Function of the membrane fusion protein, MexA, of the MexA, B-OprM efflux pump in Pseudomonas aeruginosa without an anchoring membrane.

Resistance of Pseudomonas aeruginosa to multiple species of antibiotics is largely attributable to expression of the MexA, B-OprM efflux pump. The MexA protein is thought to be located at the inner membrane and has been assumed to link the xenobiotics-exporting subunit, MexB, and the outer membrane channel protein, OprM. To verify this assumption, we analyzed membrane anchoring and localization of the MexA protein. n-[9, 10-(3)H]Palmitic acid incorporation experiments revealed that MexA was radiolabeled with palmitic acid, suggesting that the MexA anchors the inner membrane via the fatty acid moiety. To evaluate the role of lipid modification and inner membrane anchoring, we substituted cysteine 24 with phenylalanine or tyrosine and tested whether or not these mutant MexAs function properly. When the mutant mexAs were expressed in the strain lacking chromosomal mexA in the presence of n-[9,10-(3)H]palmitic acid, we found undetectable radiolabeling at the MexA band. These transformants restored antibiotic resistance to the level of the wild-type strain, indicating that lipid modification is not essential for MexA function. These mutant strains contained both processed and unprocessed forms of the MexA proteins. Cellular fractionation experiments revealed that an unprocessed form of MexA anchored the inner membrane probably via an uncleaved signal sequence, whereas the processed form was undetectable in the membrane fraction. To assure that the lipid-free MexA polypeptide could be unbound to the membrane, we analyzed the two-dimensional membrane topology by the gene fusion technique. A total of 78 mexA-blaM fusions covering the entire MexA polypeptide were constructed, and all fusion sites were shown to be located at the periplasm. To answer the question of whether or not membrane anchoring is essential for the MexA function, we replaced the signal sequence of the MexA protein with that of the azurin protein, which contains a cleavable signal sequence but no lipid modification site. The signal sequence of the azurin-MexA hybrid protein was properly processed and bore the mature MexA, which was fully recovered in the soluble fraction. The transformant, which expressed azurin-MexA hybrid protein restored the antibiotic resistance to a level indistinguishable from that of the wild-type strain. We concluded from these results that the MexA protein is fully functional as expressed in the periplasmic space without anchoring the inner membrane. This finding questioned the assumption that the membrane fusion proteins connect the inner and outer membranes.

Pseudomonas aeruginosa often infects immuno-compromised patients with cancer, dialysis, cystic fibrosis, and transplantation. Problems associated with P. aeruginosa infection are that this organism is naturally resistant to many noxious compounds such as ␤-lactam antibiotics, fluoroquinolones, chloramphenicol, and tetracycline (1). Upon exposure to antibiotics, the organism easily elevates the antibiotic resistance to a higher level than the wild type strain (2). Recent studies have suggested that both basal and elevated levels of intrinsic multiantibiotic resistance in this organism are mainly attributable to interplay between the antibiotic efflux pumps and low outer membrane permeability (3,4). Among several efflux pumps reported in P. aeruginosa, the MexA, B-OprM pump plays a central role in antibiotic resistance because it is expressed in the wild-type strain and up-regulated upon nalB mutation (5).
A MexA,B-OprM pump, encoded by the mexA,B-oprM genes, consists of two inner membrane-associated components, MexA and MexB, and an outer membrane component, OprM (6). It has been assumed on the basis of computer-aided structural prediction that MexA belongs to a membrane fusion protein family that is thought to connect the inner and the outer membranes (7,8). It has been suggested that each of MexB and OprM functions as an antibiotic-exporting component and an antibiotic exit outer membrane channel, respectively (3,4,8). Both MexA and OprM have been assumed to be lipoproteins, since they contain a consensus signal sequence for aminoterminal lipid modification, and amino-terminal amino acids seem to be blocked (6,9). However, the membrane topology and precise localization of these proteins remain obscure, except for the two-dimensional structure of the MexB protein, which spans the cytoplasmic membrane 12 times and forms 2 large hydrophilic domains extending toward the periplasmic space (10). This protein was predicted to function as the substrateexporting subunit across the cytoplasmic membrane.
Computer-aided hydropathy analysis of membrane fusion proteins suggested that they might be highly hydrophilic (7,11). Preliminary analyses of the topology of the MexA homologues, such as CzcB of Alcaligenes eutrophus and CvaA, HlyD, and AcrA of Escherichia coli, suggested that they might be largely located at the periplasmic space (12)(13)(14)(15). However, the precise membrane topology and localization of the MexA protein remained unclear. We analyzed whether or not the MexA protein can function without a lipid moiety and report that the lipid-free MexA protein was soluble in the periplasmic space and fully functional in the antibiotic export.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, and Growth Conditions-Bacterial strains and plasmids used are listed in Table I. Cells were grown aerobically in L-broth containing 10 g of Tryptone, 5 g of yeast extract, and 5 g of NaCl/liter (pH 7.2) at 37°C. pKM1 is a new version of pYZ5 (16).
Incorporation of [9, H]Palmitic Acid -Cells were grown in Lbroth overnight and diluted with 1 ml of fresh L-broth to A 600 ϭ 0.2. To this was added 1.48 ϫ 10 6 Bq of [9,10-3 H]palmitic acid (Amersham Pharmacia Biotech), and the tubes were incubated at 37°C overnight. Cells were harvested by centrifugation, washed once with 1 ml of 50 mM sodium phosphate buffer containing 150 mM NaCl (pH 7.2; 50 mM PBS 1 ), suspended in 800 l of 50 mM PBS. The cells were mixed with 200 l of 10% of octyl-␤-glucoside and 2 l of 500 mM phenylmethylsulfonyl fluoride and incubated at 4°C for 45 min. The resulting cell lysate was incubated for 135 min after adding 23 g of DNaseI and centrifuged by an Eppendorf 5412 for 10 min. The supernatant was mixed with 10 l of anti-MexA rabbit antibody (17) and protein A-agarose (Roche Molecular Biochemicals) then incubated at 4°C overnight. The protein A-agarose was washed 4 times with a solution containing 1 M NaCl, 1% octyl-␤-glucoside, 50 mM sodium phosphate (pH 7.2) and once with a solution of 1% octyl-␤-glucoside and 50 mM sodium phosphate (pH 7.2). The resulting precipitates were extracted by 20 l of Laemmli solubilizer at 100°C for 5 min, then subjected to SDS-polyacrylamide gel electrophoresis (PAGE, 10% gel), after which the dried gel was exposed to Biomax MR film (Eastman Kodak Co.).
Manipulation of the DNA-Preparation of the plasmid DNA, the restriction enzyme treatment, ligation, and transformation were carried out essentially as described by Sambrook et al. (18). A series of nested deletions of the mexA gene was obtained by exonuclease III (TOYOBO) treatment (19). Construction of mexA-blaM fusions was carried out as follows. A 1.7-kilobase NcoI/BamHI fragment encompassing the entire mexA gene was subcloned into pYZ4 pretreated with BamHI and NcoI to yield pMEXA2. pMEXA2 treated with KpnI and BamHI was incubated with exonuclease III for various lengths of time, then treated with SacI. A nested series of the deleted mexA gene was ligated with the blaM gene (coding for ampicillinase without a signal peptide) isolated by digesting pKM1 with SmaI and SacI to obtain mexA-blaM fusions. Transformants harboring the mexA-blaM fusions were screened on L-agar plates containing 12.5 g/ml kanamycin. The oligonucleotide primers used to confirm the junction between the mexA and the blaM was CTCGTGCACCCAACTGA or TGTTGAGATCCAGT-TCG for the anticoding strand.
Construction of Polyhistidine-tagged mexA Gene-We incorporated six histidines at the carboxyl-terminal glycine residue by polymerase chain reaction using forward and reverse primers, TCGAATTCTC-CGAGGTTTCCG and GGATCCCCTTGATCAGTGGTGATGGTGGTG-ATGGCCCTTGCTGTCGGTTTTCGC, respectively. Polymerase chain reaction was performed with AmpliTaq Gold (Applied Biosystems, Inc.) according to the manufacturer's instructions. Amplified products were treated with T4-DNA polymerase (New England Biolabs) and EcoRI and cloned into pBluescript II SKϩ pretreated with SmaI, EcoRI, and shrimp alkaline phosphatase (Roche Molecular Biochemicals). The nucleotide sequence was confirmed subsequently for all the polymerase chain reaction products. Next, an HindIII-EcoRI fragment (about 1,100 base pairs) covering the amino-terminal half of mexA was subcloned into the pBluescript II SKϩ (Stratagene) pretreated with HindIII, EcoRI, and shrimp alkaline phosphatase. A HindIII-BamHI fragment (about 1,570 base pairs) covering the entire mexA gene plus six histidines was subcloned into pHSG397 (20) pretreated with HindIII and BamHI (pMexA-His-397), then into pMMB67HE (21) pretreated with HindIII and BamHI to yield pMexA-His.
Site-directed Mutagenesis of Cysteine Residue-We replaced cysteine 24 with phenylalanine or tyrosine by site-directed mutagenesis using forward primers GGTCGCGATTTCGGCCCTTTCCGGGTTCGG and GGTCGCGATTTCGGCCCTTTCCGGGTACGG for phenylalanine and tyrosine, respectively, and reverse primer GACTGCAGGTAGGCG-GCATTGGCG. Polymerase chain reaction was performed with Ampli-Taq Gold (Applied Biosystems, Inc.). Amplified products were treated with T4 DNA polymerase followed by PstI digestion and subcloning into pBluescript II SKϩ digested with EcoRV and PstI. Next, the NruI-PstI fragment from the mutant mexA was subcloned into the NruI and PstI sites of mexA, contained upstream of the mexA gene through one of the two PstI sites in the mexA gene. A PstI-SmaI (about 900 base pairs) fragment containing the second PstI site was subcloned into the above plasmids pretreated with PstI and EcoRV. Next, the SacI-HindIII fragment of these plasmids containing the entire mexA gene with the respective mutation was subcloned into pMMB67EH (21) pretreated with SacI and HindIII to yield pMexA(C24F) and pMexA(C24Y), respectively. To construct histidine-tagged mexA with a mutation at residue 24, a SacI-BsaI fragment (about 1,100 base pairs) from each mutant plasmid was subcloned into pMexA-His-397 pretreated with BsaI and SacI. A SacI-BamHI fragment of each mutagenized mexA was then subcloned into pMMB67EH pretreated with SacI and BamHI.
Site-specific Insertion of the blaM Gene-To insert the blaM gene next to Leu-43 of mexA, the amino-terminal region of the mexA gene was amplified by AmpliTaq Gold (PE Applied Biosystems) using forward, TCTGAATATGGGCCATGGCG, and backward, TAAGGTAC-CAGGGTCACGATCCCGAC, primers. The amplified products were digested with KpnI and NcoI, then ligated to pYZ4 pretreated with KpnI and NcoI. The recombinant plasmid DNA was treated with KpnI, then with T4 DNA polymerase to trim the 3Ј-protruding nucleotides. This was followed by SacI digestion. Next, the blaM gene, isolated from Construction of Soluble and Membrane-bound MexAs-To construct a mexA gene in which the signal sequence was replaced with that of the azurin gene, we annealed the sense and antisense oligonucleotides, AGCTTGCCTAGGAGGCTGCTCCATGCTACGTAAACTCGCTGCGG-TATCCCTGCTGTCCCTGCTCAGTGCGCCGCTGCTGGCTGCCGAG and TCGACTCGGCAGCCAGCAGCGGCGCACTGAGCAGGGACAGC-AGGGATACCGCAGCGAGTTTACGTAGCATGGAGCAGCCTCCTAG-GCA, respectively, followed by HindIII and SalI digestion to subclone into pHSG398 (20) pretreated with HindIII and SalI (pAZU398). The DNA fragment encoding mature MexA was amplified by LA Taq (Takara, Japan) using the forward and reverse primers, CCGCTCGAGC-GGAAAAAGCGAG and CGGGATCCCCTTGATCAGCCCTTGC, respectively, according to the manufacturer's instructions. The fragment was treated with BamHI and XhoI to subclone into pAZU398 pretreated with BamHI and SalI. A BamHI-HindIII fragment containing the whole recombinant gene of this plasmid was subcloned into pMMB67HE pretreated with BamHI and HindIII to yield pAzu-MexA.
To construct a mexA gene containing the first transmembrane segment of the mexF gene, we first amplified the DNA fragment encoding mature MexA by LA Taq (Takara) with forward and reverse primers, TCCCTCGAGGGAAAAAGCGAGGCG and GGGGTACCCCTTGATC-AGCCCTTG, respectively, and subcloned it into pHSG397. Next, the first transmembrane segment of the mexF gene was amplified by LA Taq using forward and reverse primers, CCGGATCCGGCGATAGGAA-GAACCGATG and CAAGTCGACTTCCGGGTATTCGCTGATGG, respectively, and the amplified fragments treated with BamHI and SalI were subcloned into the above plasmid pretreated with BamHI and XhoI. Next, the HindIII-KpnI fragment containing the entire fusion gene was subcloned into pMMB67HE pretreated with HindIII and KpnI to yield pMexF-MexA.
Single-cell Minimum Inhibitory Concentration (MIC) of Antibiotics-E. coli XL1-Blue harboring the fusion plasmid was grown in Mueller-Hinton broth containing 0.4% (w/v) KNO 3 and 12.5 g/ml kanamycin at 37°C overnight without shaking. Cells were diluted 104-fold with Mueller-Hinton broth, and about 40 bacteria were inoculated using a microplanter (Model MIT-P, Sakuma Co., Japan) on Mueller-Hinton Agar medium (Becton Dickinson) containing 2-fold serially diluted ampicillin. The single cell MIC was defined as the lowest concentration of ampicillin that inhibited the growth of the bacteria after overnight incubation at 37°C.
Preparation of Inner and Outer Membrane Fractions-E. coli XL1-Blue harboring the fusion plasmid was grown overnight in 20 ml of L-broth, diluted with 200 ml of the same fresh medium containing 12.5 g/ml kanamycin, and rotated at 37°C for 1 h. Next, 1 mM isopropyl-1-thio-␤-D-galactopyranoside was added, and the cells were grown for an additional 4 h at 37°C. Cells were harvested by centrifugation, washed once with 20 mM sodium phosphate buffer (pH 7.2) containing 150 mM NaCl (20 mM PBS), and resuspended in 20 ml of 20 mM PBS containing 0.5 mM p-toluenesulfonyl fluoride. The suspension was subjected to sonic oscillation in an ice bath for 15 min with 24-s exposure and 36-s intermittent cooling/min (Cell Disrupter 200, Branson). Unbroken cells and large fragments were removed by centrifugation at 10,000 ϫ g for 15 min at 23°C, and the crude membrane fraction was obtained by centrifugation at 130,000 ϫ g for 1 h at 15°C. The membrane fraction was washed once with 20 mM PBS and suspended in 2.2 ml of PBS. To the membrane fraction (1.84 ml), 0.16 ml of 10% of sodium sarcosinate was added, and the mixture was incubated at 30°C for 30 min, then centrifuged at 130,000 ϫ g for 1 h at 15°C. The supernatant was used as the inner membrane fraction. The inner and outer membrane fractions of P. aeruginosa were also obtained by the method described earlier (22).

Deletion of the Lipid Moiety by Replacing Cysteine 24 with
Another Amino Acid-Since the amino-terminal region of MexA contains the lipid modification consensus sequence LSGC (9), we carried out a [ 3 H]palmitic acid incorporation experiment (Fig. 1). Fluorography showed that the extracts of TNP071 exhibited a radiolabeled protein band corresponding to the MexA, 42 kDa (Fig. 1, lane 2). In contrast, radiolabeling in the sample from the MexA deletion mutant, TNP070, was undetectable (Fig. 1, lane 1). Although the chemical nature of bonding between the fatty acid moiety and MexA was not confirmed, it is likely that the fatty acid moiety covalently attached to the MexA protein, because the lipid(s) moiety could withstand heating at 100°C for 5 min in SDS solution. It is also likely that the lipid moiety plays a role in anchoring the MexA protein to the inner membrane.
This raises the question of whether or not lipid modification at cysteine 24 is essential in the localization and function of MexA. To answer this question, we substituted cysteine 24 with phenylalanine or tyrosine (designated as C24F-MexA or C24Y-MexA, respectively). When the modified mexA gene was expressed in the mutant lacking MexA, TNP070, and analyzed by the immunoblotting method using anti-MexA antibody (17), we found an extra band(s) in addition to the band corresponding to the mature form of the wild-type MexA (Fig. 2a). In addition to this band, both mutants showed a band with a higher molecular mass than the mature form, which is most likely the MexA protein with an unprocessed signal sequence. The mutant C24Y-MexA showed a band in-between mature MexA and high molecular mass MexA, which is most likely the protein processed at another site. A cellular fractionation experiment revealed that the processed form of C24F-MexA was recovered in the supernatant fraction of the EDTA-lysozymetreated cells, whereas the unprocessed C24F-MexA appeared in the inner membrane fraction (Fig. 2b). To confirm the nature of these molecules, we constructed the histidine-tagged C24F-MexA and C24Y-MexA (designated as C24F-MexA-His and C24Y-MexA-His, respectively), and the proteins were purified by affinity chromatography using nickel-nitrilotriacetic acidagarose. Amino-terminal amino acid sequence analysis showed that the two bands in C24F-MexA-His appeared to be MQRT-PAM and FGKSEAPPPA, which were identical to the aminoterminal amino acid sequence of mature MexA and the MexA processed at glycine 23, respectively. The two bands appearing in C24Y-MexA-His were MQRTPAMRVL and KSEAPPPAQT, corresponding to the amino-terminal sequence of the mature form and the sequence after glycine 25, respectively. To assure that C24F-MexA and C24Y-MexA were free of lipid modification, we carried out the [ 3 H]palmitic acid incorporation exper- iment. As expected, the radiolabeling in both MexAs were undetectable, although the wild-type MexA expressed from the plasmid was heavily radiolabeled (Fig. 2c). These results indicated that both C24F-MexA and C24Y-MexA were not acylated. TNP070 transformed with either one of these modified mexA genes fully restored the susceptibility to azthreonam and chloramphenicol to the level of their parent strain PAO4290 (Table  II), indicating that lipid modification is not essential for the MexA function.
Function of Soluble and Membrane-bound MexA-Since the above result showed that the lipid moiety is not essential for the pump function and the processed and unprocessed forms of C24F-MexA were fractionated in the soluble and inner membrane fraction, respectively, the question arises about which form of MexA retains activity. To address this issue, we constructed the recombinant mexA genes containing a signal sequence of the azurin gene, a periplasmic protein, or an aminoterminal transmembrane segment of mexF, an inner membrane protein. When Azu-MexA or MexF-MexA was expressed in the strain lacking MexA, a single band corresponding to Azu-MexA or MexF-MexA was detectable (Fig. 3, lanes 5  and 8). To characterize the products, we analyzed the aminoterminal sequence of Azu-MexA-His, and it appeared to be AESSGKSEAP, indicating that proper processing of the hybrid MexA has occurred at the correct site.
We next localized the Azu-MexA protein by cellular fractionation and analyzed it by the immunoblotting method (Fig. 3,  lanes 6 and 7). The Azu-MexA protein was found only in the whole cell lysate but was undetectable in either the inner or outer membrane fraction. Therefore, we concluded that the Azu-MexA protein was processed, and the MexA became soluble in the periplasm. Using a similar technique, we analyzed the localization of MexF-MexA and found that the protein was only detectable in the inner membrane fraction. The molecular mass of MexF-MexA was a little larger than the wild-type MexA (Fig. 3, lanes 8 and 9), indicating that the MexF-MexA anchored the membrane via the amino-terminal hydrophobic segment of MexF. We carried out the [9,10-3 H]palmitic acid incorporation experiment using the strain expressing Azu-MexA or MexF-MexA and found that the radiolabeling in the bands corresponding to these recombinant MexAs was undetectable (data not shown). Thus, we were successful in constructing the lipid-free MexA variants localized either only in the periplasm or inner membrane.
Antibiotic susceptibility of the MexA variants was determined, and the results are shown in Table II. The TNP070 strain expressing C24F-MexA or C24Y-MexA fully restored the antibiotic susceptibility to the level of the wild-type strain. However, it is not certain whether the soluble form of MexA or the membrane-bound form is functional as the pump subunit, because these transformants expressed both forms of MexA. To ascertain this point, we tested to see if the antibiotic susceptibility of TNP070 expressed either one of Azu-MexA or MexF-MexA (Table II). Both transformants restored the antibiotic susceptibility against azthreonam and chloramphenicol similar to the strain expressing the wild-type MexA (Table II). These results unequivocally demonstrated that membrane anchoring is not essential for the function of MexA, at least for the extrusion of antibiotics.
Construction and Expression of the Reporter Gene, mexA-blaM Fusion-The results described above suggested that processed forms of the cysteine 24-modified MexA and Azu-   MexA could be soluble in the periplasmic space and functioned properly in the intact cell. Hydropathy analysis by the TOP-PRED II program (25) predicted that the MexA protein bears two hydrophobic segments at a site proximal to the aminoterminal end. One is likely to be the topology-determining signal sequence. Another is a stretch of segment immediately downstream of the signal peptide. To verify whether or not MexA contains membrane-anchoring segments, we constructed mexA-blaM fusions as described under "Experimental Procedures." We obtained 77 independent fusion genes with correct framing (Table III). The gaps between one fusion site to the nearest neighboring sites were consistently less than 21 amino acid residues, excepting that there were 33 amino acid residues between Ser-27 and Ala-60. To cover this region we constructed an additional fusion at Leu-43. Accordingly, we obtained a total of 78 clones covering the residue Gly-23 to the carboxyl-terminal end.
To ascertain proper expression of the hybrid proteins in the E. coli host, we ran SDS-PAGE of the inner membrane fraction prepared from the cells harboring the plasmids, and the protein was visualized by the immunoblotting method using an antibody raised against ampicillinase (Fig. 4). The result shown in Fig. 4 depicted that the hybrid proteins from G383-BlaM through G23-BlaM were lined up in decreasing molecular mass as the fusion sites became more distal from the carboxyl-terminal end. The size of all hybrid proteins was consistent with the size predicted from the length of the truncated MexA plus the size of the reporter protein. The size of the G23 hybrid protein was slightly larger than the S27 hybrid protein, suggesting that the sequence LSGC acts as a cleavage and lipid modification signal sequence in the E. coli host (Fig. 4, lanes 12  and 13). In fact, the [ 3 H]palmitic acid radiolabeling was positively demonstrated in isolate 77 but not in isolate 78 (data not shown). Faint protein bands with a low molecular mass below that of the expected hybrid proteins were most likely the degradative products of the hybrid proteins often seen in blaM fusion in other proteins (16,26). A faint protein band seen in Fig. 4, lane 13 with the mass about 30 kDa might be an unidentified protein derived from the E. coli host.
Localization of the Reporter Protein and Membrane Topology of the MexA Protein-The strains expressing the MexA-BlaM hybrid protein in the periplasm are expected to be ␤-lactamresistant, whereas the strains carrying the fusion in the cytoplasm will be ␤-lactam-susceptible. To ascertain the localization of the hybrid proteins, we determined ampicillin susceptibility of the cells harboring the mexA-blaM fusion gene. Table III shows that all the transformants carrying the mexA-blaM fusion exhibited the MICs of ampicillin 50 to more than 800 g/ml. The MIC of ampicillin for the host cell harboring pYZ4 without BlaM fusion appeared to be 3.13 g/ml. These   1  Gly-383  200  27  Ser-248  400  53  Asn-154  800  2  Thr-379  800  28  Phe-247  400  54  Ile-153  800  3  Asn-365  800  29  Lys-245  400  55  Lys-145  800  4  Gln-350  400  30  Glu-242  800  56  Glu-143  800  5  Ile-345  200  31  Gln-238  800  57  Leu-142  800  6  Ile-344  400  32  Asp-235  800  58  Ala-140  800  7 Lys results clearly indicated that the entire MexA polypeptide is located at the periplasm. DISCUSSION All the living organisms may be exposed to the external milieu, which may contain noxious compounds such as surfactants, hydrophobic materials, heavy metals, cytotoxic agents, and drugs. Most, if not all, of them must defend against the hazards of these noxious compounds by several means of detoxification. An efficient way to avoid suffering from such agents would be to lower the intracellular concentration of the noxious compounds by active extrusion. P. aeruginosa appeared to bear the efflux pumps, of which up-regulation renders this bacterium resistance to many antibiotics, dyes, surfactants, and organic solvents (5,6,(27)(28)(29). Since P. aeruginosa is a Gram-negative bacterium, the efflux pump assembly consists of the inner membrane efflux pump, the outer membrane exit channel, and the membrane fusion protein (3,4,8).
The MexA subunit of the MexA,B-OprM pump has been assumed to function as the membrane fusion protein, which connects the inner membrane pump, MexB, and the outer membrane exit channel, OprM (7,8,11). The MexA protein contains the lipoprotein consensus sequence at the amino terminus (9), yet the biological significance of the lipid modification has remained obscure. We addressed this issue by replacing cysteine 24 with phenylalanine or tyrosine. These MexA variants had both processed and unprocessed forms of MexA (Fig. 2a), as confirmed by analysis of the amino-terminal amino acid sequence. Strains expressing only the mutant MexA restored an antibiotic susceptibility indistinguishable from that of the wild-type strain. This result is consistent with the E. coli membrane fusion protein, AcrA, in that modification of the cysteine residue did not grossly affected the function (30). Our result demonstrated unequivocally that lipid modification was not essential for the function of MexA but did not answer the question of whether or not membrane anchoring is essential to the function of MexA.
We addressed this issue by constructing a recombinant MexA containing the signal peptide of azurin and found that the signal peptide was properly cleaved and that the mature MexA became a soluble protein (Fig. 3). Surprisingly, MexA protein without a membrane anchor fully functioned for antibiotic export. An analogous experiment has been carried out by subcloning the signal sequence-less acrA gene downstream from the OmpA signal sequence (30). Since E. coli cells harboring this recombinant gene expressed both processed and unprocessed forms of the AcrA proteins, these materials cannot be used to test whether the soluble form of AcrA is functional or not. The periplasmic localization of the entire MexA protein was confirmed by the expression of 78 MexA-␤-lactamase fusion proteins (Table III). This result is consistent with topology studies previously carried out for membrane fusion proteins: the ABC transporter in CzcB of A. eutrophus and AcrA, HlyD, and CvaA of E. coli (12)(13)(14)(15). However, the investigators constructed only 5, 10, 16, and 11 fusion protein, in CzcB, AcrA, HlyD, and CvaA, respectively (12)(13)(14)(15). This small number of constructs was not enough to account for the entire length of polypeptides with a molecular mass of about 50 kDa.
The function of MexA can be predicted from these results. Two essential functions are conceivable. (i) MexA functions as a hollow, connecting MexB and OprM, forming the antibiotic path. (ii) MexA interacts with MexB, and when MexB pumps out noxious compounds using the transmembrane proton gradient, MexA transmits this energy to OprM and opens the gate, which blocks the OprM channel. The latter possibility is analogous to the function of the TonB protein in ferric ion transport across the outer membrane (31).
We have analyzed the membrane topology of the MexB protein and reported our findings as follows (10). (i) The MexB protein spans the inner membrane 12 times, leaving both amino and carboxyl termini at the cytoplasmic side of the inner membrane. (ii) The 1st and the 4th periplasmic domains contain large polypeptides, consisting of 311 and 314 amino acid residues, respectively, and are extended largely to the periplasmic space. In contrast, the present study revealed that a large part, if not all, of the MexA is exposed to the periplasmic space. Based on these structural analyses of two inner membrane proteins of the MexA, B-OprM pump, we propose the following two topological models. It is conceivable that the MexA protein was held by the largely extended two periplasmic domains of the MexB protein, as illustrated in Fig. 5a. Based on this assumption, a tight combination of MexA and MexB is essential for interaction with OprM, energy transmission to OprM, and for outward delivery of noxious compounds. Alternatively, the MexA protein may be located over the large periplasmic loops of the MexB protein and may interact with OprM forming a hollow tube (Fig. 5b). The former model is supported by our previous result that tight coupling of MexA and MexB is essential for pump functioning (32). The latter model is consistent with the hypothesis that the MexA protein forms a noxious compound path tubing, through which the out-coming chemicals can be delivered to the outer membrane protein (8). In any case, either MexB or a combination of MexB and MexA should be able to transmit cellular energy to OprM in order to open the channel gate of OprM. Further investigations are needed to clarify the precise role of individual subunits in the proper function of the MexA, B-OprM pump. Furthermore, the present finding that MexA protein free from the membrane fully functions in the antibiotic export encourages further structural analysis of the MexA, B-OprM pump.
Acknowledgment-We are grateful to J. K. Broome-Smith at the University of Sussex, UK, for her kind gift of the plasmids pYZ4 and pKM1 and helpful discussion on the gene fusion technique.