Arsenic transport by the human multidrug resistance protein 1 (MRP1/ABCC1). Evidence that a tri-glutathione conjugate is required.

Inorganic arsenic is an established human carcinogen, but its metabolism is incompletely defined. The ATP binding cassette protein, multidrug resistance protein (MRP1/ABCC1), transports conjugated organic anions (e.g. leukotriene C(4)) and also co-transports certain unmodified xenobiotics (e.g. vincristine) with glutathione (GSH). MRP1 also confers resistance to arsenic in association with GSH; however, the mechanism and the species of arsenic transported are unknown. Using membrane vesicles prepared from the MRP1-overexpressing lung cancer cell line, H69AR, we found that MRP1 transports arsenite (As(III)) only in the presence of GSH but does not transport arsenate (As(V)) (with or without GSH). The non-reducing GSH analogs L-gamma-glutamyl-L-alpha-aminobutyryl glycine and S-methyl GSH did not support As(III) transport, indicating that the free thiol group of GSH is required. GSH-dependent transport of As(III) was 2-fold higher at pH 6.5-7 than at a more basic pH, consistent with the formation and transport of the acid-stable arsenic triglutathione (As(GS)(3)). Immunoblot analysis of H69AR vesicles revealed the unexpected membrane association of GSH S-transferase P1-1 (GSTP1-1). Membrane vesicles from an MRP1-transfected HeLa cell line lacking membrane-associated GSTP1-1 did not transport As(III) even in the presence of GSH but did transport synthetic As(GS)(3). The addition of exogenous GSTP1-1 to HeLa-MRP1 vesicles resulted in GSH-dependent As(III) transport. The apparent K(m) of As(GS)(3) for MRP1 was 0.32 microM, suggesting a remarkably high relative affinity. As(GS)(3) transport by MRP1 was osmotically sensitive and was inhibited by several conjugated organic anions (MRP1 substrates) as well as the metalloid antimonite (K(i) 2.8 microM). As(GS)(3) transport experiments using MRP1 mutants with substrate specificities differing from wild-type MRP1 suggested a commonality in the substrate binding pockets of As(GS)(3) and leukotriene C(4). Finally, human MRP2 also transported As(GS)(3). In conclusion, MRP1 transports inorganic arsenic as a tri-GSH conjugate, and GSTP1-1 may have a synergistic role in this process.

The 190-kDa MRP1 1 (gene symbol ABCC1) is a member of the ATP binding cassette (ABC) superfamily of transport proteins and was originally isolated on the basis of its elevated expression in the multidrug resistant small cell lung cancer cell line, H69AR (1). In addition to its ability to confer resistance in tumor cells, MRP1 is expressed constitutively in many nonmalignant tissues, with relatively high levels found in testes and lung (1,2). MRP1 is a primary active transporter of GSSG and GSH as well as glucuronate, GSH, and sulfate-conjugated organic anions of physiological and toxicological relevance (3). Endogenous substrates of MRP1 include the cysteinyl leukotriene LTC 4 , an important mediator of inflammatory response, and the conjugated steroids E 2 17␤G, estrone 3-sulfate, and dehydroepiandrosterone 3-sulfate (4 -8). MRP1 and the related MRP2 (gene symbol ABCC2) have also been shown to transport various xenobiotics and are key components of the so-called Phase III elimination pathways of drug metabolism (9). Several studies show that MRP1 and MRP2 can act synergistically with the phase II conjugating glutathione S-transferases (GST) to confer resistance to the toxicities of some electrophilic drugs and carcinogens (9 -11). However, several substrates of MRP1 and MRP2, including most of the natural product drugs to which they confer resistance, are not conjugated to any significant extent in vivo, but their transport is stimulated by GSH. Current evidence suggests that at least some of these drugs are co-transported with GSH across the plasma membrane (12,13).
The metalloid arsenic is an established multi-target human carcinogen and a major concern as a environmental pollutant (14). In addition, arsenic-containing compounds (e.g. As 2 O 3 ) are used in the treatment of several diseases including neoplasia and protozoal infections (15, 16). Thus, understanding the cellular mechanisms responsible for arsenic transport has both toxicological and pharmacological relevance. The ubiquitous nature of inorganic arsenic in the environment has led to the evolution of arsenic adaptation mechanisms in species ranging from bacteria to humans (17). In bacteria, yeast, and protozoa, pathways of metalloid resistance have been extensively characterized, and it has been determined that arsenic is detoxified either by extrusion from cells or by sequestration within intracellular organelles as thiol conjugates (18 -20). In mammalian cell models, MRP1 has been shown to confer resistance to arsenite (As III ) and arsenate (As V ) in a GSH-dependent man-ner (21)(22)(23)(24). However, arsenic transport by MRP1 has never been demonstrated directly, and the mechanism by which efflux occurs and the chemical nature of the transported species are still undefined.
Exposure of cells expressing MRP1 to inorganic arsenic has been shown to result in the efflux of GSH and arsenic into the tissue culture media (23). Arsenic-GSH conjugates have never been successfully isolated and identified in such systems, and this is possibly due to their unstable nature. Alternatively, as proposed by Salerno et al. (25), MRP1 could efflux arsenic in non-covalent association with free GSH. Convincing in vivo evidence indicates that MRP2 transports GSH-conjugated arsenic species into bile (26,27). Although the substrate specificities of MRP1 and MRP2 are not identical, they are very similar. Thus, MRP1 could transport arsenic in a GSH-conjugated form. The major route of arsenic excretion is urinary (60 -80% of total arsenic ingested), and very recently arsenic-GSH conjugates have been isolated from the urine of mice lacking the gene for ␥-glutamyl transpeptidase, an enzyme responsible for GSH and GSH conjugate catabolism (28). Thus, ␥-glutamyl transpeptidase is implicated in the processing of arsenic-GSH conjugates at the kidney (28).
In the present study we have investigated the form of arsenic transported by MRP1 using membrane vesicles prepared from the MRP1 overexpressing H69AR and transfected HeLa cell lines. We found that MRP1 transports As III but only in the presence of GSH, and this transport is not supported by GSH analogs that lack a free thiol group. Unexpectedly, we found that the normally cytosolic-or nuclear-localized GSTP1-1 is associated with the membrane vesicle fraction of the H69AR cell line. We present evidence that the formation of As(GS) 3 is necessary for As III efflux by MRP1 and that vesicle-associated GSTP1-1 is critical for this complex formation. Finally, the transport of As(GS) 3 by MRP1 is extensively characterized.
Derivation and culture of the MRP1 and vector control-transfected HeLa cell lines and the human small cell lung cancer cell line H69 and its MRP1-overexpressing variant H69AR have been described previously (1,31). The SV40-transformed human embryonic kidney (HEK293T) cell line was maintained in Dulbecco's modified Eagle's medium supplemented with 4 mM L-glutamine and 10% fetal bovine serum.
Chemical Synthesis of 73 As III and 73 As(GS) 3 -73 As III was prepared from 73 As V with metabisulfite-thiosulfate reagent as previously described (32). Briefly, 40 l of a reducing solution (0.1 mM NaAsO 2 , 66 mM Na 2 S 2 O 5 , 27 mM Na 2 S 2 O 3 , and 82 mM H 2 SO 4 ) was added to 40 l of carrier-free 73 As V (30 Ci) and incubated for 40 min at room temperature. The formation of 73 As III was monitored by TLC on SigmaCell type 100 cellulose plates (Sigma) developed with an isopropanol:acetic acid: water (10:1:5) solvent system (33). 73 As(GS) 3 was prepared as described previously by Delnomdedieu et al. (34) with modifications. 73 As III (final concentration 2.5 M) and GSH (final concentration 75 mM) were mixed under an argon atmosphere in HEPES/KCl buffer at a pH of ϳ3 and incubated at 4°C for Ͼ1 h. Formation of 73 As(GS) 3 was monitored by TLC as described above for 73 As V reduction, and all 73 As III was consumed and converted to 73 As(GS) 3 . Relative mobility (R F ) values were determined for starting materials and products and were consistent with previously published values (33). -The construction and expression of wild-type MRP1 (WT-MRP1), wild-type MRP2 (WT-MRP2), and the MRP1 mutants K332L, D336K, K319D, and K347D in HEK293T cells have been described previously (31,(35)(36)(37).

MRP1 and MRP2 Expression Vectors and Transfections in HEK293T Cells
Membrane Vesicle Preparation and Immunoblotting-Plasma membrane vesicles from H69, H69AR, HeLa vector, HeLa-MRP1, and transfected HEK293T cells were prepared as described, with modifications (5). Briefly, cells were homogenized in buffer containing 250 mM sucrose, 50 mM Tris, pH 7.5, 0.25 mM CaCl 2 , and protease inhibitor mixture tablets. Cells were disrupted by N 2 cavitation (a 5-min equilibration at 200 p.s.i.) and then released to atmospheric pressure, and EDTA was added to 1 mM. The suspension was centrifuged at 800 ϫ g at 4°C for 10 min, and the supernatant was layered onto 10 ml of a 35% (w/w) sucrose, 50 mM Tris, pH 7.4, cushion. After centrifugation at 100,000 ϫ g at 4°C for 1 h, the interface was removed and placed in a 25 mM sucrose, 50 mM Tris, pH 7.4, solution and centrifuged at 100,000 ϫ g at 4°C for 30 min. The membranes were washed with Tris sucrose buffer (250 mM sucrose, 50 mM Tris, pH 7.4) and then resuspended by vigorous syringing with a 27-gauge needle. Protein concentrations were determined by a Bradford assay (Bio-Rad), and aliquots of membrane vesicles were stored at Ϫ80°C.
Relative levels of MRP1 protein in membrane vesicles (1 and/or 2 g of protein) were determined by immunoblot analysis as described previously, with the human MRP1-specific mAb QCRL-1 (29). Relative levels of GSTP1-1 protein in membrane vesicles (5 g of protein) were determined by immunoblot analysis as for MRP1, with a rabbit polyclonal antibody specific for human GSTP1-1 at a dilution of 1:1000. Comparison of MRP1 and MRP2 protein expression levels was done using the polyclonal rabbit antibody MRP-1 at a dilution of 1:1000 as described previously (30). Where appropriate, relative levels of MRP1 and MRP2 expression were estimated by densitometric analysis using a ChemiImager™ 4000 (Alpha Innotech, San Leandro, CA). Equal loading of all protein was confirmed by Amido Black staining of the membranes.

73
As III and 73 As(GS) 3 Transport Studies-73 As transport assays were carried out by a rapid filtration method as previously described (5). Membrane vesicles (20 g of protein per time point) were incubated at 37°C at a final volume of 60 l (single time point) or 300 l (5 time points). Unless otherwise noted, the transport assay buffer used was HEPES (50 mM, pH 7.5), KCl (100 mM) and contained ATP or AMP (4 mM), MgCl 2 (10 mM), creatine phosphate (10 mM), creatine kinase (100 g/ml), glutathione reductase (5 g/ml), NADPH (0.35 mM), with and without GSH (3 mM) and 73 As III (30 nCi, 100 nM) or 73 As(GS) 3 (30 nCi, 100 nM). Conditions for synthesis of 73 As(GS) 3 resulted in the presence of 3 mM GSH in all transport reactions. At the indicated time points, 60 l of transport reaction mix was removed and placed in 800 l of HEPES (50 mM, pH 7.5), KCl (100 mM) buffer, filtered through glass fiber filters (type A/E; Pall Life Sciences, East Hills, NY), and washed twice, and radioactivity was quantitated by liquid scintillation counting. Transport in the presence of AMP was subtracted from transport in the presence of ATP and reported as ATP-dependent 73 As III or 73 As(GS) 3 transport. The effects of potential modulators of 73 As(GS) 3 or 73 As III transport (MRP1-specific mAbs QCRL-1 and QCRL-3 (10 g/ml), ophthalmic acid (1, 3, or 5 mM), S-methyl GSH (1, 3, or 5 mM), pH change (pH 6.5, 7, 7.5, or 8), sucrose (250 -1000 mM), E 2 17␤G (25 M), LTC 4 (1 M), GSSG (0.5 mM), As V (1 or 10 M), CdCl 2 (1 or 10 M), potassium antimony tartrate (Sb III ) (1 or 10 M)) were measured at a single time point of 3 min. The effect of exogenous purified GSTP1-1 (0.4 ng or 79 microunits/60-l reaction) on 73 As III uptake in the presence of GSH by HeLa-MRP1 vesicles was measured as above. Before use, the purified GSTP1-1 was dialyzed to remove dithiothreitol and EDTA (potential arsenic chelators) using a Slide-A-Lyzer MINI dialysis unit (Pierce) according to the manufacturer's instructions.
Kinetic parameters were determined by measuring the initial rate of 73 As(GS) 3 uptake at 7 different substrate concentrations (50 -2500 nM) at a single time point of 1 min. The mode of As(GS) 3 transport inhibition by Sb III was examined by determining kinetic parameters in the presence and absence of 5 M of this metalloid.
Confocal Microscopy-Triple staining experiments for nuclei, MRP1 (as a plasma membrane marker), and GSTP1-1 were done using the H69AR and HeLa-MRP1 cell lines. H69AR and HeLa-MRP1 cells were seeded at 2 ϫ 10 6 and 1 ϫ 10 6 cells/well, respectively, in a 6-well plate on coverslips pretreated with poly L-lysine (M r 70,000 -150,000). Thirtysix hours later the cells were fixed with 4% paraformaldehyde for 10 min and then incubated in permeabilization buffer (0.2% Triton X-100 in phosphate-buffered saline) for 5 min, washed 3 times in blocking solution (0.1% Triton X-100, 1% bovine serum albumin in phosphatebuffered saline) over 15 min, and then incubated with the GSTP1-1specific murine mAb 353-10 (1:50 dilution in blocking solution) and the MRP1-specific rat mAb MRPr1 (1:500 dilution) overnight. After washing with blocking solution 3 times over 1 h, highly cross-adsorbed Alexa Fluor 488-conjugated goat anti-mouse IgG (HϩL) and Alexa Fluor 594 goat anti-rat IgG (HϩL) (Molecular Probes, Eugene, OR) were added (1:500 dilution) with RNase A (10 g/ml) and incubated in the dark for 30 min. Cells were washed with blocking solution changed 4 times over 1 h. Nuclei were then stained with Hoechst 33342 (2.5 g/ml in phosphate-buffered saline) for 20 min, rinsed 5 times with phosphate-buffered saline, and coverslips were placed on slides with ProLong antifade mounting media (Molecular Probes). A Zeiss LSM 510 laser scanning confocal microscope (Carl Zeiss Inc., Thornwood, NY) with a Zeiss Plan-Apo 63ϫ oil (numerical aperture ϭ 1.4) objective lens was used to obtain the fluorescence images. The Hoechst 33342, Alexa 488, and Alexa 594 images were acquired consecutively using, respectively, the 364-, 488 (Ar-ion lasers)-, and 543-nm (HeNe laser) laser lines for excitation and the 385-nm long pass, 505-550-nm band pass, and 560-nm long pass filters for emission. The pinhole was adjusted so that the theoretical Z-resolution was ϳ0.7 m. Negative control experiments with Alexa Fluor 488-and/or 594-conjugated goat anti-mouse IgG (HϩL) in the absence of primary antibodies were also performed. Negative control experiments in the presence of both secondary antibodies and individual primary antibodies were also conducted to ensure no cross-reactivity between antibodies.

Uptake of As III (plus GSH) and As(GS) 3 by MRP1-enriched
Membrane Vesicles-To determine whether As III was a substrate for MRP1, ATP-dependent uptake into membrane vesicles prepared from the human small cell lung cancer cell line H69 and its MRP1-overexpressing variant, H69AR, was determined. A time course of As III uptake by the H69 and H69AR membrane vesicles is shown in Fig. 1A. ATP-dependent uptake of As III by the MRP1-enriched H69AR vesicles was extremely low and similar to uptake observed with AMP or with the non-MRP1-containing H69 vesicles. However, in the presence of GSH (3 mM), ATP-dependent uptake of As III by the H69AR but not the H69 vesicles was observed. Uptake was maximal at 3 min, at which time it reached a level of 2.6 pmol mg Ϫ1 . Similar experiments with As V were also conducted, and no ATP-dependent transport was observed in the presence or absence of GSH (data not shown).
Incubation of As III with GSH (as described under "Experimental Procedures") resulted in the spontaneous formation of As(GS) 3 as detected by TLC ( 73 As III R F ϭ 0.6; 73 As III plus GSH R F ϭ 0.01). Uptake of As(GS) 3 by the H69 and H69AR membrane vesicles was also measured (Fig. 1A). Uptake of As(GS) 3 by H69AR membrane vesicles was linear for 3 min, at which time it reached a level of 4.7 pmol mg Ϫ1 and then began to plateau. The maximum activity for As(GS) 3 uptake was almost 2-fold greater than the maximum activity observed with As III (plus GSH). Results for the H69 vesicles were similar to that of the As III time course (not shown).
Inhibition of As(GS) 3 Transport by the MRP1-specific mAb QCRL-3-Several MRP1-specific mAbs have been shown previously to inhibit transport of conjugated and unconjugated MRP1 substrates including LTC 4 , E 2 17␤G, aflatoxin B 1 -GS, vincristine, and NNAL-O-glucuronide (plus GSH) (4,5,13,30,38). When the mAb QCRL-3 (which recognizes a conformationaldependent epitope) was tested for its ability to inhibit ATP-dependent uptake of As(GS) 3 by H69AR membrane vesicles, complete inhibition was observed at a concentration of 10 g ml Ϫ1 (Fig. 1B). mAb QCRL-1, which recognizes a linear epitope in part of the linker region of MRP1 and does not inhibit transport of other MRP1 substrates (39), had no effect on As(GS) 3 transport. These results further confirm the MRP1 specificity of ATP-dependent transport of As(GS) 3 .
Ophthalmic Acid and S-Methyl GSH Do Not Support MRP1mediated As III Transport-In previous studies, it has been shown that ophthalmic acid (L-␥-glutamyl-L-␣-aminobutyrylglycine) and S-methyl GSH can substitute for GSH and support the transport of several GSH-dependent MRP1 substrates (8,13,38,40). These findings indicate that the thiol group of GSH is not required for transport of these substrates and rules out the possibility that formation of a GSH conjugate is critical for transport to occur. In contrast with other MRP1 substrates, neither ophthalmic acid ( Fig. 2A) nor S-methyl GSH (Fig. 2B) supported the uptake of As III into H69AR membrane vesicles, demonstrating that the free thiol group of GSH is required for As III transport by MRP1.
GSH-dependent As III Uptake Is pH-dependent-The stability and formation rate of As(GS) 3 are known to be greater at acidic pH (25,34), whereas MRP1 transport of LTC 4 has been shown to be relatively insensitive to changes in pH; specifically, LTC 4 transport was unchanged between pH 6.5 and 7.5 and modestly increased above pH 8 (41). To determine if transport of As III (plus GSH) is influenced by pH, transport was measured over a pH range of 6.5-8. Transport of As III (plus GSH) was similar at FIG. 1. MRP1-mediated As III (plus GSH) and As(GS) 3 uptake by inside-out membrane vesicles. A, ATP-dependent transport of As III in the presence of GSH by membrane vesicles prepared from the small cell lung cancer cell line H69 (E) and its MRP1-overexpressing variant H69AR (q) in the absence of GSH by H69 (Ⅺ) and H69AR (f) membrane vesicles and ATP-dependent transport of As(GS) 3 by H69AR (OE) membrane vesicles. Membrane vesicles (20 g of protein per time point) were incubated at 37°C in transport buffer with As III (100 nM, 30 nCi) with and without 3 mM GSH or with As(GS) 3 (100 nM, 30 nCi). B, As(GS) 3 uptake was measured after preincubation of H69AR membrane vesicles with the indicated MRP1-specific mAbs (10 g ml Ϫ1 ). The results shown are the means (ϮS.D.) of triplicate determinations in a single experiment, and similar results were obtained in at least two additional experiments. Where error bars are not visible they are smaller than the data point. acidic (pH 6.5) and neutral pH (ϳ6 pmol mg Ϫ1 3 min Ϫ1 ), whereas uptake by vesicles at more basic pH (pH 7.5 and 8) was reduced by 50% (ϳ2.8 pmol mg Ϫ1 3 min Ϫ1 ) (Fig. 3). The increased As III transport activity observed at neutral and acidic pH is consistent with the enhanced formation and stability of As(GS) 3 at acidic pH and supports the conclusion that this GSH conjugate is being formed before transport.
GSTP1-1 Is Associated with the Plasma Membrane of H69 and H69AR Cell Lines-During optimization of As III (plus GSH) uptake assay, TLC analysis indicated that under the transport assay conditions used, As(GS) 3 was formed from free As III and GSH. However, further experimentation showed that conjugate formation did not occur in the absence of membrane vesicles (data not shown). This suggested that the membrane vesicles contained a component that catalyzed the formation of As(GS) 3 . Immunoblot analysis revealed the presence of GSTP1-1 (an enzyme normally found in the cytosol and nucleus) in the membrane vesicles prepared from the H69 and H69AR cell lines (Fig. 4A). In contrast, very little GSTP1-1 was found in the vesicle fraction of HeLa cell lines stably transfected with empty pcDNA3.1(Ϫ) vector (HeLa-vector) or pcDNA3.1(Ϫ) containing MRP1 cDNA (HeLa-MRP1) (Fig. 4A).
The cellular localization of GSTP1-1 was further characterized by confocal analysis of intact H69AR (Fig 4B, top panels) and HeLa-MRP1 (Fig 4B, bottom panels) cells. A significant fraction of GSTP1-1 was localized to a region resembling the plasma membrane of H69AR, whereas diffuse, punctate cyto-plasmic staining was present in the HeLa cell line (Fig. 4B, first  column). The H69AR and HeLa-MRP1 cell lines were immunostained with the MRP1-specific mAb MRPr1 to define the plasma membrane region (Fig. 4B, second column). Merged images of MRP1 and GSTP1-1 for each cell line are shown (Fig.  4B, third column) and for cellular orientation the nuclei were stained (Fig. 4B, fourth column).
As III (plus GSH) Is Transported Only in the Presence of GSTP1-1-The ability of the H69, H69AR, HeLa-vector, and HeLa-MRP1 to transport As III (plus GSH) and As(GS) 3 was then compared (Fig. 5A). As expected, membrane vesicles prepared from the cell lines H69 and HeLa-vector, which do not express MRP1, did not transport either free As III (plus GSH) or As(GS) 3 . Consistent with the time course presented in Fig. 1A, the H69AR membrane vesicles transported As III (plus GSH) and As(GS) 3 , with an approximate 2-fold increase in As(GS) 3 transport compared with As III (plus GSH) (11 versus 5 pmol mg Ϫ1 3 min Ϫ1 , respectively). In contrast with the H69AR vesicles, very little vesicular uptake (0.4 pmol mg Ϫ1 3 min Ϫ1 ) of As III (plus GSH) was detected for the HeLa-MRP1 vesicles. However, the transport of the preformed conjugate by HeLa-MRP1 vesicles was substantial (2.5 pmol mg Ϫ1 3 min Ϫ1 ), representing an approximate 6-fold increase over the activity observed with the unconjugated As III (plus GSH). Transport of As(GS) 3 by HeLa-MRP1 was ϳ25% that of the level of transport by H69AR vesicles. This difference is most likely a result of the higher level of MRP1 present in the H69AR vesicles (Fig. 4A).
To confirm that the differences in GSTP1-1 expression levels were of functional importance for As III (plus GSH) uptake, exogenous GSTP1-1 was added to HeLa-MRP1 vesicles before transport analysis (Fig. 5B). The addition of GSTP1-1 (79 microunits) resulted in a 6-fold increase in transport of As III (plus GSH) with an activity of 2.7 pmol mg Ϫ1 3 min Ϫ1 , similar to the transport level of the preformed As(GS) 3 by HeLa-MRP1 vesicles.
Osmotic Sensitivity of As(GS) 3 Uptake-To confirm that the ATP-dependent As(GS) 3 uptake by the H69AR membrane vesicles truly represents transport into the vesicle lumen rather than surface or intramembrane binding, the effect of changes in osmolarity on vesicular uptake was examined. As(GS) 3 uptake was decreased as the concentration of sucrose in the transport buffer increased, indicating that ATP-dependent As(GS) 3 uptake by the vesicles is osmotically sensitive, as expected for a true transport process (Fig. 6A).
Kinetic Analysis of MRP1-mediated As(GS) 3 Uptake-MRP1dependent uptake of As(GS) 3 was further characterized by determining the initial rates of uptake over several concentra- tions of As(GS) 3 (Fig. 6B). According to Lineweaver-Burk analysis (Fig. 6C), the apparent K m for As(GS) 3 was 0.32 Ϯ 0.08 M, and the V max was 17 Ϯ 5 pmol mg Ϫ1 min Ϫ1 (mean Ϯ S.D. of three independent determinations).
Effect of Metalloids and Cd II on As(GS) 3 Uptake-Studies using transfected cell lines have shown that MRP1 confers resistance to the metalloids As III , As V , and Sb III but does not  5. GSTP1-1 is required for the transport of As III (plus GSH), whereas As(GS) 3 transport is independent of this enzyme. A, uptake of As III (plus 3 mM GSH) (100 nM, 30 nCi) (closed bars) and As(GS) 3  confer resistance to the transition metal Cd II (21). Thus, it was of interest to see if these compounds had any effect on the MRP1-dependent transport of As(GS) 3 . As(GS) 3 (100 nM) transport was not significantly altered in the presence of Cd II (1 and 10 M) or As V (1 and 10 M) (Fig. 7B). However, Sb III (10 M) dramatically inhibited transport of As(GS) 3 to 21% that of control levels.
The inhibition of As(GS) 3 transport was further characterized by measuring the effect of Sb III (5 M) on As(GS) 3 (75-2500 nM) uptake (Fig. 7C). Lineweaver-Burk plots showed that Sb III behaved as a potent competitive inhibitor of MRP1-mediated As(GS) 3 transport with an apparent K i of 2.8 M. This low K i value and the evidence from previous studies that MRP1 confers resistance to Sb III indicate that this metalloid is also potentially a high affinity substrate of MRP1 (21,43). Conditions for the synthesis of As(GS) 3 result in the presence of 3 mM GSH in transport assays; therefore, the formation of Sb(GS) 3 is possible, and this may well be the form inhibiting As(GS) 3 uptake.
MRP1 With Lys 332 3 Leu and Asp 336 3 Leu Mutations Does Not Transport As(GS) 3 -It has been previously reported that substitution of several ionizable amino acid residues in or proximal to the transmembrane helices of the second membrane- spanning domain (MSD2) of MRP1 resulted in dramatic changes in MRP1 substrate specificity (35,36). For example, conservative and non-conservative substitution of the lysine residue at position 332 in the predicted sixth transmembrane helix (the first transmembrane helix of MSD2) of MRP1 resulted in a substantial but selective reduction in the ability of MRP1 to transport LTC 4 and GSH, whereas MRP1 transport of estrone 3-sulfate and E 2 17␤G remained intact. In contrast, similar substitutions of the nearby aspartic acid residue at position 336 of MRP1 showed no such selectivity and resulted in a general loss of transport function. In addition, substitution of lysines at positions 319 and 347 with aspartic acid residues resulted in a selective 50% loss of GSH transport but had no effect on MRP1 transport of LTC 4 , estrone 3-sulfate, and E 2 17␤G. To determine whether these amino acid residues are critical for transport of As(GS) 3 , transport assays were done using membrane vesicles prepared from HEK293T cells transfected with pcDNA3.1(Ϫ) (empty vector), WT-MRP1, D336K-MRP1, K332L-MRP1, K319D-MRP1, and K347D-MRP1 cDNAs. The MRP1 mutant and WT proteins were expressed at similar levels (90 -110% that of WT-MRP1 levels) in HEK293T cells as determined by immunoblot analysis of the membrane vesicles (Fig. 8A). Consistent with the selective loss of LTC 4 and GSH transport by K332L-MRP1 and the general loss of transport function by D336K-MRP1, these mutants also did not transport As(GS) 3 (Fig. 8B). Despite the fact that K319D-MRP1 and K347D-MRP1 have a selectively decreased ability to transport GSH, these mutants transported As(GS) 3 at levels similar to WT-MRP1. Therefore, these data suggest that As(GS) 3 interacts with MRP1 in a similar manner as LTC 4 .
Transport of As(GS) 3 by MRP2-Previous studies comparing biliary excretion of arsenic by wild-type Mrp2 and Mrp2 deficient (TR Ϫ ) rats revealed that Mrp2 is responsible for transport of As-GSH conjugates into bile (26). To compare the transport of As(GS) 3 by MRP1 and MRP2 and confirm that human MRP2 also transports As(GS) 3 , membrane vesicles were prepared from HEK293T cells transfected with pcDNA3.1(Ϫ) (empty vector), WT-MRP1, and WT-MRP2. Immunoblot analysis of membrane vesicles prepared from transfected cells revealed that MRP2 was expressed at 1.4-fold higher levels than MRP1 (Fig. 9A). Levels of As(GS) 3 transport by MRP2 (corrected for relative expression) were 4.3 pmol mg Ϫ1 3 min Ϫ1 , and transport by MRP1 vesicles under the same conditions was similar at 3.7 pmol mg Ϫ1 3 min Ϫ1 (Fig. 9B).

DISCUSSION
Inorganic arsenic is a ubiquitous environmental pollutant and chronic exposure in humans results in formation of tumors of the skin, lung, bladder, liver, and kidney (14). The cellular pathways for metabolism and transport of inorganic arsenic are complex and likely play a fundamental role in this carcinogenic response. Furthermore, prolonged exposure to inorganic arsenic often results in cellular adaptations involving altered biokinetics, which likely impact inorganic arsenic toxicity and pharmacology. In this study we have investigated the ability of MRP1 to transport arsenic in its trivalent (As III ), pentavalent (As V ), and triglutathione conjugate (As(GS) 3 ) forms using membrane vesicles. Our results provide clear primary evidence that MRP1 transports As III in an ATP-and GSH-dependent manner, whereas As V (with or without GSH) is not a substrate or inhibitor of MRP1 transport activity. MRP1 has previously been shown to confer resistance to As III and As V when expressed in intact cells (21,22), and thus, the present results suggest that As V is reduced to As III before efflux from the cell. This was not necessarily an unexpected observation as the initial step in arsenic metabolism in both Escherichia coli and Saccharomyces cerevisiae is the reduction of As V to As III , which is catalyzed by species-specific reductases (17). The reduction of As V to As III in mammalian cells was previously considered to occur spontaneously in the presence of GSH. The recent isolation of a mammalian arsenate reductase (purine nucleoside phosphorylase) suggests otherwise (44,45); however, the in vivo relevance of this reductase is not yet firmly established (46). It has recently been proposed that the oxidation of As III to As V could be a critical detoxification step (47); however, results from our study suggest that, although more acutely toxic, As III is potentially more excretable.
Our observation that ophthalmic acid and S-methyl GSH do not support the transport of As III indicates that the free sulfur atom of GSH is required and suggests that an arsenic-GSH complex is formed. Salerno et al. (25) performed an extensive kinetic analysis of As(GS) 3 complex formation using NMR and circular dichroism in conjunction with a study of As III and Sb III accumulation in the human small cell lung cancer cell line GLC 4 and its MRP1-overexpressing variant, GLC 4 /ADR. The authors found that in vitro As(GS) 3 was formed very slowly at physiological pH, whereas efflux of arsenic and GSH from intact MRP1-overexpressing cells was relatively rapid. Thus, the authors proposed that MRP1 effluxed arsenic through a co-transport mechanism with GSH. However, the possibility that the formation of As(GS) 3 might be catalyzed by an enzyme such as GSTP1-1 was not considered. Our finding that GSTP1-1 was required for transport of As III (plus GSH) provides substantive evidence that this is likely to be the case. Overexpression of GSTP1-1 has been noted in several cell lines chronically exposed to As III , and it has been suggested that this overexpression facilitated the efflux of arsenic (24,48,49). However, the importance of GSTP1-1 in the efflux of arsenic from intact cells has been difficult to elucidate primarily because many of the compounds used to inhibit GSTP1-1 also inhibit MRP1 (e.g. ethacrynic acid) (24, 48 -51). In addition, elucidating the role of GSTP1-1 in arsenic resistance is made more complex because, in addition to its conjugating activity, GSTP1-1 regulates the mitogen-activated protein kinase pathway through binding and sequestration of c-Jun N-terminal kinase 1 (52). Thus, increased expression of GSTP1-1 could also inhibit the induction of apoptosis and confer resistance to arsenic in this manner.
Our immunoblot and confocal microscopy analysis showed that GSTP1-1 levels are enhanced at or near the plasma membrane of the H69 and H69AR cell lines. GSTP1-1 is normally a cytosolic or nuclear protein, and its close association with the plasma membrane was unexpected. The HeLa-MRP1 cell line, which contained GSTP1-1 in the cytoplasm and nuclear regions, is also resistant to arsenic, indicating that the association of GSTP1-1 with the membrane is likely to be unnecessary for resistance in the intact cell (21). However, the close proximity of conjugating enzymes in the membrane could be very important for the efficient conjugation and efflux of other labile conjugates. It has been previously reported that cytosolic GSTs are over-represented in microsomal fractions compared with cytosolic marker proteins (53). Membrane association of GSTs could be important for conjugation of hydrophobic toxins partitioned in the membrane (54). It is interesting that murine GSTA4-4, an enzyme highly efficient in the GSH conjugation of 4-hydoxyalkenals, is also closely associated with the plasma membrane (55). In addition, GSTs from the pi and the mu classes have recently been shown to be associated with goat sperm plasma membrane, where in addition to a role in detoxification, these proteins function as gamete recognition molecules (56). We have also observed that GSTP1-1 is associated with membrane vesicles prepared from HEK293T cells, 2 and therefore, this association is not restricted only to the H69 and H69AR lung tumor cells. Vesicle-based studies are often assumed to allow the isolation of the plasma membrane from, minimally, the cytoplasmic and nuclear cell compartments.
Thus, the presence of "cytosolic" GSTP1-1 and potentially other GST family members in membrane vesicle preparations could have implications for interpretation of transport data.
The kinetic parameters of As(GS) 3 transport (K m 0.32 M, V max 17 pmol mg Ϫ1 min Ϫ1 ) suggest that MRP1 is a high affinity, low capacity transporter of this metalloid conjugate. Transport of As(GS) 3 by human MRP1 is of remarkably higher affinity (ϳ1000-fold) than transport of As III by the E. coli As III /Sb IIItranslocating ATPase, ArsAB (K m 0.1 mM), or As(GS) 3 by the currently unidentified Leishmania tarentolae plasma membrane transport protein (K m 0.1-0.2 mM) (18,19). However, the overall transport efficiency (V max /K m ) was in the same order of magnitude for MRP1, ArsAB, and the L. tarentolae transporter. Two vacuolar pumps, the yeast cadmium factor protein (Ycf1p) and the L. tarentolae P-glycoprotein A (LtPGPA), are close homologues of MRP1, with 60 and 48% amino acid identity, respectively. Ycf1p and LtPGPA are well characterized transporters of As(GS) 3 and might have higher affinity for this substrate than ArsAB and the Leishmania plasma membrane transporter; however, kinetic parameters of As(GS) 3 transport by these proteins have not been reported. The differences in K m values between human MRP1 and the other transporters may possibly be related to the fact that the single cell organisms would be bathed in high concentrations of arsenic, whereas cells expressing MRP1 in mammalian tissues would be exposed to much lower concentrations. Pharmacokinetic analysis of As III distribution in human patients undergoing As 2 O 3 treatment and rodent models (57)(58)(59) suggest the K m value obtained for MRP1 is of high physiological relevance, with plasma levels ranging in the hundreds of nanomolar after dosing.
With the exception of the E. coli ArsAB transporter, which effluxes As III in its unconjugated form, all other proteins involved in the transport of arsenic have been reported to transport the As(GS) 3 conjugate (18 -20, 60). Characterization of transport in yeast and Leishmania systems included transport of As III in the presence of GSH without pre-synthesis of As(GS) 3 ; however, unlike the present study, there was no apparent requirement for catalysis of conjugate formation (19,20). The high levels of As III (100 M) and the more acidic pH (pH 7) used in these studies could have resulted in efficient spontaneous formation of As(GS) 3 . Alternatively, a GST required for As(GS) 3 formation could have been present in the yeast and Leishmania membrane vesicle preparations used. Interestingly, it has been postulated that a thiol transferase is a component required for Leishmania arsenite resistance (61).
Despite the strong in vitro data that indicate MRP1 is important for protecting cells from arsenic toxicity, some in vivo evidence is not consistent with this protective effect. For instance, Lorico et al. (62,63) report that Mrp1 (ϩ/ϩ) mice have no survival advantage over Mrp1 (Ϫ/Ϫ) mice upon acute exposure to high dose sodium arsenite. In addition, accumulation of arsenic after injection of As III was similar in all tissues examined from Mrp1 (ϩ/ϩ) and Mrp1 (Ϫ/Ϫ) animals. The authors suggested that the redundancy of transmembrane export pumps offers equivalent protection to the Mrp1(Ϫ/Ϫ) animals as the Mrp1(ϩ/ϩ) mice (62,63). An independent investigation of Mrp1(Ϫ/Ϫ) mice from the same colony revealed that 1 h after injection of As III there were no differences in urinary excretion of As III between Mrp1(ϩ/ϩ) and Mrp1(Ϫ/Ϫ) mice, suggesting that Mrp1 is not critical to urinary excretion of arsenic (28). In sharp contrast to the short exposure times and relatively high doses of arsenic used in these animal studies (28,62,63), human tumor formation is associated with long term exposure to comparatively low doses of arsenic (14). The high affinity low capacity transport of As(GS) 3 by MRP1 suggests that MRP1 would be better suited to protect tissues from low levels of arsenic and at higher concentrations may be saturated, and other protective mechanisms are required. Thus, in vivo doses of arsenic used in animal studies may overwhelm this transport protein. Therefore, MRP1 could potentially be very important for protecting tissues from the carcinogenic effects of typical low level arsenic exposure in drinking water, although chronic animal studies are required to confirm this.
In addition to MRP1 and MRP2, seven other MRP isoforms (MRP3-7, ABCC11, and ABCC12) exist, several of which are known to transport conjugated organic anions and could be involved in the protection of tissues from the toxic effects of arsenic (3,28). The response of different individuals to chronic inorganic arsenic exposure is highly variable. Polymorphisms of several genes involved in the metabolism of this "simple" inorganic molecule have been identified including enzymes involved in arsenic methylation and reduction (64). Four allelic variants of GSTP1-1 and many polymorphisms of MRP1 and MRP2 have been identified (3,(65)(66)(67). The potential effect of these genetic variations on arsenic metabolism and toxicity is largely unknown, and further studies are of critical relevance.