Scanning Cysteine Accessibility of EmrE, an H+-coupled Multidrug Transporter from Escherichia coli, Reveals a Hydrophobic Pathway for Solutes*

EmrE is a 12-kDa Escherichia colimultidrug transporter that confers resistance to a wide variety of toxic reagents by actively removing them in exchange for hydrogen ions. The three native Cys residues in EmrE are inaccessible toN-ethylmaleimide (NEM) and a series of other sulfhydryls. In addition, each of the three residues can be replaced with Ser without significant loss of activity. A protein without all the three Cys residues (Cys-less) has been generated and shown to be functional. Using this Cys-less protein, we have now generated a series of 48 single Cys replacements throughout the protein. The majority of them (43) show transport activity as judged from the ability of the mutant proteins to confer resistance against toxic compounds and from in vitro analysis of their activity in proteoliposomes. Here we describe the use of these mutants to study the accessibility to NEM, a membrane permeant sulfhydryl reagent. The study has been done systematically so that in one transmembrane segment (TMS2) each single residue was replaced. In each of the other three transmembrane segments, at least four residues covering one turn of the helix were replaced. The results show that although the residues in putative hydrophilic loops readily react with NEM, none of the residues in putative transmembrane domains are accessible to the reagent. The results imply very tight packing of the protein without any continuous aqueous domain. Based on the findings described in this work, we conclude that in EmrE the substrates are translocated through a hydrophobic pathway.

EmrE is an Escherichia coli multidrug transporter that confers resistance to a wide variety of toxicants by actively removing them in exchange for hydrogen ions (1)(2)(3)(4). EmrE is a highly hydrophobic 12-kDa protein that has been purified by taking advantage of its unique solubility in organic solvents. After solubilization and purification, it retains its ability to transport as judged from the fact that it can be reconstituted in a functional form (5). Hydrophobicity analysis of the sequence yielded four putative transmembrane domains of similar sizes. Results from transmission Fourier transform infrared measurements agree remarkably well with this hypothesis and yielded ␣-hel-ical estimates of 78 and 80% for EmrE in the organic solvent mixture (CHCl 3 , MeOH) and 1,2-dimyristoyl phosphocholine, respectively (6). Furthermore, in the 1,2-dimyristoyl phosphocholine bilayer most of the amide groups in the protein do not undergo amide-proton hydrogen/deuterium (H/D) exchange implying that most (ϳ80%) of the residues are embedded in the lipid. In addition, EmrE is the first polytopic membrane protein of this size that has been studied by high resolution NMR. A series of heteronuclear ( 1 H-15 N) two-and three-dimensional experiments as well as few triple resonance experiments were applied to the 110-residue protein and led to the assignment of 1 H, 15 N, and a large part of the 13 C backbone, as well as many of the side chain resonances. A preliminary analysis of the secondary structure, based on sequential nuclear Overhauser enhancements connectivities, deviation of chemical shifts from random coil values, and 3 J(H N H a )-coupling constants, supports a model where the protein forms four ␣-helices: residues 4 -26 (TMS1), 1 32-53 (TMS2), 58 -76 (TMS3), and 85-106 (TMS4) (7). EmrE is functional as a homo-oligomer as suggested by co-reconstitution experiments of wild type protein with three different inactive mutants in which negative dominance has been observed (8).
In this work we use the scanning cysteine accessibility method to assess the exposure of residues at various positions in the protein. The scanning cysteine accessibility method is based on the generation of mutants in which unique reactive Cys residues are implanted at desired positions. Each mutant protein is then challenged with various sulfhydryl reagents so as to assess the exposure and reactivity of the Cys residue. The scanning cysteine accessibility method was successfully used to identify residues exposed to the aqueous translocation pathway of ligand-activated channels and the cystic fibrosis transmembrane conductance regulator chloride channel (9 -11). In the case of ion-coupled transporters evidence has been provided that residues in putative transmembrane domains are inaccessible to the permeant sulfhydryl reagent NEM (12)(13)(14). It was therefore suggested that NEM could be used to identify the membrane embedded domains. However, other studies have clearly identified reactive residues in these regions (15)(16)(17)(18). The reactive residues are thought to delineate water-filled cavities that may represent the substrate translocation pathway. Indeed, in some cases the residues were shown to be located in the vicinity of the binding site of the hydrophilic substrates (16 -18).
We have previously shown that the three native Cys residues in EmrE are inaccessible to NEM and a series of other sulfhydryls (19). In addition, each of the three residues can be replaced with Ser without significant loss of activity. A protein without all the three Cys residues (CL) 2 has been generated and shown to be functional (19). Using this Cys-less protein, we have now generated a series of 48 single Cys replacements throughout the protein. The majority of them (43) demonstrate transport activity as judged from the ability of the mutant proteins to confer resistance against toxic compounds and from in vitro analysis of their activity in proteoliposomes. Here we describe the use of these mutants to study the accessibility to NEM, a well characterized sulfhydryl reagent. The study has been done systematically so that in one TMS (TMS2) each single residue was replaced. In each of the other three TMSs, at least four residues covering one turn of the helix were replaced. The results show that although the residues in putative hydrophilic loops readily react, none of the residues in putative transmembrane domains is accessible to NEM. The results suggest very tight packing of the protein without any continuous aqueous domain. In striking contrast with the findings for the other ion-coupled transporters, the results with EmrE suggest the existence of a hydrophobic pathway through which the substrates are translocated.
Mutagenesis-Mutants were obtained by polymerase chain reaction mutagenesis using the overlap extension procedure described by Ho et al. (21). For each mutation a set of two overlapping oligonucleotide primers containing the desired mutation were constructed. The outside primers were those used for the wild type EmrE (5). The template was always CL-EmrE.
Mutagenic oligonucleotides were prepared incorporating a unique restriction site to facilitate mutant identification, and in several cases this required an additional conservative mutation. Mutated DNA was identified by the acquisition of the new restriction site and sequenced (ABI PRISM TM 377, Perkin-Elmer) to ensure that no other mutations occurred during the amplification process.
Resistance to Toxic Compounds-For testing resistance to toxic compounds, cells were grown overnight at 37°C in LB-Amp medium. 5 l of serial 10-fold dilutions of the cultures (10 Ϫ1 -10 Ϫ6 ) were spotted on plates containing the desired compounds as indicated for each experiment. Growth was visualized after 24 h at 37°C. The assay was repeated at least three times for each mutant.
Expression, Purification, and Reconstitution of EmrE-E. coli JM109 transformed with the appropriate plasmid was grown in minimal medium A supplemented with glycerol (0.5%), thiamine (2.5 g/ml), ampicillin (50 g/ml), and a mixture of amino acids (MEM amino acids; Sigma). When the culture reached an A 600 ϭ 0.9, isopropyl-␤-D-thiogalactoside was added to 0.5 mM; 2 h later, the cells were harvested by centrifugation.
EmrE was extracted essentially as described (5) from 50 mg of cells (wet weight) with 550 l of chloroform:methanol (1:1) incubated for 20 min on ice. For phase separation 110 l of H 2 O were added, and the suspension was centrifuged. The upper phase was removed, and 4 volumes of acetone were added to the lower phase. After at least 1 h at Ϫ20°C, EmrE was collected by centrifugation, washed with cold acetone, and dried. For analysis in SDS-polyacrylamide gel electrophoresis, the sample was resuspended in sample buffer and analyzed in 16% gels as described (22).
For reconstitution, 4 -9 g of extracted protein was mixed with E. coli phospholipids (18 l of a suspension of 50 mg/ml in water diluted prior to the experiment in 180 l of chloroform:methanol, 1:2) and 750 l of chloroform:methanol (1:1). The suspension was dried with Argon and resuspended in a solution (60 l) containing 0.18 M NH 4 Cl, 0.015 M Tris-HCl, pH 6.9. The suspension was kept at Ϫ70°C. Before the assay, the proteoliposome suspension was thawed and sonicated in a bath-type sonicator for a few seconds until clear.
Transport Assay-Transport of [ 14 C]methyl viologen into proteoliposomes was assayed by dilution of 3 l of the ammonium chloride containing proteoliposomes into 200 l of an ammonium-free medium containing 35 M [ 14 C]methyl viologen (90 nCi), 140 mM KCl, 10 mM Tricine, and 5 mM MgCl 2 , pH 9.5. At given times the reaction was stopped by dilution with 2 ml of the same ice cold solution, filtering through Schleicher & Schull filters (0.2 m), and washing with an additional 2 ml of solution. The radioactivity on the filters was estimated by liquid scintillation. In each experiment the values obtained in a control reaction with 5 M nigericin were subtracted from all experimental points. This background was between 5 and 10% of the experimental values. The assay was done in triplicates and repeated at least three times.
Labeling with [ 14 C]NEM and Measurement of Expression Levels-Cells (50 mg of wet weight) were suspended in 500 l of a solution containing 100 mM potassium phosphate, pH 7.5, and 5 mM MgSO 4 . [ 14 C]NEM (NEN Life Science Products) was added to a final concentration of 0.5 mM (2 Ci/mol). After 20 min at room temperature the reaction was stopped by addition of 20 mM dithiothreitol. After 10 min 1 ml of above buffer was added, and the cells were collected by centrifugation and washed once with the same buffer. EmrE was extracted from the pellet as described above and analyzed in SDS gels. The amount of EmrE in each mutant was estimated from the Coomassie-stained gels. The amount of radioactivity associated with EmrE was quantitated after exposure of the gel to a Phosphorimager FUJIX BAS 100. The assay was repeated at least three times for each mutant protein.

A Vast Majority of the Residues Can Be Replaced with Cys without Loss of Activity-
The CL-EmrE protein displays significant transport activity and therefore provides a good starting point for generation of mutant proteins with single Cys replacements. In this work we report the results obtained after replacing all the amino acids (starting at position 28 and continuing up to position 57) in TMS2 and its adjacent loops ( Fig.  1). In addition, at least four mutants (one helical turn) were analyzed in each of the other TMSs and one in each hydrophilic loop (Fig. 1), a total of 48 single Cys replacements.
The activity of each of the mutant proteins has been tested in vivo and in vitro. In vivo, the resistance conferred by each of the proteins was assessed by testing the ability of cells expressing them to grow under otherwise nonpermissive conditions. This was achieved in solid media containing either ethidium (200 g/ml), acriflavine (100 g/ml), or methyl viologen (0.1 mM) in which 5 l of logarithmic dilutions of an overnight culture were spotted (Fig. 2, three right panels). Cells carrying the vector plasmid without any insert cannot grow in these media at any of the dilutions tested. Cells expressing either wild type EmrE or the CL-EmrE were able to grow at each of the dilutions. At the high dilutions (10 Ϫ3 -10 Ϫ5 ), the single colonies observed displayed a similar size indicating similar growth rates. As expected (not shown) all the cells grew to a similar degree in control plates containing only ampicillin and none of the toxic compounds. This assay provides a dynamic range to qualitatively analyze activity of the mutants generated. We assume that growth at dilutions of 10 Ϫ3 -10 Ϫ5 is a result of robust EmrE activity. Growth on ethidium is taken as the standard because, as will be seen below, in some mutants the specificity toward other substrates maybe modified.
In TMS2 and the adjacent loops, 28 of the 30 mutants generated were capable of conferring significant resistance to ethidium. Most (25 mutants) were indistinguishable from the wild type; growth was observed also at dilutions of 10 Ϫ4 or 10 Ϫ5 . Two mutant proteins (Y40C and F44C) conferred no measurable resistance. Three mutant proteins (V34C, C41, and I54C) seemed to be less effective than the wild type.
The results for mutants in the other TMS are summarized in Fig. 3. In TMS1 single unique replacements were generated distal and terminal to the essential glutamyl residue at position 14 (8). Mutant proteins I11C, L12C, A13C, V15C, I16C, G17C, and T18C were tested for function (Fig. 3) showed no activity at all in this assay. Cysteine replacements in loops 4 and 5 (L83C and H110C) exhibited full activity. The levels of resistance to different substrates are being used here to obtain a very qualitative idea of the activity of the various mutants. However, no absolute quantitation can be made because the levels of expression of EmrE do differ among the various mutants. Notwithstanding this limitation, several important conclusions can be made. Five mutant proteins are incapable of conferring measurable resistance: I11C and T18C in TMS1, Y40C and F44C in TMS2, and L93C in TMS4. Except for I11C and T18C, the mutants are not expressed to detectable levels even under full induction of expression with isopropyl-␤-D-thiogalactoside (not shown). These residues may play a role in folding and stability of the protein. All the other mutants tested, albeit expressed to diverse levels, are capable of conferring some degree of resistance to ethidium. We conclude that with the exception of the above mentioned residues, each one of the other 43 residues tested can be replaced with Cys and produces a mutant protein that maintains an activity significant enough to confer resistance to ethidium. The fact that such a large number of residues are so permissive has been observed with other transporters as well (12,16,23). It is noteworthy that even residues (such as Val 15 , Arg 29 , Pro 32 , Leu 47 , Thr 50 , Pro 55 , Ile 94 , and Gly 97 ) that are highly conserved in the family of the MiniTEXANs (1, 24) can be replaced without loss of phenotypic complementation. The conservation of these residues may be necessary for functions other than the one measured and/or for stability of the protein.
Some Replacements Modify the Specificity toward Other Substrates-EmrE is a multidrug transporter and, as such, transports a variety of toxic substrates. The specificity of all the mutants to several substrates has been compared using the phenotypic assay described above (Figs. 2 and 3). In general, growth in the presence of acriflavine yielded a pattern very similar to that displayed in ethidium with the exception of three mutants in TMS3 that displayed a decreased resistance to this toxic agent: S72C, L74C, and S75C (Fig. 3). A few strains in TMS2 (V34C, T36C, A52C, and I54C) also showed some decrease of resistance to acriflavine (Fig. 2). More dramatic differences were detected when growth was challenged with 0.1 mM methyl viologen. Mutants A48C and A52C that FIG. 2. Growth phenotype of cells expressing single Cys replacements in TMS2 and its distal and proximal loops. E. coli JM109 cells transformed with pKK56-EmrE (WT), pKK56-CL (CL), pKK223-3 (vec), or the various mutants were grown overnight at 37°C in LB-Amp medium. 5 l of 10 Ϫ1 -10 Ϫ6 dilution of the culture were spotted on a series of LB-Amp plates with various EmrE substrates or on a plate with no addition. Growth was analyzed after overnight incubation at 37°C. With the three toxicants used, growth of cells bearing pKK-56-CL or pKK56-WT EmrE was indistinguishable. At the concentrations used in this experiment, growth of pKK223-3 bearing cells was undetectable at each of the dilutions used (see three right panels for each of the compounds tested). Growth in the control plates was practically identical for all mutants tested (not shown). The heights of the bars indicate the maximal dilution at which cell growth was detected. Mutants Y40C and F44C were not expressed at detectable levels.
FIG. 1. Accessibility of EmrE residues to NEM. The secondary structure model of EmrE (5) is shown with the summary of the results described in this work. Each of the highlighted residues was replaced with Cys in a CL EmrE, and the proteins were tested for activity and challenged with [ 14 C]NEM. As a result the residues can be classified as follows: shaded circles, residues accessible to [ 14 C]NEM; black circles, residues inaccessible to [ 14 C]NEM; black triangles, residues that when replaced with Cys yielded inactive proteins or the proteins were not expressed. In addition, the boundary of TMS2 was changed from the previously suggested one (5) based on the results described here.
showed robust resistance to ethidium and acriflavine show no resistance to methyl viologen. I54C showed a very significant resistance to ethidium and acriflavine but no resistance to methyl viologen. A few others are also noteworthy: V34C, T36C, C41, A42C, L47C, Q49C, and G57C displayed 2-3 orders of magnitude less resistance to methyl viologen, relative to the others (Fig. 2).
Most Mutants Are Also Functional after Purification and Reconstitution in Proteoliposomes-All the mutant proteins have been purified and assayed for ⌬pH driven [ 14 C]methyl viologen uptake activity in proteoliposomes. Most of them (34 mutants) display significant levels of transport that range between 200 and 2000 pmol/min/mg. In addition to the five mutant proteins that do not display any activity in vivo, a few others lose practically all their activity (Ͻ100 pmol/min/mg protein) upon purification. Even though they display activity in vivo, no activity was detected with protein extracted from: L12C and A13C in TMS 1; P32C, P55C, T56C, and G57C in the loops either distal or proximal to TMS2; and S72C and L74C in TMS3. These residues may be important for the proper folding after the organic solvent treatment or for proper reconstitution into the proteoliposomes. Other assays, independent of the extraction with organic solvents, are now being developed to further test this phenomenon.
Cys Residues in Positions at Putative Transmembrane Domains Are Inaccessible to [ 14 C]NEM Labeling-We have previously shown that the activity of the wild type protein is unaffected by exposure to NEM and other sulfhydryls even at high concentrations (19). We interpreted these results as suggesting that the Cys residues are not accessible to NEM. We now confirm this conclusion by directly measuring accessibility of the Cys residues to [ 14 C]NEM and extend our studies to all the single Cys replacements generated. Each mutant protein was challenged with the radiolabel in the intact cell. It was then partially purified, and the radioactivity associated with it was measured after separation on SDS-polyacrylamide gel electrophoresis. Using this experimental protocol, neither the wild type protein nor the fully Cys-less protein (CL-EmrE) were labeled at all (Fig. 4). As expected, T28C, a mutant with a Cys residue in a putative hydrophilic loop, was fully labeled. From quantitation of the amount of radioactivity incorporated and the protein in the gel, it can be estimated that about 1 mol of [ 14 C]NEM was incorporated per mol of T28C. To demonstrate that the lack of labeling of the wild type protein is due to a structural constraint, EmrE was denatured with a mixture of 2% SDS and 8 M urea. Under these conditions all three Cys residues are fully accessible, and about 3 mol NEM are incorporated per mol EmrE. As expected, no further increase in the labeling level was observed under these conditions in either CL or T28C. These results confirm our hypothesis that the Cys residues in the wild type protein are inaccessible to NEM and allow for a detailed study of the reactivity of Cys at each one of the mutants generated.
The above experimental paradigm was used to assess the level of labeling of all the functional single Cys replacements generated in this work. The results in Fig. 5 describe the labeling of residues in TMS2 and the adjacent loops. A distinct pattern of labeling was observed in which three domains are apparent: in the first group (from Thr 28 to Ser 33 ) all the residues were labeled, although to varying degrees (between 30 and 100%). We suggest that these residues are exposed to solvent and form part of loop 2. A similar conclusion is reached with another group of residues starting at Tyr 53 and continuing to Gly 57 . We conclude that these residues form part of loop 3. A very different behavior is observed with a large group of residues inaccessible to [ 14 C]NEM. This group includes all the residues at positions 34 -52. In several mutants (V34C, T36C, A42C, S43C, L46C, and T50C), there was no measurable radioactivity associated with the protein. In the others, with two exceptions, the level was significant but always lower than 10%. The only two mutants in this domain that are labeled to levels of about 20% are Q49C and L51C.
To assess the rates of labeling, two fully accessible mutants were compared with two that are not labeled. Although both T50C and T36C are not labeled even after 2 h incubation, T28C and G57C are fully labeled after 5-10 min (data not shown).
To show that the lack of labeling is due to structural constraints, all the proteins that are inaccessible to [ 14 C]NEM in the intact cell were challenged with the label after purification and solubilization in the presence of 2% SDS. Most of the mutant proteins reacted with the sulfhydryl reagent under these conditions (data not shown). However, three of them, V34C, L47C, and A48C, were accessible only upon treatment with harsher denaturing conditions (2% SDS and 8 M urea).
A similar picture is uncovered when the accessibility of residues in the other three TMS is analyzed. The results of this study are summarized in Fig. 6. In each TMS, at least four contiguous residues were replaced with Cys and challenged with NEM. In TMS1, L12C, A13C, and I16C were nearly completely inaccessible. Low labeling of mutants V15C (13% of T28C) and G17C (20% of T28C) was observed (Fig. 6). In TMS3, mutants S72C, L73C, and L74C showed no reactivity whatsoever, whereas S75C was labeled to a low (23% of T28C) but significant degree. In TMS4, mutants I94C, C95, A96C, and G97C were not labeled to any measurable degree. As expected, residues in hydrophilic loops 4 (L83C) and 5 (H110C) were fully labeled.

DISCUSSION
Only a very small number of residues appear essential to EmrE function. In this work we describe the generation of single Cys replacements in 48 residues, about 50% of the protein, and nearly all of them (43 residues) show a measurable activity. Also, 7 of the 8 charged residues in the protein can be replaced without loss of activity. 3 Interestingly, even residues conserved in the family can be replaced without loss of activity. As mentioned above, the conservation may be necessary for functions other than the ones we are testing. Among the inactive mutants, in three cases (Y40C, F44C, and L93C) the protein is not detectable at measurable levels and therefore little can be concluded about the role of these residues on catalytic activity. It is possible that they play a role in folding, insertion, or stability of the protein. Replacement of Ile 11 and Thr 18 with Cys yields practically inactive proteins. It is interesting that both Ile 11 and Thr 18 are located at a distance of about one helical turn from the Glu 14 , the only charged residue in the putative membrane domain. Glu 14 cannot be replaced even with Asp, suggesting a central role in the catalytic cycle (8,24). 3 The role of both residues needs to be studied further by a 3 H. Yerushalmi and S. Schuldiner, unpublished results. Only two mutants were not expressed at detectable levels, Y40C and F44C. A48C was expressed at very low but detectable levels. The results shown are relative to labeling of T28C that was fully accessible to [ 14 C]NEM. In each case, level of labeling was quantitated and corrected for expression levels.
more detailed characterization of the mutant proteins and by replacements with other amino acids. In addition to Ile 11 , Glu 14 , and Thr 18 only two other residues, Tyr 60 and Trp 63 in TMS3, have been previously shown to be important for activity. Even substitution with other aromatic amino acids caused a complete loss of activity (8). The finding that most residues of ion-coupled transporters can be replaced without serious impairment of function has already been described and discussed in extensive studies of the E. coli lac permease (16). In these studies, only six positions (of 417 positions) are clearly irreplaceable. It has been suggested that the role of most of the nonessential amino acids may be to provide a structural scaffold.
In our studies, mutagenesis of two groups of residues has generated proteins with modified specificity to at least one of the substrates of EmrE. Practically all replacements in TMS3 display decreased resistance to acriflavine. In TMS2, a total of eight mutants show a significant decreased resistance to methyl viologen. They seem to cluster on two faces of the ␣-helix (Fig. 7): one group at positions 36, 47, and 54 and the other group at positions 34, 41, 48, and 52 with 42 somewhat close to this cluster. Interestingly, the first group is in the same face of two proteins that are not expressed (Y40C and F44C). These mutants provide an excellent tool for further investigation of the location of the binding contacts of substrates.
The results described in this work demonstrate that all the residues in putative membrane domains of EmrE are practically inaccessible to [ 14 C]NEM (Fig. 1). The reaction of maleimides with sulfhydryl groups involves addition of mercaptide ions in the protein to the olefinic double bond (25). Therefore low levels of labeling may result from either one of the following reasons or a combination of them: steric hindrance, lack of ionization of the sulfhydryl group, or low reactivity because of neighboring residues. NEM is a relatively hydrophobic and small maleimide that can freely cross lipid membranes. The latter contention is supported in this work by the fact that residues in all the hydrophilic loops are fully accessible. In loops 2 (residues Thr 28 to Ser 33 ) and 3 (Tyr 53 to Gly 57 ), all the residues react with NEM. The lower reactivity of some of the residues in hydrophilic loops (such as at the position of Leu 30 ) may reflect steric hindrance, because clearly its neighbors react quantitatively. Residues at positions Pro 32 and Ser 33 and Tyr 53 to Gly 56 may be at the edge of the membrane domain. In loop 4 and at the C terminus one residue was tested, and they were shown to be fully accessible to NEM.
In contrast, none of the residues in putative transmembrane domains reacted with NEM to a significant degree. In this domain some of the sulfhydryl residues face the lipid milieu, and they do not seem to be able to release their proton and therefore cannot react with NEM. As for the rest of the residues, very tight packing of the helices would prevent even relatively small reagents such as NEM from approaching.
In the case of other proteins such as the Tn-10 encoded tetracycline/H ϩ antiporter and the erythrocyte anion exchanger, evidence has been provided that residues in some transmembrane domains are inaccessible to the permeant sulfhydryl reagent NEM (12)(13)(14). It was therefore suggested that NEM could be used to identify the membrane embedded domains. However, other studies have clearly identified reactive residues in membrane domains (15)(16)(17)(18). The reactive residues are thought to delineate water-filled cavities that may represent part of the substrate translocation pathway. Also, in the extensive studies performed with the E. coli lac permease parts of the membrane embedded areas were found inaccessible to NEM. Yet, many others reacted freely with NEM (16). Indeed, both in the lac permease (16) and in UhpT (17), the NEMreactive residues were shown to be located in the vicinity of the binding site of the hydrophilic substrates. Unlike in the other transporters, in EmrE, every single TMS was shown to be inaccessible to NEM, suggesting that the substrates are not translocated through an hydrophilic pathway.
The results described in the present work cannot exclude the presence of a highly selective filter that prevents molecules other than substrates from accessing a putative aqueous pathway in EmrE. However, this possibility is refuted by Fourier transform infrared studies of EmrE in which we found that a large fraction of the amide protons do not readily exchange with solvent deuterium (6). Taken as a whole, the results described support the model of a tightly packed four helix antiparallel bundle in which the majority of the protein is well embedded in the membrane and not accessible to solvent. The boundaries of the embedded segments as estimated with this technique are in remarkably good agreement with the secondary structure determined from the NMR analysis of the protein (7).
In the case of EmrE, the substrates are quite hydrophobic, and therefore it maybe energetically more favorable to interact directly with the protein rather than permeate through a water-filled pathway. Movement of substrates through a tightly packed protein must require disruption and reorganization of the existing structure. These types of interaction have been observed in several cases. For example, NMR shows that the singular structures of soluble synthetic four helical bundles adopt a disordered array of states upon binding a hydrophobic heme cofactor. The interaction of the heme with the polypeptide is quite specific and displays a relatively high affinity in the nanomolar range (26). Tetraphenylphosphonium, a high affinity substrate of EmrE, interacts specifically also with BmrR, a transcription activator of the Bmr gene, a Bacillus subtilis multidrug transporter (27). In this protein, the tetraphenylphosphonium binds to a hydrophobic pocket with a key electrostatic component (Glu 134 ) at its bottom (28). The entrance of tetraphenylphosphonium to the binding pocket occurs after unfolding of a nine-residue ␣-helix. At present, without further structural information, we can only speculate that a similar type of binding site may exist in EmrE as well where only one negatively charged and essential Glu residue (Glu 14 ) is present in the putative transmembrane domains of the protein. The results presented above also suggest that entrance of the ligand to the binding site may require movement of parts of the protein.