Allosteric Interactions between the Two Non-equivalent Nucleotide Binding Domains of Multidrug Resistance Protein MRP1*

Membrane transporters of the adenine nucleotide binding cassette (ABC) superfamily utilize two either identical or homologous nucleotide binding domains (NBDs). Although the hydrolysis of ATP by these domains is believed to drive transport of solute, it is unknown why two rather than a single NBD is required. In the well studied P-glycoprotein multidrug transporter, the two appear to be functionally equivalent, and a strongly supported model proposes that ATP hydrolysis occurs alternately at each NBD (Senior, A. E., al-Shawi, M. K., and Urbatsch, I. L. (1995) FEBS Lett 377, 285–289). To assess how applicable this model may be to other ABC transporters, we have examined adenine nucleotide interactions with the multidrug resistance protein, MRP1, a member of a different ABC family that transports conjugated organic anions and in which sequences of the two NBDs are much less similar than in P-glycoprotein. Photoaffinity labeling experiments with 8-azido-ATP, which strongly supports transport revealed ATP binding exclusively at NBD1 and ADP trapping predominantly at NBD2. Despite this apparent asymmetry in the two domains, they are entirely interdependent as substitution of key lysine residues in the Walker A motif of either impaired both ATP binding and ADP trapping. Furthermore, the interaction of ADP at NBD2 appears to allosterically enhance the binding of ATP at NBD1. Glutathione, which supports drug transport by the protein, does not enhance ATP binding but stimulates the trapping of ADP. Thus MRP1 may employ a more complex mechanism of coupling ATP utilization to the export of agents from cells than P-glycoprotein.

The multidrug resistance protein, MRP1, 1 is believed to function as an active exporter of many conjugated organic anions from cells (1)(2)(3)(4)(5)(6)(7). Among agents that are transported are also some unconjugated compounds, including certain cancer drugs, provided that glutathione is also present (8,9). In this case, the hydrophilic conjugating compound such as glutathione may be co-transported along with the hydrophobic drug (8). As a member of the ABC super family of membrane transporters (10), MRP1 is an ATPase that is stimulated by agents that it transports (11,12). However, as yet there is little further information about how the two nucleotide binding domains of the protein act to bring about the transport event. In the case of the better studied P-glycoprotein, which belongs to a different family of this super family (13), the two NBDs have been shown to be functionally equivalent with identical ATP hydrolysis steps occurring alternatively at each domain (14,15). According to an insightful model based largely on this information, binding and translocation of the hydrophobic compound exported is controlled in an ordered fashion by the two hydrolysis steps (15). The objective of the present study was to determine whether MRP1 performs a similar symmetrical cycle of partial reactions of ATP hydrolysis to accomplish the export of conjugated organic anions. To follow the interaction of the nucleotide substrate and product of the hydrolysis reaction with the two NBDs, we have utilized photoaffinity labeling of the protein with 8-azido-ATP, which already has been shown to occur (16). Mutagenesis of key Walker motif amino acid residues revealed that like P-glycoproteins (17), both domains must be functional for nucleotide interactions to occur at either. In contrast to P-glycoprotein (14,18), however, trapping of the ADP product of ATP hydrolysis by MRP1, reflected by photolabeling with N 3 [␣-32 P]ATP, can stimulate the binding of the intact nucleotide triphosphate, which occurred exclusively at NBD1. Low concentrations of glutathione enhanced trapping of the diphosphate but not binding of the triphosphate, suggesting that its stimulation of ATPase activity reported earlier (11) is due to an effect on either catalysis or product release rather than substrate binding. These findings indicate that there is a separation of function between the two NBDs of MRP1 and allow the formulation of an allosteric model in which the distinct action of one is required for that of the other.
In Vitro Mutagenesis of MRP1 cDNA-We have employed the coding sequence of human MRP1 cDNA in the pNUT expression vector (11). The lysine residues at positions 684 (Walker A motif in NBD1, AAG to CTG) and 1333 (Walker A motif in NBD2, AAG to CTG) were mutated to leucine residues by using the QuikChange Site Directed Mutagenesis Kit. The aspartic acid residues at residues 792 and 793 (Walker B motif in NBD1, GAT to CTT) and at 1454 and glutamic acid at 1455 (Walker B motif in NBD2, GAT to CTT and GAG to CTG) were also mutated to leucine residues. To ensure that no other mutations were introduced into the cDNA during mutagenesis, fragments covering the mutations were sequenced completely and used to replace their counterparts in the wild-type MRP1 cDNA in pNUT. The mutations were verified after insertion into the pNUT expression vector.
Membrane and Vesicle Preparations-MRP1-containing membranes were prepared as follows. Cells were grown to confluence in roller bottles and washed once with 20 ml of phosphate-buffered saline. The cells were detached by incubation at 37°C for 5-10 min with 20 ml of citrate saline. Cells were collected by centrifugation at 4,000 ϫ g for 15 min. The cell pellet was washed once with membrane preparation buffer (10 mM Tris-HCl, pH 7.5, 250 mM sucrose, and 0.2 mM MgCl 2 ). The cells were resuspended in membrane preparation buffer with 1 ϫ protease inhibitors (2 g/ml aprotinin, 121 g/ml benzamidine, 3.5 g/ml E64, 1 g/ml leupeptin, and 50 g/ml Pefabloc) and equilibrated on ice for 20 min at 800 p.s.i. in a Parr N 2 cavitation bomb. After releasing the pressure, the cell homogenate was adjusted to 1 mM in EDTA. The homogenate was then diluted 5-fold with 10 mM Tris-HCl and 25 mM sucrose, pH 7.5, and centrifuged at 1,000 ϫ g for 15 min. The supernatant was overlaid on a 35% sucrose solution containing 10 mM Tris-HCl, pH 7.5, and 1 mM EDTA and centrifuged at 16,000 ϫ g for 30 min. The interface was collected, diluted 5-fold with a solution containing 10 mM Tris-HCl (pH 7.5) and 250 mM sucrose, and then centrifuged at 100,000 ϫ g for 45 min. The pellet was resuspended in a solution containing 40 mM Tris-HCl, pH 7.5, 0.1 mM EGTA and 1ϫ protease inhibitors. Membrane vesicles for drug transport experiments were prepared exactly the same way except the pellet was resuspended in a solution containing 10 mM Tris-HCl, pH 7.5, 250 mM sucrose, and protease inhibitors and passed through a Liposofast TM vesicle extruder (200 nm filter, Avestin, Ottawa, Canada).
Vanadate Preparation and Photoaffinity Labeling of MRP1 Protein-100 mM vanadate solution was prepared by dissolving 184 mg of sodium orthovanadate in 9.35 ml of water and then adding 0.65 ml of 2 M HCl. The vanadate solution was boiled for 10 min before using in the experiments. Unless otherwise indicated in the figure legend, the photolabeling was carried out in a 10 l of solution containing 40 mM Tris-HCl, pH 7.5, 2 mM ouabain, 0.1 mM EGTA, 10 mM MgCl 2 , 5 M 8-azido-[ 32 P]ATP (1 Ci), and 800 M vanadate for 10 min at 37°C. The samples were then transferred to ice and diluted with 400 l of ice-cold Tris-EGTA buffer (0.1 mM EGTA and 40 mM Tris-HCl, pH 7.5). The membranes were pelleted in a microcentrifuge in the cold room (4°C), washed again with 400 l of ice-cold Tris-EGTA buffer, resuspended in 10 l of Tris-EGTA buffer, placed on ice, and irradiated for 2 min in a Stratalinker UV Crosslinker ( ϭ 254 nm). The labeled proteins were separated on a polyacrylamide gel (7%). The amount of radioactivity incorporated into MRP1 protein was determined by electronic autoradiography using a Packard Instant Imager, and all of the quantitative evaluations of the results expressed were based on these determinations.
Membrane Vesicle Transport-ATP-dependent transport of [ 3 H]leukotriene C 4 into the membrane vesicles was assayed by a rapid filtration technique (1,5). The assays were performed with 3 g of membrane protein in a 30-l reaction volume containing 50 mM Tris-HCl, pH 7.5, 250 mM sucrose, 10 mM MgCl 2 , 200 nM LTC 4 (17.54 nCi of [ 3 H]LTC 4 ) and 4 mM ATP (or as indicated in Fig. 1 legend). After incubation at 37°C for the periods indicated, the samples were diluted with 1 ml of ice-cold 1ϫ transport buffer (50 mM Tris-HCl, pH 7.5, 250 mM sucrose, and 10 mM MgCl 2 ), and filtered through a nitrocellulose membrane (0.2 m) that had been equilibrated with 1ϫ transport buffer. The filter was then washed with 10 ml of cold 1ϫ transport buffer, air-dried, and place in 10 ml of biodegradable counting scintillant (Amersham Pharmacia Biotech). The radioactivity bound to the nitrocellulose membrane was determined by liquid scintillation counting (Beckman LS 6000SC).

Photolabeling of MRP1 with N 3 [␣-32 P]-or N 3 [␥-32 P]ATP-
MRP1 is an exporter of a variety of conjugated hydrophobic substrates (1-7) and has been shown to have ATPase activity that is stimulated by these substrates (11,12). It is assumed that the NBDs of this ABC protein are responsible for its ability to hydrolyze ATP, but the mechanism whereby these two domains act is unknown. Previous studies of ATP hydrolysis by purified MRP1 and its sensitivity to ADP implicated separate binding sites for ATP and ADP on the protein (19). To evaluate the interactions of the two nucleotides with the NBDs of MRP1, photolabeling experiments have been performed using ␣-32 Pand ␥-32 P-labeled 8-azido-ATP. Fig. 1A shows that labeling with N 3 [␣-32 P]ATP is dependent on the divalent cations, Mg 2ϩ or Mn 2ϩ . Fig. 1B demonstrates that the labeling is greatly enhanced by increasing concentrations of vanadate, consistent with the trapping of an [␣-32 P]ADP-vanadate complex after hydrolysis of MgATP or MnATP (14,20). In fact there was only a trace amount of labeling in the absence of vanadate. This small amount might reflect the binding of intact N 3 [␣-32 P]ATP. If this were the case, labeling with N 3 [␥-32 P]ATP should occur and it did (Fig. 1C). Surprisingly, however, this labeling was also strongly stimulated by vanadate. This of course cannot reflect binding of N 3 ADP-vanadate, since it is not radioactively labeled after release of the ␥-32 P label during the hydrolysis step. Instead this labeling must be due to either binding of the whole N 3 ATP molecule or phosphorylation. To explain the former possibility, i.e. that vanadate stimulates the binding of the intact N 3 ATP, it is necessary to postulate that vanadate-induced trapping of N 3 ADP, a hydrolysis product of N 3 [␥-32 P]ATP, at one NBD enhances binding of the latter compound at the other NBD. That is to say vanadate trapping of N 3 ADP at one NBD, generated by hydrolysis of N 3 ATP, allosterically stimulates the binding of N 3 [␥-32 P]ATP to the other NBD. Identification of MRP1 fragments produced by limited proteolysis using site-specific monoclonal antibodies after photolabeling enables determination of which of the NBDs plays each of these roles (see Fig. 4).
Since the amount of labeling with ␣-32 P-labeled N 3 ATP could be due to both N 3 ATP binding and N 3 ADP trapping, whereas ␥-32 P-labeled N 3 ATP measures only binding of the triphosphate, the amount of photolabeling by N 3 [␥-32 P]ATP should be much less than that by N 3 [␣-32 P]ATP under the same conditions. Fig. 1D shows that this is indeed observed. This indicates that any phosphorylation of the protein that may contribute to the labeling by N 3 [␥-32 P]ATP is substantially less than the amount of N 3 [␣-32 P]ADP trapped. The actual contribution of phosphorylation as well as the sites of binding of N 3 ADP and N 3 ATP, their interactions, and the influence of MRP1 transport substrates were investigated in subsequent experiments.
To legitimize the use of 8-azido-ATP to probe functionally relevant nucleotide interactions with MRP1, it was necessary to determine how well it supported transport compared with ATP. Fig. 1E shows that if anything the derivatized nucleoside triphosphate supported the uptake of LTC 4 by everted membrane vesicles even better than ATP. In each case, the K1 ⁄2 was of the order of 50 M, a value similar to that found by others using just ATP (5).
Influence of Transport Substrates on Photolabeling-To examine the impact of agents involved in ATP-dependent transport by MRP1 on these implied separate binding sites, the effects of increasing concentrations of doxorubicin ( Fig. 2A) and glutathione (Fig. 2B), which is required for the transport of the drug by MRP1, on photolabeling by N 3 [␣-32 P]-and N 3 [␥-32 P]ATP were tested. There was only a minor enhancement of N 3 [␣-32 P]ATP labeling at low doxorubicin concentrations (1 and 10 M) followed by inhibition at a higher concentration (100 M). Such a biphasic response was not seen with N 3 [␥-32 P]ATP labeling, although 10 and 100 M concentrations caused inhibition. Hence the drug has slightly different effects on labeling with the two azido-ATPs, possibly because of different affinities of the two NBDs for them. The mild stimulation of labeling by N 3 [␣-32 P]ATP is similar to the increased ATP hydrolysis by purified MRP1 elicited by low concentrations of this drug (11), consistent with the need for hydrolysis for the ADP trapping reflected in the N 3 [␣-32 P]ATP labeling.
Reduced glutathione had a more striking biphasic effect on N 3 [␣-32 P]ATP labeling ( Fig. 2B) with lower concentrations causing greater than a 2-fold elevation, while high concentrations inhibited almost completely. In contrast, N 3 [␥-32 P]ATP labeling was much less sensitive to reduced glutathione with concentrations up to 8 mM causing only a mild stimulation before strong but incomplete inhibition at 10 mM. Hence glutathione had a much greater influence on labeling by N 3 [␣-32 P]than by N 3 [␥-32 P]ATP, emphasizing that incorporation of radioactivity by photolabeling with the differently labeled azidonucleotides probably reflects different binding events.
Influence of ADP on Photolabeling-Since we had observed previously that low concentrations of ADP accelerated ATP hydrolysis by MRP1, while higher concentrations inhibited (19), an impact of the nucleoside 5Ј-diphosphate on the photolabeling reactions might be expected. As seen in Fig. 3 labeling with N 3 [␣-32 P]ATP was unaffected by ADP concentrations less than approximately 5 M and inhibited at higher concentrations consistent with competition for N 3 ADP trapping. In striking contrast, labeling with N 3 [␥-32 P]ATP was progressively increased by ADP concentrations up to 5 M, where nearly a 5-fold elevation occurred. Clearly, this large stimulatory effect by an ADP concentration equal to that of N 3 ATP must be due to action at a site with much higher affinity for the nucleoside diphosphate than for the triphosphate and is reminiscent of the effect of the nucleoside 5Ј-diphosphate on hydrolysis (19). Much higher concentrations of ADP were required before this stimulatory effect was overcome to yield a net inhibition, probably due to direct competition for N 3 ATP binding. It seems reasonable to conclude that the promotion of N 3 ATP binding observed here rather than N 3 ADP trapping may be responsible for the increase in the overall hydrolysis reaction (19).
Localization of ATP and ADP Binding Sites-To gain some insight into the relationship of these putative different binding sites to the NBDs of MRP1, trypsin digestion was carried out after photolabeling with N 3 [␣-32 P]-or N 3 [␥-32 P]ATP. Cleavage fragments containing NBD1 or NBD2 were identified using two different monoclonal antibodies raised against purified MRP1. 2 The first antibody, 42.4, recognizes an epitope formed by residues 723-732 between the Walker A and B motifs of NBD1 (Fig. 4A). The epitope of the second antibody, 897.2, has been localized only to the large region between residues 1248 and 1531 that contains the entire NBD2 (Fig. 4A). Western blots probed with these antibodies after partial trypsin digestion yielded distinct banding patterns (Fig. 4B). The NBD1-specific antibody, 42.4, detected three major bands of approximately 188, 119, and 53 kDa as well as several minor bands, e.g. 155 and 43 kDa (Fig. 4B, right panel). Intact MRP1 migrates at the 188-kDa position. The 119-kDa band is of the same size as a fragment detected by another antibody to MRP1, QCRL-1, after trypsin digestion of the protein (21). Although identification of the exact cleavage site responsible for its generation is complicated by the attachment of N-linked oligosaccharide chains to the most N-terminal extracytoplasmic segment (22), this band must contain NBD1. The 53-kDa band and the other minor bands detected by this antibody must also contain at least the central portion of NBD1 where the epitope is located. Since the other monoclonal antibody, 897.2, recognizes a recombinant polypeptide corresponding to the C-terminal amino acids 1248 -1531, the only major trypsin fragment it detects, in addition to the intact protein and a minor band of 155 kDa, migrating at 65 kDa must contain NBD2 (Fig. 4B, left panel).
When the protein labeled with N 3 [␣-32 P]ATP was partially digested with trypsin, five bands of 188, 155, 119, 65, and 53 kDa were labeled (Fig. 4C, lanes 1-6). The 155-kDa band can be recognized by either antibody and therefore is probably an N-terminal truncation fragment. The 119-and 53-kDa fragments were recognized by 42.4 and therefore are NBD1 fragments. The 65-kDa fragment was recognized by 897.2 and is thus an NBD2 fragment. The amount of radioactivity associated with the 65-kDa band is nearly twice that associated with 119-kDa ϩ 53-kDa bands, indicating that the product of N 3 [␣-32 P]ATP hydrolysis, N 3 [␣-32 P]ADP, is preferentially trapped in NBD2.
Four bands, all corresponding in size to those detected in Western blots probed with the NBD1-specific 42.4 antibody, were observed in trypsin digests after labeling with N 3 [␥-32 P]ATP (Fig. 4C, lanes 7-12). The 65-kDa band detectable by 897.2 was clearly not labeled. Therefore, only NBD1 seems to have been labeled by N 3 [␥-32 P]ATP. The simplest interpretation of these combined results is that binding of the intact N 3 ATP molecule, which could be reflected by labeling with either N 3 [␥-32 P]-or N 3 [␣-32 P]ATP labeling occurs only at NBD1, whereas the trapping of N 3 [␣-32 P]ADP after the hydrolysis of N 3 [␣-32 P]ATP definitely occurs at NBD2 and possibly also at NBD1. If this is the case, then the strong stimulation of labeling with N 3 [␥-32 P]ATP (occurring at NBD1) by low concentrations of ADP (binding at NBD2) should reflect a positive allosteric interaction between the two NBDs.
To further test the suggestion that N 3 [␣-32 P]ADP may be trapped at both NBDs but to a lesser extent at NBD1 than NBD2, the effects of ADP on the labeling of the tryptic fragments containing each of the domains was studied (Fig. 4D).  3), whereas that of the 119-and 53-kDa fragments reflecting NBD1 decreased only approximately 10 -20% at 5 and 10 M ADP. This indicates a much stronger inhibition by ADP of labeling of NBD2 than NBD1 and is consistent with the interpretation that there is preferential binding of ADP to NBD2 and weaker binding to NBD1 (the exclusive site of intact N 3 ATP binding).
The incorporation of 32 P radioactivity into MRP1 on incubation with N 3 [␥-32 P]ATP could occur by a phosphorylation reaction as well as by the binding of the intact nucleoside triphosphate. Indeed while there was no labeling by N 3 [␣-32 P]ATP without UV irradiation (Fig. 4E, lane 1) there was labeling by N 3 [␥-32 P]ATP (Fig. 4E, lane 2). Addition of ADP increased this labeling by less than 2-fold (Fig. 4E, lanes 3-5). UV irradiation 2 X.-b. Chang and J. R. Riordan, unpublished data.  6 and 13), 250 (lanes 7 and 14). B, quantitative comparison by electronic autoradiography. Labeling in the absence of added ADP was assigned as 100%. to promote photolabeling by the intact substrate augmented incorporation of radioactivity substantially (Fig. 4E, lane 6  versus lane 2), presumably reflecting the binding of N 3 [␥-32 P]ATP to NBD1 revealed in the trypsinolysis experiments.
This irradiation-dependent labeling was increased approximately 4-fold by the addition of 10 M ADP (Fig. 4E, lane 9  versus lane 2), again confirming the observations of Fig. 3B. Thus while some of the radioactivity incorporated into MRP1 The epitope of antibody 42.4 is a decapeptide (residues 723-732) located between the Walker A and B motifs of NBD1 (mAb1). The epitope of antibody 897.2 has been localized only to the large region between residues 1248 and 1531, which covers the entire NBD2 (mAb2). The approximate sites of two trypsin cleavages (TS1 and TS2), which can be deduced, are also indicated. B, Western blots following photolabeling and trypsinolysis. To determine whether ␣-32 P-or ␥-32 P-labeled 8-azido-ATP bound preferentially to NBD1 or NBD2, the same samples were used to do Western blots and autoradiography (C). 36 g of MRP1-containing membrane proteins was incubated in 20 l of the same incubation mixture as in previous figures (lanes 1-6 with 5 M N 3 [␣-32 P]ATP; lanes 7-12 with 5 M N 3 [␥-32 P]ATP). 5 M ADP was added to the incubation mixture with N 3 [␥-32 P]ATP to increase the amount bound. After removal from unbound nucleotide and UV irradiation, the samples were then digested in 30 l of 40 mM Tris-HCl, pH 7.4, 1 mM EDTA for 10 min at room temperature with varying amounts of trypsin. One-third of each sample was then electrophoresed for Western blotting. Fragments in lanes 1-6 were detected by probing with monoclonal antibody 897.2 (recognizing NBD2 fragments). Those in lanes 7-12 were probed with antibody 42.4 (recognizing NBD1 fragments). Trypsin to membrane protein mass ratios were: 0 (lanes 1 and 7); 1:80 (lanes 2 and 8); 1:40 (lanes 3 and 9); 1:20 (lanes 4 and 10); 1:10 (lanes 5 and 11); 1:5 (lanes 6 and 12). Arrows indicate the major trypsin digestion products detected by each of the antibodies. C, autoradiograms following photolabeling and trypsinolysis. Two-thirds of the samples from B were subjected to electrophoresis and autoradiography. Samples in lanes 1-6 were labeled with N 3 [␣-32 P]ATP and samples in lanes 7-12 with  1, 2, and 6), 1 M ADP (lanes 3 and 7), 5 M ADP (lanes 4 and  8), and 10 M ADP (lanes 5 and 9). After removing the unbound ATP, the samples in lanes 1-5 were loaded onto the gel without UV irradiation. The samples in lanes 6 -9 were UV-irradiated for 2 min at 254 nm and then loaded. from N 3 [␥-32 P]ATP is due to phosphorylation, the largest part of the stimulation by ADP is due to promotion of binding of intact N 3 ATP. Notably since the trypsin digestion after photolabeling with N 3 [␥-32 P]ATP in the presence of 5 M ADP (Fig.  4C) revealed labeling of only NBD1 fragments (119 and 53 kDa), it would appear that both phosphorylation and N 3 [␥-32 P]ATP binding are limited to NBD1. The phosphorylation reaction has not yet been extensively characterized, but it is also readily detected using [␥-32 P]ATP without the photoactivable azido group.
Interdependence of Nucleotide Binding Domains-If ATP is preferentially bound at NBD1 and ADP can be trapped at NBD2, then mutations within the Walker motifs of NBD1 and NBD2 should influence photolabeling by N 3 [␥-32 P]-and N 3 [␣-32 P]ATP. The Walker A lysines and Walker B aspartates in the NBDs were mutated as indicated in Fig. 5A. Each of these variants is expressed as fully mature glycoproteins like the wild-type except for the Walker B mutant in NBD1 (Fig. 5B). It produced only a core-glycosylated immature band indicative of biosynthetic misprocessing, and it could not be labeled by either N 3 [␣-32 P]-or N 3 [␥-32 P]ATP. Fig. 5C shows that although labeling by N 3 [␣-32 P]ATP was not abolished by the mutations, it was greatly reduced: K684L was ϳ10% of wild-type; K1333L, ϳ5%; D1454L/D1455L, ϳ15%. Therefore labeling with N 3 [␣-32 P]ATP requires both NBDs to be functionally competent. Fig.  5D demonstrates that labeling of K684L by N 3 [␥-32 P]ATP was almost eliminated and labeling of K1333L and D1454L/E1455L were decreased to ϳ10% and ϳ15% of the wild-type levels, respectively. Hence these data strongly support the concept of asymmetry but interdependence of the two NBDs of MRP1. Although most N 3 ADP trapping is at NBD2, the function of both domains is involved in this trapping, since it is nearly abolished by mutations in either. The interaction of N 3 [␥-32 P]ATP is principally with NBD1 but also requires the function of two wild-type NBDs. This apparent cooperative functioning of the two domains may increase the energetic efficiency of this drug transporter (19).
To directly determine the role of each of the NBDs in transport, membrane vesicles from cells expressing the Walker A lysine or B aspartate mutations were used to measure LTC 4 uptake. The K1333L mutation in NBD2 nearly abolished ATPdependent uptake as did the NBD2 Walker B substitution, whereas the K684L substitution reduced it by approximately half. Hence there apparently is not an exact correspondence between the effects of the mutations on photolabeling and transport. On the other hand, the function of NBD2, the predominant site of trapping of the product of ATP hydrolysis, seems also to be more critically required for transport. DISCUSSION Since the genes for ABC proteins constitute the largest super family in the genomes analyzed to date, understanding their structure and function is of great importance. Their fundamental unit consists of a nucleotide binding domain and a transmembrane domain containing multiple membrane spanning segments and the extramembraneous loops joining them. In the exporter families of eukaryotes, two such units are fused to form large membrane proteins, some of which contain additional domains as well. N-terminal of the two fundamental units of members of the MRP1 family there is a third transmembrane domain and a large cytoplasmic loop joining it to the rest of the protein that are required for transport (23). In the case of P-glycoprotein, an exporter of a broad range of hydrophobic compounds, it is generally accepted that energy for transport is derived from ATP hydrolysis at each of the NBDs in an alternating fashion (14,15), which may be at least conceptually similar to a two cylinder combustion engine. However there are preliminary indications that this equivalency of function of the two domains may not obtain for members with two other ABC proteins of a different family than the P-glycoprotein but similar to the MRPs. The sulfonyl urea receptor (SUR1) regulator of inwardly rectifying K ϩ channels is a bona fide member of the MRP1 family on the basis of sequence similarity (13), and the cystic fibrosis transmembrane conductance regulator chloride channel may be considered a member on the basis of homology of the NBDs (10). ATP hydrolysis by these two proteins has not been extensively characterized but photoaffinity labeling by 8-azido-ATP has been reported. Mu- tagenesis of Walker motif amino acids of NBD1 in SUR1 prevented trapping of the 8-azido-ADP hydrolysis product but the same changes in NBD2 did not (24). The influence of ADP on photolabeling, however, was prevented by mutating those residues in NBD2 (25). Hence the NBDs of SUR1 apparently are not functionally equivalent. Less information is available for cystic fibrosis transmembrane conductance regulator, which does bind (26) and hydrolyze ATP (27), but trapping of N 3 ATP after photolabeling with N 3 [␣-32 P]ATP has been reported to occur predominantly at NBD1 (26).
In this study we have analyzed the photolabeling of MRP1 with N 3 [␥-32 P]ATP which detects binding of the intact molecule before hydrolysis (and also phosphorylation if it occurs) and with N 3 [␣-32 P]ATP, which measures trapping of N 3 [ 32 P]ADP formed by hydrolysis as well as binding of the uncleaved substrate. Consistent with this, the amount of photolabeling on MRP1 was much greater using the ␣-32 P-than the ␥-32 P-labeled substrate. The ␣ labeling was found to be dependent on the presence of a divalent cation as is ATP hydrolysis (11,12) and augmented by vanadate, which would be expected to promote trapping of N 3 ADP as N 3 ADP-vanadate (14,20). Unexpectedly, however, it was found that labeling with the ␥-32 Plabeled compound was similarly stimulated by vanadate. This of course could not reflect N 3 ADP trapping. A previous study of ATP hydrolysis by MRP1 (19) had pointed to the presence of separate binding sites for ATP and ADP on the protein. Therefore, we postulated that enhancement of intact N 3 ATP binding to the ATP site by vanadate might occur as a consequence of an allosteric response to the increased trapping of N 3 ADP at the ADP site. This interpretation was supported by the observation that additional low concentrations of ADP caused a large increase in labeling with ␥-32 P-labeled N 3 ATP.
To physically identify these separate but interactive binding sites, fragments generated by partial trypsin digestion after photolabeling were identified by site-specific monoclonal antibodies. The results obtained showed that intact N 3 ATP bound exclusively at NBD1, whereas trapping of the hydrolysis product, N 3 ADP, occurred primarily at NBD2. These experiments are of exactly the same type as those used to show that the trapping of [␣-32 P]ADP after hydrolysis of [␣-32 P]ATP occurs to the same extent at each NBD of P-glycoprotein (14). Therefore it is reasonable to conclude that the greater sequence dissimilarity in the NBDs of MRP1 than in P-glycoprotein is reflected by greater functional distinction also. Despite this, the function of one domain is coupled to that in the other. This can be concluded from two facts. First, inactivation of either domain by mutagen- FIG. 6. Schematic representation of the interactions of ATP and ADP with the NBDs of MRP1. In the initial unoccupied molecule (far left) the squares represent the three membrane-spanning domains (MSDs). The circles represent the two cytoplasmic NBDs. The triangles represent the linker region between MSD1 and MSD2. The different shapes of the triangles or circles represent conformational changes during ATP binding and hydrolysis and conjugated anion (solute) transport. The conformation of the MSDs may also be changed during the cycle, but this is not indicated in this model. The diagram of the conjugated anion transported includes two parts, one hydrophobic (small square) and the other hydrophilic (tear drop). In this model the solute binds (step 1) to part of MSD1 (hydrophobic interaction) and the linker region between MSD1 and MSD2 (hydrophilic interaction). This suggestion is based on the finding that the MSD1 and the cytoplasmic domain linking the MSD1 to MSD2 are required for active transport of leukotriene C 4 (23). This solute binding leads to a conformational change of NBD2 and increases its affinity for ATP (step 2). This assumption is based on the observations that doxorubicin and glutathione stimulated 8-azido-[␣-32 P]ATP binding, but not 8-azido-[␥-32 P]ATP binding (Fig. 2) and that 8-azido-[␣-32 P]ADP is trapped mainly at NBD2 (Fig. 4C). The solute is then extruded from the cell during ATP hydrolysis (step 3) and release of P i (step 4). Two lines of evidence support this interpretation. First, membrane vesicles prepared from MRP1 expressing cells possess the ability to take up solute in the presence of ATP, but not AMP (Fig. 1E), i.e. ATP hydrolysis is required for transport. Second, doxorubicin and glutathione stimulate 8-azido-[␣-32 P]ADP trapping, which reflects not only ATP binding but also hydrolysis. The trapping of ADP, the product of hydrolysis, suggests that phosphate is released first so that it can be trapped by vanadate (step B). The ADP-bound form of NBD2 causes a conformational change in NBD1, which increases its affinity for ATP (steps 4 and A, step A indicates that the binding exogenously added ADP to NBD2 in an unoccupied molecule has the same effect as the ATP hydrolysis product). This interpretation is based on the observations that ADP stimulates 8-azido-[␥-32 P]ATP labeling approximately 5-fold (Figs. 3 and 4E). ATP binds to a molecule in which NBD2 is already occupied by ADP (step 5). This assumption is supported by the facts that ADP stimulates intact ATP binding severalfold (Fig.  4E),and the binding of intact 8-azido-[␥-32 P]ATP is vanadate-ependent ( Fig. 1C and step C). NBD1 transfers ␥-phosphate from ATP to the protein (step 6) and then to the ADP at NBD2 (step 7). The suggestion of P i binding at NBD1 is based on the facts that MRP1 protein is phosphorylated by 8-azido-[␥-32 P]ATP (Fig. 4E); the level of this phosphorylation is increased almost 2-fold by ADP (Fig. 4E), and 8-azido-[␥-32 P]ATP binds mainly to NBD1 fragment (Fig. 4C). The release of ADP at NBD1 leaves an empty NBD1 and an ATP-bound NBD2 (step 7). Solute binds to the MRP1 protein with an ATP already at NBD2 and then starts another cycle of ATP hydrolysis and solute transport (step 8). Steps B to D indicate that: 1) trapping of N 3 ADP is vanadate dependent (Fig. 1B); 2) binding of intact ATP (step C) is also vanadate-dependent (Fig. 1C); 3) MRP1 protein is phosphorylated (step D), ADP stimulated the level of phosphorylation nearly 2-fold (Fig. 4E) and the phosphorylation of the protein is also vanadate-dependent (step D and Fig. 1C); 4) N 3 ADP is trapped at NBD1 (step D), but the amounts of ADP trapped at NBD1 are much less than at NBD2 (steps B to D and Fig. 4C); 5) ADP competitively inhibits the trapping of N 3 ADP at NBD2, but has much less effect on NBD1 (steps B to D and Fig. 4D). esis of Walker motif residues greatly reduces the primary role of both, i.e. ATP binding at NBD1 and ADP trapping at NBD2. Second, the latter has a strong positive allosteric effect on the former. The sketch in Fig. 6 is an attempt to visualize events involved in nucleotide binding and hydrolysis as they relate to the transport of an organic anion by MRP1. Clearly the model is derived from the original one for P-glycoprotein formulated by Senior and colleagues (15). The modifications and extensions are made to accommodate the characteristics of nucleotide binding to MRP1 reported, some of which are quite different from those to P-glycoprotein. Since the two proteins do belong to separate families of the ABC superfamily, it may not be surprising that variations on the basic mechanism of the fundamental ABC units are exhibited. It will be of interest to learn if there are similar or further variations as other similar and dissimilar ABC proteins are examined. In the case of MRP1 and the closely related MRP2 further details should be discerned from studies of these purified proteins (11,12,28).