Disulfide Cross-linking Reveals a Site of Stable Interaction between C-terminal Regulatory Domains of the Two MalK Subunits in the Maltose Transport Complex*

Understanding the structure and function of the ATP-binding cassette (ABC) transporters is very important defects in ABC transporters at the of serious cystic fibrosis. the ATP-binding cassette of the maltose transporter of most other ATP-bind-ing cassettes in it an additional C-terminal regulatory domain. The published structure of a MalK dimer is elongated with C-terminal domains at opposite poles Some uncertainty exists as to whether the orientation of MalK in the dimer structure is correct. Superpositioning of the N-terminal domains of MalK onto the ATP-binding domains of an alternate ABC dimer, in which ATP is bound along the dimer interface between Walker A and LSGGQ motifs, places both N- and C-terminal domains of MalK along the dimer interface. Consistent with this model, a cysteine substitution at position 313 in the C-terminal domain of an otherwise cysteine-free

ATP-binding cassette (ABC) 1 transport systems constitute the largest family of transporters known. These transporters utilize ATP to either import or export an extremely diverse array of substrates across cell membranes. They share a common structural organization consisting of two hydrophobic transmembrane-spanning domains or subunits and two nucleotide-binding domains or subunits. Forty-eight ABC transporters have been identified in the human genome, and defects in these transporters cause a variety of diseases including cystic fibrosis, macular dystrophy, and hyperinsulinemia (1).
The maltose transporter of Escherichia coli is well characterized and serves as a model system for study of the ABC transport mechanism. The transporter complex (MalFGK 2 ) comprises four subunits; one copy each of MalF and MalG, the transmembrane subunits: and two copies of MalK, the ATPbinding subunit (2,3). A periplasmic maltose-binding protein (MalE or MBP) is also required for transport, functioning as a high affinity receptor for maltose that also stimulates the ATPase activity of the transporter by binding tightly to and stabilizing the catalytic transition state conformation of the transporter (4). The crystal structures of several of the ATPbinding subunits have been determined (5)(6)(7)(8)(9)(10)(11) as well as those of two intact transporters (12,13). The structures of the ATPbinding subunits are highly conserved and usually consist of two domains. One domain is predominantly ␤-sheet and forms a classic nucleotide-binding fold as observed in the F 1 -ATPase (14), and the second domain is ␣-helical and is specific to ABC transporters. The MalK protein differs in that it contains a third domain at its C terminus that is involved in regulatory functions, as deduced from mutational analysis (15,16). MalK binds MalT, a positive transcription factor that is required for expression of the genes involved in maltose transport and metabolism (17). Overexpression of MalK apparently sequesters MalT, preventing expression at the mal operons, whereas deletion of MalK leads to constitutive expression (18,19). Evidence of a direct protein-protein interaction between isolated MalK and MalT has been presented (17). IIA glc , an enzyme participating in the transport and phosphorylation of glucose, also appears to bind directly to MalK, inhibiting maltose transport when glucose is present by a mechanism known as inducer exclusion (15,20,21).
Biochemical evidence of cooperativity in ATP hydrolysis in many systems suggests that the two ATP-binding subunits of ABC transporters interact (22)(23)(24)(25)(26)(27). The first three structures of ABC proteins, those of transport proteins HisP from Salmonella typhimurium and MalK from Thermococcus litoralis, and of recombinase protein Rad50 from Pyrococcus furiosus (5)(6)(7) * This work was supported by National Institutes of Health Grant GM49261 and Grant Q-1391 from the Robert A. Welch Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  each reported a dimer of ATP-binding subunits. However, sites of interaction between the subunits differed in each crystal igniting a controversy over the true nature of the interaction between nucleotide-binding subunits. A Rad50-like dimer interface was again seen in the crystal of the ABC subunit LolD (MJ0796) but only when a mutation was introduced into the ATP-binding site that stabilized the dimeric conformation of this protein (11,28). In both Rad50 and MJ0796 dimers, the family signature or LSGGQ motif of one subunit makes contact with ATP bound to the nucleotide-binding or Walker A motif of the opposing subunit generating a configuration with two ATPs buried along the dimer interface (6,11). Independent biochemical evidence of a Rad50-like dimer interface in the intact MalFGK 2 transporter was obtained in our laboratory from the pattern of photooxidative cleavage induced by vanadate trapped in the position of the ␥-phosphate of ATP (29). MalK is cleaved by vanadate at both Walker A and LSGGQ, indicating that both motifs lie in close proximity to ATP. Because the LSGGQ motif is too far from the ATP-binding site in the monomer to be cleaved by vanadate, it must contact ATP across the dimer interface. A similar architecture is observed in the structure of the intact vitamin B 12 transporter BtuCD (13) but not in the structure of intact MsbA (12). Assuming that all of the ABC proteins interact with nucleotide in the same way it appeared likely that HisP, MalK, and MsbA failed to crystallize in a physiologic dimeric conformation. In this report, we rotated the MalK subunits from the T. litoralis structure to generate a Rad50-like dimer and found that the C-terminal regulatory domains of MalK, which are located at opposite poles of the T. litoralis structure well away from the dimer interface, might contact each other across the dimer interface. In support of this model, we were able to cross-link the two MalK subunits within MalFGK 2 without loss of ATPase activity or inhibition by IIA glc by introducing single cysteines into the C terminus of MalK. These data provide strong evidence of a stable interaction between C-terminal domains of MalK and new insight into the regulatory mechanism of inducer exclusion.

EXPERIMENTAL PROCEDURES
Construction of Plasmids pMR81, pMR83, and pSSB1-Plasmid pMR81 is a ColE1 plasmid encoding ampicillin resistance and expresses the malF and malG genes under control of the ptac promoter. The cysteine residues that are normally present in the wild-type MalF and MalG proteins were changed to serine residues using site-directed mutagenesis with the oligonucleotides listed in Table I. The mutagenesis of each cysteine codon was carried out individually using singlestranded DNA generated from a plasmid containing the malF and malG genes and a M13 phage packaging site. Once the mutations were made and confirmed by nucleotide sequence analysis, they were combined by fragment swapping to yield plasmid pMR81. This plasmid is fully able to complement null mutations in the malF and malG genes. Plasmid pMR83 is a P15A plasmid encoding chloramphenicol resistance and expresses the malK gene under control of the ptrc promoter; it is a derivative of the previously described plasmid pMR11 (15). The three naturally occurring cysteine residues in MalK were changed to either glycine (Cys-40) or serine (Cys-352, Cys-362) by site-directed mutagenesis with the oligonucleotides listed in Table I. The mutagenesis of each cysteine codon was carried out individually using single-stranded DNA generated from a plasmid containing an EcoRI-SalI fragment representing most of the malK gene and a M13 phage packaging site. To combine the mutations at codons 352 and 362 two rounds of mutagenesis were performed; once the C352S mutation was created it was used as a template for the creation of the C362S mutation. Once the mutations were made and confirmed by nucleotide sequence analysis, they were combined by fragment swapping to yield plasmid pMR83. This plasmid is able to fully complement a deletion of the malK gene. Finally, fragments of pMR83 and pSSGly (30) were combined to yield plasmid pSSB1, which encodes cysteine-free MalK with an N-terminal polyhistidine tag suitable for use in purification.
Construction of Plasmids Containing Single Cysteine Substitutions-Single cysteine substitutions were introduced into cysteine-free HisTagMalK by PCR using overlap extension (31). The 5Ј-3Ј sequences of the mutagenic primers are listed in Table I. PCR products were ligated into a variant of pSSB1 (pKPS2E-86) that has a unique PstI site inserted at codon 39 of malK generating a frameshift mutation. Replacement of the region containing the PstI insertion in pKPS2E-86 with the PCR product ensures, by both restriction digestion and Western blot analysis, that the PCR product has been successfully cloned into the vector. This plasmid also contains a unique NcoI site at the start of malK to facilitate subcloning. Clones were sequenced to confirm the presence of the desired mutation and the absence of other errors introduced by PCR.
Expression of the MalFGK 2 Complex-A plasmid carrying the desired malK gene was cotransformed with pMR81 into strain HN741 (32) containing plasmid pMS421 that carries the lacI q gene. Transformants that expressed both MalF and MalK as judged by Western blotting were frozen at Ϫ70°C in 8% glycerol and used to inoculate overnight cultures for larger scale preparations. Large scale preparations were grown in terrific broth (Invitrogen) containing 100 g/ml ampicillin, 20 g/ml chloramphenicol, and 50 g/ml spectinomycin until the A 600 reached 0.5-0.8, and then induced with 5-10 M isopropyl-1-thio-␤-D-galactopyranoside and grown an additional 24 h. Expression of some of the mutants was improved if the temperature was maintained at 23°C both before and after induction and if less isopropyl-1-thio-␤-D-galactopyranoside (5 M) was used in the induction. Cells were harvested and total membrane was isolated as described previously (3).
Assay of ATPase Activity-The ATPase activity of reconstituted transport complexes was measured at 37°C in 20 mM HEPES (pH 8) with 2 mM [␥-32 P]ATP, 10 mM MgCl 2 , 5 M MBP, and 0.1 mM maltose, as described previously (33). A mutant is considered to be functional if MBP stimulates the ATPase activity of the reconstituted transporter.
Disulfide Cross-linking by Cu(1,10-phenanthroline) 2 SO 4 (CuPhen)-Prior to treatment with CuPhen, purified proteins (5 g) or membranes (1 mg/ml) were treated with 10 mM DTT. The reaction mixtures (100 l) were spun through Sephadex G-50 spin columns to remove DTT, and then incubated with the indicated concentration of CuPhen for 20 min as described. Essentially the same results were obtained at 0 or 37°C. The reaction was terminated by adding 5ϫ SDS-PAGE sample loading buffer containing 5 mM N-ethylmaleimide and no reducing agent (34). The samples were then subjected to SDS-PAGE on 7.5% gels, transferred onto nitrocellulose paper, and probed with antibody against MalK (35). Blots were then incubated either with horseradish peroxidase or alkaline phosphatase-conjugated anti-rabbit antibody, and the bands were visualized by enhanced chemiluminescence (Pierce) or Western Blue stabilized substrate (Promega).
Expression and Purification of Enzyme IIA glc -PCR was used to amplify the gene encoding IIA glc from a colony of E. coli K-12, and a sequence encoding an additional six histidines was added to the 3Ј-end of the gene. The DNA was subcloned into the pET21 vector (Novagen) and transformed into BL21(DE3) for expression. Approximately 20 mg of protein, affinity-purified on Talon resin, was obtained from 1 liter of cells.
Electrophoresis-SDS-polyacrylamide gel electrophoresis was performed according to the procedure of Laemmli (36) with the indicated percentage of acrylamide. Samples in SDS-loading buffer were not heated prior to loading.
Protein Determination-Protein concentrations were determined as described previously (3) by the method of Schaffner and Weissmann (37).

Structural Alignment of MalK and the MJ0796 Dimer-
Based on recent data indicating that the MalK subunits in the MalFGK 2 transporter complex interact such that the LSGGQ motif of one subunit interacts with nucleotide bound to the Walker A of the opposing subunit (29), we have superposed the two MalK monomers A and B from the structure of the T. litoralis dimer (7) onto the structure of the MJ0796 dimer ( Fig. 1) (11). The MalK monomers are in a different conformation than the MJ0796 monomer, which appears to result from rotation of the helical domain containing the LSGGQ relative to the nucleotide-binding domain containing the Walker A (7, 10). As a consequence, alignment of the nucleotide-binding domains of MalK and MJ0796 resulted in the displacement of the helical domain and the LSGGQ motif away from the dimer interface in the MalK model (Fig. 1, compare A with E). In both MJ0796 and Rad50 dimers, serines from both the Walker A and the LSGGQ motif form hydrogen bonds with the ␥-phosphate of ATP (6,11). In our MalK dimer model (Fig. 1), the Walker A motif is similarly placed within hydrogen-bonding distance of ATP, but the oxygen of the serine of LSGGQ was either 8 (subunit B) or 12 Å (subunit A) from the corresponding oxygen in the ␥-phosphate. It has been suggested that rotation of the helical domain and movement of the LSGGQ motif into the nucleotide-binding site upon binding of ATP may be part of the mechanism of activation of ATP hydrolysis and transport (8,13,38). Hence, this open dimer model (Fig. 1A) may approximate a resting state of the MalK dimer in the MalFGK 2 transporter complex, and the structure of MJ0796 (Fig. 1E) may approximate an active state. In contrast to the crystallographic dimer of MalK from T. litoralis where the two Cterminal regulatory domains are at opposite poles of the dimer (Fig. 1F), our model, based on alignment with MJ0796, suggests that these domains may be in contact (Fig. 1B). Alignment of the MalK subunits onto the Rad50 (6) or BtuCD (13) FIG. 1. Superposition of MalK subunits A and B from T. litoralis onto the structure of the MJ0796 dimer. Swiss Protein Data Bank viewer (us. expasy.org/spdbv/) was used to perform a structural alignment of the MalK A monomer and the MalK B monomer onto a dimer of MJ0796 (11). The best fit (root mean square of 1.21 Å) was based on alignment of residues located exclusively in the nucleotide-binding domain, defined by the DOMID program (bioinfo1. mbfys.lu.se/Domid) as residues 1-87 and 159 -241 in MalK. ATP is shown in ball and stick conformation along the dimer interface and is positioned as it was in structure of the MJ0796 dimer. Nucleotide-binding domains are colored white, structures leads to the same conclusion, although the two Cterminal domains either display slight overlap (based on Rad50) or are spaced further apart (based on BtuCD) (data not shown). To test the validity of our model we placed cysteines in loops of MalK that were predicted to approach each other in the dimer, with the goal of forming an intratransporter disulfide bond that covalently links the two MalK subunits (Fig. 1, C and  D). Cross-linking experiments were performed on the intact, purified MalFGK 2 transporter complex.
Functional Characterization of Cysteine-free MalFGK 2 Transporter-The MalF, MalG, and MalK proteins that constitute the maltose transporter complex from E. coli contain seven native cysteines, but these have been removed with minimal impairment of transport function in vivo (34). The genes encoding cysteine-free versions of MalF, MalG, and MalK used in our study differ from those of Hunke et al. (34) in several minor ways. First, an N-terminal polyhistidine tag is attached to our MalK for the purpose of affinity purification; second, the source of our malK gene is E. coli rather than Salmonella (these genes are 95% identical including the locations of the three cysteines); and finally, the cysteine at position 40 of our MalK is replaced by glycine rather than serine. The cysteine-free MalF and MalG proteins are from E. coli in both systems and contain cysteine to serine substitutions at all sites. We found that the cysteine-free versions of MalF and MalG co-purified with cysteine-free HisTagMalK on metal-affinity resin (Fig. 2), as seen for the wild-type proteins although the yield of purified protein (0.6 mg/liter of cells) was substantially reduced from that generally obtained with the wild type (6 mg/liter of cells). The yield of the various mutants was generally improved using lower temperatures for growth and lower isopropyl-1-thio-␤-D-galactopyranoside concentrations (see "Experimental Procedures") that improved the detergent solubility of the overexpressed proteins. The purified cysteine-free transporter complex when reconstituted into proteoliposomes exhibits MBP-stimulated ATPase activity, a good in vitro measure of function (32). The rates of ATP hydrolysis (0.7 -2 mol/min/mg in the presence of MBP and maltose, 0.03 in the absence) generally range from 50 -100% of that seen in parallel experiments with the wildtype transporter.
Addition of Single Cysteine Substitutions into MalK-Single cysteines were individually introduced into E. coli MalK at positions 162, 295, 311, 313, and 333, residues that are predicted to be close to each other across the dimer interface in our model (Fig. 1, C and D). We also studied a transporter that contained one native cysteine at position 40 within the Walker A motif of MalK because disulfide bond formation between these residues has been reported in the P-glycoprotein ABC transporter (39). The mutant MalK proteins were coexpressed with Cys-free versions of MalF and MalG. Purification of the fully Cys-free complex and complexes containing a single Cys at either position 40 or position 313 is shown in Fig. 2A. HisTagMalK typically migrates at 45 kDa on an 11% acrylamide gel just above the MalF protein, and MalG migrates at 27 kDa (27). When 2-mercaptoethanol is omitted from the SDS loading buffer as in Fig. 2A, a higher molecular mass band migrating near 100 kDa is often seen in the purified sample. This band appears to be a dimer of MalK stabilized in SDS through disulfide bond formation, because it reacts with antibody to MalK but not MalF or MalG and is less apparent following treatment with fresh 2-mercaptoethanol or DTT (Fig.  2B). This dimer is absent in the Cys-free transporter complex. Treatment with CuPhen to promote disulfide bond formation triggers reformation of the dimer (and higher order) bands in the sample containing Cys-40 but not in the Cys-free construct. The smearing of MalK at a higher M r on the gel, most commonly seen following treatment with CuPhen, disappears if the sample is treated with DTT prior to SDS-PAGE and may result from disulfide bond formation with contaminants in the preparation or from the formation of aggregates that are not dissociated by SDS in the absence of DTT. Cys-free MalK containing the E313C substitution is notable in that MalK is depleted in the position of the monomer relative to the amount of MalF, and MalK dimer is the prominent species on the gel ( Fig. 2A). In contrast, disulfide bond formation between MalK proteins containing Cys-40 is inefficient. Quantitation of gels similar to those shown in Fig. 2, A and B, indicated that only ϳ10% of Cys-40 MalK migrates as a dimer, even following CuPhen treatment.
Disulfide Cross-linking of the C-terminal Domains of MalK-Purified transporter complexes containing single cysteines at positions 295, 311, 313, and 333 in the C-terminal domain were treated with DTT to reduce any disulfide bonds that may have formed by air oxidation during isolation and purification, desalted, and incubated with CuPhen to promote disulfide bond formation (Fig. 3A). The position of migration of MalK was shifted either partially or completely from monomer to dimer with a cysteine at position 295, 313, or 333, consistent with formation of a disulfide bond between two MalK proteins upon CuPhen treatment. Addition of DTT following CuPhen treatment brings MalK back to the position of a monomer, as anticipated for a disulfide bond (data not shown). Transporter containing the E313C substitution was selected for further experimentation because it was well expressed and cross-linking appeared to go to completion.
Our model for a MalK dimer (Fig. 1B) suggests that the 313 linkage forms between the two MalK subunits in the same transporter complex (intratransporter cross-linking); however, the formal possibility exists that cross-linking occurs between two different transporters (intertransporter cross-linking). Based on the assumption that the efficiency of inter-transporter cross-linking would be an inverse function of protein concentration, CuPhen-induced disulfide bond formation was assessed as a function of protein concentration. No decrease in the efficiency of disulfide bond formation was seen when the concentration of MalFGK 2 in detergent solution was reduced from 1.2 M down to 0.036 M suggesting that the cross-linking is occurring within the MalFGK 2 complex rather than between transporter complexes (Fig. 3B). In fact, a 2-4-fold reduction in protein concentration actually increased the efficiency of disulfide bond formation in this experiment suggesting that 50 M CuPhen concentration might be limiting at high protein concentration. Migration of fully cross-linked, or uncross-linked MalFGK 2 transporter complexes through a G3000SW XL high pressure liquid chromatography gel filtration column (Tosoh Bioscience) in 0.01% dodecyl maltoside was also compared and no difference in the time of elution was detected, as might be expected if disulfide bond formation had generated a stable transporter dimer (MalFGK 2 ) 2 . Whereas this column is able to separate soluble proteins in the range of 10,000 -500,000 Da, it should be noted that the ability of gel filtration to resolve membrane proteins of similar size in detergent micelles has not been carefully studied.
Disulfide Cross-linking of Residues in the N-terminal Domain of MalK-In our model of a MalK dimer only 6.5 Å separates the ␣-carbons of the serines at positions 162 in the N-terminal nucleotide-binding domain; hence cysteines at this position may form a disulfide bond. In the structure of the Rad50 dimer the corresponding position is the point of closest contact between identical residues in opposing subunits (6). Placement of a cysteine at this position resulted in efficient formation of a MalK dimer as judged by Western blot analysis (Fig. 4A, lanes 2 and 3); however, purified transporter complex was obtained in very low yield despite the production of large amounts of MalK protein in the membrane pellet (Fig. 4B). A positive reaction of anti-MalF antibody with the purified sample (Fig. 4A, lane 1) was used to verify that MalF copurified with the mutant HisTagMalK, as expected for an intact transporter. This MalK dimer formed spontaneously during purification and, like the E313C mutant, could be reduced with DTT and reoxidized, essentially to completion, with CuPhen treatment (data not shown). No further characterization of this preparation was performed because of the poor yield of assembled complex.
Influence of Disulfide Bond Formation on ATPase and Inducer Exclusion Activities-Because disulfide bond formation went essentially to completion with the E313C substitution, the effect of the presence of a disulfide bond on transporter function could be assessed. As in the wild-type MalFGK 2 transporter, little or no ATPase activity was detected in the oxidized transporter following reconstitution when MBP was absent (data not shown). MBP plus maltose efficiently stimulated the ATP activity of the oxidized transporter, and DTT enhanced this activity by only 30% (Table II). Similar rates of ATP  In inducer exclusion, IIA glc of the glucose phosphotransferase system inhibits maltose transport when glucose is present in the medium. Because several mutations that prevent inducer exclusion are located in the C-terminal domain of MalK (15,20) it was of interest to determine whether tethering the C-terminal domains together would interfere with this regulatory activity. In Fig. 5, the pattern of inhibition of MBP-stimulated ATPase activity by IIA glc was very similar in the E313C and cysteine-free constructs whether the disulfide bond between MalK subunits was present or absent (Fig. 5, inset). The concentration of IIA glc yielding 50% maximal inhibition of wildtype MalFGK 2 was ϳ1.5 M, comparable with the data obtained in the cysteine-free background (Fig. 5).

DISCUSSION
Repositioning of the N-terminal domains of the two MalK subunits from the T. litoralis dimer to conform to the dimeric structure of MJ0796 and Rad50, where ATP is bound along the dimer interface between the Walker A and LSGGQ motifs (6,11), creates a model for MalK in which the C-terminal regulatory domains lie in close proximity. Justification for this realignment came initially from the observation that vanadate trapped in the nucleotide-binding site of MalK mediated oxidative photocleavage of both the Walker A and the LSGGQ motifs (29). Our demonstration of efficient disulfide bond formation between cysteines placed along the proposed dimer interface in the C-terminal domain of MalK provides additional support for this arrangement of subunits, as opposed to that seen in the crystal of MalK from T. litoralis. In the structure of the T. litoralis dimer (7) the C-terminal domains are well separated (Fig. 1F). The MBP-stimulated ATPase activity of the transporter was only modestly affected, if at all, by the presence of the disulfide bond between cysteines at position 313, effectively ruling out the possibility that the dimer conformation seen in the structure of the T. litoralis protein represents some type of conformational intermediate in the translocation cycle. This possibility also seems unlikely based on the location of residues predicted to interact with the transmembrane region from the structure of BtuCD (13), which are partially obscured by the dimer interface in the published MalK structure (7).
In previous studies, disulfide bond formation was reported between MalK proteins with single cysteines placed at either position 85 or position 106 in the N-terminal domain of MalK from Salmonella (34). The cross-link at position 85, cited in support of the T. litoralis dimer (7), is also consistent with the model for dimerization depicted in Fig. 1〈. Although the ␣-carbon atoms of these residues are 32 Å apart in our MalK dimer  Fig. 1 with location of mutations highlighted in magenta (16,21). A, mutations that prevent inducer exclusion (119, 124, 228, 241, 278, 284, 302, and 322 in E. coli MalK) (viewed from side). Domains belonging to MalK subunit A are colored in darker hues, as in Fig. 1 model, they face the dimer interface and are positioned only 14 Å apart in the MJ0796 structure (11). This difference appears to be a consequence of the conformational changes that result in subunit association and activation of ATPase activity (11,13,30,38). Although 14 Å is still greater than the distance covered by a disulfide between two cysteines, protein structures may be flexible enough to allow these residues to react because a transient contact is all that is required for disulfide bond formation to occur. Disulfide bond formation between residues at position 106 is difficult to rationalize on the basis of either MalK dimer structure. Instead, as suggested by Kerr (40), it seems likely that a cysteine at this position, which is highly exposed on the presumed external surface of the dimer, has contacted MalK in an adjoining transporter complex leading to the formation of an intertransporter cross-link.
Residues at position 40 in the Walker A motif are potentially exposed at the dimer interface, but they are ϳ30 Å apart, and this distance is unchanged by the conformational changes thought to be associated with activation of ATPase activity. The side chain may also be accessible from the external surface of the dimer in a position that could permit intertransporter interaction. In the maltose transporter, this disulfide bond rarely forms with high efficiency, and we were unable to rule out the possibility that intertransporter interactions may be involved in disulfide bond formation. Cys-40 is present in the native MalK protein and may be responsible for the routine appearance of a band corresponding to a MalK dimer when wild-type MalFGK 2 is separated on SDS-PAGE gels in the absence of fresh reducing agent. Interestingly, evidence of both intra-and interdisulfide bond formation between the corresponding residues in the Walker A motifs of the P-glycoprotein has been presented, and formation of the disulfide inhibits transporter function (39). Although the structure of P-glycoprotein has not been solved, it is presumed to closely resemble other ABC proteins, suggesting that the reaction of these normally distant cysteines has trapped a nonphysiologic conformation. In contrast, the fact that the maltose transporter retains function with the two C-terminal domains tethered together justifies the use, in this instance, of disulfide bond formation as a measure of distance between residues and argues strongly that residues at position 313 are always in close proximity in the native conformation of MalFGK 2 .
The positions of mutations affecting inducer exclusion and mal gene transcription were recently mapped onto the structure of the MalK monomer (16,21). These data are reproduced in Fig. 6 in the context of our model for dimerization of MalK. Mutations that prevent inducer exclusion form clusters in both the helical domain and the regulatory domain. Whereas these clusters lie on opposite sides of the monomer, they lie on the same face of the dimer (Fig. 6A) and may define a binding site for IIA glc that spans both MalK subunits. Because the inhibitory effect of IIA glc is not lost when the two C-terminal domains are joined with a disulfide bridge, IIA glc is not inhibiting ATPase activity by preventing the association of the C-terminal regions of MalK. Rather, the binding of IIA glc across both subunits could function to prevent movement of the two Nterminal domains into closer proximity following ATP binding and thereby prevent ATP hydrolysis. One role of MBP is to stimulate ATP hydrolysis by MalFGK 2 , and we suggest that it does so by stabilizing a conformation of the transporter in which the N-terminal nucleotide-binding domains of MalK have closed upon each other, completing the two nucleotidebinding sites along the dimer interface (4,38).
Mutations affecting the regulatory interaction of MalK with the transcriptional activator MalT appear to define a binding site for MalT toward the bottom of the C-terminal regulatory domain (Fig. 6B). The ability of MalK to modulate MalT function appears to be affected by the conformation of MalK within the transporter complex because an inactive form of MalK down-modulates MalT and decreases transcription, whereas an active form of MalK does not (17). Once an in vitro assay for this regulatory function has been developed, it will be possible to determine whether MalT function is influenced in any way by cross-linking of C-terminal regulatory domains.