of which have yet to be confirmed. However, some anticipated, as well as unexpected, physiological functions of the protein have been revealed from studies of Mrp1−/− mice (139, 312, 313, 420, 455, 513, 534536). As mentioned previously, LTC4 is the highest affinity substrate known for MRP1. LTA4 is the parental compound of both conjugated and nonconjugated leukotrienes and is formed from arachidonic acid by 5-lipoxygenase (436). LTA4 is converted to LTB4, an important chemotaxin in mediating inflammatory responses, by LTA4 hydrolase (356), and to LTC4 by LTC4 synthase, which conjugates the parental leukotriene with GSH at the C6 position (549). LTC4 is produced in mast cells, basophils, eosinophils, dendritic cells, macrophages, neutrophils, platelets, kidney, and brain (336, 456). In addition, LTC4 can be synthesized in liver microsomes and endothelial cells by LTC4 synthase (452, 464). Once effluxed from the cell, LTC4 is rapidly converted to LTD4 and LTE4 by γ-glutamyltranspeptidase and dipeptidase, respectively (266). Collectively, the cysteinyl leukotrienes are referred to as the slow-reacting substance of anaphylaxis. LTD4 and LTE4 bind to G protein-coupled CysLT1 and CysLT2 receptors on target cells and are associated among other things with mediating asthma pathologies (165). Consistent with a physiological role for Mrp1 as an LTC4 transporter, leukotriene release from eosinophils and mast cells in response to IgE-mediated inflammation is reduced in Mrp1−/− mice (534). The mice also display an impaired immune response to contact sensitization. This defect led to the discovery of LTC4-dependent recruitment of dendritic cells to lymph nodes during responses to inflammation (420). Unexpectedly, the mice are also more resistant to Streptococcus pneumoniae infections (455). This has been attributed to accumulation of intracellular LTC4 in alveolar macrophages, resulting in product inhibition of LTC4 synthase and a consequential diversion of LTA4 to LTB4 production. The increased resistance to infection is likely attributable to the fact that LTB4 is a potent stimulator of phagocytic macrophages (17, 100).

Tissues that normally express relatively high levels of Mrp1, such as the testis, kidney, and oropharyngeal mucosa, are hypersensitive to etoposide in Mrp1−/− mice (534, 535). The presence of MRP1 in blood-tissue barriers such as the choroid plexus (404) also suggests that the transporter contributes to protection of sanctuary sites in the body. Consistent with this suggestion, Mrp1−/− mice display increased passage of MRP1 substrates from the blood to the cerebrospinal fluid (536). Similarly, MRP1 in the placenta may protect the developing fetus from xenobiotics and help prevent the fetal accumulation of endobiotics, such as E217βG (287, 479). Similarly, the efflux of estrone sulfate from Leydig cells by MRP1 may help to protect the testis from the feminizing effects of estrogen (395). MRP1 expression is also high in hyperplastic reactive type II pneumocytes, which proliferate in response to airborne cytotoxins and lung insults (539). Finally, there is evidence that Mrp1 expression is higher in cholestasic rats (521), suggesting that it may participate in protecting the liver from the accumulation of toxic levels of bilirubin conjugates (214, 521).

The ability of MRP1 to transport both GSH and GSSG raises the possibility that the protein contributes to maintenance of the redox state of the cell. Studies using MRP1 overexpressing cell lines and determination of GSH levels in tissues from Mrp−/− mice support this suggestion. The levels of free GSH are decreased in cells overexpressing the protein and increased in tissues from Mrp1−/− mice that normally express high levels of the protein (69, 312, 517, 518). The levels of MRP1 have also been shown to increase two- to threefold after exposure to agents that induce oxidative stress, possibly as a result of activation of transcription by a mechanism involving the antioxidant responsive transcription factor Nrf2 (166). Thus MRP1 may also contribute to GSSG efflux under conditions where GSSG production is increased (2, 96). Consistent with this proposal, Mrp1 protects rat neural astrocytes from exposure to H2O2 (185). Additionally, MRP1 has been implicated in the transport of toxic lipid oxidation products, such as 4-hydroxynonenal-GS, that are formed during periods of oxidative stress (287, 414).

As indicated above, MRP2 is responsible for the hepatobiliary excretion of organic anions into bile, including bilirubin and bile salt conjugates (4, 49, 224, 230, 382). Because of its broad substrate specificity, MRP2 is also expected to be a major contributor to the biliary disposition of many drugs and xenobiotics, as well as their conjugates (4, 105, 140, 157, 161, 231, 281, 440, 458, 487, 492, 543, 544). Because MRP2, like MRP1, transports GSH and GSSG, it has also been suggested that it plays a role in protecting cells from oxidative stress (384).


One of the most remarkable differences between the substrate specificity of MRP3 and those of MRP1 and MRP2 is its lack of transport of GSH and the poor ability of MRP3 to transport GSH conjugates (5, 182, 184, 252, 374, 557, 562, 569). Unlike MRP1 and MRP2, MRP3 shows a marked preference for glucuronidated compounds (66, 182, 184, 374, 557, 562, 569), consistent with its proposed role in protecting the liver from accumulation of bile salts and other potentially toxic conjugated compounds. Moreover, in addition to conjugated bile salts, MRP3 transports monovalent bile salts such as cholate, taurocholate, and glycocholate that are not substrates of either MRP1 or MRP2 (66, 181, 184, 559, 562). This has prompted the suggestion that MRP3 contributes to the enterohepatic circulation of bile salts (107, 181, 252, 474). However, Mrp3−/− mice show no symptoms attributable to such a defect (33, 560). In addition to being responsive to cholestatic conditions, expression of MRP3 may be coordinated with induction of phase I enzymes by xenobiotics that activate a number of different signal transduction pathways (317). For example, the induction of Mrp3 by microsomal enzyme inducers has been associated with increased hepatic efflux of conjugates of common hepatotoxic drugs, such as acetominophen, into blood (469). More recently, Mrp3 has been shown to be almost totally responsible for the basolateral efflux of acetaminophen glucuronide from the liver (319). Studies with knock-out mice have also revealed that inactivation of Mrp3 has a major effect on the efflux of morphine glucuronides from the liver via sinusoidal membranes resulting in a major decrease in circulating levels of morphine-3-glucuronide (319). Overall, the regulation and substrate specificity of Mrp3 suggests that one of its likely physiological functions is to protect the liver from accumulation of a variety of hepatotoxic xeno- and endobiotics (66, 107, 183, 184, 252, 447, 474, 557, 561563, 569).

E. MRPs 4 and 5

Like MRP1, -2 and -3, MRP4 and MRP5 are organic anion transporters (61, 62, 337, 508, 537, 558). However, they differ from the MRP1-like proteins in their ability to transport nucleoside and nucleotide analogs, as well as cyclic nucleotides (61, 93, 215, 392, 407, 508, 532, 537). The ability of MRP4 and MRP5 to transport cyclic nucleotides has prompted speculation that they contribute to modulation of intracellular cAMP and cGMP levels (43, 61, 215, 418). However, some controversy exists concerning their affinities for these nucleotides, and several lines of evidence suggest that they do not have a major influence over intracellular cAMP and cGMP concentrations (261, 532). Nevertheless, the colocalization of MRP5 and phosphodiesterase 5 in smooth muscle cells of the genitourinary tract is intriguing, particularly when coupled with the observation that the phosphodiesterase inhibitors, such as silfenadil and trequinsin, also inhibit MRP5-mediated cGMP efflux (359). This raises the possibility that the protein may contribute to the efficacy of drugs such as sildenafil and trequinsin in raising cGMP levels in genitourinary smooth muscle cells.

It has also been suggested that MRP4 and MRP5 are involved in determining extracellular levels of cyclic nucleotides. MRP4 is expressed at relatively high levels in the kidney proximal tubules where it could contribute to urinary excretion of cAMP and cGMP (508). More recent studies suggest a broader role for the protein in organic anion excretion and that MRP4 may be responsible for the excretion of some organic anions formerly attributed to MRP2. MRP4 transports p-aminohippurate (PAH), which is typically used to assess organic anion transport in renal proximal tubular cells, and does so with considerably higher affinity than MRP2 (471). Furthermore, PAH excretion is normal in mutant rats lacking MRP2, suggesting that MRP4 may be the primary physiological transporter responsible (471). Whether or not GSH is required for transport of cyclic nucleotides and possibly other substrates by MRP4 and MRP5 remains unresolved, although it has been reported that MRP4 cotransports GSH with bile salts (419). Efflux of cAMP by MRP4 was initially reported to be inhibited by depletion of intracellular GSH with agents such as buthionine sulfoximine (BSO), and overexpression of the protein was associated with a decrease in GSH levels (261). Polarized MRP5 transfected cells were also reported to efflux GSH (537). However, subsequent studies using transfected, nonpolarized HEK cells failed to confirm a GSH requirement for cAMP transport by either protein (532).

MRP4 can also transport prostaglandins PGE1 and PGE2, neither of which are transported by MRP1, MRP2, MRP3, and MRP5, with relatively high affinity (408). The transport of PGE1 and PGE2 is also potently inhibited by other prostaglandins and thromboxane B2, as well as several nonsteroidal anti-inflammatory drugs (NSAIDS), suggesting that the release, as well as the synthesis of prostaglandins, may be blocked by at least some of these widely used drugs.

Given the initial characterization of the substrate profile of MRP4, it was unexpected to find that hepatic expression of the protein is upregulated with that of MRP3 during cholestasis (14, 102). Furthermore, the levels of MRP4 are even more profoundly affected than those of MRP3 when expression of the bile salt export pump (BSEP/ABCB11) is severely diminished (454). The decrease in BSEP expression is also accompanied by an increase in both serum and urinary bile acids, suggesting that MRP4 might be involved in the efflux of these compounds in the liver and the kidney. Based on the ability of various bile salts to inhibit the transport of E217βG by MRP4, the protein appears to have a relatively high affinity for sulfated bile salts and steroids (558). Consequently, it is possible that MRP4, together with MRP3, contributes to a shift to urinary excretion of bile salts and other organic anions during conditions when biliary excretion is compromised.


MRP6 has been shown to transport a number of GS-conjugated organic anions in vitro that are also transported by other MRP1-like MRPs, including LTC4, S-(2,4-dinitrophenyl)glutathione, and N-ethylmaleimide S-glutathione (NEM-GS), and the protein is also inhibited by relatively nonspecific anion transport inhibitors, such as probenecid and indomethacin (32, 202). However, glucuronidated substrates of the other transporters, such as E217βG, appear not to be MRP6 substrates (32). In addition, MRP6 transports the cyclic peptide endothelin receptor antagonist BQ-123 but not endothelin-1 itself (32, 316) and has been reported to confer low-level resistance to various epipodophyllotoxins and anthracyclines (32). The majority of PXE-causing defects map to NBD2 of the protein, suggesting that the disease is associated with loss of its transport function rather than altered substrate specificity (277279). The prevalent expression of the protein in the liver and kidney suggests the possibility of a defect in systemic disposition of an as yet unidentified substrate. A potentially interesting association between decreases in sulfated glycosaminoglycans (GAGs) in the urine has been observed in PXE patients and to a lesser extent in healthy carriers of the disease (315). Whether the decrease in urinary GAGs is attributable to loss of MRP6 function and is in some way linked causatively to PXE has not been established.

G. MRPs 7 and 8

Little is known of the specificity of MRP7 for potential endogenous substrates. At present, the protein has been shown to transport E217βG in vitro, but with relatively low affinity compared with MRP1, MRP2, MRP3, and LTC4 but also poorly compared with the other three proteins (63). Among the shorter MRPs, MRP8 appears to have an exceptionally broad substrate specificity. In addition to the ability to transport cyclic nucleotides, as shown with MRP4 and MRP5, MRP8 is also able to transport glutathione, glucuronide, and sulfate-conjugated substrates such as LTC4, E217βG, and estrone-3-sulfate as shown for MRP1 and MRP2. In addition, MRP8 shares with MRP3 the ability to transport monoanionic bile acids, such as glycocholate and taurocholate, and like MRP3 is expressed in liver, raising the possibility that it might contribute to bile acid homeostasis (64).


A. Drug Resistance Profiles

The drug resistance profiles of the MRPs have been described extensively in recent reviews, and only a brief overview is provided here (78, 96, 161, 256, 257, 287).

Historically, human ABC drug efflux pumps, typified by P-gp/MDR1, have been associated with resistance to natural product-type cytotoxic agents (10, 219). Although initial characterization of MRP1 revealed a considerable overlap between the drug resistance profiles of the two proteins, some important distinctions were also observed (96, 161, 256, 287). Notable among these differences was the ability of MRP1 to confer resistance to certain arsenical and antimonial oxyanions and its relatively poor ability to confer resistance to taxanes (73, 482, 553).

The MRP1-like MRPs all share to a greater or lesser extent the ability to confer resistance to natural product-type drugs, presumably reflecting their evolution as a protective mechanism against xenobiotics encountered in the diet and the environment (161, 256, 287). However, studies using transfected cell lines have revealed that some interesting examples of substrate overlap and complementarily exist among these proteins (32, 73, 85, 118, 146, 187, 192, 252, 268, 337, 482, 533, 557, 569). The drug resistance profiles of MRP1 and MRP2 are most similar with respect to natural product type drugs and are relatively broad (231). Both proteins confer resistance to anthracyclines, Vinca alkaloids, and epipodophyllotoxins, and both confer only very low levels of resistance to taxanes (73, 85, 118, 146, 198, 482, 553). Transport of several of the natural product drugs by both MRP1 and MRP2 is also stimulated by GSH (85, 301, 413). A major distinction between MRP1 and MRP2 is the ability of the latter to confer low-level resistance to platinum-based drugs (85, 240). In addition to anticancer drugs, MRP2 but not MRP1 is able to transport HIV protease inhibitors such as saquinivir and indinavir (197).

MRP3 appears to have a narrower drug profile than either MRP1 or MRP2 and confers resistance to epipodophyllotoxins, but not anthracyclines or Vinca alkaloids (252, 557, 569). MRP3 also differs from MRP1 and MRP2 in that its ability to confer resistance to epipodophyllotoxins is not GSH dependent (557). As indicated above, MRP3 is a very poor transporter of GS-conjugates and does not transport GSH, which may explain its inability to confer resistance to other classes of natural product type drugs (252). The drug resistance profiles of MRP6 and MRP7 have been less extensively characterized (32, 192, 316). However, MRP6 has been reported to confer modest levels of resistance to etoposide and teniposide (3- to 4-fold) and low levels of resistance to anthracyclines and cisplatin (32). Like MRP3, MRP6 does not confer resistance to Vinca alkaloids (32). MRP7 appears unique among the MRPs in conferring resistance preferentially to taxanes, such as docetaxel and paclitaxel (192). The protein also confers low levels of resistance to Vinca alkaloids, but not anthracyclines or epipodophyollotoxins (192). Despite the overlapping resistance profiles of the MRP1-like MRPs, selection in natural product drugs to which several of these proteins can confer resistance appears overwhelmingly to result in increased expression of either MRP1 or P-gp/MDR1. Why among the MRPs, MRP1 is preferentially overexpressed is not fully understood. However, the widespread basal expression of MRP1 in numerous cell types and the ability of relatively small changes in levels of the protein to confer detectable resistance to some drugs may be important contributing factors.

As with the MRP1-like MRPs, the shorter MRPs that have been characterized (MRP4, MRP5, and MRP8) share overlapping, but not identical, drug resistance profiles (61, 154, 268, 337, 407, 453, 537). Among these proteins, only MRP4 has been shown to be overexpressed in response to drug selection (453). The overexpression of MRP4 in drug-selected human T-lymphoid CEM cells provided the first example of a human ABC transporter capable of conferring resistance to nucleoside analogs, such as azidothymidine monophosphate (AZT), 9-(2-phosphonylmethoxyethyl)adenine (PMEA), and Lamivudine (453). In addition to these anti-HIV drugs, MRP4 has been shown to confer resistance to the antiviral drug Gancyclovir (1). MRP5 also confers resistance to PMEA, and both proteins increase resistance to the thiopurines thioguanine and 6-mercaptopurine (537). However, neither MRP4 nor MRP5 appears to confer substantial resistance to anticancer nucleoside drugs, such as gemcitabine and cytarabine (407). MRP5 confers resistance to 5-fluorouracil (392), whereas MRP4 does not (407). MRP8 also confers resistance to 5-fluorouracil as well as other fluoropyrimidines (154). Like MRPs 4 and 5, MRP8 also confers resistance to PMEA (154). Thus, while the MRP1-like MRPs are associated with resistance to natural product drugs, the shorter MRPs primarily confer resistance to nucleoside and nucleotide analogs. Nevertheless, members of the two subsets of proteins share the ability to confer resistance to the antimetabolite methotrexate and camptothecin-like topoisomerase I inhibitors, such as irinotecan. MRP1, -2, -3, -4, and -8 all confer resistance to short-term exposure to methotrexate, and all have been shown to transport the drug in its unmodified nonglutamylated form (187, 252, 558). MRP5 has also been shown to confer resistance to methotrexate and to be capable of transporting not only the unmodified drug but also methotrexate di- but not triglutamate (533). In addition, MRP1, MRP2, and MRP4 have been found to increase resistance to irinotecan and its active metabolite SN-38 (193, 370). Studies with Mrp4−/− mice indicate that Mrp4 may protect the brain from accumulation of topotecan (270).

Although most commonly associated with resistance to “traditional” cytotoxic drugs, the MRPs, notably MRP1, may confer resistance to more recently developed cytotoxic peptides, antimetabolites, and immunoconjugates. Thus MRP1 has been shown to confer resistance to the cytotoxic peptide N-acetyl-Leu-Leu-norleucinal (ALLN) and a toxic peptide derivative, which is based on a Thr-His-Thr-Nle-Glu-Gly backbone conjugated to butyl and benzyl groups (4A6) (98). In addition, MRP1 also appears to confer resistance to the naturally occurring bicyclic peptide depsipeptide FK228 that has shown promise in phase II clinical trials of cutaneous T-cell lymphoma. FK228 interferes with several transduction pathways and has most recently been demonstrated to be a potent inhibitor of histone deacetylase (344). Interestingly, FK228 is a prodrug that is activated intracellularly by glutathione-dependent reduction of a disulfide bond. Whether it is the parental compound or its reduced form that may be transported by MRP1 is not presently known (542). MRP1 has also been implicated in resistance to the immunoconjugate gentuzumab-ozogamicin, a promising agent in the treatment of acute myeloid leukemia (523). The conjugate consists of an antibody targeted to CD33 that is conjugated via an acid-hydrolyzable linker to a derivative of the antitumor antibiotic calicheamicin-β1, which is released following internalization and entry into lysosomes (269). Finally, there is some evidence that MRPs’ capability of transporting methotrexate may limit the uptake of recently developed antifolates, such as 10-deazaminopterin (233, 467, 533).

B. Inhibitors and Reversing Agents

Given the broad substrate specificity of the MRPs, the definition of an inhibitor becomes to some extent application dependent, since although there are a growing number of examples of compounds that bind to these proteins without being transported, many substrates also compete reciprocally for transport. Furthermore, in many cases it is not known whether the “inhibitor” is a substrate or not. Experimentally, inhibitors have been used as a means of implicating MRPs in resistance to a variety of drugs, or in the transport of a wide range of xeno- and endobiotics. To date, the inhibitors in general use have been of relatively low affinity and specificity. In some cases, they were originally designed for extracellular targets and, as a consequence, may have relatively low membrane permeability. One example is MK571, which was designed as an LTD4 receptor antagonist. MK571 is an effective inhibitor of MRP1, MRP2, and MRP4, but its use with intact cells requires concentrations in excess of 5 μM (60, 138, 419, 507). Other compounds have been previously characterized as inhibitors of other ABC transporters, such as the sulfonylurea glibenclamide. Glibenclamide is an inhibitor of SUR1/ABCC8, but it also inhibits MRP1 and some other ABC transporters that are not members of the C branch of the superfamily, such as ABCA1 (29, 48, 77, 163, 385). Some compounds that inhibit MRPs, for example, probenecid, are even less specific and are general inhibitors of organic anion transport (22, 145, 517). There has also been considerable interest in the potential of dietary flavonoids such as genistein and quercetin, as well as synthetic flavonoids such as flavopiridol, to inhibit both MRP1 and MRP2 (106, 188, 211, 282, 284, 516). This is both because of their potential as reversing agents and also because of possible drug interactions that may occur as a result of dietary intake of these compounds or their use as alternative therapies involving these compounds.

In addition to inhibitors that have been used for experimental purposes, there has been considerable interest in developing novel compounds and treatment regimens that may prevent or reverse clinical multidrug resistance mediated by MRP1 (reviewed recently by Boumendjel et al., Ref. 44). Some strategies designed to specifically inhibit MRPs as opposed to other drug transporters exploit the GSH dependence of transport of some cytotoxic drugs and the fact that GSH conjugates are not in general substrates for proteins such as P-gp/MDR1 and BCRP (161, 287). One approach has been to block GSH synthesis with the γ-glutamylcysteine synthetase inhibitor BSO (138, 150, 338, 518, 554). In addition, a number of GSH peptidomimetics, some of which were developed as glutathione S-transferase (GST) inhibitors, also inhibit transport by MRP1, as do various protease-resistant GSH derivatives (50, 372). GS-conjugates, some of which are GST inhibitors, such as GS-ethacrynic acid, are also good substrates for MRP1 and potential competitive inhibitors of the transport of other pharmacologically active compounds (44, 50, 51, 555).

Several compounds that were developed originally as inhibitors of P-gp/MDR1 have also been shown to inhibit MRP1. These include the quinoline derivative MS-209 (354) and the pipecolinate derivative VX-710 (biricodar) (141, 142). The latter compound appears to inhibit not only P-gp/MDR1 and MRP1, but also BCRP (341). The naturally occurring polyhydroxylated sterol Agosterol A can also inhibit both P-gp/MDR1 and MRP1, and an azido-derivatized analog of the compound has been used to photolabel both proteins (see also sect. ixB) (13, 343, 409411). Although multifunctional agents such as VX-710 hold the appeal that they may simultaneously inhibit multiple drug transporters, they may also increase the potential for changes is pharmacokinetics that potentially limit tolerable drug dosages, and their IC50 values for the individual transporters may vary considerably. In part because of these considerations, several attempts have been made to develop high-affinity and high-specificity MRP1 inhibitors (44). One such series of inhibitors, comprised of pyrrolpyrimidine analogs, has been described, some of which show considerable specificity for MRP1 as opposed to other drug transporters, with IC50 values in the submicromolar range (526, 527). However, these compounds display some ability to inhibit drug metabolizing cytochrome P-450s such as CYP3A (527). At present, the most highly specific MRP1 inhibitors described are based on tricyclic isoxazoles, and these were discovered by high-throughput screening. LY465803 and a closely related photoactivatable derivative, LY475776, inhibit MRP1 with EC50 values in the range of 10–80 nM (327, 366, 367, 396, 463). These compounds do not inhibit either MRP2 or BCRP, and their affinity for P-gp/MDR1 appears to be ∼100-fold lower than that for MRP1 (88, 463). Their ability to bind to MRP1 is also dependent on GSH or certain of its analogs (see sect. ixB) (88, 327, 396).


A. Background

Despite overwhelming evidence from in vitro and intact animal studies of the ability of ABC drug transporters to increase resistance to a wide variety of cancer chemotherapeutic agents, with few exceptions their contribution to clinical drug resistance remains poorly defined. To a considerable degree, this may be attributable to the challenges of designing and implementing informative clinical trials. Ideally such trials would 1) utilize reliable assays to determine the profile and levels of transporter expression in patient tumors before and after treatment, 2) correlate outcome with transporter expression, and 3) demonstrate improved outcome following confirmed inhibition of individual transporters. A number of trials of P-gp/MDR1 reversing agents that attempted to meet these goals have had disappointing results (25, 45, 121, 126, 127, 275). However, interpretation of the outcome is confounded by the fact that the existence of alternative transporters was unknown or not assessed in the patient population. Furthermore, earlier P-gp/MDR1 reversing agents were of relatively low specificity and affinity and in some cases were found to have significant pharmacokinetic effects that required reduction in dosing of the chemotherapeutic agent(s) used. With at least one second generation P-gp/MDR1 reversing agent, PSC833 (a nonimmunosuppressive derivative of cyclosporine), inhibition of other ABC transporters involved in hepatic clearance, such as MRP2 and the bile salt transporter BSEP/ABCB11, as well as cytochrome P-450s, such as CYP3A, may have been a contributing factor (25, 275, 501). More recently, high-affinity highly specific, P-gp/MDR1 specific reversing agents have been developed. One of these, zosuquidar (LY335979), has shown minimal pharmacokinetic effects, combined with confirmed inhibition of P-gp/MDR1 in recent phase I trials involving solid and hematological malignancies (130, 143, 437). At present, the outcome of phase II and phase III trials of zosuquidar is not known. To date, there have been no comparable trials of MRP specific reversing agents. However, studies of clinical samples have revealed widespread expression of the MRPs in a variety of tumor types, particularly in the case of MRP1 (see below). The possible involvement of MRPs and other transporters in clinical drug resistance has been the subject of recent reviews (75, 275, 276). Consequently, we have focused primarily on studies that either indicate a significant association between transporter expression and negative disease outcome and/or that are supported by relevant biological studies.

B. MRP1 and Solid Tumors

The widespread expression of MRP1 in normal tissues provides an additional challenge to assessing the implications of its presence in clinical samples. This is a particular problem when using approaches such as RT-PCR and immunoblotting with tumor samples that may contain varying amounts of normal tissue (28, 168). Nevertheless, these approaches, combined with immunohistochemistry using well-characterized MRP1-specific monoclonal antibodies (128, 129, 175, 176, 178, 243, 446, 539), have provided convincing evidence of elevated expression of MRP1 in a variety of solid tumors, including such common cancers as lung, breast, and prostate. The strongest case for a functional role of MRP1 in the clinical resistance of solid tumors is provided by several consistent studies of lung cancer.

1. Lung cancer

Frequent expression of high levels of MRP1 has been found in non-small cell lung cancer (NSCLC), which accounts for >75% of lung cancer cases (363, 484, 539). NSCLC, unlike small cell lung cancer (SCLC), is inherently multidrug resistant, and moderate to high levels of expression of MRP1 have been found in a high proportion of both untreated adenocarcinoma and squamous cell carcinomas, the two major forms of NSCLC (267). High levels of MRP1 have been correlated with a higher grade of differentiation of NSCLC, particularly adenocarcinoma (363, 484, 539). Although this would suggest that the protein is more prevalent in less aggressive tumors that might be expected to have a better prognosis, two early studies concluded that elevated MRP1 levels are a predictor of poor response to treatment with drugs known to be substrates of the protein (376, 377). A recent, more extensive study of NSCLC not only confirmed the expression profiles found previously, but also found that the level of MRP1 was a highly significant negative indicator of response to chemotherapy and overall survival (38).

The frequency of MRP1 expression in untreated SCLC is lower than in NSCLC. When present, the protein appears to be limited to small foci of cells within the tumor and the tumor periphery, rather than the more uniform expression observed in NSCLC (38, 539). Nevertheless, MRP1 positivity in untreated tumors has been found to be predictive of poor response to chemotherapy in two studies (200, 259). The restricted expression of MRP1 may be correlated with the fact that SCLC is generally responsive to initial chemotherapy but almost always manifests itself as drug-resistant disease upon relapse. For ethical reasons, tumor material is rarely collected from relapsed SCLC patients. Consequently, the levels of MRP1 in relapsed drug-resistant SCLC have not been firmly established. However, a small study in which longitudinal samples were obtained did find a posttreatment increase in MRP1, as well as several other drug resistance markers (255).

2. Breast cancer

Several independent studies indicate that MRP1 expression is a negative prognostic marker for some types of breast cancer. Three of them concluded that MRP1 positivity of early-stage breast cancer was associated with shorter times to relapse after postsurgical adjuvant chemotherapy (364, 365, 431). The results of these investigations have recently been supported by a larger analysis of MRP1 expression in over 500 premenopausal women with early-stage breast cancer (124). This latest study also indicates a strong association between expression of MRP1 and reduced time to relapse, as well as reduced overall survival.

The levels of MRP1 in tumors of relapsed breast cancer patients who received presurgical adjuvant chemotherapy have consistently been found to be higher than in the initial tumor (122). However, the increases in MRP1 levels were not predictive of response to chemotherapy. Analysis of MRP1 expression in axillary lymph nodes of breast cancer patients also indicated that levels of protein were higher in the metastases than in the primary tumor, while the reverse was the case for P-gp/MDR1 (573). The drug regimen used for adjuvant chemotherapy in the above studies is comprised of cyclophosphamide, 5-fluorouracil, and methotrexate. A number of MRPs, including MRP1, have been shown to confer resistance to methotrexate (62, 187, 252, 557). In addition, increased expression of MRP1 in transfected MCF7 breast cancer cells results in a moderate increase in resistance to cyclophosphamide (346). To date, only MRP5 and MRP8 have been shown to confer resistance to 5-fluorouracil (154, 392). However, MRP5 is only able to transport the metabolized form 5′-fluoro-2′-deoxyuridine, and it is not known if this is also true for MRP8. MRP8 has also been reported to be expressed in breast cancer (34). However, its expression was not determined in the studies described.

3. Prostate cancer

MRP1 overexpression has been documented as a resistance mechanism in prostate cancer cell lines selected by exposure to doxorubicin (552). The protein is expressed at moderate levels in normal prostatic epithelium and at high levels in prostatic intraepithelial neoplasia, as well as in prostatic adenocarcinoma (75). The levels of MRP1 have been reported to increase with cancer stage and invasiveness (485) and to be positively associated with mutant p53 status of the tumors, an association that has also been observed in NSCLC (259, 485). The clinical correlation between high levels of MRP1 expression and p53 mutation is supported by in vitro studies of the regulation of the human and murine MRP1/Mrp1 genes that have shown that wild-type p53 is a strong suppressor of MRP1/Mrp1 transcription (351, 486, 525). The fact that the commonly used antiandrogen flutamide and its active metabolite hydroxyflutamide are effectively effluxed by MRP1 overexpressing cells raises the possibility that MRP1 may contribute to the development of hormonally refractory forms of the disease for which there is currently no effective systemic therapy (153). To date, there has been one phase II trial of the effect of combining mitoxantrone and prednisone with the dual P-gp/MDR1 and MRP1 inhibitor VX-710 (Biricodar, Incel) in the treatment of hormone refractory prostate cancer. The results of the trial showed little if any effect of the addition of the inhibitor (400). However, the ability of MRP1 to confer resistance to mitoxantrone has not been firmly established. Although some drug-selected cell lines that overexpress MRP1 have been reported to be cross-resistant to mitoxantrone (162, 342, 450), studies with MRP1 transfected cells have yielded contrary results (73). Consequently, additional trials with other drug regimens that include more firmly established MRP1 substrates might be justified.

4. Neuroblastoma

Neuroblastoma is the most common extracranial solid tumor in children (335). One of the strongest negative molecular indicators of outcome in neuroblastoma is amplification of the NMYC gene (82). The expression of MRP1 has been reported to be positively correlated with NMYC amplification and to be an independent, negative prognostic indicator (324, 368). Studies of MRP1 expression in neuroblastoma cell lines support a role for NMYC as a positive regulator of MRP1 transcription. In addition, modulation of NMYC levels influences resistance to drugs that are known MRP1 substrates, and VX-710 sensitizes neuroblastoma cell lines that express MRP1 to a similar range of drugs (155, 324, 369, 547).

C. MRP1 and Hematological Malignancies

There have been a number of analyses of MRP1 expression in both acute and chronic leukemias with discordant results. As with solid tumors, the clinical significance of MRP1 expression in leukemic cells is complicated by the fact that the protein is expressed in all normal hematopoietic cell lineages. In acute myeloblastic leukemia (AML), MRP1 expression has generally not been found to be prognostic of response to chemotherapy (239, 391, 499, 509). However, it may be correlated with stage in AML (391, 509) and possibly overall survival (123). One of the challenges to interpretation of the results of studies such as those cited is the possible presence in leukemic cells of multiple transporters. For example, one study of AML found that the combined presence of MRP1 and P-gp/MDR1 was predictive of resistance to treatment while each transporter alone was not (271). Similarly, MRP1 has not been found to be prognostic in acute lymphocytic leukemia (ALL) (87, 239, 441). In contrast, expression of MRP1 in chronic lymphocytic leukemia has generally been found to be high and may be of clinical significance (79, 220, 362).

D. Clinical Relevance of Other MRPs

Assessment of the contribution of other MRPs to clinical resistance is at a very early stage, and there is presently little evidence of an association between expression and treatment response or disease outcome. A survey of the presence of MRP2 in various cancers using tissue microarrays found that the protein was expressed at varying frequency and levels in renal, gastric, breast, lung, colon, and ovarian carcinomas (438). Unlike MRP1, expression of MRP2 has been causatively linked to cisplatin resistance in a number of in vitro studies (85, 229, 230, 293, 333), and increased expression of MRP2 has been associated with cisplatin resistance in human colorectal carcinoma (174). Elevated expression of MRP2 and MRP3 has also been reported in hepatocellular carcinoma (40, 357). Together with other factors, MRP2 expression levels were found to be predictive of response in ovarian cancer (332), while other studies of ovarian cancer have found an unfavorable outcome to be linked to MRP1 and MRP3 levels (373). Posttreatment increases in MRP1, MRP2, and MRP3 expression have also been reported in bladder cancer (491), and increased expression of MRP3, together with MRP1, has been observed in human glioma (156). Recently, MRP3 was reported to be overexpressed in pancreatic carcinoma (248) and to be associated with poor outcome and a failure to respond to chemotherapy in childhood ALL and AML, respectively (477). Among the short MRPs, MRP4 has been reported to be a negative prognostic factor in neuroblastoma and MRP5 to be overexpressed in pancreatic carcinoma (248, 370), while MRP8 and a mRNA encoding a possibly truncated form of MRP9 have been detected at high levels in breast cancer (34, 35).


A. ATPase Activities of Purified ABCC Proteins

The ATPase activities of purified ABCC proteins such as MRP1, MRP2, and CFTR are about two orders of magnitude lower than reported for some prokaryotic ABC transporters and P-gp/MDR1 (7, 11, 158, 291, 320, 325, 326, 402, 406). Consequently, it has proven difficult to study the hydrolytic properties of these proteins using crude membranes. Native human MRP1 has been purified from H69AR cells, and a histidine-tagged form of the protein has been purified from transfected baby hamster kidney cells and in P. pastoris (57, 58, 320, 325, 326, 540). When reconstituted into proteoliposomes, purified MRP1 has an ATPase activity of 5–10 nmol/mg protein. In addition to a much lower Vmax value than P-gp, MRP1 displays a severalfold higher affinity for ATP with a Km of 100–300 μM (325, 326, 461). A similar Vmax value was obtained for the ATPase activity of purified reconstituted MRP2 (158). Possible reasons for the relatively low ATPase activities of the ABCC proteins when compared with some other ABC proteins have been provided by studies of the ATP binding and hydrolysis characteristics of the individual NBDs of CFTR (6, 7, 24, 132, 488), MRP1 (136, 194196, 353, 386, 387, 412, 530, 548, 572), and SUR1 (334, 503). For the purposes of this review, we have focused on studies of MRP1 that illustrate functional differences between the protein’s NH2- and COOH-proximal NBDs. However, the results of these and other studies reveal extensive similarities among MRP1, CFTR, and SUR1 that can likely be extrapolated to other C branch proteins.

B. ABC NBDs Functional Cooperatively

A fundamental characteristic of all ABC transporters studied to date is that their ATPase activity is highly dependent on cooperative interactions between the two NBDs. Although there have been reports describing the ATPase activity of purified, soluble forms of the NBDs from a number of eukaryotic ABC proteins, including ABCC proteins, these data are difficult to interpret given the established interdependence of the two domains in the intact protein. The reason for this interdependence was initially suggested by the crystal structure of the soluble ABC-like protein Rad50, which dimerizes in the presence of ATP (189, 190). These studies were subsequently supported by elucidation of the structures of bacterial ABC transporters such as MsbA from Vibrio cholera (vcMsbA) (56) and BtuCD (297). The crystal structures of Rad50, vcMsbA, and BtuCD indicate that the Walker A and Walker B motifs of one NBD cooperate with the C motif of the apposing NBD, to form two composite nucleotide binding sites (NBSs) (Fig. 7A). Thus involvement of both NBDs in the formation of each NBS provides an explanation for the cooperativity observed during studies of the transport cycle of ABC proteins such as P-gp/MDR1 and some bacterial transporters (19, 92, 173, 253, 309, 448). For ease of discussion, we have followed a convention when referring to ATP binding or hydrolysis by a specific NBD, of ascribing the function to the NBD that contributes the Walker motifs to the NBS.

FIG. 7.

A: the ATP binding pocket based on the crystal structure of HlyB NBD dimer. The figure shows views of an ATP molecule in one of two nucleotide binding pockets based on the crystal structure of the HlyB dimer. Views were chosen to illustrate the interaction between the nucleotide and the conserved Walker A and Walker B motif of one NBD molecule and the signature “C” sequence from its partner (551). The figure was generated and edited using PyMol to expose the nucleotide binding pocket. B: 2-dimensional schematic of a closed NBD dimer with two bound ATP molecules. The schematic shows a possible, generic, “head-to-tail” MRP NBD dimer model that is based on conserved structural elements from known ABC protein crystal structures with α-helices and β-sheets shown as cylinders and arrows, respectively. NBD1 is shown in blue, and NBD2 is in green. The α-helix in NBD2 designated as 7’ corresponds to the 13-amino acid region between the Walker A and ABC signature sequence that is missing from NBD1. Regions believed to be important for NBD-NBD communication are shown in transparent boxes.

C. Stoichiometry of ATP Hydrolysis and Substrate Transport

Studies of ABC proteins from prokaryotes where the NBDs are structurally identical and of P-gp/MDR1 have supported a general model of the transport cycle in which NBDs are functionally equivalent and hydrolysis of ATP occurs alternately at each NBS (262, 361, 443, 459, 460, 512). The fact that the NBDs of P-gp/MDR1 can be exchanged without loss of function provides strong support for the model (26, 199). However, the stoichiometry of ATP hydrolysis and substrate transport has not been fully resolved. In the case of P-gp, considerable evidence exists to support a model in which hydrolysis of ATP at either NBS results in transport of one molecule of substrate (459). A more recent variation of this model proposes that the binding and hydrolysis of one ATP molecule drives a “power stroke” in which the protein shifts from a high- to low-affinity substrate binding state with the concomitant transport and release of one molecule of substrate (443). Hydrolysis of a second ATP is then required to reset the protein in a high-affinity state for the next transport cycle. In this model, it remains unclear whether binding and hydrolysis of ATP at each of the two NBSs is dedicated to a different step in the transport cycle of proteins such as P-gp/MDR1 and the prokaryotic transporters, or whether the process is stochastic. Although many studies support the proposal that the NBDs of these proteins are truly functionally identical, some evidence suggests that the position of the NBD may influence its function (26, 199). Until relatively recently, it had also been commonly assumed that the hydrolysis of ATP drove the conformational changes in the protein required for transport. However, recent studies, including those of ABCC proteins, have provided strong evidence that it is ATP binding rather than hydrolysis that converts the protein from a high- to low-affinity substrate binding state (173, 328, 329, 386, 424). The differences between these two models notwithstanding, it is clear that C branch proteins do not fit readily with a model in which the NBDs are functionally equivalent. Unlike P-gp, the two NBDs differ considerably with respect to their ability to bind and hydrolyze ATP (136, 194, 353). In the case of MRP1, there is considerable evidence that each NBD is responsible for distinct steps in the transport cycle and that the shift from a high- to low-affinity substrate binding state involves the ordered rather than stochastic binding of ATP to the two NBDs (136, 194, 353).

D. Experimental Approaches Used to Study ATP Binding and Hydrolysis

Much of the evidence of functional differences between the NBDs of MRP1 has come from ATP binding and ADP trapping experiments with photoactivatable derivatives of ATP such as γ- or α-[32P]8-azido-ATP derivatives (136, 194, 353). At 4°C, these derivatives can be used to examine nucleotide binding under conditions that minimize hydrolysis, and the bound nucleotide can be covalently linked to one or the other NBD by ultraviolet irradiation. Comparison of the results obtained with α- and γ-32P-labeled derivatives allows determination of whether the cross-linked nucleotide is ATP or ADP (125, 136, 194, 353, 387, 530). At physiological temperatures, labeled azido-ADP generated by hydrolysis can be trapped in the presence of orthovanadate or beryllium fluoride, either in a form believed to mimick a posthydrolytic transition state or an ATP-binding ground state, respectively (125, 505). Two approaches have been used to selectively investigate nucleotide binding and trapping by each NBD. The first relies on labeling of the protein followed by cleavage at a protease hypersensitive site in the cytoplasmic linker connecting the NH2-proximal NBD to MSD2 (Fig. 5) (176, 194, 353). The two major fragments produced can then be separated by SDS-PAGE. Alternatively, dual-expression vectors have been used to coexpress stoichiometrically equivalent amounts of two protein fragments, similar to those produced by limited trypsinolysis, which associate with very high efficiency to form a functional LTC4 transporter (136).

With the use of the experimental approaches described above, it has been shown that Mg2+-dependent ATP binding at 4°C occurs primarily at NBD1, while trapping of ADP at higher temperatures in the presence of orthovanadate or beryllium fluoride occurs predominantly at NBD2 (136, 194, 242, 353, 387, 530). These and other studies have revealed that the affinity of NBD1 for ATP is two- to threefold higher than that of NBD2 (548) and that the latter is the major, perhaps only, site of ATP hydrolysis. Several studies have concluded that NBD1 of proteins such as MRP1 (136, 194, 353, 572), CFTR (7, 24, 37), and SUR (334) may be incapable of ATP hydrolysis. However, at least in the case of MRP1, there is also evidence to the contrary (see below) (386). The nonequivalence of the two NBDs is supported by several additional observations. Thus, although ATP binding and hydrolysis at NBD2 is highly dependent on ATP binding to NBD1, binding of ATP by NBD1 is much less dependent on the functional state of NBD2. For example, binding of ATP by NBD1 does not decrease when the ATP binding or ATPase activity of NBD2 is eliminated by mutation of essential residues in its Walker A or Walker B motifs (136, 194). Furthermore, the expression of soluble forms of the two NBDs of MRP1 has shown that NBD1 is able to bind ATP with relatively high affinity in the absence of NBD2, while no binding could be detected under similar conditions with soluble NBD2 (136, 538). These observations indicate that relatively tight binding of ATP can occur at NBD1 independently of interactions with NBD2 while binding of ATP by NBD2 is strongly dependent on the binding of ATP to NBD1.

Further evidence that the two NBDs of MRP1 are not functionally equivalent comes from the effect of various mutations on substrate transport, as determined by in vitro studies with inside-out membrane vesicles. Typically, mutations of conserved amino acids in the Walker A and Walker B motifs of NBD1 that are known to eliminate ATP hydrolysis decrease transport of LTC4 by ∼50–70%, depending on the nature of the mutation, while comparable mutations in NBD2 essentially inactivate the protein (136, 194, 387). It appears likely that the partial activity of the NBD1 mutant proteins may be attributable to retention of some level of ATP binding by the mutant NBD that enables a reduced level of ATP binding and hydrolysis by NBD2. In support of this suggestion, conversion of the conserved Walker A Lys residue in NBD1 to a neutral amino acid or to Arg results in a partially active protein (136, 194, 387), while substitution with a negatively charged Asp residue almost completely eliminates MRP1 LTC4 transport activity (387).

E. Functionally Important Structural Differences Between ABCC NBDs

What are the structural features that contribute to the nonequivalency of the two NBDs? Sequence comparisons of ABC protein NBDs reveal that, along with the Walker A/B and signature C elements, there are other motifs, namely, the Q, D, and H (switch) loops, that contain highly conserved Glu, Asp, and His residues, respectively (Fig. 4). NBD crystal structures have been solved for a number of bacterial ABC proteins besides, including HisP (201), MJ0796 (550), MalK (104), HlyB (449, 551), GlcV (514), BtuD (297), and MsbA (55, 56, 415), as well as for the human proteins TAP1 (137) and, most recently, both the wild-type CFTR NBD1 and the NBD containing the common Δ508 mutation (290). Although there are some regions of unique architecture among these proteins, a common configuration of eight α-helices (α1–8) and nine β-sheets (β1–9) is observed (Fig. 7B) (with the exception of MJ0796, which lacks the final 2 conserved helices). These L-shaped domains are functionally separated into two arms: arm I, an F1-like ATP-binding unit comprised of the Walker A and Walker B elements, and arm II, containing the C motif (172). Alignments of the ABC NBDs indicate a high level of tertiary structural conservation in the nucleotide binding pocket. The adenine moiety of ATP generally stacks against a conserved aromatic residue near the end of, or following, β1, while the side chains of residues in the linker region between β1 and β2 stabilize the ribose group (104, 137, 201, 290, 514, 550). Residues in the Walker A motif typically hydrogen bond with the α-, β-, and γ-phosphates, and the negatively charged residues in the Walker B motif coordinate Mg2+ binding and/or stabilize ADP in the binding pocket through interactions with water (Fig. 7A). Similarly, Asp and His in the D and H loop, respectively, make contacts with water that stabilize the binding of nucleotide, while the conserved Gln residue in the Q loop interacts with the catalytic Mg2+ and attacking water molecule. Comparison of nucleotide free, ADP-bound, and ATP-like AMP-PNP· Mg2+ bound NBD structures in ABC NBDs such as GlcV suggests that significant structural changes occur on nucleotide binding (514). These conformational changes primarily involve the Walker A motif and the Q loop and are thought to be transmitted to the MSDs through the sequence adjacent to the latter (297).

The most obvious difference between the two NBDs of the MRP-related transporter proteins and CFTR is the previously mentioned, highly conserved, apparent deletion of 13 amino acids between the Walker A motif and the Q loop in NBD1. On the basis of the general structure illustrated in Figure 7B, this would eliminate a β-sheet as well as an α-helix located between β4 and β5 found in approximately half of the structures of the ABC proteins determined to date (137, 201, 449, 550). Whether the foreshortening of the region between the Walker A motif and the Q loop contributes to the relatively high affinity with which NBD1 of CFTR and MRP1 binds ATP is not clear (7, 136). The insertion of a 13-amino acid sequence from NBD1 of P-gp/MDR1 that corresponds to the region missing in MRP1 eliminated high-affinity ATP binding and changed the conformation of the domain, since a conformation-dependent monoclonal antibody (QCRL-3) against this region no longer recognized the protein (136, 180). However, the most recent crystal structure of CFTR NBD1 indicates that the atomic contacts established by ATP are similar to those found in other ABC NBDs with known crystal structures (290). Clearly, additional features are important for the observed differences in ATP binding and hydrolysis exhibited by the two NBDs of the C branch proteins.

As mentioned earlier, in most ABC NBDs, there is a Glu residue immediately following the Walker B motif that serves as a catalytic base for cleavage of the γ-phosphate group (345, 472, 515). Although Glu is present at the appropriate location in NBD2 of the ABCC proteins, the corresponding amino acid in NBD1 is Asp with the exception of CFTR in which it is Ser (72, 386). The consequences of mutating the Asp and Glu residues in NBD1 and NBD2 of MRP1 have illustrated the importance of these residues in the catalytic cycle and have provided useful information about the probable transport mechanism of the protein. Studies of other ABC proteins (201, 345, 506) suggest that conversion of Asp to Glu in MRP1 NBD1 would be expected to increase the ATPase activity of this domain and, based on an alternating sites model of catalysis, might enhance the rate of substrate transport. Paradoxically, such a mutation decreases LTC4 transport activity by ∼80% (386). As expected, the Asp to Glu mutation markedly increases cleavage of the γ-phosphate of ATP by the mutant NBD1. However, it also results in vanadate-independent trapping or occlusion of nucleotide and a failure to release the ADP produced. The mutation also results in a major decrease in ADP trapping at NBD2, suggesting that occupancy of NBD1 by ADP prevents hydrolysis at NBD2. The decrease in hydrolysis at NBD2 under these circumstances would be predicted by the alternating sites model of hydrolysis, since it is postulated that only one NBS can be occupied by ADP at any given time (459). The reciprocal Glu to Asp mutation in NBD2 also inactivates the protein and results in tight binding of nucleotide by the mutant NBD2 that is also vanadate independent. In this case, the bound nucleotide was found to be a mixture of ATP and ADP, suggesting that the mutation had not completely eliminated cleavage of the γ-phosphate. The NBD2 Glu to Asp mutation also increased ATP binding and ADP trapping at the associated wild-type NBD1. This is consistent with other evidence suggesting that the binding of ATP by NBD2 stimulates ATP binding by NBD1 (136, 194). However, it also reinforces the possibility that the native NBD1 of MRP1, like the Glu to Asp NBD2 mutant, has some ability to hydrolyze ATP, albeit low when compared with the native NBD2 (386).

In addition to the atypical Walker B motif in NBD1, the C signature in NBD2 of MRP1 and other ABCC proteins is unusual (72, 147, 387). The signature sequence in NBD1 of ABCC proteins conforms well to the canonical C signature and contains the conserved core LSGGQ motif. However, this core motif in NBD2 of MRP1, CFTR, and SUR1 is LSVGQ, LSHGH, and FSQGQ, respectively (3, 72, 417). Furthermore, the more extended signature sequence normally contains highly conserved Arg and Ser residues at positions 8 and 10, relative to the start of the core. Although these residues are present in the NBD1 signature, the Arg is replaced by Leu and the Ser by Cys in NBD2 of most of the ABCC proteins (386, 387). In P-gp, the conserved Ser residue has been shown to be required for hydrolysis (309, 496). Since the NBD2 signature sequence would be involved in hydrolysis of ATP by NBD1, it appears highly likely that these variations contribute to the low or absent ATPase activity of this NBD. As with several other ABC proteins, studies of the signature sequences in MRP1 indicate that they are not required for ATP binding (387, 412, 466, 489, 490, 496). However, mutation of conserved Gly residues in the signature sequence of NBD1 and NBD2 inactivate or partially inactivate LTC4 transport, respectively (387, 412). Although binding of azido-ATP appears to be unaffected, the C signature mutations prevent the transition to a low-affinity substrate binding state (387, 490). This is presumed to be a consequence of a failure to form the correct interface in the NBD closed dimer (387, 390). Similar observations have been made with P-gp/MDR1 (489, 496).

F. High- and Low-Affinity Substrate Binding States

A number of experimental approaches have demonstrated that ABC transporters such as P-gp/MDR1 and LmrA shift from a high- to low-affinity substrate-binding state in the presence of ATP (295, 402, 403, 442, 512, 519, 524). This change in affinity is most apparent in the presence of ATP and orthovanadate or beryllium fluoride, both of which markedly increase occupancy of the protein by ADP. The decrease in substrate binding is presumed to be a consequence of conformational and positional changes in the NBDs involved in formation of a closed NBD dimer that are transmitted to the MSDs. The consequential reorientation of TM helices then results in the substrate being transferred from the high-affinity site to which it initially bound, to a low-affinity site from which it would be released. At present, there is little evidence to demonstrate that this actually involves physical transfer of substrate from one site in the protein to another, rather than a decrease in the affinity of the initial binding site and the release of the substrate into a vestibule open to the extracellular space.

The presence of substrates and inhibitors and, in some cases, compounds that bind but are not transported, also influences ATP binding and hydrolysis by proteins such as P-gp/MDR1 and MRP1 (282, 321, 386, 394, 402, 403, 442, 524). Studies with purified proteins indicate that the activation of P-gp/MDR1 ATPase activity can be quite substantial (9, 402, 461). In the case of MRP1, substrates such as LTC4 appear to have less effect and typically stimulate ATPase activity 1.5- to 2-fold at most (57, 325, 326). The presence of GSH or GSH-independent substrates such as LTC4 modestly stimulates the binding of ATP to NBD1 of MRP1, which, in turn, stimulates the binding and hydrolysis of ATP by NBD2 (136, 194, 282). In a number of ABC transporters, substrate binding has been shown to result in changes in accessibility of the NBDs to proteases and chemical modification (232, 254, 294, 322, 323, 355, 473). Fluorescence spectroscopy has provided strong evidence that the binding of some substrates to P-gp/MDR1 results in alterations that are transmitted from the MSDs to the NBDs (294296, 473). In a study of MRP1 using protease accessibility and attenuated total reflection-Fourier transform infrared spectroscopy (ATR-FTIR), it was concluded that binding of drugs that are transported in a GSH-dependent manner did not result in conformational changes in the cytosolic portions of the protein (322). However, such changes were elicited by GSH and resulted in increased nucleotide binding/hydrolysis (322, 520). Precisely how this stimulation occurs has not been completely defined but presumably involves a conformational change in the MSDs of the protein that facilitates initial binding of ATP and subsequent formation of a closed NBD dimer in the presence of ATP.

Investigation of the steps involved in shifting MRP1 from high- to low-affinity states has been facilitated by the availability of photoactivatable, natural substrates, such as LTC4, and various mutant proteins, such as those described above, that have permitted substrate binding to be examined with the protein effectively being locked in various states of nucleotide occupancy. The results of these studies indicate that conversion of the Walker B catalytic Glu residue in NBD2 to Asp markedly potentiates the shift from high- to low-affinity LTC4 binding in the presence of ATP or the poorly hydrolyzable analog adenosine 5′-O-(3-thiotriphosphate) (ATPγS) (386). As indicated above, this mutation results in impaired release of ATP and ADP from NBD2 and increases ATP binding at NBD1. Thus formation of the low-affinity state appears to involve binding of ATP to both NBDs and persists as long as NBD2 is occupied by either ATP or ADP. Conversely, mutation of the atypical Walker B Asp residue in NBD1 to Glu results in stable occupancy of NBD1, primarily by ADP, and decreases ATP binding and trapping at NBD2 (386). It also prevents the conversion of the protein from a high- to low-affinity state in the presence of ATP or ATP plus vanadate. Further evidence that the decrease in substrate affinity occurs when both NBDs are occupied by ATP is provided from a double mutant in which the Asp and Glu residues were switched (386). The mutant protein is completely inactive with respect to LTC4 transport but binds ADP and ATP avidly at NBD1 and NBD2, respectively. In the presence of ATP or ATP plus vanadate, this mutant remains in a high-affinity LTC4 binding state. However, ATPγS is able to promote the formation of the low-affinity state, although less effectively than with the wild-type protein. Thus occupancy of the mutant NBD1 by ADP prevents formation of the low-affinity state. It is tempting to speculate that this mimics the step involved in resetting the wild-type protein in a high-affinity state, i.e., as a result of the hydrolysis of ATP at NBD1. However, the possibility remains that this step may simply involve dissociation of ATP from NBD1.

G. Interactions Between MSDs and NBDs

How conformational changes are transduced between the NBDs and MSDs of the ABC transporters remains poorly understood. There is considerable evidence that conformational changes of the NBDs both within and between the two domains upon nucleotide binding and hydrolysis result in significant reorientation of at least some of the TM helices in their associated MSDs (136, 194, 308, 327, 386, 394, 425, 429). Similarly, changes in the orientation of certain TM helices may be capable of influencing the functional interactions between the NBDs that are required for ATP hydrolysis (218, 242, 310, 468, 570). It has been suggested based on the crystal structures of bacterial transporters such as MsbA and BtuCD that intracellular domains in contact with, or connected to, both TM helices and the NBDs may serve as “bridges” that transduce signals between the NBD and TMDs (55, 56, 297).

BtuCD is a vitamin B12 transporter from Escherichia coli that consists of two identical MSDs (BtuC) and two identical NBDs (BtuD) (103, 297). The structures of the MSDs of BtuCD are unusual in that they appear to contain 10 helices, which complicates extrapolation of the structure to other ABC proteins (297). However, a dihelical structure in CL3 of BtuC, located between the sixth and seventh TM helices, makes multiple side-chain contacts with BtuD residues around the Q loop, and to Met and Gln residues downstream of the Walker A and the ABC signature motifs, respectively. The lipid A transporter MsbA from both E. coli and V. cholera has a more typical structure with six TM helices in each MSD (55, 56). In these proteins, NBD-MSD contacts occur between CL1, which connects the second and third TM helices of each MSD and the Q loops in each NBD. These two CLs each form a trihelical ”U“-like structure that has been suggested to provide a pivot point on which the associated NBDs may rotate during hydrolysis (52, 55). Although there is no extended homology between BtuC CL3 and MsbA CL1, both loops contain a conserved Gly motif that is found in other ABC proteins, including MRPs, suggesting that these domains may be functionally conserved (297). The equivalent regions in MRP1 correspond to CL4 located between TMs 7 and 8 in MSD1 and CL6 between TMs 13 and 14 in MSD2, respectively (Figs. 1 and 5). There is some experimental evidence to suggest that MRP1 CL6 communicates with the NBDs, since nonconservative mutation of Asp1084, predicted to be at or near the cytoplasm-membrane interface of TM14, to Gln severely decreases binding and vanadate-induced trapping of azido-ATP at NBD2 indicating reduced hydrolysis of ATP (568). This mutation also prevents the transition of the protein from high- to low-affinity substrate binding states in the presence of ATP or ATPγS. However, mutation of Pro1060 and Pro1068 predicted to disrupt CL6 secondary structure had little effect on activity, and MAb MRPm5 directed against a peptide epitope in this loop does not inhibit LTC4 transport (242, 243). On the other hand, mutations of a conserved Pro residue (Pro1150) in CL7 connecting TM15 and TM16 affects both the substrate specificity and catalytic activity of MRP1 (242). The substrate specificity of MRP1 is also affected by mutations of two conserved Tyr residues in this region (77).

The cytoplasmic regions directly connecting MSDs and NBDs are also likely to be important for transmitting conformation changes between membrane and cytosolic domains (56). There is strong experimental evidence from studies of P-gp/MDR1 that the binding of ATP results in a change in the predicted tilt of TM6, which is linked directly to NBD1 (429). In MRP1, the corresponding TM is TM11, and it was found recently that Ala substitution of a residue, Asn590, located in this helix caused a global decrease in MRP1 transport activity by decreasing ATP binding to NBD1 (570). Nonconservative mutation of other nearby residues in TM11, Arg593, Phe594, and Pro595, also causes a global elimination of transport activity and LTC4 binding (52, 160, 242). Interestingly, Arg593 of MRP1 corresponds to Arg347 of CFTR, mutation of which has been correlated with a mild form of cystic fibrosis (83). This mutation has been shown to decrease the ATPase activity of CFTR three- to fourfold (238). Given the many functional similarities between the cooperative interactions of the NBDs of MRP1 and CFTR, the CFTR Arg to Glu mutation may also be affecting ATP binding to NBD1. Thus the regions linking NBDs and MSDs may transmit conformational changes both from the NBDs to the MSDs, and vice versa.

H. The Transport Cycle of MRPs

A model of the hypothetical transport cycle of MRP1 is illustrated in Figure. 8. The various steps shown in Figure 8 can be summarized as follows. Step 1) Binding of substrate (or GSH plus GSH-dependent substrate) to a high-affinity site(s) induces conformational changes that enhance binding of ATP to NBD1. Step 2) The initial binding of ATP by NBD1 stabilizes the interaction between NBDs by establishing contacts with the C signature of NBD2, facilitating the binding of a second molecule of ATP. Step 3) Binding of the second ATP completes formation of the closed NBD dimer and also causes conformational changes in NBD2. The combined positional and conformational changes resulting from ATP binding are transmitted to the MSDs resulting in a decrease in the affinity for substrate. Step 4) The protein maintains a low-affinity state following hydrolysis of ATP at NBD2, as long as NBD1 is occupied by ATP and ADP has not been released by NBD2. Although the steps in the model leading to formation of the low-affinity substrate binding state are strongly supported by experimental data (57, 136, 386, 387, 394, 572), how the protein resets for another cycle remains more speculative and hinges on whether or not NBD1 is catalytically active. Step 5) If NBD1 lacks ATPase activity, resetting of the protein might occur after the release of only ADP from NBD2, or require release of ADP from NBD2 and ATP from NBD1. Step 6) However, it may be premature to exclude the possibility that hydrolysis of ATP by NBD1 is required to reset the protein. As mentioned above, low levels of orthovanadate-dependent trapping of ADP by wild-type NBD1 can be detected, and mutations that increase ADP trapping at NBD1 lock MRP1 in a high-affinity substrate binding state, even when ATP is bound by NBD2 (136, 386). Finally, given that the ATPase activity of purified MRP1 is ∼100-fold lower than that of similar preparations of a protein such as P-gp/MDR1 (320, 325, 326, 402), it remains possible that the rate-limiting step in the transport cycle is the hydrolysis of ATP by NBD1.

FIG. 8.

Model of the hypothetical transport cycle of MRP1. Illustrated are hypothetical steps in the transport cycle of MRP1 and possibly other MRPs. MSD1-NBD1 and MSD2-NBD2 are shown in blue and gray, respectively. MSD0 is not shown both for simplicity and because it is not required for transport of at least some substrates, including LTC4. A description of the various steps in the model and the experimental evidence supporting it is provided in section ix, E, F, and H.


A. Experimental Approaches Used to Investigate Substrate Recognition

Several complementary approaches have been taken to identify regions and specific amino acids in the MRPs that are important for substrate recognition and transport. To date, these studies have focused on MRP1-like proteins, primarily MRP1 itself and to a lesser extent MRP2 and -3. Extensive use has been made of in vitro transport studies with vesicles prepared from the plasma membranes of transfected mammalian and insect cells expressing wild-type and mutant MRPs (134, 161, 274, 287, 299). This experimental approach has been extremely informative with respect to identification of potential substrates, inhibitors, and modulators of the proteins and has provided a powerful complement to the use of intact cells for drug efflux or chemosensitivity assays. In addition, transport studies with membrane vesicles containing mutant MRPs have allowed identification of regions or amino acids that are critical for overall activity, as well as substrate selectivity, and stable expression of the transporter. A major reason why this approach has been so informative is attributable to the relatively hydrophilic characteristics of many MRP substrates compared with those of drug transporters such as P-gp/MDR1. This has facilitated kinetic analyses of a substantial number of substrates and mutant MRPs. The in vitro transport studies have also been complemented by photoaffinity labeling with several azido derivatized and intrinsically photolabile compounds to identify regions of the protein predicted to be in or close to substrate binding sites (90, 89, 225, 273, 327, 394, 396, 409, 540). Although far from complete, sufficient information has now been obtained to develop a tertiary structure model of the disposition of a considerable number of amino acids involved in determining MRP1 substrate specificity and/or transport activity.

B. Photoaffinity Labeling Studies

Partial proteolysis has been used to identify regions of MRP1 that are cross-linked to several photoactivatable compounds (89, 90, 225, 327, 394). After cleavage, labeled fragments can be identified using well-characterized MRP1 MAbs or antibodies against exogenous epitopes that have been introduced at specific locations in the protein. Alternatively, advantage has been taken of the ability to coexpress various fragments of MRP1 that associate to reconstitute an active transporter and that can readily be separated by gel electrophoresis (134) (see sect. viiiD). Three different patterns of labeling have emerged.

The earliest attempts to localize substrate binding sites in MRP1 used azido derivatives of the quinoline N-(hydrocinchonidin-8′-yl)-4-azido-2-hydroxybenzamide (IACI) and iodoaryl azido-rhodamine 123 (IAARh123) (89, 90). Proteolytic digestion and epitope mapping indicate that IACI and IAARh123 cross-link to a similar extent to one or more amino acids between Ser542-Arg593 in TMs 10 and 11 and Cys1205-Glu1253 in TMs 16 and 17 (89, 91). These regions correspond topologically to locations in P-gp/MDR1 that have been previously implicated in substrate binding (47, 101, 148, 541). Unmodified LTC4 has also been used to photolabel MRP1 (272, 299, 540). Both coexpression and partial proteolysis studies indicated that LTC4 cross-linked strongly to MSD1 and only weakly to MSD2 (394). This profile of labeling differs significantly from that obtained recently using an arylazido derivative of LTC4 that appears to predominantly label MSD2 in a region that includes TMs 16 and 17 (226, 540). The reason for the differences between the labeling patterns of LTC4 and the derivative appears likely to be attributable to the presence of the bulky arylazido group that affects the affinity of the latter compound for MRP1, since, as indicated below, most mutations in TMs 16 and 17 have little effect on LTC4 transport and TM17 (1228–1248) can be replaced entirely with an alanine helix without loss of leukotriene transport (23). Recent improvements in MALDI TOF mass spectrometry methods have allowed the more complete analysis of integral membrane proteins and identification of their ligand binding sites. For example, Ecker et al. (111) have shown by MADLI TOF mass spectrometry that ATP binding increases the accessibility of the fifth TM helix of the bacterial ABC transporter LmrA to labeling by a photoactive ligand while ATP hydrolysis has an opposite effect. Human P-gp/MDR1 has also been analyzed by mass spectrometry, and ligand-binding regions have been identified (112). Recently, this methodology has been applied to MRP1, and comparisons of the mass spectra of the peptides generated by in-gel protease digestions of purified unlabeled MRP1 and MRP1 photolabeled with LTC4 have identified six candidate LTC4-modified fragments that include the COOH-proximal region of CL3 (amino acids 260–274), TM6 (amino acids 320–331), TM7 (amino acids 372–385), TM10 and its cytosolic juxtamembrane region (amino acids 546–553), and TM17 and its cytosolic juxtamembrane region (amino acids 1233–1255 and 1248–1264) (540). To a significant extent, the sequence assignments of the LTC4-modified peptides are consistent with other LTC4 photolabeling and transport studies of wild-type and mutant MRP1 proteins, but less so with studies using the arylazido derivatized LTC4. However, the apparent binding of LTC4 to the COOH-proximal region of CL3 (amino acids 260–274) is not consistent with earlier findings (529), and further investigations using more sophisticated mass spectrometry methods for fragmentation studies of the LTC4-labeled peptides are clearly needed.

A third pattern of photolabeling has been observed with two compounds that require GSH in order for them to bind to MRP1. The first is a polyhydroxylated sterol, agosterol-A, that has been shown to reverse P-gp/MDR1- and MRP1-mediated resistance in vitro (13). In contrast to P-gp/MDR1, the binding of azidoagosterol-A to MRP1 is GSH dependent, and contrary to the behavior of LTC4, the predominant site labeled is in MSD2 (409). Mutational studies have provided evidence that the region to which cross-linking occurs involves TM helices 16 and 17 (410, 411). A similar profile of GSH-dependent cross-linking is observed with a highly potent inhibitor of MRP1 (327, 366, 396). The compound LY475776 is a tricyclic isoxazole that inhibits MRP1-mediated LTC4 transport in a GSH-dependent fashion, with a EC50 in the 50 nM range (88, 327, 366, 396). Inhibition is completely dependent on the presence of GSH (or certain of its analogs), as is the ability of the compound to photolabel MRP1, which is adversely affected by mutations in TM17.

The contrast between the labeling profiles obtained with LTC4 and the two GSH-dependent compounds suggested that the predominant labeling of MSD1 by LTC4 might be attributable to the GSH moiety. Consistent with this possibility, the labeling of MRP1 by azidophenacyl-GSH, which has been shown to substitute for GSH in vesicle transport studies, is very similar to that of LTC4 (396). Although apparently not directly labeled by either unmodified LTC4 or azidophenacyl-GSH, very weak labeling of CL3 has been reported with a different arylazido derivative of GSH (225). The significance of this low level of labeling is unclear, but CL3 residues 204–280 are clearly important for transport activity and trafficking of MRP1 (see sect. ivB) (20, 21, 529). This region has also been shown to be essential for binding of both unmodified LTC4, azidophenacyl-GSH, azido AG-A, and LY475776 (394, 396, 409, 529).

C. Mutagenesis Studies

Regardless of the diverse structures of the photoactivatable compounds used to label MRP1 and the differences in their labeling profiles, competition studies suggest that they bind to mutually exclusive sites on the protein (89, 90, 327, 396, 409). This observation is consistent with many in vitro vesicular transport studies demonstrating reciprocal competition between structurally unrelated substrates, with some exceptions (281). Mutational studies have also identified individual amino acids that are important for the transport of a range of diverse substrates, supporting the notion of a common binding site or pocket (52, 76, 77, 159161, 241, 242, 281, 289, 410, 468, 565, 567, 568, 570). However, numerous examples exist of amino acid residues, mutation of which alters transport of some substrates and not others (76, 77, 159, 160, 207, 241, 242, 285, 289, 565568). In addition, apparent positive cooperativity has been observed between certain substrates, or between substrates and compounds that may bind to the protein but not be transported (281, 282, 284, 299303, 413, 518). Thus the experimental evidence, which is reviewed below, suggests that substrates establish multiple, often but not always, overlapping interactions with amino acid residues that collectively form a relatively large binding pocket. If these interactions induce conformational changes in MRP1, as appears likely, it can be envisaged how in some instances the binding of one compound may facilitate binding of another by unmasking additional contact points. Alternatively, positive cooperativity may result from the binding of one compound masking residues that disfavor interaction with another substrate or ligand. Finally, it should be noted that several residues have been identified that, rather than being important for the activity or substrate specificity of MRP1, play a critical role in the stable expression of the transporter in mammalian cell plasma membranes (84, 160, 241, 242, 468).

To some extent, mutagenesis studies of the MRPs have been guided by previous investigations of the location of functionally important residues in P-gp/MDR1, as well as the characterization of naturally occurring mutations in CFTR/ABCC7 and other ABCC proteins that underlie genetic disorders. Major differences in substrate specificity also exist between human and nonprimate, mammalian MRP1 orthologs (314, 371, 482), which has aided identification of amino acids critical for the transport of some substrates (371, 483, 565, 566). Finally, systematic mutation of amino acids with side chains predicted to contribute to substrate binding (e.g., polar, ionizable, and aromatic residues), or to influence α-helical geometry (e.g., Pro) or membrane insertion during biogenesis (e.g., aromatic and basic residues) has illustrated the importance of these residues with respect to substrate specificity, overall activity, and stable expression of the transporters in mammalian cell membranes (52, 76, 77, 159, 160, 207, 241, 242, 281, 285, 289, 410, 468, 565568, 570).

As discussed in section ixB, studies of P-gp/MDR1 indicate that TM6 and TM12 play major roles in determining its substrate specificity. In MRP1-like MRPs, the corresponding TMs are 11 and 17. TM17 is relatively highly conserved, exceptionally so between MRP1 and MRP3 (Fig. 9). Mutational studies of TM17 in MRP1, MRP2, and MRP3 have revealed multiple polar and/or aromatic residues and basic residues that have pronounced effects on substrate specificity, with respect to various classes of natural product drugs and conjugated organic anions, such as E217βG and LTC4, as well as folic acid analogs such as methotrexate and leucovorin (206, 207, 374, 432, 468, 565, 567, 569). In MRP1, the majority of these residues are located in a region of TM17 predicted to reside in the inner leaflet of the membrane and at the membrane-cytosol interface (Fig. 9). Although conservative or nonconservative substitutions of some residues, such as Tyr1236 and Thr1241, affect transport of only certain substrates, conservative mutations of others, such as Tyr1243, Thr1242, and Trp1246, alter transport of multiple, structurally unrelated compounds. Whether this is because the latter residues make specific contacts with a wide array of substrates or because their mutation perturbs the architecture of the drug-binding pocket is not yet known. The highly conserved Trp1246 in MRP1 may also play a critical role in defining the position of the helix-cytosol interface, possibly by forming stabilizing pi-cation interactions with Arg1249 at position i + 3 (207, 410, 468). Surprisingly, most mutations in TM17 of MRP1 (defined as residues 1228–1248) have a negligible effect on the transport of LTC4. For example, mutation of Trp1246 to Cys, Ala, Phe, or Tyr eliminated E217βG and NNAL-O-glucuronide transport, resistance to natural product drugs, and binding of the GSH-dependent inhibitor LY475776, but had little effect or no effect on LTC4 transport (207, 281, 327). Thus residues in TM17 appear to be critically involved in the recognition and transport of various natural product drugs and glucuronide conjugates, but not for the high-affinity binding or transport of LTC4. Consistent with this notion, transport studies have shown that unmodified natural product drugs compete more effectively for E217βG transport than for LTC4 transport, despite the fact that the two organic anion conjugates compete reciprocally with each other (298, 299, 349). More direct evidence is provided by the recent demonstration that TM17 (residues 1228–1248) of MRP1 can be replaced by a stretch of Ala residues without loss of LTC4 transport activity (23). However, LTC4 transport by the TM17-poly(Ala) mutant is no longer inhibitable by E217βG. Nevertheless, it should be noted that when residues in the COOH-proximal portion of the TM17 α-helix that extends beyond position 1248 into the cytoplasm (e.g., Arg1249) are mutated, loss of LTC4 binding and transport is observed (410, 468), consistent with such residues having a role in maintaining an active conformation of the substrate binding pocket of MRP1 rather than interacting directly with its substrates.

FIG. 9.

Comparison of the effects of mutating comparable polar amino acid residues in transmembrane (TM) 17 of human MRP1 and MRP3 on substrate specificity. Illustrated are the positions of comparable polar residues in TM17 of human MRP1 and MRP3 and the effects of their mutation on the ability of the proteins to transport or confer resistance to various organic anions and drugs. The model of TM17 of MRP1 is derived from the energy-minimized model of MSD1 and MSD2 described in text (sect. xiA) and Figure 10 (52). The helix was then “mutated” at four specific locations (Thr-1242, Tyr-1243, Leu-1247, Val-1248). Triton version 3.0 was used to create an approximate model of MRP3 TM17. PyMol was used to highlight and color the specific mutations on the two helices. As can be seen from the inset, which shows the predicted position of TM17 in the plasma membrane, the residues that influence substrate specificity are predominantly in the predicted inner leaflet regions of the helices. Of particular note are the substantial differences between the two proteins with respect to residues involved in conferring resistance to VP-16 (etoposide) and in transport of E217βG.

In contrast to TM17, TM6 of MRP1 is relatively poorly conserved among ABCC family members. In addition, the corresponding TM1 in P-gp/MDR1 has not been implicated by mutagenesis studies to be important for the activity of this drug transporter. However, TM6 of MRP1 contains significantly more polar residues than does TM1 of P-gp/MDR1, and consequently, it was proposed that this might favor interaction of MRP1 with its more hydrophilic substrates (457). Consistent with this idea, several charged residues in TM6 of MRP1 have been demonstrated to be critical for overall transport activity in one case (Asp336) and in particular, for LTC4 binding and transport in two others (Lys332 and His335) (159, 160). Thus both conservative and nonconservative substitutions of Asp336 eliminate binding and transport of a broad range of organic anions. In contrast, comparable substitutions of Lys332 cause a selective loss of LTC4 transport by MRP1. GSH transport is also reduced, but transport of E217βG, methotrexate, and estrone sulfate is unaffected. A similar substrate-selective loss of transport activity is observed when His335 (which is on the same face of the TM6 helix as Lys332) is mutated to Glu, Gln, or Leu. Confirming the critical importance of TM6 for binding and transport of LTC4 by MRP1 is the recent observation by Bao et al. (23) that replacement of TM6 (amino acids 320–337) with a poly(Ala) chain, in contrast to a similar replacement of TM17, eliminates LTC4 transport.

Finally, as mentioned previously, six single amino acid substitutions have been described thus far which substantially reduce or eliminate expression of MRP1, at least in mammalian cells. These residues are located throughout the primary structure of the protein. They include nonconservative substitutions of Trp142 predicted to be in TM4 of MSD0 (241), Asp430 at the membrane-cytosol interface of TM8 in MSD1 (160), Asp792 in NBD1 (84) and in MSD2, Pro1113 in the extracellular loop connecting TM14 to TM15 (242), and Arg1202 and Glu1204 at the membrane-cytosol interface of TM16 (468). In some cases, culturing the transfected cells harboring the mutant MRP1 at reduced temperatures (conditions where the proofreading machinery for monitoring protein folding is less stringent) restores MRP1 protein expression levels. This suggests that the mutations may cause misfolding and destabilization of the transporter leading to enhanced degradation, as observed for some mutant CFTR proteins (242, 468). In other instances where MRP1 protein expression is abrogated by opposite charge substitutions, expression is unaffected by substitutions with more neutral or same-charge residues. However, the expressed mutants are not necessarily active. For example, MRP1 mutants in which Arg1202 in TM16 is replaced with Lys, Gly, or Leu are expressed but not when the substituting amino acid is Asp. Similarly, the Lys substituted mutant of the nearby Glu1204 is poorly expressed while mutants with neutral (Leu) and same-charge (Asp) substitutions at position 1204 are expressed at levels comparable to wild-type MRP1. However, the expressed Arg1202 mutants possess wild-type MRP1 transport activity while the neutrally substituted Glu1204 mutant is inactive, despite the fact that it retains the ability to bind LTC4. The Glu1204 mutant also exhibits a dramatic increase in vanadate-induced trapping of ADP at NBD2, indicating that the catalytic activity of the transport is compromised (468). The distinct phenotypes associated with mutations of the highly conserved Arg1202 and Glu1204 are presumably caused by perturbations in the α-helical geometry of TM16 that contribute (depending on the substituting amino acid) to misfolding of MRP1 and, in the case of the neutral Glu1204 Leu mutant, disruption of the signaling between the TMs that comprise the substrate translocation pathway through the membrane and NBD2.

D. Is Amino Acid Sequence Conservation Predictive of Substrate Specificity?

What is the implication of the sequence conservation of regions such as TM17 with respect to the ability of MRP1, -2, and -3 to transport common substrates? Although TM17 of MRP1 and MRP3 are identical in 19 of 21 amino acids, mutation of conserved polar residues has strikingly different effects on the overall activity and substrate specificity of the two proteins (207, 374, 567, 569). For example, elimination of the hydrogen bonding capability of polar residues, Tyr1232, Ser1229, Ser1231, and Ser1233, predicted to be in the outer leaflet region of TM17 in MRP3 had major effects on both overall transport activity and specificity (569). Despite the complete conservation of this region in MRP1, the only mutation that affected function was the conversion of Tyr1236 to Phe that selectively decreased resistance to vincristine (567) (Fig. 9). Similarly, mutation of the conserved Trp residue predicted to be at the membrane cytoplasm interface of TM17 in MRP1, -2, and -3 had different effects on the transport of the common substrate E217βG (206, 207, 374). Conservative and nonconservative substitutions of this amino acid in MRP1 and MRP3 decreased and increased transport of the conjugated estrogen, respectively, while in MRP2, only nonconservative substitutions had an effect. Thus, even in highly conserved regions that are clearly important for substrate recognition and transport, the correlation of structure and function is complex and cannot be predicted on the basis of primary amino acid sequence.

A number of amino acid residues in TM11 of MRP1 are also involved in determining the overall activity and substrate specificity of the protein (52, 160, 242, 570). As found with TM17, most of these mutation-sensitive residues are located within the predicted inner leaflet region of the helix or at its cytoplasmic interface. However, unlike TM17, mutations of at least four residues, Asn590, Arg593, Phe594, and Pro595, affected LTC4 binding and transport, as well as the transport of other substrates (52, 160, 242, 570). Conservative substitutions of Phe594 with Tyr or Trp had selective effects on substrate specificity, suggesting that it may be involved in direct interaction of MRP1 with its substrates (52). Nonconservative (Glu) and neutral (Leu) substitutions of Arg593 reduced transport of multiple substrates, whereas this was not the case for the Lys-substituted mutant (160), indicating the positive charge at this position likely plays a role in maintaining the architecture of the substrate binding pocket. Similarly, only Ala substitution of Asn590 (a “cavity”-creating substitution) adversely affected MRP1 activity, while replacing this residue with Asp or Gln had no effect, suggesting that the polar side chain of Asn590 may be involved in interhelical interactions that influence the conformation of the protein in the vicinity of the binding pocket (570).

In addition to TMs 11 and 17, the substrate specificity of MRP1 is affected by mutation or naturally occurring variation of residues in several other TM helices, including TMs 6, 7, 8, 9, 10, 14, 15, and 16 as well as some of the TMs connecting cytoplasmic loops (most notably CL7 between TM15 and TM16) (76, 77, 160, 169, 241, 242, 411, 468, 565, 566, 568). Mutational studies of MRP2, although relatively limited, have also identified functionally important residues in TMs 6, 11, 14, and 16 (208, 432) in addition to TM17 (206). For example, mutation of the Arg586 in TM11 of MRP2 was found to selectively decrease transport of GSH conjugates (208). However, this was not the case for the corresponding Arg593 in TM11 of MRP1 where a comparable mutation affected the transport of a broad range of organic anion substrates as well as binding of LTC4 (160).

One of the most striking examples of a major alteration in substrate specificity resulting from single amino acid variation came from the functional characterization of mammalian MRP1 orthologs. MRP1 is relatively highly conserved among mammals, and the human protein exhibits 88, 86, 92, and 98% sequence identity with the mouse, rat, dog, and macaque proteins, respectively (144, 314, 371, 481). However, with the exception of macaque MRP1, the other orthologs fail to confer resistance to anthracyclines and are poor transporters of E217βG (144, 314, 371, 482). The lack of anthracycline resistance has been traced to the presence of a Gln rather than Glu residue in TM14 (Glu1086 in human MRP1), while the poor E217βG transport seems attributable in large part to the presence of Ala rather than Thr in TM17 (Thr1242 in human MRP1) (565, 566). Surprisingly, the common ability to confer resistance to vincristine and VP-16 was lost following reciprocal substitution of the Glu or Gln residue in TM14 of the human and murine proteins, respectively (565). However, resistance could be rescued by a second mutation of the Thr or Ala residue in TM17 that restored the “pairing” of Glu and Thr or Gln and Ala. Thus an apparently conserved function of the human and mouse proteins depends on the pairing of two nonconserved amino acids. Residues in TM14 of MRP2 and MRP3 have also been implicated in determining the substrate profiles of both proteins (209). MRP3 differs from both MRP1 and MRP2 in its ability to transport monovalent bile salts such as taurocholate (184). Mutation of Arg1096 in TM14 of rat Mrp2 to Leu, as found at the corresponding position of rat Mrp3, resulted in acquisition of the ability to transport taurocholate, while introduction of a basic amino acid in place of Leu in Mrp3 resulted in a loss of both taurocholate transport and transport of the common substrate E217βG (209).

Certain types of amino acids have been systematically mutated in MRP1 (and to a lesser extent in MRP2), because the chemical and physical properties of their side chains suggest that they are likely to be involved in the overall architecture and flexibility of the translocation pathway of the protein and/or in direct interactions of the transporter with its substrate(s). Pro residues have the potential to kink α-helical elements and, because of their low intrinsic hydrogen bonding capacity, may increase the flexibility of TMs (439). As a consequence, Pro hinges and other features of TM helix geometry may be important in the structural responses to conformational changes that occur during the binding and transport of MRP1 substrates (210, 242). When the 12 Pro residues in MSD0 and CL3 were individually replaced with Ala, relatively few substantial changes in MRP1 organic anion transport activity were observed (210). In contrast, 14 of the 18 Pro residues in MSD1 and MSD2 were mutation sensitive (242). Interestingly, while there are 9 Pro residues in each of the second and third MSDs, their distribution is quite asymmetric. Thus the majority of Pro residues in MSD1 occur in the predicted TMs, whereas this is not the case in the third MSD. Moreover, the Pro residues in the COOH-proximal MSD are significantly more conserved among ABCC family members than are those in MSD1.

Consistent with their predicted importance in maintaining the structure of a functional transporter, single Ala substitutions of seven of nine Pro residues in MSD1 and five of nine in MSD2 reduced transport of at least some organic anion substrates by 50% or more (242). The mutation-sensitive Pro residues in MSD1 were all found in the predicted TMs, whereas this was not the case for MSD2. One particularly interesting phenotype was observed after Ala substitution of the highly conserved Pro1150 that is located in the cytoplasmic loop (CL7) connecting TM16 to TM17. This mutant exhibited a substantial increase in E217βG and methotrexate transport, while transport of other organic anions was reduced or unchanged. The increase in E217βG transport by the Pro1150 mutant was not only associated with an increase in uptake affinity (Km) but also with an increase in the apparent affinity of the mutant transporter for ATP. Orthovanadate-induced trapping of ADP by the mutant protein was also dramatically reduced, indicating the ability of NBD2 to hydrolyze ATP is significantly compromised. Together, these observations suggest that the structural integrity of CL7 is important for direct interactions of at least some substrates with the transporter, as well as coupling transport with conformational changes in NBD2 necessary for its catalytic activity (242). It is worth noting that mutations of some of the conserved Pro residues in CFTR/ABCC7 and MRP6 ABCC6 are associated with cystic fibrosis and PXE, respectively, indicating that the functional importance of these residues is apparently conserved.

The side chains of aromatic and polar aromatic amino acids are potentially capable of establishing a variety of bonding interactions with various MRP1 substrates, as well as influencing the position of TM helices in the membrane through a variety of inter- and intrahelical interactions and interactions with membrane phospholipids (52, 77, 206, 207, 241, 289, 374). Trp residues occur at a greater frequency in most ABCC family members than they do in transporters belonging to other ABC subfamilies, including P-gp/MDR1. Furthermore, mutation of all 11 Trp residues of P-gp/MDR1 revealed that none of them is essential for the transport function of this protein (260). In contrast, of the 30 Trp residues in MRP1, mutations of 6 of them, which are moderately to highly conserved and are found in or near TMs (including Trp1246 in TM17 described earlier), result in substantial changes in transport activity or substrate specificity (241). Thus Ala substitution of Trp445 (TM8), Trp553 (TM10), and Trp1198 (TM16) eliminated or dramatically reduced transport levels of a broad range of organic anion substrates. On the other hand, Ala substitutions of Trp361 (TM7) and Trp459 (TM9) caused more moderate alterations. More conservative substitutions of Trp445, Trp553, and Trp1198 resulted in substrate-selective retention of transport in the case of Trp445 and Trp1198 but not Trp553 in TM10. It would appear that while these mutation-sensitive Trp residues may play a role in determining the position of certain TM helices in MRP1, the side chains of each of these residues also contribute in their own distinct way to the transport activity and substrate specificity of the transporter.

Finally, as already alluded to above, systematic substitutions of charged amino acids have implicated a significant number that are important for MRP1 activity. These mutation-sensitive ionizable residues are distributed throughout the primary structure of the protein, and the mutant phenotypes observed are quite variable. Thus the phenotypes associated with nonconservative mutations of basic and acidic residues in MRP1 range from a total loss of protein expression (e.g., Asp430, Asp792, Arg1202, Glu1204) (84, 160, 468) to a substrate-selective loss (e.g., Lys332, Arg433) (76, 159, 160) or general loss of substrate binding and transport activity (e.g., Asp336, Arg1197, Arg1249) (159, 160, 410, 468) to a loss of transport activity but retention of substrate binding (e.g., Asp1084, Glu1204) (468, 568). At present, it is not possible to predict the phenotype that will ensue from mutation of a charged residue based on its location in the primary structure of MRP1 or its degree of conservation among other ABC proteins. This should change as more information about the three-dimensional structure of the transporter becomes available.

Studies of charged residues in MRP2 and MRP3 are not as extensive as those of MRP1 and, since many of the substrates of these transporters are anionic, have focused primarily on the role of basic amino acids. Mutation of Lys and Arg residues in TMs 6, 11, 13, 14, 16, and 17 of MRP2 that are conserved either in MRP1 and MRP3, or only in MRP1, identified Lys325 (TM6) and Arg586 (TM11) as being important for transport of glutathione conjugates such as LTC4, but not compounds conjugated with glucuronide or sulfate (208). Thus, as observed with MRP1, mutations identified to date that affect transport of GSH conjugates appear to primarily involve TMs in MSD1. Overall, the data are consistent with the notion that the binding pocket contains regions that interact primarily with the anionic moieties of conjugated substrates (or GSH) and others are involved in binding the hydrophobic components of the substrate.


A. Molecular Modeling

In the absence of a high-resolution crystal structure, it is often challenging to interpret the results of mutational studies. This is because of the difficulty of ascertaining whether a specific mutation directly affects molecular contacts with substrate or alters specificity indirectly, either by changing the architecture of the substrate binding pocket or affecting the ability to undergo the conformational changes involved in substrate transport. In lieu of a high-resolution crystal structure, atomic models of the tertiary structure of the two core MSDs of MRP1 have been derived by homology modeling and molecular dynamics simulations and used to indicate the possible disposition of residues in the protein that have been shown to affect substrate specificity and overall transport activity (52). These atomic models are based on the crystal structures of the bacterial lipid transporter MsbA from V. cholera and E. coli, as well as a model of P-gp/MDR1 that has incorporated structural data derived from various sources, including cysteine-scanning mutagenesis (55, 56, 478). Because MsbA and P-gp/MDR1 share a relatively high level of homology, P-gp/MDR1 was modelled initially on the MsbA structure. The derived P-gp/MDR1 model was then manipulated to accommodate electron microscopic data obtained from two-dimensional crystals and single particles, as well as disulfide cross-linking data from cysteine scanning mutagenesis (305308, 422, 424). Although the sequence similarity between MsbA and MRP1 is low by comparison with P-gp/MDR1, it was possible to identify two regions of homology, corresponding to residues 325–596 in MSD1 and 1019–1249 in MSD2 (52). These regions were used to derive energy-minimized models of MSDs 1 and 2 of MRP1 (Fig. 10). Thus the structure lacks MSD0 and CL3, as well as some other CL loop regions and the NBDs and the COOH-terminal region of the protein.

FIG. 10.

Three-dimensional cartoon of MRP1. The figure is based in part on the energy minimized model of MSD1 (blue) and MSD2 (green) of MRP1 described in Reference 52. NBD1 (blue) was modeled by threading onto the crystal structure of NBD1 of CFTR (290), and NBD2 (green) was threaded onto HlyB (449). The threaded NBDs were then aligned with the crystal structures of the NBD dimer of MJ0796 and the structure of MsbA from Vibrio cholerae (472). Walker A, Walker B, and the C signature are shown in red, yellow, and orange, respectively. The 5 TM helices of MSD0 are depicted in red. However, the tertiary structure of MSD0 and the manner in which it interacts with the remainder of the protein is completely unknown.

The best models predict that several polar aromatic residues previously shown by mutational studies to influence substrate specificity (Trp553 TM10, Trp1198 TM16, and Trp1246 TM17 ) (207, 241, 567) are located close to the membrane cytosol interface with their side chains projecting toward a chamber formed by the TM helices of the two MSDs. Based on the model, Phe594 (TM6) was predicted to be located in the inner leaflet region with its side chain also projecting into the chamber (52). The predicted position of the Phe residue suggested that it might be a component of an “aromatic basket” formed together with the previously identified Trp and Tyr residues. Indeed, mutation of Phe594 to Ala drastically reduced transport of four different substrates tested and eliminated photolabeling by LTC4 (52). Conservative substitutions with Trp or Tyr on the other hand differentially affected transport of specific substrates. Thus mutation to Trp decreased transport of E217βG and methotrexate but not estrone sulfate and GSH, while the Tyr mutation had the opposite effect. The selective alterations in transport resulting from the two conservative mutations suggest that Phe594 may interact directly with at least some substrates and are consistent with the possibility that the entrance to the translocation pore may be defined by a ring of aromatic residues. The model also predicts that the majority of residues in TMs 6, 11, 14, 16 and 17, shown by mutational studies to influence substrate specificity, are located on faces of their respective helices that form the lining of the putative translocation pathway through the membrane. Thus the model is consistent with the possibility that substrates establish interactions primarily with amino acid side chains lining the inner leaflet region of the translocation pathway. Although in agreement with data derived from a number of mutagenesis studies, the models remain to be validated by other experimental approaches such as cysteine-scanning mutagenesis and high-resolution studies of the three-dimensional structure of the protein in nucleotide and substrate-bound states.

B. Electron Crystallographic Studies

Electron crystallographic studies of P-gp/MDR1, CFTR, and MRP1 have provided three-dimensional structures with resolution limits of 8, 20, and ∼22 Å, respectively (423, 426, 427). All three proteins display some twofold pseudosymmetry with a ring of protein surrounding an electron-dense cavity or barrel of variable size and shape, which is presumed to correspond to the translocation pore. The structure of MRP1 is presently less refined than those of the other two proteins, but analysis of single particles revealed some deviation from twofold pseudosymmetry in the form of two small electron-dense regions on the periphery of the ringlike structure (423). One of these regions may be attributable to the third NH2-terminal MSD. However, individual TM helices are not resolvable, and the current images of MRP1 provide no information about the manner in which the TM helices of MSD0 may be packed, or how they might interact with the core of the protein. Unlike P-gp/MDR1 and CFTR, the two-dimensional crystals of MRP1 had a unit cell that consisted of a protein dimer (423). However, this arrangement may be a crystallization artifact, and although some previous biophysical and biochemical studies of MRP1 and CFTR have suggested that their native, functional unit is a dimer (115, 399, 475, 564), the evidence remains inconclusive for both proteins.


The discovery and characterization of the MRP family of transporters has dramatically increased appreciation of the potential for energy-dependent efflux pumps to contribute to clinical drug resistance encountered in the treatment of a number of diseases. It has also served to illustrate the challenges of, and the opportunities for, circumventing this resistance. From a broader perspective, members of the MRP family are clearly involved in protection of many tissues and organs against a vast array of xenobiotics and their metabolites. The proteins are also major determinants of the distribution and disposition of both physiological and pharmacological substrates in the body. Consequently, interest in the pharmacological implications of naturally occurring genetic variations or polymorphisms in these proteins is increasing rapidly (78) (see At a more fundamental level, studies of the MRPs have contributed significantly to our understanding of sometimes subtle but significant variations in the mechanisms by which ABC proteins couple ATP binding and hydrolysis to substrate translocation across the lipid bilayer. Such studies have also provided insights into the evolutionary mechanisms involved in generating “substrate diversity” among these related proteins, while retaining a level of redundancy that appears to be an essential characteristic of many biological systems. With the increasing success in determining the crystal structures of prokaryotic ABC transporters, there is reason to be optimistic about achieving a similar objective for one or more of the mammalian ABCC proteins in the not too distant future. This will allow the integration of a wealth of biochemical information available from existing structure-function studies into a coherent picture of the molecular structure and dynamics of these fascinating transport proteins.


Address for reprint requests and other correspondence: R. G. Deeley, Cancer Research Institute, 10 Stuart St., Suite 300, Queen’s University, Kingston, Ontario, Canada, K7L 3N6 (e-mail: deeleyr{at}