Subfamily C of the human ABC (ATP-binding cassette) superfamily contains nine proteins that are often referred to as the MRPs (multidrug-resistance proteins). The ‘short’ MRP/ABCC transporters (MRP4, MRP5, MRP8 and ABCC12) have a typical ABC structure with four domains comprising two membrane-spanning domains (MSD1 and MSD2) each followed by a nucleotide-binding domain (NBD1 and NBD2). The ‘long’ MRP/ABCCs (MRP1, MRP2, MRP3, ABCC6 and MRP7) have five domains with the extra domain, MSD0, at the N-terminus. The proteins encoded by the ABCC6 and ABCC12 genes are not known to transport drugs and are therefore referred to as ABCC6 and ABCC12 (rather than MRP6 and MRP9) respectively. A large number of molecules are transported across the plasma membrane by the MRPs. Many are organic anions derived from exogenous sources such as conjugated drug metabolites. Others are endogenous metabolites such as the cysteinyl leukotrienes and prostaglandins which have important signalling functions in the cell. Some MRPs share a degree of overlap in substrate specificity (at least in vitro), but differences in transport kinetics are often substantial. In some cases, the in vivo substrates for some MRPs have been discovered aided by studies in gene-knockout mice. However, the molecules that are transported in vivo by others, including MRP5, MRP7, ABCC6 and ABCC12, still remain unknown. Important differences in the tissue distribution of the MRPs and their membrane localization (apical in contrast with basolateral) in polarized cells also exist. Together, these differences are responsible for the unique pharmacological and physiological functions of each of the nine ABCC transporters known as the MRPs.
Cellular efflux of xenobiotics is often mediated by members of the ABC (ATP-binding cassette) superfamily of transmembrane proteins which use the energy of ATP binding and hydrolysis to perform their functions. Three mammalian ABC proteins, P-glycoprotein (ABCB1), ABCG2 [also known as BCRP (breast cancer-resistance protein)] and MRP1 (multidrug-resistance protein 1) (ABCC1), were originally discovered because of their ability to confer resistance to anti-neoplastic agents in cultured tumour cells . In recent years, however, these and other xenobiotic-transporting ABC proteins have also been shown to be important determinants of the disposition (tissue distribution) and elimination of many clinically relevant drugs. Thus the ABC drug-efflux proteins contribute to the pharmacokinetic profiles, efficacy and toxicity (side effects) of a large number of therapeutic and diagnostic agents, as well as chemicals found in the environment and diet. Equally, if not more importantly, many of these drug-transporting ABC proteins also mediate the cellular efflux of a broad range of physiological metabolites with pivotal roles in signalling pathways in the cell.
Phylogenetic analyses have divided the 48 human ABC genes into seven subfamilies: A–G. Subfamily C contains nine drug transporters that are often referred to as the MRPs or ABCC proteins (Figure 1A and Table 1) . The MRP/ABCC proteins are found throughout Nature, including in plants, marine organisms and unicellular eukaryotes where they carry out many important functions. However, this chapter focuses on human MRPs with reference to their mammalian orthologues where relevant.
Of the nine MRP/ABCC proteins, four of them, i.e. ABCC4, ABCC5, ABCC11 and ABCC12 (the ‘short’ MRP4, MRP5, MRP8 and ABCC12 respectively), have a typical ABC structure with four domains comprising two MSDs (membrane-spanning domains) (MSD1 and MSD2), also known as TMDs (transmembrane domains), each followed by an NBD (nucleotide-binding domain) (NBD1 and NBD2) (Figure 1B). ABCC1, ABCC2, ABCC3, ABCC6 and ABCC10 (the ‘long’ MRP1, MRP2, MRP3, ABCC6 and MRP7 respectively) have five domains with the extra domain, MSD0, located at the N-terminus of these transporters. Current evidence supports a topology model where MSD1 and MSD2 [which each contain six TMHs (transmembrane helices)] form the translocation pathway through which substrates cross the membrane (Figure 1C). It also supports a model where the two NBD proteins associate in a head-to-tail orientation to form a ‘sandwich dimer’ that comprises two composite NBSs (nucleotide-binding sites) [3,4] (Figure 1C). These NBSs are each composed of the highly conserved Walker A and Walker B motifs from one NBD and the ABC active transport signature motif from the other NBD, and together are responsible for the binding and hydrolysis of two molecules of ATP. The NBDs of ABC proteins are generally very highly conserved; however, the NBDs of the ABCC/MRP-related proteins contain some sequence variations that are likely to be responsible for some of the differences in how members of this subfamily of transporters interact with ATP and the products of its hydrolysis [5,6].
Substrate translocation for the MRPs, as for all mammalian ABC proteins, is a multistep process which begins with substrate recognition and binding to a high-affinity conformation of the transporter on one side of the membrane, followed by conformational changes in the transporter that allow substrate translocation across the membrane and then substrate release on the opposite side of the membrane (Figure 2A). The transporter then once again assumes its high-affinity conformation so that another round of transport can take place. These processes are obligatorily coupled to the binding and hydrolysis of ATP, as well as the release of ADP/Pi, although the precise details of this coupling have not yet been fully elucidated .
A large number of molecules are transported across membranes by the MRPs (Figure 2B and Table 2). Some of the MRPs share a limited degree of overlap in substrate specificity (at least in vitro), but differences in transport kinetics are often substantial. Important differences in the tissue distribution of the MRPs and their membrane localization (apical in contrast with basolateral) in polarized epithelial and endothelial cells also exist (Table 3). Together, these differences are responsible for the unique pharmacological and physiological functions of each of the individual MRP transporters.
The long MRPs
MRP1 (gene symbol ABCC1) was the first of the drug-transporting ABCC proteins to be cloned during investigations into the cause of multidrug resistance in human lung carcinoma cells . The ABCC1 mRNA expressed in these lung cancer cells was predicted to encode a 1531-amino-acid 170 kDa protein. However, in mammalian cells, the protein is both N-glycosylated and phosphorylated, and thus the mature transporter typically exhibits an electrophoretic mobility of ~190 kDa. Depending on the cell type in which MRP1 is expressed, its corresponding protein band in gels is often quite diffuse, owing to heterogeneity of glycosylation.
Pharmacological and physiological roles of MRP1
It is well established that, in vitro, MRP1 can confer resistance to many widely used anti-neoplastic drugs. These include conventional cytotoxic agents such as doxorubicin, vincristine, etoposide and methotrexate, as well as some of the newer ‘targeted’ agents that modify various signal transduction pathways (e.g. tyrosine kinase inhibitors) [6,7]. MRP1 mRNA and/or protein has also frequently been detected in tumour samples from patients, but its overall contribution to clinical resistance is still not well defined. This is also true of the related P-glycoprotein and ABCG2 transporters, and is largely due to the well-documented difficulties of accurately measuring the amount and activity of the transporters in tumour compared with normal tissues in rigorously designed clinical trials. Nevertheless, MRP1 is currently considered to be the most clinically relevant of the MRPs with respect to drug resistance in cancer, a major obstacle to successful chemotherapy. In addition to tumour cells, MRP1 contributes to drug and xenobiotic disposition in normal cells and thus one of its important roles is that of tissue defence .
MRP1 mRNA and protein is found in all organs, but at relatively higher levels in certain tissues (e.g. testes or lung) (Table 3). The tissue distribution of MRP1 is consistent with its role in limiting the penetration of certain cytotoxic agents at a number of blood–organ interfaces and thus MRP1 contributes to so-called pharmacological sanctuary sites in the body, such as the blood–brain barrier and the blood–testis barrier . The most direct evidence for a protective role for MRP1 has come from studies of mice in which the Abcc1 gene has been disrupted. Although these knockout mice are viable and fertile, they display increased chemosensitivity in certain tissues such as the seminiferous tubules, the intestine, the oropharyngeal mucosa and the choroid plexus [8,9].
Shortly after the demonstration that MRP1 could confer multidrug resistance, it was discovered that MRP1 was also a high-affinity transporter of the pro-inflammatory cytokine LTC4 (leukotriene C4) . LTC4 is an arachidonic acid derivative conjugated to glutathione (GSH) that is involved in asthmatic and allergic reactions. It is exported from cells after its synthesis and, together with its metabolites leukotriene D4 and leukotriene E4, it exerts its biological effects by acting on CysLT (cysteinyl leukotriene) receptors present on the surface of various target cells. Abcc1−/− mice display a diminished inflammatory response consistent with a defect in LTC4 efflux, providing confirmatory in vivo evidence that a key physiological role of MRP1 is to mediate LTC4 export .
In addition to LTC4, MRP1 transports many other structurally diverse GSH-conjugated organic anions, some of which are endogenous metabolites and others which are the products of Phase II xenobiotic metabolism [1,7] (Table 2). Thus it seemed that MRP1 was (at least one of) the ubiquitous ATP-dependent GSH conjugate efflux pumps proposed previously by Ishikawa . However, MRP1 is much more versatile than this because it can also transport organic anions conjugated to glucuronate and sulfate. Again, some of these are endogenous metabolites (e.g. oestradiol glucuronide, oestrone sulfate), whereas others are conjugates of xenotoxins such as the glucuronide conjugate of the tobacco-specific carcinogen NNAL [4-(methylnitrosoamino)-1-(3-pyridyl)-1-butanol] . Despite its ability to transport a range of important physiological metabolites, there are no known human diseases associated with a lack of functional MRP1.
Many compounds have been identified as substrates of MRP1 (and other MRPs) using an in vitro assay that measures the ATP-dependent uptake of a radiolabelled form of the compound into inside-out membrane vesicles prepared from cells overexpressing MRP1  (Figure 3A). Such vesicular uptake assays avoid the technical difficulties of quantifying ATP-dependent efflux of hydrophilic molecules (the preferred substrates of the MRPs) from intact cells. Nevertheless, it should be noted that the in vivo relevance of vesicular transport of many molecules shown in vitro to be MRP substrates has not yet been established. The Abcc1−/− mice provide an extremely important, but not ideal, model for in vivo substrate testing because of the marked differences in the substrate specificity of MRP1 from primates and non-primates . Thus, whereas MRP1 from humans and macaque monkeys can transport anthracycline drugs (e.g. doxorubicin) and oestradiol glucuronide, this is not the case for MRP1 from rat, mouse, dog or cow, despite the fact that the amino acid sequence similarity is >90% for all species.
Finally, an unusual aspect of MRP1 is its complex interactions with the reducing tripeptide GSH (γ-Glu-Cys-Gly) . Among the many functions of this cellular antioxidant is its critical role in protecting cells from the deleterious effects of oxidative stress. GSH is also required for Phase II xenobiotic metabolism to form hydrophilic GSH conjugates which are then exported by MRP1 (or MRP2). GSH itself is a low-affinity substrate of MRP1 (and MRP2), whereas the pro-oxidant glutathione disulfide (GSSG) is a relatively higher-affinity substrate , although the consequences of these transport activities on redox homoeostasis are still not fully understood. Of relevance, however, is the observation that GSH levels in some tissues of Abcc1−/− mice are elevated 2-fold . Recent reports also suggest that GSH efflux in response to cell damage as part of an apoptotic signalling pathway may be mediated by MRP1 .
Some xenobiotics, including verapamil or bioflavonoids such as apigenin, stimulate MRP1-mediated GSH efflux without being transported themselves [7,15]. In contrast, the Vinca alkaloid vincristine markedly enhances the transport of GSH (and vice versa), and thus, in this case, GSH appears to be co-transported with (or cross-stimulates transport of) the drug without conjugate formation. MRP1-mediated transport of the conjugates NNAL-O-glucuronide and oestrone sulfate is also enhanced by GSH; however, in contrast with vincristine transport, GSH only stimulates the process and is not itself transported [1,7]. The biological activities of GSH are typically attributed to the proton-donating properties of the thiol (SH) group of its central cysteine residue. However, this is not the case for its stimulatory effects on MRP1 activity because non-reducing analogues such as ophthalmic acid or S-methyl-GSH can functionally substitute for GSH . GSH (and some analogues) can cause changes in the conformation of MRP1, but how these changes relate to its transport-stimulating activity is not known .
Structure of MRP1
Most evidence supports a topology of MRP1 comprising 17 TMHs divided among its three MSDs (MSD0, MSD1 and MSD2) (Figure 1B), but the precise arrangement of these domains is not yet known. Structural analyses of MRP1 and other mammalian ABC proteins by X-ray crystallography or other biophysical methods pose a significant challenge, mainly due to the large size of these polytopic membrane proteins which makes it difficult to isolate the large amounts of purified active protein typically required for such studies. Recent moderate-resolution electron microscopy structural studies have been able to resolve the TMHs of MRP1, although their precise arrangement with respect to one another is still unclear . Current models of MRP1 are based on a relatively high-resolution structure of the bacterial (Staphylococcus aureus) ABC transporter Sav1866 in its nucleotide-bound form [3,18]. These models are limited in that they represent only the four-domain core structure of MRP1 because Sav1866 (or any other bacterial transporter) does not contain a domain corresponding to MSD0. In addition, these models reflect only a ‘snapshot’ of one conformation (nucleotide-bound, substrate-free) of the many assumed by MRP1 during its complex transport process.
Molecular determinants of MRP1 substrate transport and protein expression
The technique of site-directed mutagenesis has been used strategically to discover specific regions and amino acids of MRP1 which are critical for: (i) substrate specificity, (ii) proper folding and assembly ensuring stable expression at the plasma membrane, and (iii) coupling of substrate transport to the binding and hydrolysis of ATP [1,7,19].
As might be expected, amino acids important for the substrate specificity of MRP1 are frequently located in the TMHs, particularly those in MSD1 and MSD2 which form the substrate-translocation pathway through the membrane. A notable example of such a mutation-sensitive residue is Lys332 found in the first TMH of MSD1, TMH6 . When this basic amino acid is replaced by either an oppositely charged or neutral amino acid, binding and transport of LTC4 is essentially eliminated, whereas oestradiol glucuronide transport remains unchanged . The amphipathic C-terminal TMH17 also contains a number of polar amino acids important for substrate specificity. For example, substitutions of Trp1246 eliminate oestradiol glucuronide transport and drug resistance, but have little effect on LTC4 transport .
Differences in the substrate and inhibitor selectivity of various MRP1 mutants has led to the conclusion that MRP1 contains at least three classes of substrate/modulator-binding sites: one that requires TMH17-Trp1246 (and probably other polar TMH17 residues), one that requires TMH6-Lys332 (and other ionizable TMH6 residues), and a third that requires neither of these amino acids . It appears likely that that each substrate (or modulator) establishes its own unique set of atomic contacts in a multipartite substrate-binding pocket of MRP1, although it is clear that certain substrates share at least some common binding determinants . Unequivocal identification of these atomic contacts should be forthcoming as the methods for high-resolution structural analyses (including MS) of large membrane proteins continue to improve .
MRP2 was first cloned from rat liver in 1996 using a strategy that took advantage of its sequence similarity to human MRP1 (reviewed in ). Before this, MRP2 was known as cMOAT (canalicular multispecific organic anion transporter), largely based on functional studies of mutant rat strains deficient in the biliary efflux of organic anions such as bilirubin glucuronide . These mutant rats were determined to harbour disabling mutations in Abcc2. The human, rabbit, mouse and canine orthologues were cloned soon thereafter, and, as for all of the MRPs, they show a high degree of amino acid identity with one another (77–83%) . Although MRP2 is similar in size and topology (and presumably structure) to MRP1, it contains additional sequence motifs in the cytoplasmic region linking MSD0 to MSD1, and at the C-terminus of the transporter that confer distinct properties on the protein with respect to its plasma membrane trafficking and its ability to participate in functional interactions with other proteins in the cell . Like MRP1, MRP2 functions as an ATP-dependent organic anion efflux pump. However, MRP2 is distinct from MRP1 and most other MRPs in that it is found exclusively on the apical membrane of polarized epithelial and endothelial cells, predominantly those in the liver, kidney and intestine, where it is particularly well situated to play a role in the elimination, as well as in the oral bioavailability of drugs, xenotoxins and their metabolites  (Table 3).
MRP2 and Dubin–Johnson syndrome
Dubin–Johnson syndrome is an autosomal recessive disorder caused by mutations in the ABCC2 gene. Both nonsense and missense mutations have been described and, in many cases, these result in severely diminished levels or absence of the protein. Individuals with Dubin–Johnson syndrome frequently have increased serum-conjugated bilirubin levels , consistent with the fact that biliary elimination of bilirubin glucuronides (which are MRP2 substrates) is impaired. Although those with Dubin–Johnson syndrome are mostly asymptomatic, neonates can present with cholestasis, and, during pregnancy, overt hyperbilirubinaemia can lead to jaundice in women. The mutant Abcc2 rat strains mentioned above have served as models for the study of this human disease.
Physiological and pharmacological roles of MRP2
In the liver, MRP2 primarily functions to mediate the efflux of bile acids and GSH, which helps in biliary homoeostasis. Many GSH, sulfate and glucuronide conjugates have been identified as MRP2 substrates, some of which are listed in Table 2. Prominent among these are a variety of oestrogen conjugates which suggests a role for MRP2 in sex steroid homoeostasis and/or cytoprotection. Whereas MRP1 and MRP2 both transport oestradiol glucuronide, the kinetics differ substantially . Thus it appears that MRP2 contains two similar, but non-identical, ligand-binding sites: one site from which substrate is transported and a second site (allosteric) that regulates the affinity of the transport site for the substrate. Oestradiol glucuronide appears to bind to both sites, but this is not the case for all ligands that bind to MRP2.
In vitro, MRP2 is similar to MRP1 in its ability to confer resistance to a spectrum of natural product anticancer drugs. Distinct from MRP1, however, MRP2 can confer resistance to platinum-containing drugs [1,6,24,27]. The clinical relevance of these in vitro observations, however, is debatable since MRP2 expression in tumour samples has not been associated with response to chemotherapy or clinical outcome. In contrast with its questionable role in drug-resistant tumours, MRP2 has a well-established role in the biliary elimination of conjugated metabolites of numerous drugs and other xenobiotics (Table 2). Glucuronide conjugates of acetaminophen and diclofenac are substrates of MRP2, as are glucuronide conjugates of several carcinogens, at least in vitro [1,24,28,29].
Structure–function studies of MRP2
As shown for MRP1, mutational analyses of MRP2 point to a particularly crucial role for charged amino acids in the TMHs of MSD1 and MSD2 as determinants of substrate specificity, although the number of such studies is relatively limited . Two highly conserved uncharged residues, Trp1254 at the cytoplasmic interface of TMH17 and Pro1158 in the cytoplasmic loop connecting TMH15 to TMH16, are also important for activity [31,32]. However, the consequences of mutating these amino acids in MRP2 compared with the analogous Trp1246 and Pro1150 in MRP1 (and Trp1242 and Pro1147 in MRP3) are dramatically different [31–33]. Substantial differences in the molecular environments of these residues in the two transporters probably account for the marked differences in the substrate specificities of the mutants. These studies demonstrate that correlation of the structure to the function of the MRPs is complex and cannot be predicted on the basis of primary amino acid sequence alone.
Vectorial (transcellular) transport
As noted for MRP1, most quantitative information regarding the substrate specificity and affinity of MRP2 has been obtained using vesicular transport assays (Figure 3A). Some interspecies differences exist with respect to the substrate specificity and substrate affinity of MRP2 and its non-primate orthologues , and current evidence suggests that animal (at least rat) pharmacokinetic data are often poor predictors of pharmacokinetics in humans. This limitation provided the impetus to develop cell culture systems that more closely approximate polarized cells directly involved in drug disposition (found in intestine, kidney and liver) and restricted distribution to tissue ‘sanctuaries’ (blood–tissue barriers). Such polarized cells have apical and basolateral membranes separated by a tight junction. Drug transporters tend to be asymmetrically localized in these two membranes which facilitates transcellular (i.e. vectorial) transport of a drug or metabolite from the apical to basolateral (and basolateral to apical) side of a cell. Thus double- and multiple-transfected polarized cell lines have been generated which express one or more human uptake transporters [e.g. OATP1B1 (organic anion transporter polypeptide 1B1); gene symbol SLCO1B1] in the basolateral membrane, and efflux transporters human MRP2 (and/or MRP4) in the apical membrane (although uptake transporters can be present here as well) [29,35] (Figure 3B). Such cell lines are proving useful as models of hepatobiliary and renal drug transport.
MRP3 (gene symbol ABCC3) and MRP1 have the highest degree of sequence similarity among the MRPs (Table 1), and, like MRP1, MRP3 is expressed on the basolateral membranes of polarized cells. MRP3 is expressed predominantly in the small intestine, the kidney (distal tubules) and the pancreas (Table 3) . It is also highly expressed in two of the three zones of the adrenal cortex where its function remains unknown. Despite its sequence similarity to MRP1 (and MRP2), MRP3 has a very different (and more limited) substrate profile. The most striking difference is the very low affinity and capacity of MRP3 to transport GSH . Unlike MRP1 (and MRP2), MRP3 also does not require this tripeptide for efflux of the anticancer drug etoposide. Like MRP2, MRP3 displays complex transport kinetics (in vitro) with some of its conjugated substrates. This suggests that the protein contains both a transport-binding site and a modulatory site, although these sites have not yet been defined at the molecular level .
There are no diseases associated with mutations in the human ABCC3 gene and, consistent with this, Abcc3−/− mice are healthy and show no apparent phenotype until challenged with certain pharmacological agents [36,38,39]. Although a helpful model for pharmacological studies (see below), these mice have not yet provided any major insights into the physiological role of MRP3. However, because MRP3 expression on the hepatic sinusoidal membrane is often up-regulated under cholestatic conditions (bile duct ligation in rats and cholestatic disease in humans), it is a commonly held view that MRP3 serves as a ‘backup’ system for the removal of toxic liver metabolites via blood when MRP2-mediated transport into bile is impaired [36,40]. This concept is well supported by studies of Abcc2/Abcc3 double-knockout mice . Thus MRP3 transports its substrates over the basolateral sinusoidal membrane of liver and gut towards the circulation for subsequent elimination in urine.
In an attempt to identify the physiological substrate(s) of MRP3, Borst and colleagues developed a targeted metabolomics approach based on the premise that the abundance of MRP3 substrates (glucuronide conjugates) in plasma and urine should be reduced in Abcc3−/− compared with wild-type mice . Thus plasma and urine from these mice were screened for compounds containing a glucuronic acid moiety by MS. In this way, glucuronide conjugates of several plant-derived dietary phyto-oestrogens were identified as substrates of MRP3 . This unbiased methodological approach for substrate identification may prove useful in studies of the less well characterized ABCC6 and MRP7 (see below).
Relevant pharmacological examples of MRP3 substrates include the glucuronide conjugates of morphine and acetaminophen. The case of morphine is particularly interesting because the two glucuronide conjugates that are formed during its metabolism in humans differ profoundly in their pharmacological activity. Thus morphine 6-glucuronide is a potent analgesic like the parent drug, whereas morphine 3-glucuronide is not. Relative to their wild-type counterparts, Abcc3−/− mice showed increased levels of morphine 3-glucuronide in liver and bile, whereas plasma levels and excretion via the urine were decreased . Associated with these pharmacokinetic differences, the antinociceptive potency of the bioactive morphine 6-glucuronide was also decreased in the Abcc3−/− mice. These observations suggest that, because of its influence on the disposition of morphine and its active metabolite, individual variations in MRP3 expression could contribute to the differences in morphine pharmacokinetics and efficacy observed in human populations .
The human ABCC6 gene is located within 9 kb of the ABCC1 gene on chromosome 16, suggesting that the two genes may have arisen from a gene-duplication event (Table 1). The structural organization of the two genes is similar, and the encoded proteins are also comparable in size and topology. Initial in vitro studies suggested that the 190 kDa ABCC6 protein (which is found on the basolateral membrane of polarized endothelial and epithelial cells) was an organic anion transporter with properties similar to those of MRP1 and MRP3, with the notable exception that it transported glucuronide conjugates (e.g. oestradiol glucuronide) rather poorly. However, despite their genetic proximity and the similarities of their in vitro properties, the in vivo functions of ABCC6 and MRP1 are now known to differ dramatically. Thus, quite unexpectedly, it was discovered that recessive mutations in ABCC6 are responsible for a rare human genetic disorder known as PXE (pseudoxanthoma elasticum) [43,44].
PXE is a late-onset progressive disease characterized by aberrant ectopic mineralization of soft connective tissues such as the skin, as well as in the eyes and the cardiovascular system, with the most devastating complication in many patients being the eventual loss of visual acuity in the third or fourth decade of life. The connective tissue pathology of PXE has been recapitulated in Abcc6−/− mice, as have the cardiac dysfunction and skin manifestations seen in humans afflicted with this disease [44,45]. Well over 300 ABCC6 mutations, including numerous point mutations, deletions and insertions, have been described in patients with PXE, which is presumed to reflect the instability of the ABCC6 region of chromosome 16. Unfortunately, only experimental therapies are presently available to treat this disease.
ABCC6 is expressed predominantly in the liver and kidney, and it is rather curious that the pathology of PXE does not directly affect these tissues. The precise role of ABCC6 in mineralization processes remains obscure, although several intriguing hypotheses are currently being investigated . Most evidence supports the ‘metabolic hypothesis’ which postulates that, in the absence of ABCC6 activity, there is a deficiency of circulating factors or metabolites required to maintain normal mineralization under calcium and phosphate homoeostatic conditions . Recent skin grafting and other experiments utilizing Abcc6−/− mice have been particularly valuable in providing mechanistic insights into the defects underlying PXE . For example, skin from a Abcc6−/− mouse did not develop mineralization when grafted on to a wild-type Abcc6+/+ mouse, but the skin from a wild-type mouse showed mineralization after grafting on to a Abcc6−/− mouse, consistent with the conclusion that circulating factors in the blood of the recipient mouse were regulating the degree of mineralization of the graft, irrespective of the graft genotype.
MRP7 (gene symbol ABCC10) was the last of the long MRPs to be described and remains the least well characterized. Although its general architecture appears similar to that of the other long MRPs, MRP7 is distinct primarily due to substantial sequence divergence in its first MSD (MSD0). As a result, MRP7 lacks the highly conserved N-glycosylation sites found in this region of MRP1–MRP3 and MRP6 [46,47]. Whether or not this difference contributes to any functional differences is not yet known.
The tissue and membrane localization of MRP7 has not yet been fully characterized, in part because of the scarcity of specific high-affinity antibodies. The physiological role(s) of MRP7 is unknown, and Abcc10−/− mice have not yet been described. In vitro, MRP7 transports a broad range of prototypical MRP organic anions (including LTC4 and oestradiol glucuronide), each with its own set of kinetic parameters (affinity and capacity), as do other MRPs. MRP7 can also confer resistance to several classes of natural product anticancer and antiviral drugs, including antimitotic agents (vincristine or docetaxel) and certain nucleoside analogues  (Table 2). The resistance spectrum associated with MRP7 expression is thus narrower than for MRP1 and MRP2, and more similar to that for MRP3 since it does not include resistance to anthracyclines (e.g. doxorubicin). There are some recent reports that certain small-molecule inhibitors of tyrosine kinases (e.g. imatinib) interact with MRP7. Despite these interesting laboratory findings, however, there is still little evidence that supports a role for MRP7 in clinical drug resistance or sensitivity in cancer patients.
The short MRPs
Because they lack the MSD0 characteristic of the long MRPs, the short MRPs (MRP4, MRP5, MRP8, ABCC12) have a more typical ABC structure with just two MSDs and two NBDs (Figure 1B). MRP4 (gene symbol ABCC4) is the best characterized of the short MRPs, and, curiously, its closest homologue is the CFTR (cystic fibrosis transmembrane conductance regulator) (ABCC7), a Cl− channel regulated by cAMP. However, no ion channel activity has ever been ascribed to MRP4. On the other hand, sequence alignments place MRP5 (gene symbol ABCC5), MRP8 (gene symbol ABCC11) and ABCC12 into a separate subcluster of the ABCC/MRP subfamily (Figure 1A). Comparatively little is known about these three ABCC proteins, and, consequently, they are considered together below.
Human ABCC4 is highly polymorphic, although no genetic diseases have been firmly linked with mutations in this gene. However, the absence of MRP4 protein has recently been associated with a selective defect in ADP storage in platelet δ-granules which in turn is associated with prolonged bleeding times and bleeding diathesis . MRP4 is present at low levels in all normal tissues, with substantially higher levels found in the prostate. It has also been implicated in the aggressiveness of some tumours, including prostate tumours and neuroblastoma [50,51].
MRP4 was first functionally identified as a transporter of adefovir, a nucleoside monophosphate antiviral agent . However, uptake studies in MRP4-enriched isolated membrane vesicles and efflux studies using intact MRP4-transfected cells, as well as studies in Abcc4−/− mice, have established that the substrate specificity of this transporter extends well beyond this class of drugs [6,36,38,53]. In addition to its own unique substrate-specificity profile (see below), MRP4 is unusual because of its ability to localize to either basolateral or apical membranes in polarized cells, depending on the tissue where it is found. For example, in prostate tubuloacinar cells and hepatocytes, MRP4 localizes to the basolateral membrane, whereas it is found at the apical membrane in renal proximal tubules and the luminal side of brain capillaries [53,54]. The mechanisms underlying the tissue-specific localization of MRP4 are only partly understood and appear to involve interactions with different adaptor and scaffolding proteins in the different cell types .
Pharmacological and physiological roles of MRP4
In addition to adefovir and other antiviral agents, MRP4 also confers resistance to anticancer agents including thiopurine analogues, methotrexate and topotecan [6,53,56,57]. Thiopurine analogues and most nucleoside-based antivirals require intracellular phosphorylation before they can exert their pharmacological activity. For example, 6-mercaptopurine is metabolized to the monophosphate nucleotide 6-thio-IMP, which in turn inhibits several enzymes of de novo purine nucleotide synthesis. Thus MRP4 confers resistance to thiopurine bases (and by inference nucleoside analogues), by effluxing the anionic phosphate metabolites rather than the parent compounds.
Much of the evidence demonstrating the importance of MRP4 in drug disposition/elimination has come from studies of Abcc4−/− mice. Although these mice exhibit no discernible phenotype in the absence of drugs, topotecan (a topoisomerase I inhibitor) accumulation in both brain tissue and in cerebral spinal fluid is enhanced in Abcc4−/− mice, reflecting the dual localization of MRP4 at the basolateral membrane of the choroid plexus epithelium, and at the apical membrane of the endothelial cells of the brain capillaries . Renal elimination of many drugs is also reduced in Abcc4−/− mice, and a key protective role for MRP4 in the kidney is suggested [53,58]. Common NSAIDs (non-steroidal anti-inflammatory drugs) (e.g. celecoxib) inhibit transport by MRP4 which may contribute to clinically significant kidney toxicity when cytotoxic agents such as adefovir or methotrexate are co-administered with NSAIDs [53,59]. Abcc4−/− mice are also more sensitive to the haemopoietic toxicity of thiopurines . In humans, a non-synonymous polymorphism in ABCC4 encodes an inactive transporter and it may be that the increased sensitivity of some Japanese patients to thiopurines reflects the greater frequency (>18%) of this polymorphism in the Japanese population .
Equally important to its drug (and drug metabolite)-transporting function, MRP4 mediates the cellular efflux of several endogenous metabolites that play key roles in signalling pathways involved in processes such as cell proliferation, differentiation, cell-cycle progression and apoptosis. Notably, MRP4 transports cAMP (and cGMP), molecules that relay external signals to downstream protein kinases [58,60]. Because of the relatively low affinity of MRP4 for cAMP (and cGMP), its relevance in regulating intracellular levels of cyclic nucleotides has been questioned . However, the growing evidence that cyclic nucleotide signalling is highly compartmentalized suggests that MRP4 may be more involved in regulating local microdomain levels rather than whole-cell concentrations of cAMP . Several eicosanoids, including PGE2 (prostaglandin E2), are also substrates of MRP4 [36,59]. Best known as a mediator of pain and inflammation, PGE2 has also been implicated in the development of some tumours as well as the stimulation of their growth and angiogenesis, and response to cytotoxic chemotherapy [62,63].
MRP5, MRP8 and ABCC12
There are no known diseases associated with mutations in ABCC5, and the phenotype of the Abcc5−/− mouse has not yet been reported in the literature. Although ABCC5 mRNA is present at low levels in most tissues, the tissue distribution of the MRP5 protein is only partially characterized, in part due to the limited availability of suitable antibodies for definitive immunolocalization studies [36,54,64] (Table 3). Relatively higher levels of MRP5 have been detected in the brain capillary endothelial cells, pyramidal neurons and astrocytes, as well as in smooth muscle cells of various tissues in the human genitourinary system [54,64]. However, the physiological role of MRP5 in these tissues is unknown. MRP5 has been reported to localize to the basolateral membrane of some polarized epithelial cells, but, in the brain capillary endothelial cells, it localizes to the apical membrane [36,54] (Table 3). It may be that, like MRP4, MRP5 is capable of dual-membrane localization, but there are presently too few data to support this conclusion just yet.
There are some discrepancies in the literature with respect to the reported substrate specificity and affinity of MRP5 which have not yet been resolved . Xenobiotic substrates identified by in vitro studies include methotrexate, the antiviral agent adefovir, and various other nucleoside and nucleotide monophosphate analogues or metabolites [36,65] (Table 2). For example, MRP5 confers resistance to the important anticancer drug 5-FU (5-fluorouracil) by mediating the cellular efflux of its cytotoxic active monophosphate metabolite, 5-dUMP, rather than the parent drug itself .
Physiological metabolites transported by MRP5, at least in vitro, include folic acid and glucuronide conjugates (e.g. oestradiol glucuronide), as well as GSH and GSH conjugates (e.g. LTC4) . Cyclic nucleotides, mainly cGMP, are also transported by MRP5, a property it shares with MRP4 and MRP8 . It has been speculated that MRP5 may act as a cGMP ‘overflow’ pump . Thus it is conceivable that, when intracellular levels of cGMP are elevated either by induction of cGMP biosynthesis or by inhibition of phosphodiesterases, normal cGMP levels could be restored via efflux by MRP5. However, direct evidence supporting this role for MRP5 is still lacking. Similarly, the in vivo relevance of the ability of MRP5 to transport the therapeutic agents mentioned above has not yet been established.
Human ABCC11 is located on chromosome 16q12.1 and, whereas multiple mRNAs are transcribed from this gene, the full-length MRP8/ABCC11 transporter is predicted to be 1382 amino acids with a typical four-domain ABC structure [68,69]. There is no evidence yet to suggest that any of the putative polypeptides encoded by the shorter ABCC11 transcripts would be functional transporters. ABCC11 is unusual among the ABCC family members in that no orthologous genes have been found in mammals, except for primates, and thus Abcc11 is clearly not an essential gene. However, the discovery that a single nucleotide polymorphism in ABCC11, 538G>A, determines the type of earwax (cerumen) secreted by the ceruminous apocrine glands has revealed a physiological function for this transporter. The 538G>A polymorphism causes a non-conservative arginine substitution of Gly180 in the first TMH of MRP8 which appears to disrupt the glycosylation and stability of the protein . The 538A/A genotype corresponds to the dry earwax type common in East Asian populations, whereas the 538G/A and 538G/G genotypes correspond to the wet earwax type more common in populations of European and African origin. Individuals with the 538G/G and 538G/A genotypes also suffer from axillary osmidrosis (armpit secretions of foetid sweat), a condition rarely seen in 538A/A individuals . Thus evidence to date suggests that accumulation of glandular secretions into vacuoles and granules before their release is a physiological function of MRP8. However, the molecular identity of these secretions is not yet known.
In contrast with its closest relative MRP5, full-length MRP8 localizes to apical membranes in stably transfected polarized epithelial cells. ABCC11 mRNA has been found in a variety of tissues, but the multiplicity of ABCC11 transcripts makes it difficult to make any correlations with respect to MRP8 protein expression patterns [68,69]. The MRP8 protein has been detected in axons of neurons in the human central and peripheral nervous systems using immunofluorescence microscopy, and it has been proposed that it may mediate the efflux of neuromodulatory steroids such as dehydroepiandrosterone 3-sulfate .
In membrane vesicles prepared from transfected cells, MRP8 can transport a wide range of compounds including cGMP and cAMP (such as MRP4 and MRP5), as well as conjugated organic anions such as LTC4, and sulfated and glucuronidated steroids (such as MRP1–MRP3) [6,47,71,72] (Table 2). Whether or not any of these molecules are physiological substrates of MRP8 is not known. In transfected cell systems, MRP8 can confer resistance to methotrexate as well as 5-FU by effluxing its active metabolite 5-FdUMP (5-fluoro-2′-deoxyuridine-5′-monophosphate, a property it shares with MRP5 [47,66]. However, to date, there is no convincing evidence that MRP8 plays a role in resistance or sensitivity to these anti-metabolites in cancer patients.
ABCC12 is located on the same chromosome as ABCC11, and the two genes are oriented in a tandem fashion just 20 kb apart, suggesting that they probably arose from a gene-duplication event [68,69]. Also like ABCC11, multiple splicing variants of ABCC12 exist, with the longest full-length mRNA transcript predicted to encode a protein of 1359 amino acids . ABCC12 transcripts are most abundant in testes, but only a minor fraction of these are able to encode a full-length transporter. This profile is generally shared with both the rat and mouse Abcc12 orthologues [68,69]. When full-length human ABCC12 is ectopically expressed in human embryonic kidney cells, the 150 kDa protein is not glycosylated, and rather than localizing to the plasma membrane like other MRPs, it is found predominantly in the endoplasmic reticulum . The recombinant ABCC12 protein does not transport any of the organic anions transported by the other MRPs in vesicular transport assays, nor does it confer resistance to cytotoxic agents in intact cell assays. Thus the substrate specificity of ABCC12 is unknown, and, for this reason, it is not referred to here as an ‘MRP’. Because full-length ABCC12 appears to be expressed only in testicular germ cells and sperm, it has been suggested that ABCC12 may play a role during the latter part of the male meiotic prophase, the development of spermatids and/or possibly in sperm function . At present, however, there is no experimental evidence supporting these proposed functions. The generation and characterization of Abcc12−/− mice may prove useful in determining the physiological role of this elusive transporter.
When MRP1 was first cloned in 1992, it was difficult to imagine that within less than 9 years, eight additional homologues of this transporter would be known. Considerable effort has been expended dissecting out the similarities and differences among the MRP/ABCC transporters, and, although much has been learned about some, others remain poorly characterized. MRP1, MRP2 and MRP4 are currently the best studied MRPs, and both important pharmacological and physiological functions have been described for them. MRP1 remains the most likely of the MRPs to be clinically relevant in drug-resistant tumours. Its role in mediating inflammatory responses through its ability to efflux LTC4 is also firmly established. MRP2 and MRP4, on the other hand, appear to contribute significantly to the elimination of xenobiotics and their metabolites into the bile and urine respectively. MRP4 is further noted for its ability to transport important signalling molecules such as cAMP and PGE2. MRP3 also has a role in drug metabolite elimination, but it seems to be more secondary. In contrast with MRP1-4, the in vivo physiological substrates of MRP5, ABCC6, MRP7, MRP8 and ABCC12 have still not been identified. It is clear that the connective tissue disorder PXE is caused by mutations in ABCC6, but, until the physiological substrate of this transporter is known, the underlying pathogenesis of this disease will remain a puzzle. Further work is also needed to better understand the substrates and functions of MRP5, MRP7, MRP8 and ABCC12. Finally, the involvement of GSH in the transport mechanisms of a few, but not all, MRPs is a distinctive feature of the MRP/ABCC subfamily, even if the physiological implications of this involvement are still unclear.
Differential plasma membrane trafficking (apical in contrast with basolateral membranes) in polarized epithelial and endothelial cells contributes to the specialized functions of several MRPs. However, little is known about the mechanisms that regulate the amount of a given MRP/ABCC on the plasma membrane, although it seems certain, at least in the case of MRP2 and MRP4, that interactions with intracellular scaffolding/chaperone proteins are involved. It is also known that the interindividual variation in basal transport activity of a particular MRP can be substantial (up to 80-fold), but, again, relatively little is known about what determines the actual amounts of MRP/ABCC proteins expressed in different tissues, or how these levels might be affected by exposure to xenobiotics, stress or disease conditions.
Finally, high-resolution crystal structures of both the short and long MRP/ABCC proteins are needed to better understand their complex transport mechanism(s), and precisely how substrate binding and translocation is coupled to ATP binding and hydrolysis. Such structures would also enhance understanding of how the MRP/ABCC proteins can recognize such a vast yet MRP-selective array of endogenous and xenobiotic substrates. In the case of the long MRPs, such studies should also reveal how their extra MSD0 functionally interacts with the core four-domain structure of these transporters.
• The nine MRP/ABCC transporters comprise a subfamily of the ABC superfamily of proteins that utilize the power of ATP binding and hydrolysis to move their physiological and pharmacological substrates across the plasma membrane
• Mutations in two of the MRP/ABCC genes, ABCC2 and ABCC6, cause hereditary human diseases (Dubin–Johnson syndrome and PXE respectively)
• Some MRP/ABCC transporters are believed to either play a role in resistance to anticancer and antiviral drugs in patients with resistant disease, and/or influence drug efficacy and toxicity through their ability to modulate disposition or elimination.
• Many MRP/ABCC proteins transport organic anions that can be formed either from endogenous molecules (such as LTC4 or PGE2) or by conjugation of drug metabolites. However, the in vivo substrates of some MRP/ABCC transporters are still not known.
• The substrate specificities of some MRP/ABCC transporters overlap to some degree, although differences in transport kinetics are often substantial. Significant species differences in the substrate specificity of MRP/ABCC transporters have also been noted.
• Differences in tissue-expression patterns and membrane localization (apical in contrast with basolateral) in polarized epithelial cells contribute to the unique pharmacological and physiological functions of each MRP/ABCC transporter.
This work has been supported by grants to S.P.C.C. from the Canadian Institutes of Health Research.
- © The Authors Journal compilation © 2011 Biochemical Society