Properties of P-glycoprotein with Mutations in the “Catalytic Carboxylate” Glutamate Residues*

It is known from earlier work that two conserved Glu residues, designated “catalytic carboxylates,” are critical for function in P-glycoprotein (Pgp). Here the role of these residues (Glu-552 and Glu-1197 in mouse MDR3 Pgp) was studied further. Mutation E552Q or E1197Q reduced Pgp-ATPase to low but still measurable rates. Two explanations previously offered for effects of these mutations, namely that ADP release is slowed or that a second (drug site-resetting) round of ATP hydrolysis is blocked, were evaluated and appeared unsatisfactory. Thus the study was extended to include E552A, -D, and -K and E1197A, -D, and -K mutants. All reduced ATPase to similar low but measurable rates. Orthovanadate-trapping experiments showed that mutation to Gln, Ala, Asp, or Lys altered characteristics of the transition state but did not elimi-nate its formation in contrast e.g. with mutation of the analogous catalytic Glu in F 1 -ATPase. Retention of ATP as well as ADP was seen in Ala, Asp, and Lys mutants. Mutation E552A in nucleotide binding domain 1 (NBD1) was combined with mutation S528A or S1173A in the LSGGQ sequence of NBD1 or NBD2, respectively. Synergistic effects were seen. E552A/S1173A had extremely low turnover rate for ATPase,

Expression of P-glycoprotein (Pgp) 1 in plasma membranes of cells causes resistance to multiple, structurally diverse drugs in cancer and AIDS patients (for reviews, see Refs. 1 and 2). Pgp is a member of the ABC transporter family. It confers drug resistance by coupling ATP binding and hydrolysis at two cytoplasmic nucleotide binding domains (NBDs) to drug export via two transmembrane domains. The first mechanistic model for drug transport (3) proposed that ATP is hydrolyzed alter-nately at the two NBDs with collapse of the chemical transition state and release of product P i providing the driving force for drug export via coupled long range conformational changes between the NBDs and the transmembrane domains. This model has gained acceptance as a general framework with recently suggested modifications and refinements (4,5). Defining the mechanism of Pgp in detail is an important goal because it relates to conceiving strategies to disable or circumvent the action of Pgp in patients and to understanding the mechanism of ABC transporters generally.
Biochemical and mutagenesis studies have shown that the NBDs of Pgp act cooperatively (3) and suggested that they alternate between "open" and "closed" conformations (6,7). We have proposed that ATP initially binds to each of the NBDs while Pgp is in the open conformation, and then the NBDs approach each other to form an integrated structure, the closed conformation, in which catalytic side chains required to stabilize the ATP hydrolysis transition state complex are made available from both NBDs. After hydrolysis of ATP coupled to drug movement from inner membrane leaflet to the outer side of the membrane, the NBDs dissociate to re-form the open conformation and then release the ADP product (6,7). Data showing binding of ATP at two sites in Pgp with relatively weak affinity (8) has provided evidence for an open conformation, while evidence for a closed conformation consisting of interdigitated NBDs has come from cross-linking studies (9 -11) and from a recent study in which combined mutagenesis of the Pgp NBDs produced an occluded conformation that binds ATP and ADP very tightly (see below). In other ABC transporters there is extensive evidence for dimerization of the NBDs, producing a conformation that binds nucleotide at the interface between two NBDs. Examples include x-ray structural data (12)(13)(14)(15), photocleavage studies (16), and gel filtration studies of NBD subunit dimers (17)(18)(19). Mutagenesis of specific conserved Glu residues termed "catalytic carboxylates" facilitated NBD dimerization in several of these studies while at the same time reducing ATPase activity to very low rates.
In several, diverse ATPase enzymes Glu residues have been implicated from their location in x-ray structure models of the catalytic sites to play a critical role in ATP hydrolysis by immobilizing, polarizing, and stereochemically orienting a water molecule in position for in-line nucleophilic attack at the ␥-phosphate of ATP (20). Prominent early examples were recA protein (21) and F 1 -ATPase (22). Biochemical evidence has supported this concept. For example, in F 1 -ATPase the relevant Glu has been shown to be directly involved in bond formation in the transition state complex (23)(24)(25), and modulation of the carboxylate side chain location by substitution with Asp or iodoacetic acid-reacted Cys gravely impaired catalysis (26). Beginning with the x-ray structure of HisP NBD subunit (27) and later in several other x-ray structures, it was apparent that a conserved Glu residue immediately adjacent to and downstream of the conserved "Walker B" Asp residue might provide the same catalytic carboxylate function in ABC transporter NBDs.
A mutagenesis screen of carboxylate residues by Urbatsch et al. (28) underscored the importance of the highly conserved, putative catalytic carboxylate residues Glu-552 and Glu-1197 in mouse MDR3 Pgp. Mutagenesis of either residue to Gln yielded Pgp with strongly impaired drug transport and ATPase activities, although mutant proteins did show orthovanadate (V i )-induced trapping of 8-azido-ADP upon incubation with 8-azido-ATP and V i , suggesting that at least one turnover of 8-azido-ATP hydrolysis as far as the P i release step was occurring. Urbatsch et al. (28) proposed that in the mutants release of product ADP was unusually slow (28). Later studies from the same laboratory indicated that mutants E552Q and E1197Q induce asymmetric conformations of the two NBDs (29,30). Sauna et al. (31) studied the equivalent conserved glutamates in human MDR1 Pgp, introducing Gln and Ala mutations. Consistent with the earlier data, both drug transport and ATPase activity were strongly impaired in mutant proteins, and V i -induced trapping of 8-azido-ADP was seen. However, these workers differed from Urbatsch et al. (28) as to the impact of the mutations, concluding that a second round of ATP hydrolysis, required to "reset" the drug-binding sites after an initial outward transport-generating ATPase turnover, was blocked. Sauna et al. (31) also found that when Ala or Gln mutations were introduced simultaneously into both NBDs, the resultant double mutants showed V i -independent occlusion of 8-azido-ADP upon incubation with 8-azido-ATP along with loss of ATPase and drug transport.
In a recent study we further characterized nucleotide binding and hydrolysis properties of pure mouse MDR3 Pgp in which Glu residues at both positions 552 and 1197 were simultaneously substituted with Ala, Gln, Asp, or Lys (32). Using a sensitive radioactive assay, we found that the mutants displayed real but very low ATPase activity and that inability to form the normal transition state, rather than slow ADP release, was the primary defect. Both E552Q/E1197Q and E552A/E1197A mutants were able to bind MgATP and MgADP with high affinity, suggesting that they were trapped in the closed conformation of Pgp, which, we propose, occurs immediately before the hydrolytic step (see above).
Therefore it was apparent that mutations of the catalytic carboxylate glutamate residues in Pgp have the potential to clarify several aspects of mechanism. Here we present detailed studies of effects of mutations Glu to Gln, Ala, Asp, or Lys, present singly in either NBD, using pure mouse MDR3 protein and the naturally occurring ligands MgATP and MgADP. First we evaluated two previous suggestions, namely that ADP release (28) or a second round of ATPase turnover (31) is defective in Gln mutants. Second, we extended the studies using Ala, Asp, and Lys mutants. Third, we combined the mutations S528A and S1173A, involving mutations of the conserved Ser residue in the "LSGGQ" ABC signature sequence of NBD1 or NBD2, with the catalytic carboxylate E552A mutation of NBD1 to assess the effects of mutations "in cis" or "in trans" at the putative NBD dimer interface. . Other mutations at positions Glu-552 and Glu-1197 were generated by PCR mutagenesis as described previously (32).

Materials
Mutants E552A/S528A and E552A/S1173A were generated by transferring a BglI-SmaI fragment containing the S528A mutation or an SpeI-SnaB1 fragment containing S1173A (33) into pHILD-mdr3.6-His6 containing E552A (32). All mutant proteins contained a C-terminal His 6 tag and were expressed in P. pastoris and purified as described previously (32,33). Protein concentration was determined by reference to a standard calibrated by amino acid analysis as described previously (33).
Activation of Pgp by DTT and Lipids, Assay of ATPase Activity, and Centrifuge Column Elution of Pgp-These procedures were carried out as described previously (32,33). Briefly Pgp was activated by preincubation with 8 mM DTT and E. coli lipids at ratio of 50:1 (for ATPase measurement) or 2:1 (for nucleotide-trapping experiments). Release of [ 32 P]P i from [␥-32 P]ATP was determined using the charcoal (Norit) adsorption assay using 0.5-1 g of activated Pgp. When present, verapamil (150 M) was added directly to the ATPase assay buffer. K m(MgATP) and K i(MgADP) were determined as in Refs. 32 and 33.
Trapping of Nucleotide by V i -Trapping of nucleotide by Pgp after preincubation with [␣-32 P]ATP or [8-14 C]ADP and V i utilized centrifuge column elution as described previously (32,33). Briefly the trapping procedure entailed preincubation of Pgp (5-10 g) with 200 M [␣-32 P]ATP or 100 M [8-14 C]ADP, 200 M V i , and 2 mM MgSO 4 , with or without 150 M verapamil, at 37°C for 20 min (with ATP) or 120 min (with ADP) followed by centrifuge column elution to separate bound from free ligands and counting of eluates to determine bound nucleotide. Identification of the trapped nucleotide (ATP versus ADP) was by TLC analysis (32). Rate of onset of trapping was carried out at room temperature to slow the reaction sufficiently to allow accurate determination. Rate of release of trapped nucleotide was measured by allowing centrifuge column eluates to incubate for varied times at 37°C and then passing the eluates through a second centrifuge column to determine remaining bound nucleotide. Where sequential rounds of V i trapping were carried out the procedure was modified as follows. In the first round of V i trapping, 90 g of activated Pgp was incubated in 40 mM Tris-HCl, pH 7. Samples were placed on ice for 2 min, and then 2 ϫ 100-l aliquots were passed through centrifuge columns at 4°C that were pre-equilibrated with 40 mM Tris-HCl, pH 7.4, 0.1 mM EGTA, 10% glycerol, 10 mM DTT. The eluates were combined and quickly counted by the Cerenkov method, and a 10-l sample was removed for protein quantitation. MgATP was added to a concentration of 1 mM. The eluates were then incubated at 37°C for 180 min (wild type and E552Q) or 90 min (E1197Q) to allow release of trapped nucleotide. Next samples were placed on ice for 2 min followed by passage through a centrifuge column equilibrated as above. After quickly counting by the Cerenkov method, 5 l was removed for protein quantitation. For the second round of V i trapping, the remaining sample was incubated under conditions identical to the first round of trapping in 100-l total volume at 37°C for 20 min. Samples were then placed on ice for 2 min followed by passage through centrifuge columns at 4°C. Eluates were counted by the Cerenkov method, and protein was quantitated. It was noted that Pgp (wild type or mutant) lost activity if DTT or ATP was omitted at any of the centrifuge column steps.

RESULTS
As remarked in the Introduction, there is agreement that the conserved Glu residues (Glu-552 and Glu-1197 in mouse MDR3 Pgp) that lie immediately adjacent and C-terminal to the Walker B Asp residues in the NBDs are catalytically important in P-glycoprotein. However, authors of previous studies differed as to the proposed mechanism by which mutations to Gln impaired function (28,31). Here we further studied the roles of these Glu residues. First we re-evaluated the earlier conclusions. Second, we made mutations to Ala, Asp, and Lys, and we compared data obtained with these mutants. Third, we combined the E552A mutation in the N-terminal NBD with a mutation Ser to Ala in the LSGGQ ABC signature sequence of either the N-terminal or the C-terminal NBD. Pure mouse MDR3 Pgp obtained from P. pastoris grown in fermenter culture was used throughout. Yields of all mutant proteins were the same as for wild-type Pgp. All purified proteins showed a single band on SDS gels as for example in Ref. 33.

Effects of E552Q and E1197Q Mutations
ATPase Activity-Pgp containing E552Q and E1197Q mutations in the N-and C-terminal NBDs, respectively, was purified and activated with lipids and DTT. ATPase activity was measured using [␥-32 P]ATP in a sensitive charcoal adsorption assay. Results are shown in Fig. 1, A and B (note differences in vertical axes). Calculated turnover rates are given in Table I with wild-type enzyme assayed under the same conditions shown for comparison. Duplicate purifications of each mutant enzyme gave the same results. Contrary to previous results (28,31) we found that both E552Q and E1197Q mutants undergo multiple turnovers during the time course of the assay (contrast these data with the zero or close to zero activity of mutant E552A/S528A shown in Fig. 6B also reproduced in duplicate Pgp purifications). Moreover there was apparent "asymmetry" in that the mutation in the C-terminal NBD (E1197Q) was more impairing than the one in the N-terminal domain (E552Q) in regard to turnover and drug stimulation. As with wild type, both mutant enzymes showed activation and then inhibition when titrated with increasing concentrations of verapamil (data not shown). The concentration of verapamil yielding maximal activity was ϳ150 M for all three enzymes. V i potently inhibited wild-type and both mutant enzymes. IC 50 values were: wild-type, 3 M; E552Q, 0.1 M; E1197Q, 0.2 M (data not shown). The enhanced apparent affinity for V i in mutants is consistent with previous work (30). K m values were calculated from plots of ATPase activity versus MgATP concentration ( Fig. 2) and are given in Table I. Since for technical reasons the maximal ATP concentration used was 2 mM, which was insufficient to reach saturation in Fig. 2, A and B, these values are apparent only. However, they are useful for comparisons, showing that both mutants had lower K m values than wild type with E1197Q producing a larger effect.
Determination of K i(ADP) - Fig. 2 shows ATPase activity in the presence of varied concentrations of MgADP from which K i(MgADP) was calculated (Table I). It is seen that the E552Q and E1197Q mutants were not very different from wild type, indicating that ADP binding affinity is not significantly altered by the mutations.
Release and Formation of V i -trapped ADP- Fig. 3 shows release of [␣-32 P]ADP after prior trapping with V i and [␣-32 P]ATP. Rates and half-times for release are shown in Table II (columns three and four). The data for wild type (t1 ⁄2 ϭ 77 min) are in good agreement with previous reports (32,33). Release of ADP was faster in the E1197Q mutant (t1 ⁄2 ϭ 17 min) than in wild type, whereas E552Q (t1 ⁄2 ϭ 60 min) was similar to wild type. Formation of V i -trapped ADP was faster in both mutants as compared with wild type (Table II, column one) consistent with more potent inhibition of ATPase by V i (above). Trapped V i -ADP is thought to mimic a transition state of ATP hydrolysis, therefore one may conclude that the mutations have altered the nature of the transition state but have not abolished it.
Sequential Rounds of V i Trapping of Nucleotide-In Fig. 4, V i trapping was carried out using [␣-32 P]ATP, and the stoichiometry of trapped ADP was measured ("1 st Round Trapping"). In wild-type and both mutant proteins it was close to 1 mol/mol of Pgp. Trapped ADP was then released by extended incubation (wild type and E552Q, 3 h; E1197Q, 90 min). DTT, glycerol, and ATP were included in centrifuge column buffers to avoid enzyme denaturation. Stoichiometry of nucleotide remaining bound is shown ("After Releasing 1 st Trap"). A second round of V i trapping with [␣-32 P]ATP was carried out, and it was seen that the stoichiometry of trapped nucleotide was again 1 mol/ mol of Pgp. The same results were seen with mutant E1197A. Previously a similar experiment was reported with human Pgp using 8-azido-ATP, and it was concluded that Gln mutants differed from wild type in being unable to undergo a second round of trapping (31). As shown here, this was not the case when the natural ATP was used with mouse Pgp.   2, A and B).
Summarizing this section, the data show that the E552Q and E1197Q mutant mouse Pgp carry out multiple ATPase turnovers but at a much reduced rate compared with wild type. Affinity for MgADP is not altered, and the rate of release of V i -trapped ADP is not slowed, suggesting that deceleration of product ADP release is not the cause of functional impairment. The facts that multiple turnovers of ATPase occur and that sequential trapping of ADP by V i is not impaired suggest that the explanation that the mutations prevent a second ATPase turnover required to reset the drug sites is also unsatisfactory.

Effects of Ala, Asp, and Lys Mutations at Residues Glu-552 and Glu-1197
To further study the role of the catalytic Glu residues we introduced Ala, Asp, and Lys at positions 552 and 1197 in mouse MDR3 Pgp. ATPase activities of the pure proteins were measured as in Fig. 1, and turnover rates are given in Table III. In each case activity was linear for 4 h as in Fig. 1. Activities of  these mutants were very low, and drug stimulation was much reduced. The results, especially with the Asp mutants, reveal a strict requirement for the wild-type Glu carboxyl for normal catalysis. 2 Asymmetry between effects of mutations in N-and C-terminal NBDs was seen to some extent with Lys but not with Ala and Asp mutants. Apparent K m(MgATP) values were determined (Table III); again asymmetry was not apparent in Ala or Asp mutants. When incubated with V i and [␣-32 P]ATP, all of the mutants trapped significant amounts of radioactive nucleotide (Fig. 5A). Thin layer chromatography was carried out to determine the nature of the trapped nucleotide. In wild type there was zero ATP trapped under these experimental conditions, confirming previous data (32,34). Both E552Q and E1197Q behaved similarly to wild type; however, in the other mutants a substantial fraction (26 -47%) of the trapped nucleotide was in the form of ATP (data not shown). It should also be mentioned that whereas the amounts of nucleotide retained after centrifuge column elution in the absence of V i were negligible in wild-type, Gln, Asp, and Lys mutants, with E552A and E1197A, V i -independent retention amounted to 0.4 and 0.15 mol of nucleotide/ mol of Pgp, respectively, of which ϳ75% was ATP (data not shown). Half-times for release of the V i -trapped nucleotide are given in Table II. In all cases, release occurred at a faster rate than in wild type. Rates of onset of nucleotide trapping in the presence of V i and [␣-32 P]ATP were measured at room temperature to slow the reaction and are given in Table II. This analysis revealed that the Lys mutants trapped nucleotide much more slowly than wild type and that the other mutants were generally similar to wild type. Fig. 5B shows [ 14 C]ADP trapping in the presence of V i for each of the mutants. In all cases trapping of the nucleotide was strongly stimulated by verapamil, showing that none of the mutations affected communication between the catalytic sites and the drug-binding sites. The Gln, Ala, and Asp mutants trapped significant amounts of ADP, but the Lys mutants were much less able to do so. If V i was omitted, there was negligible trapping of ADP in any of the mutants.
Summarizing this section, each of the mutants tested displayed low but real ATPase and ability to trap nucleotide after preincubation with V i plus ATP. Surprisingly Asp and Lys mutants were not much different in these properties from Gln and Ala mutants. Release of trapped nucleotide was generally faster than in wild type with the single exception being E552Q. The Lys mutants were less able to trap ADP added directly with V i , and this was consistent with a significantly reduced rate of onset of trapping measured in Lys mutants with V i plus ATP. Asymmetry between effects of the same mutation placed in either NBD1 or NBD2 was seen in some instances, notably with Gln mutants, but was the exception rather than the rule. In both Ala mutants, there was measurable V i -independent retention of nucleotide, which, together with the tendency to retain ATP as well as ADP, was redolent of the behavior of the E552A/E1197A mutant described previously (32).

Effects of E552A/S528A and E552A/S1173A Mutations
Residues Ser-528 and Ser-1173 occur in the highly conserved LSGGQ ABC signature sequences in NBD1 and NBD2 of Pgp, respectively. These Ser residues are known from x-ray crystal- lography of ABC transporters (12,14) and photocleavage experiments (16) to lie close to the ␥-phosphate of ATP bound in catalytic sites. Ser-1173 is expected to project into the same ATP-binding site of which Glu-552 is part, whereas Ser-528 is expected to project into the ATP-binding site of which Glu-552 is not part. S528A and S1173A mutations alone have mild effects, reducing ATPase by only 26% (33). We previously found that the combination mutant S528A/S1173A showed synergistic effects, producing much stronger inhibition of ATPase than was seen with either mutation singly (33), and that the combination mutant E552A/E1197A showed synergistic effects, producing a conformation of Pgp that was able to tightly bind ATP and ADP (32). Thus, it was of interest to test effects of the combinations E552A/S528A and E552A/S1173A. Fig. 6B shows ATPase activities of these two purified mutant proteins compared with E552A in Fig. 6A. Turnover rates are given in Table III (last two lines). With E552A/S528A significant turnover was not measurable. Duplicate preparations of pure enzyme gave identical results. This emphasizes that we can in fact detect a zero or very close to zero turnover rate in a mutant Pgp. The rate in E552A/S1173A Pgp was extremely low, similar to that previously seen in E552A/E1197A and lower than that in S528A/S1173A. Duplicate purifications of E552A/S1173A gave the same result. Fig. 7 shows retention of nucleotide by these mutants after incubation with [␣-32 P]ATP with or without V i and then passage through a centrifuge column. With V i present both mutant proteins retained significant amounts of nucleotide (close to 1 mol/mol of Pgp in the presence of verapamil); however, TLC analysis (not shown) revealed that the retained nucleotide was a mixture of ATP and ADP (E552A/S528A, 48% ATP; E552A/S1173A, 40% ATP) as compared with zero ATP in wild type. 3 Also there was considerable retention of nucleotide even in the absence of V i . Rates of formation of the retained nucleotide species were slower than wild type in both mutants, whereas nucleotide release was faster (Table II). Release of retained nucleotide was well fit by a single exponential curve (not shown).
Both mutant proteins showed trapping of [ 14 C]ADP in the presence of V i (Fig. 8) with E552A/S1173A approaching 1 mol/ mol and E552A/S528A showing lower amounts. Verapamil increased the level in both cases, showing that communication between NBDs and drug-binding sites was intact. In the absence of V i negligible trapping of [ 14 C]ADP occurred.
Summarizing this section, combination of S528A or S1173A with the E552A mutation did produce synergistic effects, notably to strongly inhibit ATPase activity, to foster V i -independent occlusion of nucleotide, and to increase the proportion of ATP versus ADP occluded.

DISCUSSION
Our long term goal is to understand the molecular mechanism of ATP hydrolysis and its coupling to drug transport in Pgp. In this study we investigated further the role of two conserved Glu residues (Glu-552 and Glu-1197 in mouse MDR3 Pgp) that are known to be critical for function in Pgp (28,31) and are thought to function as catalytic carboxylates in the ABC transporter family (35). In addition, we combined the mutation E552A at the catalytic carboxylate position in NBD1 with single Ser to Ala mutations at the conserved Ser of the LSGGQ ABC signature motif in either NBD1 or NBD2 to study synergistic interaction of these residues.
Studies of the E552Q and E1197Q Mutations-It was concluded from previous work (28,31) that these mutations in mouse or human Pgp eliminated ATPase activity. However, here we showed that this is not the case; rather the activity is reduced to a low but significant level. Although the activities reported are low, we are confident that they are not due to contaminants on the following grounds. First, duplicate purifications of the same mutant enzyme consistently gave the same activity. Second, a mutant with zero or very close to zero activity was found (see below) that was also reproducible in duplicate purifications. Third, the activity of E552Q and E1197Q mutant proteins showed the same biphasic response to verapamil (activation at lower concentration and inhibition at higher concentration) as seen in wild-type Pgp.
It had previously been proposed (28) that retardation of ADP release could be the cause of loss of ATPase activity in these mutants. Here we evaluated this proposal experimentally by measurement of K i(MgADP) and the rate of release of V i -trapped ADP. Neither of these experiments reproduces exactly the physiological situation, for example K i(MgADP) is expected to measure affinity of the open NBD dimer, and release of V itrapped ADP is much slower than release of ADP in the normal catalytic pathway. On the other hand, if the mutations did cause tenacious retention of product ADP, to the extent necessary to achieve the impairment of ATPase activity seen, then in comparisons of mutant versus wild-type enzymes, one might reasonably expect K i(MgADP) to be significantly decreased, which did not occur, or the rate of release of V i -trapped ADP to be significantly slowed, whereas it was actually unchanged (E552Q) or accelerated (E1197Q). Therefore we do not feel that retardation of ADP release provides a satisfactory explanation for the functional impairment in the mutants.
A different explanation for the effects of these mutations was derived from work with human Pgp and 8-azido-ATP in Ref. 31 where it was proposed that the mutations prevented a second round of ATPase turnover required to reset the drug-binding site. However, we found here with mouse E552Q and E1197Q Pgp that multiple turnovers of ATP hydrolysis occurred, and sequential rounds of V i -ADP trapping could be achieved. Thus, in reference to mouse Pgp, the explanation offered for functional impairment caused by these mutations in Ref. 31 appears unsatisfactory.
We did see asymmetric behavior of the two mutations, in that, for example, E552Q had 2.6 times higher ATPase than E1197Q, and K m(MgATP) was Ն5.2-fold higher. This is in accord with previous findings of asymmetric effects on NBD1 versus NBD2 produced by these mutations (29,30). Both mutants were able to trap V i -ADP when preincubated with ATP and V i , consistent with previous observations using 8-azido-ATP (28,31). Thus, in contrast to the situation in F 1 -ATPase (23,24) substitution of the catalytic Glu carboxyl side chain by Gln did not abolish formation of the transition state complex. The transition state complex of the mutants is not the same as in wild type, however, as evidenced by a changed IC 50 for V i inhibition and changes seen in rates of formation and release of trapped V i -ADP. It is also interesting that the turnover rate for the double mutant E552Q/E1197Q was 0.01 s Ϫ1 (32), which is considerably lower than that of either E552Q or E1197Q (Table  I). Thus the presence of one intact NBD moderately improves turnover. Proteolysis studies had shown that V i -8-azido-ATP trapping occurs in both NBDs (30,31) in the single mutants; thus both NBDs appear to hydrolyze 8-azido-ATP.
Studies of E552A, E1197A, E552D, E1197D, E552K, and E1197K Mutations-Substitution by Ala should effectively preclude any orientation or polarization of the attacking water by the residue at this position. Substitution by Asp could retain such properties but move the water significantly further from the ␥-phosphate of ATP. Substitution by Lys should fail to polarize the attacking water correctly and disrupt charge balance within the catalytic site. It was somewhat surprising therefore that in terms of their effects on ATPase turnover, all six mutants behaved similarly (Table III), showing a low residual activity. In Ala and Asp mutants, K m(ATP) was reduced as in Gln mutants (above). The results emphasize the strict requirement for Glu at this position with Asp failing to provide significant partial activity. It appears that in the absence of the Glu side chain other residues in the catalytic site take over as (far less efficient) catalytic residues. Possible candidates are Gln-471 and Gln-1114, the "Q-loop" residues, as originally suggested in Ref. 27. The NBD asymmetry seen in the Gln mutants was not seen with Ala or Asp mutants. Therefore asymmetric effects seen previously (29,30) may be limited to particular mutants and may not be generalized to wild-type Pgp.
Overall the effect of the mutations was to change characteristics of the transition state but not to abolish it. V i -trapping experiments showed some differences between the mutants in stoichiometry and in rates of formation and release of V itrapped nucleotide. In terms of stoichiometry, the mutants behaved somewhat similarly to wild type with even Lys mutants trapping significant amounts of V i -ADP when preincubated with V i and MgATP, although less trapping was seen when preincubated with V i and ADP. In all cases release of V i -ADP was faster than in wild type (Table II); rates of formation of trapped V i -ADP after preincubation with V i and ATP were more variable. One notable feature was that E552K and E1197K mutants showed slow onset of trapping of V i -ADP from V i and MgATP as if the MgATP association rate were slow or a conformational change to the closed (dimer) state were impaired by the introduction of positive charge.
As noted under "Results," after preincubation with V i and ATP there was an increase in the proportion of ATP that was retained after centrifuge column elution by Ala, Asp, and Lys mutants. This, together with the V i -independent retention of nucleotide noted in the Ala mutants, reveals a tendency toward the behavior seen in the E552Q/E1197Q, E552A/E1197A, E552D/E1197D, and E552K/E1197Q "double" mutants (32). Our conclusion from that work was that the double mutants arrested the catalytic cycle at a step immediately after ATP binding with the NBDs in a closed (putative dimeric) state in which nucleotide became occluded. This occluded state could not proceed efficaciously to the normal transition state. It would seem possible that the single mutations studied here engender the same effect, to a lesser degree.
It may be noted that in diverse ABC transporters, various mutations at the catalytic carboxylate (Glu) position have uniformly been seen to reduce ATPase activity to very low levels, examples being Asp, Ala, Cys, Gln, and Ser in BmrA (36); Gly in KpsT (37); and Asp in HisP (38) as well as Gln or Ala in numerous other cases. The usual explanation proffered is that the wild-type Glu residue represents the "catalytic base." In contrast to what was found in F 1 -ATPase (23,24), however, mutations of the catalytic Glu to Gln do not appear to abolish transition state formation in Pgp, at least as measured by V i -ADP trapping. Our data with Pgp suggest that alterations to the integrity of the transition state, a tendency toward partial arrest in a closed state, and the slow onset of V i -ADP trapping (in Lys mutants) may be related to a different, common explanation for reduced ATPase turnover, i.e. that the mutations impair rapid formation of an NBD dimer interface that normally occurs upon ATP binding and/or distort that interface when it does form thus altering characteristics of the closed conformation in which the transition state normally forms. This explanation regarding Pgp may be applicable to other ABC transporters.
Combinations of Mutation E552A with S528A or S1173A Produce Synergistic Effects-In x-ray structures of dimeric ABC transporter NBDs, the side chain -OH of conserved Ser in the LSGGQ signature sequence is close to the ␥-phosphate of bound ATP (12,14,15). Also bound V i -ADP induces photocleavage at this position (16). The relevant Ser residues in Pgp are Ser-528 (NBD1) and Ser-1173 (NBD2). A previous study showed that while mutation of either Ser to Ala reduced ATPase by 26%, combined mutation S528A/S1173A produced 96.4% inhibition and reduced k cat /K m by 100-fold, providing clear evidence for a synergistic effect of the combined mutations (33). Also the single mutations E552A and E1197A produced ATPase turnover rates of 0.075 and 0.066 s Ϫ1 (Table III),