INTRODUCTION

The product of the malE gene of Escherichia coli, the periplasmic maltose-binding protein (MBP),¹ binds maltose and other maltodextrins with high affinity (1). Ligand-bound MBP interacts with a membrane-bound complex of the MalF, MalG, and MalK proteins (2, 3) to take up maltodextrins via a transmembrane channel formed by MalF and MalG (4, 5, 6). The NH₂- and COOH-terminal domains of MBP appear to bind to MalG and MalF, respectively (7). Certain mutations in malF and malG allow cells lacking MBP to transport maltodextrins (8, 9). These mutant MalF-MalG (MalFG) complexes can stimulate ATP hydrolysis by MalK in the absence of MBP (6), suggesting that the binding of MBP to the periplasmic face of the MalFG complex normally activates MalK in the cytoplasm.

Ligand-bound MBP interacts with the Tar chemotactic signal transducer to initiate an attractant response to maltose (10, 11, 12, 13, 14). Tar is proposed to form a homodimer (15, 16). Genetic (12, 13, 17) and structural (18) evidence indicates that the attractant L-aspartate binds at the dimer interface of the periplasmic domain of Tar, contacting residues in both subunits. Recent data from intergenic complementation studies (53) is consistent with the prediction from computer modeling (19) that MBP also contacts both subunits of Tar.

MBP is thought to function as a monomer in which the globular NH₂- and COOH-terminal domains are connected by a flexible hinge (20, 21). Short maltodextrins, like maltose and maltotriose, are buried in a cleft between the two domains when MBP closes (20). The nonreducing end of longer maltodextrins hangs out of the cleft, and bulky, nonphysiological ligands like cyclomaltoheptaose may prevent closure of the cleft almost completely when they bind to MBP (22). A bound maltodextrin forms extensive hydrophobic and hydrogen bonds with residues on each wall of the cleft to stabilize the closed conformation. To achieve the closed conformation, one domain of MBP undergoes a 35° rotation and an 8° lateral twist in the hinge relative to the other domain (21).

A working hypothesis is that when MBP assumes its closed conformation, residues in the two domains are brought into the correct spatial relationship to interact with complementary sites on the MalF-MalG complex or with the Tar dimer (14, 19, 20). However, since MBP (23, 24) and other binding proteins (25, 26, 27, 28) equilibrate between open and closed forms in the presence or absence of ligands, it is uncertain whether MalFG and Tar first interact with the open, closed, or some intermediate form of MBP.

We have introduced mutations into a plasmid-borne malE gene to create the amino acid substitutions G69C and S337C in the normally cysteine-free MBP. Based on the crystal structure of the ligand-bound, closed form of MBP (20),² we anticipated that cysteines at these positions could form an interdomain disulfide bridge. We have determined that this cysteine-substituted MBP does form such an intramolecular cross-link. Growth experiments, transport measurements, and chemotaxis assays strongly suggest that this cross-linking occurs in the periplasm in vivo. The double mutant MBP does not function in maltose transport or chemotaxis, and it inhibits, in a dose-dependent fashion, the function of wild-type MBP in transport and chemotaxis. This result demonstrates that the cross-linked protein, which we infer is in the closed conformation, retains biological activity and competes with wild-type MBP for interaction with MalFG and Tar in vivo.

MATERIALS AND METHODS

Strains and plasmids used are listed in Table I. Strain YZ8 (malE444; Ref. 14) was used as the host for mutagenized plasmids. Strain CJ236 containing plasmid pCJ105 (29) was used to produce single-stranded DNA for site-directed mutagenesis. Strains YZ9, YZ11, and YZ12 were constructed by introducing the malE genes from strains RP437 (30), MM187, and MM188 (31) into strain YZ8, with selection for growth on minimal maltose plates. All strains with a YZ designation gave constitutive, high level expression of the chromosomal mal regulon because they contained the malT^c1 (34) allele. Genetic manipulations were performed according to Miller (35).

Table I. Strains and plasmids Name Relevant genotype Origin Strains CJ236 dut-1 ung-1 relA thi-1/FCJ105 Camr Ref. 29 MM187 malTc1 malE16-1 Ref. 31 MM188 malTc1 malE18-1 Ref. 31 YZ8 malTc1 malE444 Ref. 14 YZ9 YZ8 malE+ This study YZ11 YZ8 malE16-1 This study YZ12 YZ8 malE18-1 This study Plasmids pBR322 Ampr, Tetr Ref. 32 pJF1 pBR322 Ampr malE+ oriM13 Ref. 33

Media were made according to Miller (35). Luria broth was used for routine growth of cells. Cells for growth curve determinations and for osmotic shock were grown at 37 °C with vigorous swirling in M69 minimal medium containing 0.4% (w/v) maltose and 1 µg/ml thiamine. Glycerol was added to minimal medium at 0.2% as indicated. For cells grown under ``reducing'' conditions, cells were incubated at 37 °C in stationary test tubes in the same medium containing, in addition, 2 mM DTT and overlaid with several mm of mineral oil to exclude atmospheric oxygen. To make semisoft agar for swarm plates, Bacto-Agar (Difco) was added to minimal maltose medium to 0.3%. Ampicillin (50 µg/ml), kanamycin (100 µg/ml), and tetracycline (10 µg/ml) were added to liquid or solid media as required.

Mutagenesis of plasmid pJF1 (containing the malE⁺ gene and the single-stranded origin from phage M13) or malE mutant plasmids derived from it was performed with the method of Kunkel (36). The mutagenized plasmid DNA was introduced into strain YZ8, selecting for Amp^r. Colonies of transformed cells were screened on maltose swarm plates. If mutant swarm phenotypes were observed among the transformants, plasmid DNA was prepared from isolates having those phenotypes, and the DNA was sequenced over the mutagenized region to confirm the presence of the expected mutation. If no mutant phenotypes were observed, plasmid DNA was prepared from 10 random isolates and sequenced over the mutagenized region to identify mutant plasmids. Plasmids containing two mutations were constructed by using a single mutant plasmid as starting material for a second round of mutagenesis.

For analytical work, periplasmic proteins from cells in exponential or overnight cultures grown in minimal glycerol medium were released by osmotic shock as described by Koshland and Botstein (37), with the modifications of Walker and Gilbert (38).

Strain YZ8, expressing either wild-type or double mutant (G69C/S337C) MBP, was grown to an A₆₀₀ of 0.7 in medium containing 0.5 × Luria broth and 0.5 × M63 medium (35) containing 0.4% (w/v) maltose, 0.2% glycerol, 1 µg/ml thiamine, and 50 µg/ml ampicillin. Periplasmic proteins were released by osmotic shock as described by Kondo et al. (39). MBP was isolated from the shock fluid by amylose affinity column chromatography (40). DTT at 1 mM was included in the buffers during purification of the mutant MBP to maintain it in the reduced state to promote its binding to amylose. Protein concentrations were determined as described previously (5), using the method of Schaffner and Weissmann (41).

SDS-PAGE analysis (42) was carried out using 10% acrylamide. The sample buffer was 0.32 M Tris-HCl (pH 6.8) with 8% SDS (w/v), 40% glycerol (v/v) and 0.01% bromphenol blue. For gels run under reducing conditions, -mercaptoethanol was added to a final concentration of 5%. Native gels were prepared according to Laemmli (42), except that SDS was omitted from all solutions. In crude osmotic shock fluid, MBP was detected by immunoblot analysis (43) using polyclonal anti-MBP rabbit antiserum (New England Biolabs) at a 10,000-fold dilution in 5% nonfat dry milk. The rabbit immunoglobulin was visualized with enzyme-conjugated goat anti-rabbit secondary antibody (Bio-Rad) using appropriate reagents (nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate; Boehringer Mannheim).

Minimal maltose swarm plates were inoculated from fresh colonies with toothpicks and incubated for 16 h at 32 °C. Maltose capillary assays were carried out according to Adler (44) with cells grown at 35 °C with vigorous swirling in glycerol-minimal medium A (35) containing 0.1% casamino acids (Difco). Cells were harvested at an A₅₇₈ of 0.5 and used at a final density of about 2 × 10⁶ cells/ml. Results were normalized to the aspartate responses of cells prepared from the same culture.

Cells for maltose transport were assayed in cells grown in the same way as cells for capillary assays. The V_max and apparent K_m for maltose uptake were determined as described previously (31).

RESULTS

In order to cross-link the NH₂- and COOH-terminal domains of MBP with a disulfide bridge, we had to introduce suitable pairs of cysteine residues into the protein. We used two criteria to evaluate the suitability of residues for substitution by cysteine. The first was that the distance between the -carbons of the two residues to be substituted had to be 8 Å or less in the crystal structure determined for the closed form of MBP (Fig. 1; 20). The second was that each singly substituted MBP had to function normally in maltose transport and chemotaxis, as assessed by its ability to support the formation of wild-type swarms in maltose semisolid agar (Fig. 2).

malE malT

Among the various combinations of double substitutions that were tested, only transformants receiving a plasmid encoding the G69C/S337C substitutions failed to swarm normally. All further experiments were carried out with this double mutant protein.

We detected the cross-linked form of MBP by performing SDS-PAGE analysis of osmotic shock fluid from stationary phase cells of strain YZ8 (malE) containing various MBP-encoding plasmids (Fig. 3). On gels run without reducing agent, >80% of the MBP from the double mutant strain ran in a band of lower mobility (about 5 kDa higher apparent molecular mass) than wild-type MBP. (The intramolecularly cross-linked form of MBP, in which the 370-residue polypeptide is held together by a disulfide bond between residues 69 and 337, might be expected to migrate aberrantly during SDS-PAGE.) In contrast, singly substituted G69C or S337C MBP had the same mobility as wild-type MBP under these conditions. The wild-type protein, both single mutant proteins, and the double mutant protein formed bands at the same position when SDS-PAGE was performed under reducing conditions. The G69C/S337C protein produced the more slowly migrating species whenever reducing agent was omitted from the sample and running buffers.

We also prepared samples by boiling cells directly in sample buffer lacking reducing agent or by lysing cells with a freeze-thaw procedure in the absence of reducing agent. These samples were subjected to SDS-PAGE and immunoblotted with anti-MBP antibody. Strains producing the G69C, S337C, and G69C/S337C proteins all showed dozens of bands at the molecular mass of MBP and larger (data not shown), whereas the strain producing wild-type MBP yielded a single band of the expected molecular mass. This result suggests that each of the cysteine-containing mutant proteins could nonspecifically cross-link to many different proteins. We think this cross-linking occurs upon cell disruption, since proteins from the cytoplasm or inner membrane that contain reduced sulfhydryls are mixed with Cys-substituted MBP under transiently reducing conditions followed by rapid oxidation.

The double mutant protein was purified from a periplasmic fraction obtained by osmotic shock. Purification was carried out under reducing conditions by affinity chromatography, using a cross-linked amylose resin. After dialysis in the absence of reducing agent, the double mutant protein still resolved into two bands during SDS-PAGE under nonreducing conditions, but it ran as a single band comigrating with purified wild-type MBP under reducing conditions (Fig. 4, a and b). On native gels in the absence of reducing agent, wild-type MBP migrates as a single band, and the G69C/S337C protein migrates as two bands of slightly higher mobility (Fig. 4, c and d), with the upper band being the more intense. When the proteins were exposed to 10 mM DTT before and during electrophoresis, the double mutant protein still migrated as two bands, although their position was shifted upward and the two bands were of equal intensity.

To quantify the growth defect caused by the G69C/S337C substitutions, we compared the growth in minimal maltose medium of strain YZ8 expressing wild-type, G69C, S337C, or G69C/S337C MBP (Fig. 5). Cultures of cells expressing either of the single mutant proteins had growth rates indistinguishable from those of cells producing wild-type MBP. In contrast, cells expressing the double mutant protein hardly grew, failing to reach an A₅₉₀ of 0.2 even after 8 h of growth at 37 °C. Strains producing G69C/S337C MBP grew identically to wild-type cells in minimal glycerol medium. Thus, the defect for growth in minimal maltose medium results from an inability to utilize maltose rather than some more general impairment of growth in minimal medium.

To test whether the growth defect associated with G69C/S337C MBP was due to the protein being in a closed form, we wanted to determine whether the protein functioned if its disulfide bridge was reduced in vivo. To prevent oxidation of the DTT used as a reducing agent in this experiment, cells were grown in stationary test tubes in minimal maltose medium covered with a layer of mineral oil to exclude atmospheric oxygen. Growth of strain YZ8 expressing plasmid-encoded wild-type MBP was similar in the absence or presence of 2 mM DTT (Fig. 6). Identical results were obtained with strain YZ8 expressing G69C or S337C MBP (data not shown). In contrast, 2 mM DTT substantially improved the growth of strain YZ8 expressing G69C/S337C MBP (Fig. 6), although even with DTT the growth rate was considerably lower than that of strain YZ8 expressing the wild-type or single mutant proteins. The failure of DTT to fully rescue the ability to grow on maltose could be due to incomplete reduction of the disulfide or a general pleiotropic effect of the double mutant MBP.

To demonstrate that the growth defect of cells expressing the doubly cysteine-substituted MBP was due to decreased ability to take up maltose, we determined the V_max and apparent K_m of maltose transport in the presence and absence of DTT (Table II). Cells for transport assays were grown in minimal glycerol medium. (The presence of the malT^c1 allele in the YZ strains ensured that the genes of the mal regulon were expressed at a high, constitutive level even in the absence of maltose.) The maltose K_m values in cells expressing the wild-type, single mutant, and double mutant proteins were similar, but the V_max value for the double mutant strain was decreased about 4-fold compared with the other strains. A 30-min preincubation with 2 mM DTT had little effect on the V_max or K_m values for the wild-type or single mutant strains, but such treatment increased the V_max for the double mutant an average of 4.5-fold.

Table II. Maltose transport in strains producing cysteine-substituted MBP The values of the Vmax and apparent Km of maltose uptake were determined for duplicate or triplicate samples on different days. (The Mal control containing plasmid pBR322 was assayed only once). Each determination is given to show the day-to-day variation. Assays and kinetic parameters of transport were determined as described under ``Materials and Methods.'' Transport was measured at about 25 °C in the absence (DTT) or 20 min after the addition (+DTT) of 2 mM DTT. NA, not applicable. MBPa Vmaxb Vmax Km Km  DTT +DTT +DTT/DTT  DTT +DTT +DTT/DTT µm Wild type 7.3 6.9 1.0 0.59 0.62 1.1 16.5 15.8 0.9 1.7 1.5 0.9 5.8 7.0 1.2 0.93 1.2 1.3 G69C 14.0 13.6 1.0 2.3 3.0 1.3 8.5 12.3 1.4 0.95 2.1 2.2 S337C 8.3 11.4 1.4 0.80 1.5 1.9 10.3 14.8 1.4 1.9 3.0 1.6 G69C/S337C 2.6 9.9 3.7 1.4 3.1 2.2 1.6 8.4 5.4 0.59 1.3 2.2 None 0 0 NA NA NA NA a  MBP was encoded by a plasmid-borne malE gene expressed in strain YZ8 (malE). b  Vmax × min1.

The capillary assay for chemotaxis, unlike the swarm plate assay, allows a chemotactic response to maltose to be measured without requiring that maltose be metabolized. Thus, transport defects do not interfere with chemotaxis in this assay, in which chemical gradients are created by diffusion from the mouth of the capillary. Nevertheless, strain YZ8 expressing plasmid-encoded G69C/S337C MBP was totally defective in maltose taxis (Fig. 7), whereas either G69C or S337C MBP supports wild-type maltose taxis. The procedure for capillary assays did not allow us to test whether DTT restored the ability of G69C/S337C MBP to function in maltose taxis.

The cross-linked form of MBP should be structurally similar to the closed conformation of maltose-bound MBP. If the closed form of MBP initiates interaction with MalFG and Tar, overexpression of the G69C/S337C protein should interfere with the transport and chemotactic functions of wild-type MBP. When wild-type MBP is present in amounts lower than those of fully induced malE⁺ cells, such inhibition should be more pronounced. We therefore utilized two malE signal-sequence mutations that decrease levels of periplasmic MBP by a known amount (31).

The effect of G69C/S337C MBP on the swarm phenotypes of strains YZ9, YZ11, and YZ12 (with 100, 23 and 11%, respectively, of the constitutive level of chromosomally encoded MBP) is shown in Fig. 8. The consequences of expressing G69C/S337C MBP were most severe in strain YZ12 and least severe in strain YZ9. Similar results were obtained when the growth of these strains was monitored in minimal maltose liquid medium (Fig. 9). Finally, coexpression of plasmid-encoded G69C/S337C MBP with wild-type MBP from the chromosome led to complete inhibition of the chemotactic ability of strain YZ9 to accumulate in maltose-containing capillaries (Fig. 10).

To confirm that the increased growth inhibition conferred by the plasmid expressing G69C/S337C MBP was correlated with an increased ratio of cross-linked to normal protein, we prepared immunoblots (Fig. 11) of osmotic shock fluid from exponential phase cultures growing in minimal glycerol medium. The amount of cross-linked MBP remained about constant, whereas the amount of noncross-linked MBP decreased in the predicted order: YZ9 > YZ11 > YZ12 > YZ8.³ The relative amounts of cross-linked MBP in shock fluid from cells in exponential phase (Fig. 11) and from cells from overnight, stationary phase cultures (Fig. 3) are similar.

DISCUSSION

The specificity of interaction of bacterial substrate-binding proteins with membrane-protein complexes responsible for transport and chemoreception has focused on the large conformational changes of the binding proteins as they go from their open to closed forms. An attractive model is that docking of the binding proteins with their membrane counterparts involves residues in both the NH₂-terminal and COOH-terminal domains of the binding proteins. In the open form of the binding protein these residues are not in the right spatial relationship to interact simultaneously with their docking partners. The large conformational change that occurs as the binding protein assumes a closed form would then bring these residues into the correct juxtaposition to interact with the membrane partner. For MBP, this model is supported by structural (20, 21), genetic (7, 12, 14), and computer modeling (19) studies.

Recent work from several laboratories (25, 26, 27, 28) has suggested that binding proteins are in equilibrium between closed and open forms in the absence or presence of ligand. Rather than triggering closure of the binding proteins, as had originally been proposed (45), recent data suggest that hydrogen bonding and hydrophobic stacking interactions between the ligand and the cleft-exposed faces of the two domains shift the equilibrium in favor of the closed conformation for the two binding proteins, MBP and the lysine-arginine-ornithine-binding protein, for which both open and closed structures have been solved (20, 21, 46). Techniques used to detect the closed/substrate-free form of periplasmic binding proteins include fluorescence changes of the C4-dicarboxylate-binding protein (DctP) of Rhodobacter capsulatus (25, 26); trapping by conformation-specific antibodies with the histidine-binding protein (HisJ) of Salmonella typhimurium (27); and crystallization of the closed/substrate-free form of the galactose/glucose-binding protein (MglB) of S. typhimurium (28).

The ligand-free form of binding proteins is not functionally inert. In a reconstituted histidine transport system employing purified HisJ and right-side-out membrane vesicles containing HisQMP, it was found that excess substrate-free HisJ inhibited transport (47). Also, the best fit to the data for maltose transport in Escherichia coli as a function of the concentration of maltose and periplasmic MBP (31) was obtained when the theoretical model considered that the ligand-bound and the ligand-free forms of MBP compete for interaction with MalFG (23, 24).

It has usually been assumed that binding proteins initiate interaction with membrane transport proteins or chemotactic signal transducers when the binding proteins are in the closed conformation. However, the available data do not, in our view, rule out a priori that a binding protein, already associated with ligand, may have to assume an open form that can then ``clamp down'' on the membrane components. Ligand would then stabilize this complex. One scenario requiring such a mechanism would be that an element on a membrane protein might sterically interfere with binding of the closed form of MBP. This element might be accommodated in, or even contribute to, the ternary complex. In the case of transport, such an ``intrusive'' element could even facilitate subsequent opening of the binding protein to allow substrate release. Wolf et al. (48) demonstrated that mutant HisJ proteins defective in closing fail to interact normally with the HisQMP membrane transport complex. However, although this result is consistent with a requirement for the closed form in binding to HisQMP, the defect in closing could also interfere with a ``clamping down'' process of the type just described.

Interdomain disulfide cross-bridges have been used before to study the properties of sulfate-binding protein (49) and galactose/glucose-binding protein (50) in vitro. MBP containing introduced cysteines at residues 69 and 337 (G69C and S337C) spontaneously forms such cross-bridges in the periplasmic space. This circumstance has allowed us to test whether such a cross-linked binding protein can bind to membrane transport and chemotaxis components in vivo.

Besides G69C/S337C MBP, we constructed two other double mutant MBPs: S233C/P298C and S233C/A301C. In either mutant protein, if cross-bridges formed they would link the NH₂-terminal and COOH-terminal domains of MBP at the other end of the cleft from the 69-337 cross-bridge (Fig. 1). We confirmed the presence of the S233C, P298C, and A301C substitutions by DNA sequencing, but no bands with a mobility different from that of wild-type MBP were observed during SDS-PAGE analysis (data not shown). We inferred that the cross-linking did not occur. We learned later that the 233 and 301 residues may be too far apart in the closed form of MBP to cross-link efficiently.⁴ The 233 and 298 residues should be close enough to allow cross-linking,⁴ but the proline to cysteine substitution at residue 298 may distort local secondary structure, thus precluding cross-linking between Cys²³³ and Cys²⁹⁸. The distortion cannot be global, however, since the P298C and S233C/P298C proteins retain nearly wild-type function (data not shown).

G69C/S337C MBP does not function in maltose transport (Fig. 5, Table II) or maltose chemotaxis (Fig. 7). The slow growth and residual transport observed in strains expressing G69C/S337C MBP probably depend on the fraction of the protein that remains noncross-linked (Fig. 3). Reduction of the disulfide bridge in vivo restores transport function to G69C/S337C MBP (Fig. 6, Table II), demonstrating that the two introduced cysteine residues per se do not seriously impair MBP function. Until the structure of the cross-linked form of the protein is determined, we cannot be certain what exact relationship its structure bears to that of the closed form of ligand-bound wild-type MBP.

The conclusion that there is a good match between the structure of the cross-linked form of MBP and the closed, ligand-bound form of wild-type MBP is supported by the observation that G69C/S337C MBP is negatively co-dominant for growth on maltose and for maltose chemotaxis (Figs. 8, 9, 10). This result establishes that the cross-linked protein does not have a null phenotype, because it competes with wild-type MBP for MalFG and Tar. The specificity of the competition is demonstrated by the finding that the negative dominance of G69C/S337C MBP is greater when the amount of wild-type MBP is reduced by malE signal-sequence mutations (Figs. 9 and 10).

Three lines of evidence strongly suggest that G69C/S337C MBP forms an interdomain disulfide bridge in the periplasm. 1) Oxidizing agents other than atmospheric oxygen were not required to produce the cross-linked protein. 2) Singly cysteine-substituted mutant proteins and the double mutant protein formed nonspecific cross-links to many proteins when cells were lysed under nonreducing conditions, but only the double mutant protein was cross-linked in periplasmic fractions prepared by osmotic shock. 3) The purified G69C/S337C protein yielded the same band of about 45 kDa apparent molecular mass during SDS-PAGE that was seen in osmotic shock fluid. This last result rules out the possibility that the more slowly migrating form of G69C/S337C MBP is cross-linked to another membrane or periplasmic protein.

Because the purified double mutant protein migrates at a molecular mass close to that of wild-type MBP in both native and denaturing gels (Fig. 4), the disulfide cross-link appears to be intramolecular, as we predicted, rather than intermolecular. The slight differences in mobility of the double mutant protein on native gels may reflect the presence of two conformations of MBP, even after reduction. One might predict the protein could assume the following forms: 1) an open, reduced conformation without maltose bound; 2) a closed, reduced form with maltose bound; 3) a closed, oxidized form with maltose bound; 4) a closed, oxidized form without maltose bound. Any or all of these forms could migrate somewhat differently on a native gel. Ligand-bound (closed) and ligand-free (open) forms of lysine-arginine-ornithine-binding protein are resolved by high pressure liquid chromatography (51). Furthermore, maltose bound to some fraction of the purified double mutant MBP could prevent the reduced form of the protein from opening, favoring rapid reoxidation and reformation of the cross-link. The separation of mutant, but not wild-type, liganded and unliganded forms may reflect that the cysteine substitutions alter the equilibrium thermodynamics or the kinetics of MBP opening and closing. Clearly, more work will be required to resolve these alternatives.

An important, and still unanswered, question is whether the cross-linked G69C/S337C MBP must harbor maltose in its substrate-binding cleft in order to compete effectively with wild-type MBP. Jacobson et al. (49) showed that an interdomain cross-linked form of G46C/S129C sulfate-binding protein has reduced interdomain flexibility and exhibits a much slower dissociation of sulfate. However, the Cys⁶⁹-Cys³³⁷ cross-bridge is at one end of the cleft (Fig. 1), and it is possible that limited movement between domains still occurs at the other end of the cleft, allowing maltose to enter and leave the binding site. We wanted to make an MBP with two cross-bridges, one at each end of the cleft, which presumably would be truly ``locked shut.'' Unfortunately, as described above, we have not yet identified another pair of cysteine residues that could provide this second cross-link.

Cross-linked MBP was observed in shock fluid prepared from cells grown in glycerol-minimal medium (Fig. 3), in which periplasmic maltose concentrations should be low or nonexistent. Also, preliminary investigations indicate that strains expressing G69C/S337C MBP show increased methylation of Tar in cells grown in minimal glycerol medium with or without maltose, whereas cells producing wild-type MBP show increased methylation of Tar only when they are grown with maltose.⁵ Thus, the double mutant protein appears to be able to induce an attractant signal via Tar without its having been exposed, at least intentionally, to maltose. We plan to measure the kinetics of maltose binding of purified cross-linked G69C/S337C MBP. It will be more difficult, although worthwhile, to determine rigorously whether cross-linked G69C/S337C MBP isolated from cells grown in maltose-free or maltose-replete medium is bound to maltose.

We think that the cross-linked MBP will be a useful tool for further investigations of maltose transport and chemotaxis, both in vivo and in vitro. An advantage of our system for such studies is that any decrease in maltose affinity caused by chemotaxis-defective or transport-defective malE mutations should not influence the outcome, thus simplifying the analysis.

Treptow and Shuman (7) reported the isolation of malE missense mutations causing specific defects in maltose transport. Some of these have already been shown to be co-dominant (52). Mutations in malE that cause transport defects can be introduced into the G69C/S337C mutant, and the ability of the triple mutant protein to inhibit transport mediated by wild-type MBP can be determined. If the triple mutant protein no longer inhibits, then the transport-defective mutation probably blocks MBP-MalFG binding. If the triple mutant protein still inhibits, then the mutation may interfere with the ability of MBP to initiate transmembrane signaling by MalFG rather than binding. Cross-linked MBP may also prove useful for measuring MBP-stimulated activities of the MalFGK transport complex in vitro (5, 6). Can, for example, binding of cross-linked MBP to MalFG stimulate ATP hydrolysis by MalK, and if so, how many rounds of ATP hydrolysis does the interaction trigger?

We have accumulated a set of malE mutations that disrupt maltose chemotaxis (14), but analogous to the situation with the transport-defective malE mutations, we do not know whether the mutations interfere with binding of MBP to Tar or with generation of the chemotactic signal after the mutant MBP has bound to Tar. We can create triple-mutant MBP species similar to those described above for analyzing transport and ask if the third mutation eliminates the negative dominance for chemotaxis imposed by G69C/S337C MBP. We can also determine whether the triple mutant protein can stimulate increased methylation of Tar in the absence of maltose. The lack of an in vitro assay for MBP binding to Tar, largely because of the apparent low affinity of MBP for Tar (estimated in vivo at 250 µM; Ref. 31), requires us to resort to another method to examine the structure/function relationships of the MBP-Tar interaction. We think that studying the effects of particular mutations on the negative-dominant phenotype of G69C/S337C MBP constitutes such a viable alternative.