INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Pseudomonas aeruginosa is an opportunistic human pathogen characterized by high-level intrinsic and mutationally acquired resistance to multiple antibiotics, for which multidrug efflux systems have emerged as key mechanisms. A number of multidrug efflux systems, including MexAB-OprM (12, 35, 36), MexCD-OprJ (34), MexEF-OprN (18), and MexXY-OprM (1, 27, 46), have been identified in this organism; two of these, MexAB-OprM and MexXY-OprM, contribute to intrinsic resistance in wild-type cells (1, 21, 46). Hyperexpression of MexAB-OprM occurs in nalB-type multidrug-resistant mutants, while overexpression of MexXY (AmrAB) correlates with acquired aminoglycoside resistance in this organism (1, 21, 46). Expression of MexCD-OprJ and MexEF-OprN is so far restricted to nfxB (34)- and nfxC (18)-type multidrug-resistant mutants, respectively. These tripartite efflux systems include an inner membrane, presumed drug-proton antiporter of the resistance-nodulation-cell division efflux family (39) (MexB, MexD, MexF, and MexY), a periplasmic but cytoplasmic membrane-anchored membrane fusion protein (15) (MexA, MexC, MexE, and MexX), and an outer membrane efflux protein (15) (OprM, OprJ, and OprN) (31). Apparently, antibiotics entering the cytoplasm or, perhaps, the cytoplasmic membrane are captured by the pump and extruded directly into the medium (31).

The OprM protein was first reported in 1992 to be associated with acquired multiple antibiotic resistance in P. aeruginosa (26). A component both of the MexAB-OprM and MexXY-OprM multidrug efflux systems, the protein plays a role in resistance to several antibiotics, including aminoglycosides (as part of MexXY-OprM) (1), tetracycline, chloramphenicol, quinolones, -lactams, novobiocin, macrolides, trimethoprim, and organic solvents (as part of MexAB-OprM) (19, 21-23, 26). Hybridization studies with an oprM probe have revealed that OprM is highly conserved in all serotypes of P. aeruginosa (3), highlighting its probable significance vis-à-vis intrinsic and acquired antibiotic resistance in this organism. Expected to function in the extrusion of drugs across the outer membrane, OprM has been implicated as a channel-forming protein. Recently, channel-forming activity has been demonstrated in planar lipid bilayer membranes, although the channels observed did not seem compatible with export of the diversity and size of compounds known to be substrates of the MexAB-OprM pump (47). Still, the probable involvement of TonB, a cytoplasmic membrane energy-coupling protein, in MexAB-OprM-mediated drug efflux and resistance (51) may facilitate channel opening. This protein is known, for example, to interact with outer membrane receptors for ferric siderophore complexes, promoting conformational changes necessary for ligand passage through the receptor channel (5, 28).

OprM homologues have been identified in a variety of gram-negative bacteria (33) and include OprJ and OprN of P. aeruginosa (18, 34), SrpC of the Pseudomonas putida SrpABC pump (17), OpcM of the Burkholderia cepacia CeoAB-OpcM pump (6), and SmeC of the Stenotrophomonas maltophilia SmeABC pump (GenBank accession number AF173226; X.-Z. Li, L. Zhang, and K. Poole, Abstr. 100th Gen. Meet. Am. Soc. Microbiol., abstr. A-33, 2000). Predicted to be lipoproteins, these all possess a characteristic lipoprotein box; indeed, acylation of OpcM has been demonstrated (6).

This study was undertaken to examine the significance of acylation for OprM function, as well as the functional importance of the highly conserved regions of the OprM family. The numerous deletions constructed as part of this process were then used to assess the accuracy of a topology model for OprM based on the recently solved crystal structure of TolC, an OprM homologue (20). This trimeric protein is comprised of monomers that span that membrane four times, producing a 12-membrane-spanning -barrel within the outer membrane with an extensive periplasmic -helical barrel (20).

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains, plasmids, and growth media. Bacterial strains and plasmids used in this study are listed in Table 1. Bacteria were routinely cultured in Luria-Bertani (LB) broth (1% [wt/vol] Difco tryptone, 0.5% [wt/vol] Difco yeast extract, 0.5% [wt/vol] NaCl) at 37°C with shaking (180 rpm), except for susceptibility testing, for which cultures were not shaken. In some instances, 2× TY broth (1.6% [wt/vol] Difco tryptone, 1% [wt/vol] Difco yeast extract, 0.5% [wt/vol] NaCl) was used as the growth medium. Escherichia coli strains carrying plasmids were cultivated in the presence of the appropriate antibiotic: ampicillin (100 µg/ml), chloramphenicol (30 µg/ml), kanamycin (50 µg/ml), or tetracycline (10 µg/ml). P. aeruginosa strains harboring pVLT31 and its derivatives were cultivated in the presence of tetracycline (10 µg/ml).

DNA methodology. Basic DNA procedures, including restriction endonuclease digestions, ligations, transformations, and agarose gel electrophoresis, were performed as described by Sambrook et al. (40). The alkaline lysis method (40) or a plasmid midi kit (Qiagen Inc., Missassauga, Ontario, Canada) was used to isolate plasmids from E. coli DH5. The genomic DNA of P. aeruginosa was extracted by the method of Barcak et al. (2). DNA fragments used in cloning were extracted from agarose gels using Prep-A-Gene (Bio-Rad Laboratories, Richmond, Calif.) according to the manufacturer's instructions. E. coli cells were made competent by the CaCl₂ method (40) or, when supercompetent cells of E. coli were required, the method of Inoue et al. (14). Oligonucleotides used in this study (Table 2) were designed and assessed using a Primer Design software package (version 2.01; Scientific & Educational Software, Durham, N.C.) to minimize secondary structure and dimer formation. All oligonucleotides were chemically synthesized by Cortec DNA Services Inc., Queen's University, with the exception of oprm68xz and oprm69xz, which were synthesized by MWG Biotech Inc. (High Point, N.C.). Nucleotide sequencing was carried out using universal or custom primers by Cortec DNA Services (Queen's University) or the Laboratory Services Division, University of Guelph, Guelph, Ontario, Canada). Nucleotide sequence was analyzed using the PCGene software package (Intelligenetics Inc., Mountain View, Calif.) and DNAMAN (Lynnon Biosoftware, Vaudreuil, Quebec, Canada).

Construction of the oprM deletion mutants. To delete oprM in P. aeruginosa ML5087 and its derivatives, the gene replacement suicide vector pK18mobsacB carrying an oprM deletion (pXZL04) was constructed as previously described (44). Following transformation of E. coli S17-1, pXZL04 was mobilized into P. aeruginosa ML5087 via conjugation (34). Briefly, E. coli S17-1 carrying pXZL04 was grown overnight in LB broth in the presence of kanamycin with shaking at 37°C, and the P. aeruginosa strains were grown in LB broth overnight without shaking at 42°C. Overnight donor and recipient cultures (100 µl of each) were mixed, pelleted by centrifugation (3 min in a microcentrifuge), and resuspended in 50 µl of LB broth. The cell suspensions were spotted on LB agar and incubated overnight at 37°C. The bacterial cells on LB agar were resuspended with 1 ml of LB broth, and dilutions were plated onto agar plates supplemented with kanamycin and tetracycline (to counterselect E. coli). Transconjugants carrying a copy of pXZL04 in the chromosome were recovered from kanamycin-tetracycline-containing LB agar plates and streaked onto LB agar containing 10% (wt/vol) sucrose. Sucrose-resistant colonies, which had lost pK18mobsacB sequences (kanamycin sensitive) and carried either an unaltered wild-type copy of oprM or the oprM deletion, were thus recovered. Potential oprM deletion mutants were initially screened for antibiotic susceptibility, a phenotype typical of OprM-deficient P. aeruginosa. The presence of the oprM deletion was then confirmed using PCR as described previously (52) with genomic DNA and primers oprmp1 and oprm21xz (Table 2). The absence of OprM in these mutants was also confirmed by Western immunoblotting with an antiserum specific to OprM (see below).

Cloning of the native oprM gene. The wild-type oprM gene (GenBank accession number L23839) of ca. 1.5 kb was previously cloned into pT7-7 and pVLT31, yielding pKPM-1 and pKPM-2, respectively (48). Examination of the genomic oprM sequence of P. aeruginosa (Pseudomonas genome project [http://www.pseudomonas.com]) revealed that the cloned oprM gene of pKPM-1 and pKPM-2 differed from the chromosomal counterpart at the 3' end, with the last 66 bp of native oprM (corresponds to 22 residues in OprM) replaced by 42 bp (14 residues in OprM) from the phagemid originally used to clone the gene. Still, the OprM protein encoded by pKPM-2 (48) was fully functionally and restored drug resistance to OprM-deficient strains (44, 47). To obtain a clone of the wild-type oprM gene, PCR was used to amplify 800 bp at the 3' end of the native oprM gene from the chromosome of P. aeruginosa PAO1 (or ML5087), which was then used to replace the corresponding segment of oprM on pKPM-1. Amplification was carried out using primers oprm14xz (anneals upstream of the BamHI site in oprM) and oprm21xz (contains a HindIII site) (Table 2) in a reaction mixture containing 50 ng of chromosomal DNA, 40 pmol of each primer, 0.2 mM each deoxynucleoside triphosphate, 2 mM MgSO₄, 10% (vol/vol) dimethyl sulfoxide, and 2 U of Vent DNA polymerase in 1× Thermo reaction buffer (New England Biolabs; NEB). The mixture was heated for 2 min at 94°C and then amplified by 30 cycles of 1 min at 94°C, 1 min at 56°C, and 2 min at 72°C. The PCR products were purified using a Qiaquick PCR purification kit (Qiagen) according to the manufacturer's instructions. After digestion with BamHI and HindIII, the PCR product was cloned into BamHI-HindIII-restricted pKPM-1, which removed the incorrect 3' sequence of oprM, yielding plasmid pXZL33. Following confirmation of the correct nucleotide sequence and expression of OprM from pXZL33, an oprM-containing XbaI-HindIII fragment of pXZL33 was cloned into XbaI-HindIII-restricted-pVLT31 and -pTZ19U, yielding pXZL34 and pXZL40, respectively.

PCR-based site-directed mutagenesis of oprM. A number of deletion, insertion, and point mutations were introduced into oprM by using an approach based on PCR. To construct the A12-R98 deletion derivative, two PCRs were carried out, resulting in the amplification of sequences upstream (using primers oprm1xz and oprm68xz) and downstream (using primers oprm69xz and oprm21xz) of the deletion endpoints. The PCR mixtures were formulated as above except that 10 ng of pXZL33 was used as the template. The mixtures were heated for 2 min at 94°C and then subjected to 30 cycles of 1 min at 94°C, 1 min at 52°C, and 0.5 min (for the 100-bp upstream product) or 1.2 min (for the 1.2-kb downstream product) at 72°C. Following purification and digestion with XbaI and SstI (100-bp product) or SstI and HindIII (1.2-kb product), the PCR products were cloned into XbaI-HindIII-restricted pTZ19U via a three-piece ligation procedure to yield pXZL134. The deletion was confirmed by nucleotide sequencing, liberated on a ca. 1.3-kb XbaI-HindIII fragment, and cloned into XbaI-HindIII-restricted pVLT31 to yield pXZL137. The C1G OprM mutation was also constructed by amplifying oprM in two parts. The region upstream of the mutation site was amplified with primers oprm1xz and oprm6xz (Table 2), and that downstream was amplified with primers oprm7xz (a mutagenic primer that changes the TGC cysteine codon to the GGC glycine codon) and oprm2xz (Table 2), using reaction conditions described above. The 140-bp upstream fragment was digested with XbaI and PstI and cloned into XbaI-PstIII-restricted pRK415. This construct was then digested with PstI-HindIII to permit cloning of the PstI-HindIII-digested 1.4-kb downstream product. The resultant vector, which carries an intact oprM gene with a PstI site near the glycine codon, was digested with PstI and treated with T4 DNA polymerase (NEB) at 12°C for 20 min (40). Vector recircularization with T4 DNA ligase yielded pXZL11, which carried the desired C1G oprM mutation. The C1G mutant oprM gene of pXZL11 was then recovered on a 1.5-kb XbaI-HindIII fragment and cloned into XbaI-HindIII-digested pT7-7 to produce pXZL13. Since the oprM gene used here originated from pKPM-1, in which the 3' sequence of oprM differed from that of the native oprM (above), the C1G oprM-containing XbaI-SalI fragment (ca. 300 bp at the 5' end of oprM) of pXZL13 was used to replace the XbaI-SalI fragment of the native oprM gene of pXZL40, resulting in pXZL141. The C1G mutated oprM gene of pXZL141 was then liberated on an XbaI-HindIII fragment and cloned into XbaI-HindIII-restricted pVLT31 to yield pXZL146.

Phagemid-based site-directed mutagenesis of oprM. Several deletion and most substitution mutations were constructed using a Muta-Gene phagemid in vitro mutagenesis kit (Bio-Rad) according to the manufacturer's instructions, with some modifications. To obtain a single-stranded copy of oprM as required for the mutagenesis protocol, E. coli CJ236 cells harboring the oprM-containing phagemid pXZL40 (pTZ19U::oprM) were cultured in chloramphenicol- and ampicillin-supplemented 2× TY broth to an optical density at 600 nm of 0.3. The helper phage M13K07 was then added at a multiplicity of infection of ca. 20; the cells were incubated for an additional 1 h, at which time kanamycin (70 µg/ml) was added. Following incubation overnight, bacteria were removed by centrifugation (17,000 × g for 30 min at 4°C), and the supernatant was recovered and recentrifuged. DNase-free RNase A (Sigma) was added to the supernatant at a final concentration of 10 µg/ml, and the supernatant was incubated at room temperature for 30 min. Then 15 ml of 3.5 M ammonium acetate-20% (wt/vol) polyethylene glycol 8000 was added to 45 ml of the phage-containing supernatant, and the mixture was incubated on ice for 30 min. Phage particles containing single-stranded pXZL40 were recovered by centrifugation (17,000 × g for 15 min at 4°C) and resuspended in high-salt buffer (300 mM NaCl, 100 mM Tris-HCl [pH 8.0], 1 mM Na₂EDTA). Single-stranded phagemid DNA was prepared from the phage particles as described elsewhere (40) and dissolved in 10 mM Tris-HCl-1 mM EDTA (pH 8.0).

OprM expression in P. aeruginosa. To assess expression of mutated oprM genes in P. aeruginosa, the oprM-containing pVLT31 derivatives were mobilized from E. coli DH5 into OprM-deficient P. aeruginosa strains K1110 and K1113 via triparental mating (44). The P. aeruginosa transconjugants were selected on LB agar containing either tetracycline (10 µg/ml) and chloramphenicol (15 to 25 µg/ml) (for K1110 or K1113 derivatives) or tetracycline (10 µg/ml) and imipenem (0.5 µg/ml) (for K1113 derivatives only). For the preparation of whole cell extracts, cells were grown overnight at 37°C in LB broth in the presence of the appropriate antibiotics, diluted 1:20 in 5 ml of LB broth, and incubated for 4 to 5 h with shaking. One milliliter of cell culture was then harvested by centrifugation (15,000 × g for 3 min), and the cell pellet was resuspended in 100 µl of 50 mM sodium (or potassium) phosphate buffer (pH 7.2). Inner and outer membranes were prepared from cell envelopes following their separation on sucrose density gradients or by differential sarkosyl solubilization as described (30, 50). Expression of OprM derivatives in E. coli and P. aeruginosa was analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (25) and Western immunoblotting using a polyclonal anti-OprM antiserum (23, 52).

Antibiotic susceptibility assay. The susceptibility of P. aeruginosa strains harboring mutant oprM plasmids (see above) to various antimicrobial agents was tested by a twofold serial broth dilution method, with an inoculum of 5 × 10⁵ bacteria/ml. The MIC was reported as the lowest antimicrobial concentration inhibiting visible growth after 18 to 20 h of incubation at 37°C.

Protein secondary structure prediction and sequence analysis. Prediction of OprM and TolC secondary structure involved the use of the Jpred² server (http://jura.ebi.ac.uk:8888/) (7, 8). Alignment of the OprM and TolC amino acid sequences was carried out using the PCGene (Intelligenetics) PALIGN program and the DNAMAN sequence analysis software (Lynnon Biosoft). Multiple alignment of the OprM homologues was carried out using DNAMAN.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Operation of mutant OprM proteins in P. aeruginosa. Alignment of OprM with its homologues in P. aeruginosa, P. putida, B. cepacia, and S. maltophilia identified a number of regions with a high degree of sequence conservation (Fig. 1). To assess their importance, if any, for OprM function, we deleted several of these conserved regions and assessed the effect on activity in the OprM-deficient strains K1110 and K1113. The latter strain is a nalB derivative that hyperexpresses MexA and MexB and thus provides for a greater enhancement of antibiotic resistance (compared to K1110) in the presence of functional OprM. One of the more striking examples of sequence conservation involved an LGGGW sequence near the extreme C terminus of OprM and its homologues (Fig. 1). C-terminal truncations which removed this and all downstream OprM sequence (pXZL66) abrogated OprM expression (Fig. 2, lane 19), as did an in-frame deletion of the LGGGW (residues 446 to 450) sequence only (pXZL76; 13 in Fig. 3) (Fig. 2, lane 18). In contrast, deletion of sequence downstream of LGGGW (N451 to A468; pXZL67; 14 in Fig. 3) yielded a protein (Fig. 2, lane 20) which restored the antibiotic resistance of P. aeruginosa strains K1110 (Table 3) and K1113 (Table 4). Still, mutations in individual residues within this region, including L446 (Fig. 2, lanes 33 and 34), G448 (Fig. 2, lanes 35 and 36), and W450 (Fig. 2, lane 37) failed to affect OprM expression or activity (Tables 3 and 4).

View larger version (151K): [in this window] [in a new window]   FIG. 1.   Multiple alignment of OprM and its homologues in Pseudomonas spp. and related organisms. Residues conserved in all proteins are highlighted in black, while those conserved in 75 and 50% of the indicated proteins are highlighted in dark and light gray, respectively. Proteins examined include SrpC of the SrpABC efflux system of P. putida (accession number AF029405), SmeC of the SmeABC efflux system of S. maltophilia (accession number AF173226), OprJ of the MexCD-OprJ efflux system of P. aeruginosa (accession number U57969), OpcM of the CeoAB-OprcM efflux system of B. cepacia (accession number U38944), and OprN of the MexEF-OprN efflux system of P. aeruginosa (accession number X99514). The lipoprotein box is overlined and occurs downstream of putative lipoprotein signal sequence. Numbers at the right show the position of the last residue in each line within the mature protein sequences, where the Cys residue within the lipoprotein box (indicated by an arrow) is the first amino acid of the mature proteins. Alignment was carried out using the DNAMAN software package (Lynnon Biosoft).

View larger version (53K): [in this window] [in a new window]   FIG. 2.   Immunodetection of OprM proteins expressed by oprM P. aeruginosa strain K1110 (A) and nalB oprM P. aeruginosa strain K1113 (B) harboring the indicated plasmids. Whole cell extracts were separated by SDS-PAGE, electroblotted, and developed with antibodies to OprM. The OprM derivative encoded by each of the plasmids is indicated in parentheses.

View larger version (37K): [in this window] [in a new window]   FIG. 3.   Alignment of OprM and TolC, highlighting the structural features of TolC and putative membrane-spanning regions of OprM. Exact matches (|) and conserved changes (.) are indicated. Regions of TolC existing as -sheet () and -helix () are highlighted below the TolC sequence (derived from and enumerated as described in reference 20), and the membrane-spanning -sheets are highlighted in bold. Regions of both TolC and OprM predicted to be -sheet by secondary structure prediction programs on the Jpred2 server are underlined. OprM deletions constructed and tested in this study are highlighted above the OprM sequence, with shaded boxes representing deletions which yielded a functional protein, filled boxes representing deletions which yielded a nonfunctional protein, and open boxes representing deletions which failed to yield an OprM protein. The endpoints of the large but partially functional N-terminal deletion (A12-R98; pXZL137) are indicated with vertical arrows directed upward, and a nonfunctional C-terminal truncation (L446-A468; pXZL66) is indicated with a vertical arrow directed downward. The site of insertion of a nine-amino-acid HA tag is also highlighted. Numbers at the right represent the position of the last amino acid residue in each line within the mature OprM sequence, where the acylated Cys residue is residue 1. Numbering of TolC residues is relative to the precursor form of the protein.

View this table: [in this window] [in a new window]   TABLE 3.   Antibiotic resistance of oprM P. aeruginosa K1110 expressing mutant OprM proteinsa

View this table: [in this window] [in a new window]   TABLE 4.   Antibiotic resistance of nalB oprM P. aeruginosa K1113 expressing mutant OprM proteinsa

Influence of N-terminal acylation on OprM activity. OprM and its homologues possess a region at the N terminus, termed the lipoprotein box, immediately downstream of an N-terminal signal sequence (Fig. 1). A conserved Cys residue within this box is the first amino acid of the mature proteins and the deduced site of acylation for this family of presumed lipoproteins; indeed, acylation of OprM has been demonstrated (N. Bianco and K. Poole, unpublished data; T. Nakae, A. Nakajima, Y. Sugimoto, and H. Yoneyama, Abstr. 100th Gen. Meet. Am. Soc. Microbiol., abstr. A-28, 2000). To assess the importance of acylation for OprM expression and function, we mutated the Cys residue (to Gly) and examined expression and activity of the C1G mutant OprM (pXZL146) in K1110 and K1113. P. aeruginosa harboring pXZL146 produced a readily detectable OprM protein (Fig. 2, lane 21), and this plasmid enhanced the antibiotic resistance of K1110 and K1113 strains carrying it. To further assess the significance of OprM acylation, the lipoprotein signal (including lipoprotein box) of OprM was replaced by the type I signal sequence of OprF, the major outer membrane protein of P. aeruginosa(pXZL147). Again, the OprM derivative carrying the OprF signal was well expressed (Fig. 2, lane 22), particularly in K1113 (Fig. 2B, lane 22), and highly active in both K1110 (Table 3) and K1113 (Table 4). Cell fractionation studies also confirmed that the C1G and OprF-OprM proteins were present in the outer membrane (data not shown). Thus, loss of the acylation site of OprM did not obviate OprM activity.

Topological model of OprM. The recently proposed topological model that describes OprM as a -barrel of 16 transmembrane domains (47) is likely incorrect, in light of the recently solved crystal structure of TolC (20), an OprM homologue (ca. 20% identical [Fig. 3]). Accordingly, OprM likely spans the outer membrane only four times, with both the N and C termini occurring within the periplasm, and displays a substantial periplasmic component. To assess the accuracy of such a model for OprM, the Jpred² secondary structure prediction server was used to assess the secondary structure of this protein. Initially, however, the server was applied to TolC, to determine how closely the predicted secondary structure matched the crystal structure. Intriguingly, the various prediction programs of Jpred² correctly identified TolC as a protein that was substantially -helical (data not shown) and accurately identified three of the four membrane-spanning -sheets (S2, S4, and S5 [Fig. 3]). The first membrane-spanning -sheet (S1) was only weakly predicted (Fig. 3). Similarly, Jpred² identified OprM as a predominantly -helical protein (data not shown) with three putative membrane-spanning -sheets located roughly where they would be predicted following alignment of TolC and OprM (opposite S2, S4 and S5 [Fig. 3]). Alignment of the N- and C-terminal halves of OprM, which confirmed the homology of these two halves (23.7% identity and 9.8% conserved changes) and the likely gene duplication involved, also confirmed that the putative membrane-spanning regions S2 and S5 were equivalently placed within the N- and C-terminal halves of the protein (Fig. 4). Moreover, the region within the N-terminal portion of OprM which aligns with S4 of the C-terminal half is bracketed by sequences whose deletion (3 and 4 [Fig. 3]) is tolerated (see below) and thus unlikely to correspond to membrane-spanning regions. As this region also overlaps sequences of OprM that align with S1 of TolC (Fig. 3), it is the only reasonable candidate for the first membrane-spanning -sheet of OprM. In light of the internal homology, the equivalent placement of the putative membrane-spanning regions within the N- and C-terminal halves of OprM supports the existence of these regions as membrane-spanning domains in OprM (Fig. 4) and support OprM adopting a structure reminiscent of TolC.

View larger version (34K): [in this window] [in a new window]   FIG. 4.   Internal alignment of OprM, highlighting the homology between the N-terminal (OPRMN) and C-terminal (OPRMC) halves of the protein. Exact matches (|) and conserved changes (.) are indicated. Putative membrane-spanning -sheet regions are underlined in bold and enumerated as for TolC. Numbers at the right represent the position of the last amino acid residue in each line within the mature OprM sequence, where the acylated Cys residue is residue 1.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The OprM family of outer membrane multidrug efflux proteins shares a number of regions of sequence conservation, several of which were targeted here for mutagenesis. While the expectation was that these would be important for function, only two of nine permissive deletions (G199-A209 and A278-N286) impaired OprM function, and none of the highly conserved residues that were mutated in these regions was essential for activity. Still the very highly conserved LGGGW sequence near the extreme C terminus of OprM appears to be especially critical for expression and, thus activity of this protein. Consistent with these results, deletion of the final 70 amino acids of OprM was previously shown to obviate expression of the protein (47). Intriguingly, the originally reported oprM gene possessed an anomalous sequence at its 3' end provided by the phagemid vector into which oprM had been cloned, although the resultant OprM protein was fully functional. This protein lacked the final 22 amino acids of native OprM (replaced with 14 phagemid-derived amino acids), confirming results presented here that the extreme C terminus of OprM is dispensable for activity. Perhaps importantly, however, the nature of the LGGGW sequence was retained, being changed to LWGG in this altered oprM gene. The conservation of a tryptophan residue in this region may be of some importance. A conservative substitution (W450Y) failed to alter expression or activity of OprM, while additional substitutions were unsuccessful.

Lipoproteins typically possess a signature lipoprotein box near the N terminus, immediately following a signal sequence (37). An absolutely conserved cysteine residue within this box in OprM and its homologues, though the site of acylation of these lipoproteins, was dispensable, suggesting that acylation is not a prerequisite for function. Similarly, MexA, though also a lipoprotein, need not be acylated to be functionally active (49). The role of these lipid tails, then, in the assembly or activity of the MexAB-OprM efflux pump is unclear.

In light of the recently solved crystal structure of TolC, which describes the protein as a trimeric "channel-tunnel" that spans both the outer membrane (as a -barrel) and the periplasm (as a -helical barrel), it is reasonable to assume that OprM, a TolC homologue, adopts much the same conformation. Such a structure contrasts with the recently published model for OprM, which describes it a monomeric -barrel comprised of 16 membrane-spanning segments (47). Consistent with the TolC model and in contrast to the monomeric -barrel model, however, recent analysis of the OprM sequence using a neural network approach predicted that a large portion of the protein would be periplasmic (11; A. Ferguson, personal communication). Using a variety of secondary structure prediction tools, the predominantly -helical nature of OprM (and TolC) was confirmed, as was the placement of putative -sheet regions that could form membrane-spanning domains. These agreed quite nicely with predictions based on alignment of the TolC and OprM sequences (Fig. 3). Interestingly, too, OprM, like TolC (20), displays internal homology, further supporting OprM adopting a TolC-like conformation. Using the TolC model, then, it was possible to assess the implications of the various deletions constructed in this study and to use this information to validate the TolC model for OprM. Surprisingly, the highly conserved LGGGW encompassed by 13 does not correspond to any significant structural feature in TolC, occurring as it does in a region of OprM that aligns with TolC between -helices H8 and H9. Although one cannot be sure about the exact placement of these helices in OprM, much of the region encompassed by 13 is conserved between OprM and TolC, suggesting that it does not lie within OprM helix H8 or H9. While the reason for the nonpermissive nature of this deletion is unknown, the fact that deletion of downstream residues does not affect expression (or function) of OprM indicates that an H9 is unlikely to be important for OprM function or assembly. As such, the lack of expression of a 13 OprM derivative cannot be due to an effect on topology of this downstream helix.

The region from G199 to A209 of OprM (encompassed by 7 in Fig. 3) corresponds to the beginning of H4 in TolC, although the proximity of this region to H3 and the possibility of some variation regarding the exact positioning of comparable helices in OprM suggests that either of these -helices may be affected in OprM G199-A209. The fact, however, that other deletions in the putative H3 -helix (e.g., 5 and 6 in Fig. 3) are nonpermissive suggests that 7 may well affect H4. Why this would abrogate function is unclear, although the placement of this region of TolC at the extreme periplasmic end of the protein (facing the cytoplasmic membrane) suggests it may, in OprM, be important for interaction with MexA or MexB. Alternatively, since the N-terminal end of H4 contacts H7 in forming the periplasmic -helical barrel of TolC (20), 7 could disrupt this contact in OprM, perhaps affecting proper barrel formation or assembly.

10, encompassing A278 to N286 of OprM, corresponds to the C terminus of H6 in TolC (Fig. 3) which, together with the downstream -helix H8, forms a pseudocontinuous helix that spans the length of the -helical barrel of TolC (20). Intriguingly, the region encompassed by this deletion corresponds to a region in the TolC -helix H6 that contacts -helix H7 (20). Quite possibly, then, this deletion disrupts such an interaction in OprM, again compromising proper barrel formation.

With the exception of 11 (F317 to F326), which likely disrupts a membrane-spanning domain of OprM and thus is expected to be nonpermissive, the nonpermissive deletions identified in this study do not correspond to expected membrane-spanning regions of the protein. Interestingly, however, they all target the long helices, dubbed H3 and H7 in TolC. Given their importance in formation of the periplasmic -helical barrel of TolC (they extend the length of the barrel), their truncation might prevent assembly of the periplasmic barrel domain, yielding an improperly folded and highly unstable protein. Moreover, in contrast to the permissive deletions in H4 and H6, which likely affect only one set of contacts, truncations in H3 and H7 would, by virtue of their shortening these helices, disrupt the positioning of residues downstream of the deletion, perhaps impacting on multiple contacts with other helices. This, too, might render the protein unstable.

The ability to delete the first ca. 100 amino acids of OprM and retain some function, though surprising, was reminiscent of the ferrichrome receptor FhuA (4). It is now known that the N terminus of this and other ferric siderophore receptors encompasses a TonB-responsive cork domain that blocks the receptor channel from the periplasmic side of the protein (10, 24). As MexAB-OprM is apparently dependent on TonB (51), it may be that the N terminus of OprM similarly functions as a TonB-responsive cork. Given the unique structure predicted for OprM, however, it likely looks and functions somewhat differently than it does in FhuA. This would explain both the lack of a TonB box (the site on traditional TonB-dependent outer membrane receptors [e.g., FhuA] that interacts with TonB) in OprM and its homologues and the differential effect of TonB mutations on drug efflux versus iron transport activity in P. aeruginosa (Q. Zhao and K. Poole, submitted for publication), since TonB would have to interact differently with OprM than with a ferric siderophore receptor. It is also interesting that OprM possesses an N-terminal extension not shared by TolC (Fig. 3) and TolC appears not to be TonB dependent (at least antibiotic resistance, and therefore AcrAB-TolC-mediated resistance, is not compromised in an E. coli tonB mutant [Q. Zhao and K. Poole, unpublished data]). Whether the additional N-terminal residues are associated with a TonB dependence of MexAB-OprM activity or simply relate to the acylation and membrane anchoring of the N terminus of OprM (TolC is not a lipoprotein) is, however, uncertain. The results of our study are in consonance with those of Hancock's group (46a).