Implication of Porins in β-Lactam Resistance of Providencia stuartii*

An integrative approach combining biophysical and microbiological methods was used to characterize the antibiotic translocation through the outer membrane of Providencia stuartii. Two novel members of the General Bacterial Porin family of Enterobacteriaceae, named OmpPst1 and OmpPst2, were identified in P. stuartii. In the presence of ertapenem (ERT), cefepime (FEP), and cefoxitin (FOX) in growth media, several resistant derivatives of P. stuartii ATCC 29914 showed OmpPst1-deficiency. These porin-deficient strains showed significant decrease of susceptibility to β-lactam antibiotics. OmpPst1 and OmpPst2 were purified to homogeneity and reconstituted into planar lipid bilayers to study their biophysical characteristics and their interactions with β-lactam molecules. Determination of β-lactam translocation through OmpPst1 and OmpPst2 indicated that the strength of interaction decreased in the order of ertapenem ≫ cefepime > cefoxitin. Moreover, the translocation of these antibiotics through OmpPst1 was more efficient than through OmpPst2. Heterologous expression of OmpPst1 in the porin-deficient E. coli strain BL21(DE3)omp8 was associated with a higher antibiotic susceptibility of the E. coli cells to β-lactams compared with expression of OmpPst2. All our data enlighten the involvement of porins in the resistance of P. stuartii to β-lactam antibiotics.

Providencia stuartii is an opportunistic pathogen involved in community-acquired as well as hospital-acquired infectious diseases. Clinical strains of P. stuartii are mostly isolated from urinary tract infections of patients with long-term indwelling urinary catheters and, in fewer cases, from respiratory and skin infections (1,2). P. stuartii is reported as one of the most resistant species in the family of Enterobacteriaceae (3). P. stuartii strains show high levels of resistance to the majority of antibiotic classes but were found to remain susceptible to carbapenems (3,4). P. stuartii produces a chromosomally encoded cephalosporinase, AmpC, which causes the natural resistance to aminopenicillins and narrow-spectrum cephalosporins (5). The productions of different extended-spectrum ␤-lactamases (6,7,8,9,10,11) and metallo-␤-lactamases (12,13,14) have been reported in association with resistance to carbapenems in Providencia spp. Other enzymatic mechanisms of antibiotic resistance identified in P. stuartii include acetyl aminotransferases targeting aminoglycoside antibiotics and an integronencoded erythromycin esterase involved in the resistance to macrolides (15,16). Moreover, mutations of the gyrA gene leading to a modification of target site were described in resistance to fluoroquinolones (17). In contrast, little is known about the involvement of membrane proteins in antibiotic resistance of P. stuartii. One study suggested a role of outer membrane porins in antibiotic permeability for Proteus, Morganella, and Providencia strains by selection of mutants resistant to cefoxitin, a cephalosporin of the second generation (18). However, the authors suggested that the three genera might have a common characteristic of producing a single porin in the outer membrane.
The membrane permeability of P. stuartii toward antibiotics is of our interest and the two major outer membrane porins, named OmpPst1 and OmpPst2, were characterized in this study. The correlation between porin expression and ␤-lactam permeation was studied. An approach combining microbiology and electrophysiology was used to analyze the selectivity of the Providencia porins toward different ␤-lactams and to decipher the molecular basis of the antibiotic flux through the porin channels. Finally, we rationalized our results by comparing the sequences and the homology modeled structures of the Providencia porins to both Escherichia coli OmpC and OmpF.

EXPERIMENTAL PROCEDURES
Bacterial Strain and Growth Conditions-The type strain P. stuartii ATCC 29914 was received from Pasteur Institute, Paris. The identity of the strain was confirmed by API 20E test (bioMérieux, Marcy l'Etoile, France) and 16 S RNA sequencing. For routine cloning, E. coli strain DH5␣ was used (19). For expression of porins, E. coli strain BL21(DE3)omp8 (⌬lamB ompF::Tn5 ⌬ompA ⌬ompC) was used (20). Bacteria were grown at 37°C in either Luria-Bertani (LB) or Mueller-Hinton (MH) medium (Difco Laboratories, Detroit, MI). Antibiotics were added to the media at final concentrations of 100 g/ml for ampicillin and 25 g/ml for kanamycin when needed.
Antibiotic Susceptibility Tests-The MIC values were determined in triplicate by a standard 2-fold broth dilution method using MH broth according to the CLSI guidelines. The results were scored after 18 h of incubation at 37°C and classified according to the Antibiogram Committee of the French Society for Microbiology. The antibiotics tested comprised different chemical classes including ␤-lactams (imipenem, ertapenem, cefepime, cefpirome, ceftazidime, cefoxitin); phenicol (chloramphenicol) and fluoroquinolone (sparfloxacin).
Extended Spectrum ␤-Lactamases (ESBL) and Metallo-␤lactamases (MBL) Tests-All P. stuartii strains were controlled with extended spectrum ␤-lactamases (ESBL) and MBL tests. The detection of ESBLs was performed using the standard double-disk synergy test on Mueller-Hinton agar as described before (4). Disks containing cefepime (30 g), ceftazidime (30 g), cefotaxime (30 g), and aztreonam (30 g) were placed around a disk containing 30 g of augmentin (20 g of amoxicillin and 10 g of clavulanate). The increase of the inhibition diameter toward augmentin disk was considered to be ESBLpositive. The strains were also tested for the presence of MBL enzymes by the double-disk synergy test (DDST) and the combination disc test (CDT) with imipenem and EDTA as described before (22). The positive control of MBL presence was a clinical Enterobacter cloacae 8 -1072 strain harboring the VIM metallo-␤-lactamase which was available in our laboratory. A synergistic inhibition area between imipenem and EDTA disks together with an increase of inhibition ring diameter of combinatory disk were supposed to be MBL-positive.
Preparation of Outer Membrane Proteins-The outer membrane proteins of P. stuarii strains were prepared from exponential-growth-phase cultures grown in 100 ml of LB broth. The cell pellets were treated by ultrasound and intact cells were removed by centrifugation at 10,000 rpm. The membrane fraction was separated by ultracentrifugation at 100,000 ϫ g. The membrane fraction was treated with 0.15% sodium N-lauryl sarcosine to solubilize the cytoplasmic membrane (23). The outer membrane fraction was obtained after a second ultracentrifugation.
SDS-PAGE and Western Blotting-The proteins were analyzed on SDS-polyacrylamide gels (12% acrylamide, 0.1% SDS) using Laemmli buffer. Separate gels were processed to electrotransfer the proteins onto nitrocellulose membrane in the presence of 0.05% SDS. The immunodetection with polyclonal antibodies directed against OmpF and OmpC was carried out as described previously (24).
Isolation of Providencia Porins-The extraction and purification of OmpPst1 and OmpPst2 porins were carried out as previously described with minor modifications (25). Briefly, cultures of E. coli BL21(DE3)omp8 harboring pGOmpPst1 or pGOmpPst2 were grown in LB broth containing ampicillin (100 g/ml) and kanamycin (25 g/ml). Expression of the porins was induced during the exponential growth phase by addition of IPTG to a final concentration of 0.4 mM. Cells were harvested 6 h after induction. The cells were disrupted using an EmulsiFlex-C3 high-pressure homogenizer (Avestin Europe, Mannheim, Germany). SDS was added to the cell suspension to a final concentration of 2% (v/v), followed by an incubation with gentle stirring at 60°C for 1 h. Bacterial envelopes were collected by centrifugation at 22,000 rpm for 1 h. The pellet was pre-extracted with 20 mM phosphate buffer (pH 7.4) containing 0.125% octyl-polyoxyethylene (octyl-POE, Alexis, Läuflingen, Switzerland) to remove proteins from the envelopes without solubilizing the porin. The extract was centrifuged again and the pellet was resuspended in 20 mM phosphate buffer (pH 7.4) containing 3% octyl-POE to solubilize the porin. The suspension was incubated at 37°C for 1 h with rigorous shaking at 250 rpm. The insoluble materials were removed by centrifugation (40,000 rpm, 40 min) and the porin-containing preparations were concentrated using Amicon Ultra-15 centrifugal filter devices with a molecular weight cutoff from 30 kDa (Millipore, Schwalbach, Germany). Final porin dilutions for bilayer measurements were prepared using 20 mM phosphate buffer, pH 7.4, containing 1% octyl-POE.
Determination of N-terminal Amino Acid Sequences and Mass Spectrometry-The protein sequencing service was done at IBSM-CNRS, Plate-forme Protéomique, Marseille, France. Preparations of Providencia porins were resolved by SDS-PAGE and electrotransferred with Tris-borate buffer onto a PVDF Immobilon TM -PSQ membrane (Millipore, St. Quentin en Yvelines, France). After Ponceau staining, the bands corresponding to the porins were excised. The N-terminal amino acid sequences were determined by Edman degradation of 5 cycles using Procise 494 sequencer. To investigate which porin disappeared under the antibiotic pressure, the respective protein band of the parental strains at about 39 kDa was analyzed by mass spectrometry. The outer membrane fractions of the parental ATCC strain and the resistant derivatives were solved on SDS-PAGE and stained with Coomassie Blue. The porin bands were excised and treated with trypsin for 8 h. Mass spectrometer was a Microflex II (Brucker, Bremen, Germany).
Single Channel Conductance Measurements-Virtually solvent-free planar lipid membranes were formed using diphytanoylphosphatidylcholine (DPhPC) (Avanti Polar Lipids, Alabaster, AL) according to the Montal-Mueller technique (26). The measurements were carried out with buffer containing 1 M KCl, 20 mM MES, pH 6.0. OmpPst1 or OmpPst2 porins from a 5 ng/ml solution in 150 mM KCl containing 1% octyl-POE were added to the cis side compartment (contacted by the ground electrode). Incorporation was achieved by stirring after addition and applying a 50 -200 mV transmembrane voltage. Ertapenem, cefepime, and cefoxitin were added at concentrations of 1 mM to 10 mM to investigate their permeation rates through the porin channels. Electrical recordings were made through a pair of Ag/AgCl electrodes (World Precision Instruments, Berlin, Germany), attached to an Axon Instruments 200B amplifier with a capacitive head stage, digitized by an Axon Digidata 1440A digitizer, computer controlled by Clampex 10.0 software (all by Axon Instruments, Foster City, CA). The Clampfit software (Axon Instruments) was utilized to analyze the recording data.
Porin Homology Modeling-The MODELLER suite (27) was used to build structural models of OmpPst1 and OmpPst2. The modeling was based on the structures of OmpF and OmpC (28,29) used individually or together (multiple template method) as templates. The alignments between the target sequences and the structure(s) of OmpF and OmpC were obtained using SALIGN (27). The resulting sequence identities are reported in supplemental Table S1. For each alignment, ten structural models were generated by the MODELLER program, from which the "representative model" was defined as the one that minimized both the overall "Modeler objective function" and the Dope score evaluation function. Further, to select the "final best model", the "representative models" from each independent strategy were compared using absolute normalized Z-scores given by the PROSA method. The Z-scores (supplemental Table S1) indicated that the best models for the Providencia porins were obtained when using OmpF and OmpC together as templates. This is expected as the multiple sequence alignment revealed a better global coverage of the target sequences (OmpPst1/OmpPst2) with parts of the target sequence alternatively well aligned by OmpF and OmpC (Fig. 4).

RESULTS
Antibiotic Susceptibility of P. stuartii ATCC 29914-MIC assays were carried out to analyze the antibiotic susceptibility of P. stuartii ATCC 29914 toward ␤-lactam antibiotics (Table 1). Chloramphenicol and the fluoroquinolone sparfloxacin were included as representative controls for different antibiotic classes. P. stuartii ATCC 29914 was susceptible to almost all antibiotics tested. However, we observed a decreased susceptibility of this strain to imipenem (MIC value of 2 g/ml). Clavulanate, a ␤-lactamase inhibitor used in combination with amoxicillin (commercial name Augmentin), was used to detect extended-spectrum ␤-lactamases. However, the test was negative with no synergy effect observed between Augmentin and other ␤-lactams (supplemental Fig. S1A). Similarly, the chelator EDTA was used as metallo-␤-lactamase inhibitor. The detection for metallo-␤-lactamase was also negative with no modification in the inhibition zone around imipenem disks in the presence of EDTA (supplemental Fig. S1, B and C).
The derivatives of P. stuartii ATCC 29914 selected in the presence of increasing concentrations of ertapenem, cefepime, or cefoxitin were tested for antibiotic susceptibility using MIC assays (Table 1). Table 1 presents the results of the strains selected at the highest concentration of ertapenem, cefepime, or cefoxitin used. The results indicate a clear correlation between the decrease of antibiotic susceptibility and the porin deficiency (Fig. 1). The strains E0.5, F0.5, and FX32 showed the maximal increase in resistance to ␤-lactam antibiotics. Compared with the parental strain, the MIC values of E0.5, F0.5, and FX32 with ertapenem increased 256-fold, 32-fold, and 16-fold, respectively. For the late generation cephalosporins cefepime and cefpirome, the MIC of strain E0.5 increased 4 -8-fold, whereas the MICs of F0.5 and FX32 increased 128 -256-fold. A similar increase of resistance to cefoxitin was observed in all three resistant derivatives E0.5, F0.5, and FX32 (64 -128-fold). In contrast, the susceptibility to chloramphenicol and sparfloxacine was unchanged in all tested variants.
MICs were also determined in the presence of phenylarginine-␤-naphthylamide (PA␤N), a well-known efflux pump inhibitor acting on members of the RND efflux pump family of Gram-negative bacteria (30,31). No significant difference in the MIC values were observed in the presence of PA␤N suggesting that PA␤N-sensitive efflux activity does not play a role in the level of antibiotic susceptibility of P. stuartii ATCC 29914 and the resistant derivatives to ␤-lactams. To investigate whether an enzymatic mechanism might be developed and associated with the decrease of ␤-lactam susceptibility, clavulanate as ␤-lactamase inhibitor (32,33) and EDTA as metallo-␤-lactamase inhibitor (22) were used in the MIC assay. However, no effects of clavulanate or EDTA on the susceptibility of the resistant strains were detected ( Table 1).
Analysis of Outer Membrane Protein Profiles-The outer membrane analysis indicated the loss of a porin with a size of about 39 kDa in the ertapenem-resistant strain E0.03, in the cefepime-resistant strain F0.13, and in the cefoxitin-resistant strain FX4 as well as in the later successive resistant derivatives (Fig. 1). The analysis of the outer membrane profiles indicated that the porin deficiency occurred in the presence of ␤-lactams at the concentrations above the MIC of the parental strain (Table 1). Polyclonal antibodies directed against the E. coli porins OmpF and OmpC were used to detect the presence of Enterobacteriaceae porins in P. stuartii (24). Both OmpF and OmpC antibodies recognized a porin band at 39 kDa from P. stuartii outer membrane fraction (Fig. 1, lower case). The traces of immunodetected porin of strains F0.13 to F0.5 and FX4 to FX32 demonstrated a very weak level of porin expression.
We investigated whether the reduced porin expression in adaptation to the treatment with ertapenem, cefepime, or cefoxitin was a result of a reversible down-regulation of gene expression or due to irreversible mutations. The resistant strains E0.5, F0.5, and FX32 were grown in successive LB cultures in the absence of antibiotics and their outer membranes were analyzed by SDS-PAGE. The production of the porin was not restored suggesting that the porin deficiency in the resistant strains was not associated with a reversible regulation event (supplemental Fig. S2).
As two genes encoding general bacterial porins are present in the genome sequence of P. stuartii, we used mass spectrometry to identify which of them was involved in antibiotic susceptibility. The outer membrane fractions of the parental ATCC strain and the resistant derivatives were solved on SDS-PAGE and stained with Coomassie Blue. The porin band which disappeared in the resistant strains was excised from the lane of the parental ATCC strain as well as strains E0.02, F0.06, and FX1. Mass spectrometry after trypsin digestion of the isolated proteins and the BLAST searches against the protein data base indicated that the porin, which disappeared from the outer membrane fraction in the presence of ␤-lactams was OmpPst1 (supplemental Table S2). The major protein band at about 40 kDa (above the protein band missing in the resistant derivatives) was identified to be OmpA.
Role of OmpPst1 and OmpPst2 in Antibiotic Susceptibility-The porin-deficient E. coli strain BL21(DE3)omp8 was used to express OmpPst1 and OmpPst2 of P. stuartii ATCC 29914 ( Table 2). The expression of Providencia porins increased the susceptibility of the producing cells to most ␤-lactam antibiotics which demonstrated the passage of ␤-lactams through P. stuartii porins. The MIC values of cefepime for E. coli BL21(DE3)omp8 expressing OmpPst1 were 3-fold less than the MIC values of E. coli BL21(DE3)omp8 expressing OmpPst2. Similarly, the MICs with cefpirome, ceftazidime, and cefoxitin  omp8 strain were isolated and analyzed by SDS-PAGE with or without heat modification. The denatured proteins migrate at about 39 kDa whereas the non-denatured proteins showed a size of about 110 kDa, ϳ3 times bigger than the denatured forms, suggesting the typical trimeric structure of enterobacterial porins (Fig. 2). Furthermore, the trimeric structures of OmpPst1 and OmpPst2 were also confirmed by black lipid bilayer assays with a typical three step gating at a threshold voltage of 200 mV for OmpPst1 and 50 mV for OmpPst2 (Fig. 2).
The isolated porins were also subjected to N-terminal sequencing with five cycles by Edman degradation technique. The 5 amino acids identified at the N termini in both OmpPst1 and OmpPst2 were AEVYN corresponding to the porin mature sequences. The result indicated the cleavage of the signal sequence and integration of OmpPst1 and OmpPst2 into the outer membrane when expressed in E. coli.
Electrophysiological Studies of Antibiotic Permeation through OmpPst1 and OmpPst2-Planar lipid bilayer technique was used to study the translocation of antibiotics through OmpPst1 and OmpPst2 at a single molecular level. As shown in a previous study, the interaction of ␤-lactams with single porin channel results in a transient blockage of ion currents that is time-resolvable (23). OmpPst1 showed a single channel conductance of 2.7 Ϯ 0.3 nS in 1 M KCl, pH 6.0 with a threshold potential of channel closure between 150 -200 mV ( Fig. 2A), whereas OmpPst2 showed a single channel conductance of 3.4 Ϯ 0.3 nS with a gating potential of about 40 mV (Fig. 2B). Statistical analysis of the data revealed that OmpPst2 is a highly voltagesensitive channel that closes at very low transmembrane voltages. Usually, the gating of enterobacterial porin channels in lipid bilayer occurs at voltages above 150 mV. However, OmpPst2 showed gating and subconductance states even at lower transmembrane voltages (Ͻ40 mV). A comparison of the characteristics of OmpPst1 and OmpPst2 with the E. coli porins OmpF and OmpC is shown in Table 3.
To analyze the molecular interactions between antibiotics and porin channels, we measured the fluctuations in the ionic currents through single trimeric channels after addition of an antibiotic. Addition of ertapenem resulted in frequent ion current blockages in both OmpPst1 and OmpPst2 reflecting strong antibiotic-channel interactions. However, ertapenem produced more ion current blockages with OmpPst1 than of OmpPst2 (Fig. 3). Addition of cefepime and cefoxitin also resulted in ionic current fluctuations with an increase in the background noise but the blockage events were not as strong as with ertapenem. Furthermore, the frequency of blockage events was higher for OmpPst1 than for OmpPst2. The average residence time of ertapenem in both channels was calculated to be approximately 150 s, whereas the residence time of cefepime and cefoxitin was only 70 -80 s. The average residence times were the same, independent whether the antibiotic was added to the cis or trans side of the lipid membrane and it did not depend on the concentration of the antibiotic.
The kinetics of antibiotic transport through channels can be derived from average residence times and number of binding events (23). The flux of antibiotics through the channel is proportional to the k on rate (entrance rate) calculated from the number of binding events (Table 4). For 1 mM of ertapenem, added to the cis side of the membrane, the flux was calculated to be 5 molecules per second per OmpPst1 monomer and 3 molecules per second per OmpPst2 monomer. In the case of cefoxitin and cefepime, the flux was calculated to be 2 molecules per second OmpPst1 monomer and 1 molecule per second per OmpPs2 monomer. These results clearly indicate that translocation of antibiotics through OmpPst1 is more efficient than

Sequence Analysis and Comparative Modeling of OmpPst1 and
OmpPst2-A search with the BLASTP program of the National Center for Biotechnology Information using the amino acid sequences of OmpPst1 and OmpPst2 showed specific hits for Gram-negative porins, which belong to the outer membrane channel superfamily. A comparison of the mature amino acid sequences of OmpPst1 and OmpPst2 with the E. coli general diffusion porins OmpF and OmpC revealed about 50% sequence identity ( Fig. 4 and supplemental Table S1). The mature amino acid sequences of OmpPst1 and OmpPst2 shared 76% identity together. OmpPst1 protein has 352 amino acids and shares 84 and 75% identity respectively to hypothetical proteins PROSTU_01774 and PROSTU_03464 of P. stuartii ATCC 25827 of which the genome was sequenced. OmpPst2 mature sequence has 343 amino acid residues and shares 76 and 100% identity respectively to PROSTU_01774 and PROSTU_03464 of P. stuartii ATCC 25827.
The sequence alignment using ClustalW2 revealed the conservation of typical secondary structure of enterobacterial porins with 16 ␤-strands, as well as 8 short periplasmic turns and 8 loops of variable lengths in-between the strands (Fig.  4). Various differences in the ␤-strands as well as those corresponding to insertions and deletions were found in the extracellular loops between OmpPst1, OmpPst2, OmpC, and OmpF. Regarding the domains involved in pore function, the L3 loop that forms the constriction region of the channel, is strongly conserved among the enterobacterial general porins (34). The L3 sequence is well conserved between OmpPst1 and OmpPst2 (80% identity); however, this domain is importantly modified (about 60%) when compared with L3 loops of OmpC and OmpF (Fig. 4). The supposedly corresponding antigenic sites in L3 loop of OmpPst1 (DVFPLW-GADTMA) and OmpPst2 (DVLPLWGADTMD) contain 7 and 6 amino acid substitutions (underlined), respectively.  A homology modeling of OmpPst1 and OmpPst2 using the structures of E. coli OmpF and OmpC as templates was carried out. Models based on OmpF can be seen in supplemental Fig. S3. The residue positions were numbered accordingly with OmpF as reference. Together with the sequence comparison (Fig. 4), the substitutions of important residues were highlighted. In OmpPst1, the important differences were M38D, M114V, E117L, F118W, A123M, R167A. Similarly, for OmpPst2 the most significant residue differences are K16Q, M38D, K80Q, E117L, F118W, R167L, R168D.

DISCUSSION
The outer membrane porins of P. stuartii have been studied at the molecular level to determine their role in antibiotic translocation. Our microbiological and electrophysiological analysis demonstrated the interaction of Providencia porins with ␤-lactams.
P. stuartii porins play the primary role in response to antibiotic stress. Previously, Mitsuyama et al. (18) showed that cefox-itin stress selected porin-deficient mutants in Proteus, Morganella, and Providencia. In this study, we focused on ␤-lactam antibiotics that are clinically used for treatment of Gram-negative bacterial infections and have been reported with increasing resistance phenomenon in various bacteria. P. stuartii ATCC 29914 produces two porins that cross-reacted with antibodies directed against the E. coli porins OmpF and OmpC. Selection of mutants resistant to the ␤-lactam antibiotics ertapenem, cefepime, and cefoxitin resulted in a significant decrease of porin expression (Fig. 1). Furthermore, these mutants that exhibited a porin expression deficiency showed an increase of MIC values for several ␤-lactams not associated with the enzymatic production such as ␤-lactamases and/or metallo-␤-lactamases (Table 1). These data indicate that changes in the composition of the outer membrane, especially the porins, play a key role in the acquired resistance of P. stuartii to ␤-lactams.
The two porins OmpPst1 and OmpPst2 of P. stuartii were characterized to be trimeric channels. They showed about 50% identity to members of the General Bacterial Porin family of Enterobacteriaceae. The analysis and homology modeling of the amino acid sequences in comparison to enterobacterial nonspecific porins, such as OmpF and OmpC (28,29), showed a high conservation of the typical porin structure with 16 ␤-strands, 8 periplasmic turns, 8 extracellular loops. However, the sequence analysis revealed significant modifications of known key residues, pointing in the lumen of the channel, such as those located in the L3 loop which forms the constriction eyelet inside the porin channel. The different composition of the L3 loop could also explain why the F4 antibody is able to detect the constriction loop L3 of most members of the General Bacterial Porin family of Enterobacteriaceae, but fails to detect the L3 loop of Providencia isolates (34). The L3 sequence is conserved between OmpPst1 and OmpPst2 suggesting an adaptation/role associated with the membrane permeability of P. stuartii. Whether a special conformation of the L3 loop would be involved in the low level susceptibility of Providencia to ␤-lactam antibiotics needs to be further investigated and clarified.
The electrophysiological characterization of Providencia porins in planar lipid bilayers revealed the conductance of OmpPst1 to be 2.7 nS, which is quite similar to E. coli OmpC, whereas OmpPst2 with 3.4 nS is an intermediate between OmpF and OmpC ( Fig. 2 and Table 3). The gating behavior of OmpPst2 is unusual: it showed a closing of the porin channel at transmembrane voltage below 40 mV. Typical enterobacterial porins show gating at voltages above 150 mV. The origin of this unusual gating behavior of OmpPst2 is unclear.
The analysis of the interaction between ␤-lactam molecules and the porins using single-channel measurements in planar lipid bilayers demonstrated the involvement of OmpPst1 and OmpPst2 in antibiotic transport. The strength of interaction decreased in the order of ertapenem Ͼ Ͼ cefepime Ͼ cefoxitin (Fig. 3). The interactions depend on the molecular structures of the antibiotic and, in particular, on their surface properties (35,36). Furthermore, the affinity site of the porin channel for an antibiotic depends on the nature and the conformation of the residues that are exposed in the lumen of the channel (37), especially, at the porin constriction region (38). OmpPst1 showed higher permeation rates for ␤-lactams in the bilayer as well as in the MIC assays than OmpPst2 (Tables 2 and 4). Previous studies have shown a correlation between the interaction of antibiotics with the porin channels and their translocation efficiency (35,39,40). Here, we observed that ertapenem had a higher affinity to the porin channels of Providencia in comparison to the cephalosporins. It is important to mention that OmpPst2, despite having a larger conductance, showed lower permeation rates for the tested antibiotics than OmpPst1. Moreover, OmpPst1 was found to be mutated in the resistant derivative strains with irreversible expression deficiency ( Table 2 and supplemental Fig. S2). All these data suggested a prominent role of OmpPst1 in the antibiotic uptake of P. stuartii. Further investigations are required for a better understanding of the structures and functions of the two porins of P. stuartii.
It is also important to note that no clear information was obtained in this study on the expression of OmpPst2 in P. stuartii cells under laboratory conditions or in the presence of ␤-lactam antibiotics. Previously, some quiescent porins were discovered in bacteria that are deficient of major porins such as OmpN and OmpK37 in E. coli and K. pneumoniae (41,42). Whether ompPst2 is a silent gene, further studies on the regulation of this gene should be carried out in more detail.
The sequence alignment and the proposed homology models of OmpPst1 and OmpPst2 suggested a high conservation of the anti-parallel ␤-barrel transmembrane domains of Gram-negative bacterial porins. However, compared with OmpF and OmpC, the extracellular loops such as L1, L4, L5, and L6 seemed to be more diverse in space with various lengths in OmpPst1 and OmpPst2. This may be predictable for more flexible dynamic movements of these loops at the surface of Providencia porins. Especially, an insertion of peptide sequence of 8 amino acids (VTSEGDSY) was observed in L5 loop region of OmpPst1. This phenomenon was observed in some enterobacterial isolates that were multiresistant to antibiotics (43).
In previous studies, the molecular mechanism of antibiotic permeation through OmpF was deciphered (35). The use of molecular simulations revealed that the bottleneck for antibiotic translocation stems from the difficulty of overcoming the constriction region of the porin (which presents a reduced size and a strong electrostatic field). Two regions, above and at the constriction zone of OmpF, were identified that form specific affinity sites for the antibiotics. We used this information to predict residues important for the channel properties of the Providencia porins, especially to compare them with the residues involved in the antibiotic translolation through E. coli OmpF. In the case of OmpPst1, we found that (i) Arg-167 of OmpF, identified as an important residue at the extracellular entrance of the OmpF channel, is replaced by an alanine in OmpPst1; (ii) Met-38 and Phe-118, which form a hydrophobic pocket above the constriction region, are substituted by an aspartic acid and a tryptophan residue, respectively; (iii) Glu-117, a key charged residue of the constriction region is replaced by a hydrophobic leucine. Similarly, in the case of OmpPst2, (i) Arg-167 and Arg-168, which affect the basic character at the entrance of the OmpF channel, are replaced by leucine and aspartic acid, respectively; (ii) Phe-118 and Met-38, which affect the hydrophobic character of the region above the constriction zone, are replaced by tryptophan and aspartic acid, respectively; (iii) Lys-80 and Lys-16, which were identified as residues involved in the basic staircase in OmpF, are both replaced by glutamine; (iv) Glu-117, which is a key charged residue of the constriction region of OmpF, is substituted by leucine.
The substitutions identified in Ompst1 and Ompst2 could disrupt the "basic ladder" found in OmpF and OmpC. In OmpF and OmpC, four basic residues, Lys-16, Arg-42, Arg-82, and Arg-132, form a positively charged cluster in the ␤-barrel wall facing the negatively charged loop L3 (29). This organization forms a strong transversal electric field in the constriction eyelet of the channel (44,45). We hypothesize that these changes might explain the ␤-lactam translocation rate as well as the voltage-sensitivity of OmpPst2. Since the L3 loop is well con-served between OmpPst1 and OmpPst2, the substitution K16Q in OmpPst2 might be very important for the different electrophysiological properties between the two Providencia porins. Although the proposed models are based on a high level of homology, they should be the subject of further experimental verifications, such as mutagenesis of the key residues. It may also be interesting to analyze the differences between the Providencia porins and E. coli OmpF/OmpC in more detail as previous studies have shown that even single amino acid exchanges can drastically alter the selectivity of a porin and affect the conformation of the porin channel (46,47,48).
The findings that OmpPst2 has a higher conductance but a lower antibiotic flux than OmpPst1 confirm that the substrate translocation is a complex process that depends on local interactions and not only on the size or dimension of the pore (35,49). Carbapenem antibiotics that are often used to treat resistant Gram-negative bacteria in intensive care units are of our further interest to investigate their molecular interaction during the passage through various enterobacterial porin channels. An attractive approach in future studies would be to obtain a molecular description of the antibiotic translocation by combining computer simulations with experimental mutagenesis of residues predicted by the simulations as important for channelantibiotic interactions.