Membrane Efflux and Influx Modulate both Multidrug Resistance and Virulence of Klebsiella pneumoniae in a Caenorhabditis elegans Model

Abstract

Cross-resistance to cefoxitin (FOX), chloramphenicol (CMP), and quinolones (nalidixic acid [NAL]) related to a putative efflux system overexpression has recently been reported for Klebsiella pneumoniae. The potential impact of this multidrug resistance (MDR) on the virulence of K. pneumoniae was evaluated in the Caenorhabditis elegans model. For 2 of the 3 MDR clinical isolates studied, a significant increase in acrB transcription was found in comparison with their antibiotic-susceptible revertants. ATCC 138821 and MDR, revertant, and derivative strains with altered porin expression were studied. Strains proved or suspected to overexpress an efflux system were significantly more virulent than the ATCC and revertant strains (time to kill 50% of nematodes [LT₅₀] in days: 3.4 to 3.8 ± 0.2 versus 4.1 to 4.4 ± 0.3, P < 0.001). Inversely, strains with altered porin expression were significantly less virulent, independently of the expression level of efflux system (LT₅₀ = 5.4 to 5.6 ± 0.2, P < 0.001). Altered porin expression did not change MICs of CMP and NAL but did those of FOX (4 to 16× MIC) and ertapenem (16 to 64× MIC). The strains with a normally or an overexpressed efflux system that received the β-lactamase CTX-M-15 became more widely resistant without modification of their virulence potential, suggesting that balance between resistance and virulence is dependent on the type of resistance mechanisms. In conclusion, this study shows that the expression of both efflux systems and porins is a key factor not only for antibiotic resistance but also virulence potential in K. pneumoniae.

Klebsiella pneumoniae, which was recognized over 100 years ago as a cause of community-acquired pneumonia and 20 years ago as a cause of community-acquired pyogenic liver abscesses, is also a common cause of nosocomial infections that range from mild urinary tract infections to severe respiratory tract infections and bacteremia (15, 30). Moreover, K. pneumoniae has been found to produce various plasmid-mediated enzymes which confer resistance to most β-lactams, particularly the extended-spectrum cephalosporins and more recently carbapenems (2, 22, 24). Treatment of serious infections caused by extended-spectrum β-lactamase (ESBL)-producing K. pneumoniae bacteria is difficult because these organisms are frequently resistant to various other families of antibiotics, including fluoroquinolones (28). Fluoroquinolone resistance in K. pneumoniae arises through specific mutations within the target proteins DNA gyrase and topoisomerase IV and also through a lower uptake of quinolones because of efflux system overexpression (3, 18). Ruzin et al. showed that the resistance to quinolones related to overexpression of the AcrAB efflux pump of K. pneumoniae was associated to resistance to other antibiotics, including chloramphenicol, erythromycin, tetracycline, and also tigecycline, a recently commercialized molecule (32). As for us, we showed that cross-resistance to cefoxitin, chloramphenicol, and quinolones that we observed in bacteremia K. pneumoniae isolates was abolished when these antibiotics were tested in the presence of the efflux pump inhibitor phenylalanine arginine β-naphthylamide (PAβΝ), alone for chloramphenicol and fluoroquinolones and in combination with cloxacillin for cefoxitin (27). Interestingly, Padilla et al. also have recently found a moderate increase in the cefoxitin MIC concomitantly to resistance to chloramphenicol and ciprofloxacin in derivatives of K. pneumoniae strain 52145 harboring knockouts of the gene acrR encoding the repressor of the efflux system AcrAB (25). They also have shown that the K. pneumoniae AcrAB efflux system is involved in resistance to the host antimicrobial peptides present in the lung, one of the first barriers of the innate immune system against infections. This finding strongly suggests the participation of efflux systems of K. pneumoniae in its virulence, as did the results provided by Coudeyras et al. that showed that the potential efflux pump EefABC of K. pneumoniae conferred an advantage to this bacterial species to colonize the digestive tract in a murine model (4).

Taking into consideration these previous findings, we studied the role of the AcrAB efflux system in the cross-resistance to cefoxitin, chloramphenicol, and quinolones observed in three genetically different bacteremia K. pneumoniae isolates that we have previously published (27). We also evaluated the virulence of these clinical isolates in comparison with isogenic strains susceptible to all antibiotic families and/or with porin alteration in the Caenorhabditis elegans nematodes whose innate immune system mimics that of the human being.

MATERIALS AND METHODS

Bacterial strains.

Three sets of K. pneumoniae strains were studied. Each set comprised four isogenic strains. One strain was the original strain corresponding to a bacteremia isolate previously published and shown to have a cross-resistance to cefoxitin, quinolones, and chloramphenicol. As this cross-resistance was demonstrated to be abolished in the presence of PAβN, it was considered to be related to an efflux system overexpression (27). This type of strain was called KPBj E+. The second strain was a derivative of the original strain that had spontaneously become susceptible to the three antibiotic families. This type of reverted strain was called KPBj Rev. The two remaining strains were mutants of the original and the reverted strains, with porin alteration and therefore called KPBj E+ P− and KPBj Rev P−, respectively. All these strains were compared in the C. elegans model with K. pneumoniae strain ATCC 138821 and strains KPBj1 E+ and KPBj1 Rev into which a plasmid harboring a gene encoding the ESBL CTX-M-15 was introduced by transformation. These two types of strains were called KPBj1 E+ ESBL and KPBj1 Rev ESBL, respectively. Escherichia coli strain TE1 producing CTX-M-15 was used to purify the plasmid encoding this ESBL (21).

Selection of mutants with ertapenem.

A recent report has shown that ertapenem is an appropriate antibiotic to select K. pneumoniae strains with porin alteration (10). Thus, this molecule was used to generate such mutants from strains KPBj E+ and KPBj Rev by using the procedure previously described by Miller et al. (19). Briefly, 5 μl of an overnight culture were transferred to 9 ml of fresh Mueller Hinton (MH) broth containing ertapenem (Merck Sharp & Dohme-Chibret, Clermont-Ferrand, France) at 0.25× MIC for the strain tested and incubated for 18 h at 37°C. Such growth conditions were repeated until mutants grew on MH agar plates containing 4× the original MIC, on which overnight culture aliquots were plated between each broth culture cycle.

Preparation of outer membrane proteins.

Outer membrane proteins were prepared from exponential-growth-phase cultures in MH broth by ultrasonic treatment, and ultracentrifugation and differential solubilization of the cytoplasmic material with sodium lauryl sarcosinate (0.3%), as previously described (9), followed. The same protein amounts of the final preparations were electrophoresed on a sodium dodecyl sulfate-polyacrylamide gel (acrylamide, 10% [wt/vol]; SDS, 0.1% [wt/vol]).

Immunocharacterization of outer membrane proteins.

Exponential-phase bacteria in Luria Bertani (LB) broth were centrifuged, and the pellet was solubilized in boiling buffer at 96°C (17). Total cell protein (optical density at 600 nm [OD₆₀₀] = 0.01, corresponding to equal amounts of protein per well) was loaded onto an SDS-polyacrylamide gel (10% polyacrylamide, 0.1% SDS). Proteins were electro-transferred onto nitrocellulose membranes in transfer buffer. An initial saturating step was performed overnight at 4°C with Tris-buffered sodium (TBS; 50 mM Tris-HCl [pH 8.0], 150 mM NaCl) containing skimmed milk powder (10%). The nitrocellulose sheets were then incubated in TBS containing skimmed milk powder (5%) and Triton X-100 (0.2%) for 2 h at room temperature in the presence of polyclonal antibodies (1:2,000 dilution) directed against denatured OmpA and against denatured OmpC and OmpF porins known as OmpK36 and OmpK35 in K. pneumoniae, respectively (17, 27, 34). The detection of antigen-antibody complexes was performed with alkaline phosphatase-conjugated AffinitiPure goat anti-rabbit IgG antibodies (17).

Transformants producing CTX-M-15.

Plasmid DNA encoding CTX-M-15 was extracted from E. coli strain TE1 and then purified using the Qiafilter plasmid midi kit (Qiagen, Courtaboeuf, France) according to the manufacturer's recommendations. This plasmid was electroporated into competent strains KPBj1 E+ and KPBj1 Rev. Transformants were selected on LB agar plates containing 4 mg/liter of cefotaxime. The presence of CTX-M-15 in the transformants growing in the presence of cefotaxime was checked by the double-disk synergy test (13).

Strain typing.

As each KPBj Rev strain was spontaneously derived from a KPBj E+ strain that was stored at −80°C in LB broth (Difco, Paris, France) containing 10% glycerol, the clonal relatedness of these two strains was checked by typing them using the random amplified polymorphic DNA (RAPD) method as previously described (11).

Real-time RT-PCR for analysis of acrB expression.

Analysis of the mRNA (transcription) levels of the acrB gene was performed according to the method previously described by Doumith et al. (6). Briefly, total cellular RNA was extracted using the RNeasy minikit (Qiagen) and treated with RNase-free DNase (Qiagen) for 30 min at 37°C, after which a second step of purification was performed. RT-PCR was carried out in a LightCycler using the one-step LightCycler RNA Master SYBR green I kit (Roche Applied Science, Meylan, France) according to the manufacturer's protocol. The specificity of the generated products was tested by melting-point analysis. Amplifications were performed in triplicate from three different RNA preparations. Cycle threshold (C_T) values of the target acrB gene were compared with the C_T values of the housekeeping rpoB gene, chosen as an endogenous reference for normalizing the transcription levels of the target gene. Strain ATCC 138821 was used as a control, and the normalized relative expression of the acrB gene was determined for each strain according to the following formula: 2^−ΔΔCT, where ΔΔCT = (CT_acrB − CT_rpoB)_{studied strain} − (CT_acrB − CT_rpoB)_{control strain}. Primers used for acrB were 5′-CGATAACCTGATGTACATGTCC-3′ and 5′-CCGACAACCATCAGGAAGCT-3′ and for rpoB 5′-AAGGCGAATCCAGCTTGTTCAGC-3′ and 5′-TGACGTTGCATGTTCGCACCCATCA-3′.

Antibiotic susceptibility.

Susceptibility to nalidixic acid, ciprofloxacin, levofloxacin, chloramphenicol, cefoxitin, cefotaxime, ceftazidime, and ertapenem was determined by either the agar dilution method or the agar disk diffusion method. MIC determination, replicated three times, was performed and interpreted following the recommendations of the French Antibiogram Committee (http://www.sfm.asso.fr/nouv/general.php?pa=2).

Nematode killing assay.

The C. elegans model has been developed to study host-pathogen interactions and to identify the basic pathways well conserved during evolution that are linked to microbial pathogenesis. In this test, the studied bacterium is presented as food to the nematodes instead of E. coli strain OP50, an avirulent strain that is their usual food in the lab. Ingestion of the bacterium studied by the worms results in an infection and ultimately death of the worms (16). The time required by the bacterium studied to kill the worms compared with the life duration observed when the worms are fed E. coli strain OP50 is an indirect marker of virulence potential of the bacterium studied. The C. elegans infection assay was carried out as described by Lavigne et al. (16) using the Fer-15 mutant line, which has a temperature-sensitive fertility defect. Fer-15 was provided by the Caenorhabditis Genetics Center that is funded by the NIH National Center for Research Resources (NCRR). To synchronize the growth of worms, eggs were collected using the hypochlorite method. Overnight cultures of K. pneumoniae strains in the nematode growth medium (NGM) were harvested, centrifuged, washed once, and suspended in phosphate-buffered saline (PBS) solution at pH 7.0 at a concentration of 10⁵ CFU/ml. NGM agar plates were inoculated with 10 μl of bacterial suspension and incubated at 37°C for 8 to 10 h. Plates were brought back to room temperature and seeded with L4-stage worms (20 to 30 per plate). Plates were then incubated at 25°C and scored each day for live worms under a stereomicroscope (Leica MS5). A worm was considered dead when it no longer responded to touch. Worms that died as a result of having been trapped by the wall of the plate were excluded from the analysis. At least three replicates repeated five times were performed for each selected strain. The 50% lethal time (LT₅₀) and 100% lethal time (LT₁₀₀) corresponded to the time (in days) required to kill 50% and 100% of the nematode population, respectively.

In vitro and in vivo strain growth.

For in vitro experiments, bacteria grew in LB broth. An aliquot (1 ml) of an overnight culture of each strain was transferred in 99 ml and cultured at 37°C under shaking conditions. Bacterial growth was monitored by change in the OD₆₂₀ and determined twice an hour over a period of 10 h. Growth curves were performed in triplicate for each strain, and the OD₆₂₀ measurements were averaged for each time point in order to plot growth curves. In vivo experiments consisted of measuring the number of bacteria within the C. elegans digestive tract 72 h after ingestion, as described by Garsin et al. (8) Three replicates were performed for each strain. The antibiotic susceptibility of each strain recovered from the digestive tract of the nematodes was determined by the disk agar diffusion method and compared to that of the corresponding strain given as food to the nematodes.

Statistical analysis.

To compare the entire survival curves in nematode killing assays, we used a Cox regression. In order to perform pairwise comparison between two different strains, we used a log rank test. The analysis was carried out using SPSS 6.1.1 (SPSS Inc., Chicago, IL). The Kruskal-Wallis test was used to compare the in vitro and in vivo bacterial growth of the different strains. The chi-square test was used to compare the mean values of the normalized acrB transcription.

RESULTS

Clonal relatedness of KPBj E+ and KPBj Rev strains.

Each KPBj E+ strain and its corresponding revertant KPBj Rev strain showed strictly identical RAPD profiles, confirming the clonal relatedness of the two strains (data not shown).

Outer membrane protein profile.

Growth cycles of strains KPBj E+ and KPBj Rev in the presence of ertapenem allowed for selection of mutants for which ertapenem MICs were 16 to 64 times higher (Table 1 ). As indicated in Fig. 1, the ertapenem derivatives exhibited a similar porin-deficient phenotype. No porin was detected in strains KPBj1 Rev P−, KPBj1 E+ P−, KPBj3 Rev P−, KPBj3 E+ P−, and KPBj5 Rev P−, and a very weak signal was obtained with strain KPBj5 E+ P− (Fig. 1A; Table 1). This lack of porin was observed with antisera that recognized OmpK35 (Fig. 1A) and OmpK36 (data not shown). However, the signal corresponding to OmpA, a structural protein of the bacterial outer membrane, was conserved (Fig. 1B). The SDS-PAGE assay showed that the major modification that occurred in the outer membrane protein profile of the ertapenem mutants consisted of the absence of a major outer membrane protein (Fig. 1C).

TABLE 1.

Antibiotic susceptibility of the different K. pneumoniae strains^a

Strain	Putative overexpressed efflux	Membrane protein content		MIC (mg/liter)
		Porins	OmpA	NAL (≤8, >16)	CIP (≤0.5, >1)	LEV (≤1, >2)	CMP (≤8, >8)	CTX (≤1, >2)	CAZ (≤1, >8)	FOX (≤8, >32)	ETP (≤0.5, >1)
ATCC 138821	−	+	+	4	0.03	0.03	2	0.06	0.125	2	0.007
KPBj1 E+	+	+	+	32	0.5	1	128	0.125	0.25	8	0.007
KPBj1 Rev	−	+	+	2	0.015	0.03	2	0.06	0.125	2	0.007
KPBj1 E+ P−	+	−	+	32	0.5	1	128	0.25	0.125	32	0.5
KPBj1 Rev P−	−	−	+	2	0.03	0.06	4	0.25	0.125	32	0.5
KPBj1 E+ ESBL§	+	+	+	32	1	0.5	128	128	32	8	0.125
KPBj1 Rev ESBL	−	+	+	4	0.03	0.03	2	128	16	2	0.03
KPBj1 E+ P− ESBL	+	−	+	32	2	1	128	256	32	32	4
KPBj1 Rev P− ESBL	−	−	+	4	0.06	0.06	4	256	32	32	2
KPBj3 E+	+	+	+	32	0.25	0.5	128	0.06	0.25	4	0.007
KPBj3 Rev	−	+	+	2	0.015	0.03	4	0.06	0.125	2	0.007
KPBj3 E+ P−	+	−	+	32	0.5	0.5	128	0.25	0.25	32	0.25
KPBj3 Rev P−	−	−	+	2	0.03	0.015	4	0.25	0.25	32	0.25
KPBj5 E+	+	+	+	8	0.06	0.125	8	0.125	0.25	8	0.015
KPBj5 Rev	−	+	+	2	0.015	0.06	2	0.06	0.125	2	0.007
KPBj5 E+ P−	+	±	+	8	0.125	0.125	8	0.25	0.25	32	0.25
KPBj5 Rev P−	−	−	+	2	0.03	0.06	2	0.125	0.125	16	0.25

^aNAL, nalidixic acid; CIP, ciprofloxacin; LEV, levofloxacin; CMP, chloramphenicol; CTX, cefotaxime; CAZ, ceftazidime; FOX, cefoxitin; ETP, ertapenem; §, ESBL, CTX-M-15; +, present; −, absent; ±, very weak. Breakpoints according to the French Antibiogram Committee are given in parentheses following respective antibiotics.

FIG. 1.

Analysis of outer membrane proteins in Klebsiella pneumoniae. (A) Porin OmpK35 immunodetection with antiserum directed against denatured OmpF family porin. (B) OmpA immunodetection with antiserum directed against denatured OmpA. Lanes: 1, ATCC 138821; 2, KPBj1 Rev; 3, KPBj1 Rev P−; 4, KPBj1 Rev ESBL; 5, KPBj1 E+; 6, KPBj1 E+ P−; 7, KPBj1 E+ ESBL; 8, KPBj3 Rev; 9, KPBj3 Rev P−; 10, KPBj3 E+; 11, KPBj3 E+ P−; 12, KPBj5 Rev; 13, KPBj5 Rev P−; 14, KPBj5 E+; 15, KPBj5 E+ P−. (C) Outer membrane proteins of strains KPBj3 stained with Coomassie blue after SDS-PAGE analysis: a, b, c, d, and e, molecular mass standards of 105 kDa, 84 kDa, 50 kDa, 33 kDa, and 29 kDa, respectively. Arrow indicates the position of the major outer membrane protein absent in strains KPBj3 Rev P− and KPBj3 E+ P−.

Transcription of gene acrB.

The relative transcription levels of gene acrB for strains KPBj E+ and KPBj Rev are shown in Fig. 2. There was a significant increase in acrB transcription levels for strains KPBj1 E+ and KPBj3 E+ compared with strains KPBj1 Rev and KPBj3 Rev, respectively. Such a difference was not observed with strains KPBj5. The introduction of a plasmid harboring the bla_CTX-M-15 gene did not modify the difference in acrB expression levels between strain KPBj1 E+ and strain KPBj1 Rev.

FIG. 2.

Relative mRNA expression level of gene acrB. Mean and standard deviation (SD) of acrB expression relative to that of control strain ATCC 138821 (white bar) are indicated with gray bars for strains KPBj Rev and with black bars for strains KPBj E+. p, statistically significant difference; ns, not significant difference.

Antibiotic susceptibility.

As indicated in Table 1, the spontaneously reverted KPBj Rev strains issued from KPBj E+ strains exhibited a restoration of full susceptibility to cefoxitin, quinolones, and chloramphenicol. Both KPBj E+ and KPBj Rev ertapenem mutants with porin alteration showed levels of susceptibility to quinolones and chloramphenicol similar to those of the parental strains but a significant increase in cefoxitin MICs (4 to 16×) and ertapenem MICs (16 to 64×) (Table 1). Strains KPBj1 E+, KPBj1 Rev, KPBj1 E+ P−, and KPBj1 Rev P− susceptible to cefotaxime and ceftazidime became resistant to these antibiotics in the presence of CTX-M-15 (Table 1) without susceptibility changes, with regard to the other antibiotics tested except for ertapenem. Ertapenem MICs were 16 and 4 times higher for strains KPBj1 E+ and KPBj1 Rev producing CTX-M-15, respectively, and 8 and 4 times higher when these CTX-M-15-producing strains had porin alteration (Table 1).

Strain virulence.

In the nematode killing assay, all KPBj Rev strains showed a virulence profile similar to that of strain ATCC 138221. Inversely, all KPBj E+ strains were significantly more virulent (LT₅₀ in days, 3.4 ± 0.2 to 3.8 ± 0.2; LT₁₀₀ in days, 7.0 ± 0.1 to 7.2 ± 0.8) than strain ATCC 138221 (P < 0.01) and KPBj Rev strains (LT₅₀, 4.1 ± 0.2 to 4.4 ± 0.3; LT₁₀₀, 8.3 ± 0.7 to 8.6 ± 0.6) (P < 0.001). Concerning the mutants with altered porin expression, they showed similar LT₅₀ and LT₁₀₀ values independently of the expression level of the efflux system in the original strains (for KPBj E+ P−, LT50, 5.4 ± 0.2 to 5.6 ± 0.2; LT₁₀₀, 10.6 ± 0.6 to 10.8 ± 0.8 versus, for KPBj Rev P−, LT₅₀, 4.9 ± 0.2 to 5.3 ± 0.2; LT₁₀₀, 10.0 ± 0.5 to 10.3 ± 0.7). These mutants were significantly less virulent than strain ATCC 138221 and than both strains KPBj Rev and KPBj E+ (P < 0.001). Introduction of the plasmid-mediated CTX-M-15 enzyme into strains KPBj1 E+ and KPBj1 Rev did not modify the virulence level of these strains. Thus, strain KPBj1 E+ ESBL (LT₅₀, 3.6 ± 0.3; LT₁₀₀, 7.1 ± 0.6) was significantly more virulent than strain KPBj1 Rev ESBL (LT₅₀, 4.3 ± 0.3; LT₁₀₀, 8.5 ± 0.5) (P < 0.001).

In vitro and in vivo bacterial growth.

The growth curves determined by culturing the strains in a rich medium did not show significant differences between the four types of strains (KPBj E+, KPBj Rev, KPBj E+ P−, and KPBj Rev P−) in each of the 3 sets of strains (KPBj1, P = 0.65; KPBj3, P = 0.44, and KPBj5, P = 0.39) (data not shown). Identical results were obtained, with regard to the in vivo bacterial growth: CFU counts of the different strains in the digestive tract of nematodes measured 72 h after ingestion varied by around 10⁶ without significant difference (data not shown).

DISCUSSION

Several Gram-negative human pathogens, including Pseudomonas aeruginosa, Salmonella enterica, and extraintestinal pathogenic E. coli, have been shown to kill C. elegans when presented to the nematodes as a source of nutrient (1, 5, 16). In this study, we showed that K. pneumoniae can also kill C. elegans. To evaluate this K. pneumoniae behavior, we used three bacteremia clinical isolates previously shown to have a multidrug-resistant (MDR) phenotype related to a putative overexpressed efflux system. Sequencing analysis of the genome of three K. pneumoniae strains has indicated that this species encodes several efflux systems involved in the MDR phenotype (7, 23, 36). In the present study, RT-PCR experiments focused on the acrB gene allowed us to detect a slightly but significantly increased transcription of this gene in two of the three clinical isolates in comparison with their derivatives having spontaneously lost the MDR phenotype, suggesting an overexpression of the AcrAB efflux pump, which was not previously detected by immunoblot experiments (27). Involvement of another efflux system to explain the MDR phenotype of the third strain is our current hypothesis. However, to definitely link the MDR phenotype to the increased expression of the AcrAB pump, it would be necessary to study acrB knockouts derived from our clinical isolates.

The virulence of the three MDR clinical isolates was compared with that of derivatives having spontaneously lost the MDR phenotype and with derivatives in which porin expression was shown to be altered. This procedure allowed us to investigate the role played by efflux and porin expression in the colonization process of the C. elegans nematode and in MDR, respectively.

The overexpression of an efflux pump, proved for two and putative for one of our clinical isolates, was correlated with an increase in the virulence index of K. pneumoniae as measured by the kinetics of nematode killing. This result confirms the previous one recently reported by Padilla et al. (25). The presence of an active efflux pump is clearly a key parameter during the infectious process, possibly related to the protection of the pathogen against natural host responses, such as antimicrobial peptides (25). A more original finding provided by our study is the effect of porin deficiency on virulence. In the ertapenem derivatives, we observed pleiotropic negative porin expression, e.g., a lack of OmpK35 and OmpK36, and not only a shift in the porin expression as previously reported for several enterobacterial species (26). Moreover, as the signal corresponding to OmpA, an architectural protein of the bacterial outer membrane (17) was conserved, the membrane alteration was focused on the family of nonspecific trimeric porins (26). This porin lack was associated with a decrease in the virulence index of K. pneumoniae strains, as reflected by the significant modification of the kinetics of the killing of the nematodes even in strains that overexpressed an efflux pump. Involvement of porins in bacterial virulence has been mentioned by some teams (12, 29, 31, 35). However, this is the first time that a correlation between porin expression and level of virulence has been demonstrated for K. pneumoniae. Porins expressed by the bacteria under conditions present in the host body contribute to the stability of the outer membrane (35) and also participate in the bacterial defense against the chemical immune system, e.g., defensins and antimicrobial peptides (29). Moreover, they also play a role in bacterial adhesion and invasion (12, 31). In addition to the role of efflux pumps in the colonization, it is now strongly suggested that the membrane transporters are active partners of bacterial virulence in K. pneumoniae.

The three strains with a demonstrated or putative overexpressed efflux system showed a cross-reduced susceptibility to cefoxitin, chloramphenicol, and quinolones. The altered porin expression derivatives of these strains and of their susceptible revertants conserved the parental level of resistance to chloramphenicol and quinolones. Inversely, the resistance levels to cefoxitin and ertapenem were increased when strains had altered porin expression. This increase was, moreover, with similar amplitude, for strains with or without an overexpressed efflux system. Thus, the efflux system appears to be the key mechanism for MDR, with regard to chloramphenicol and quinolones, whereas porin alteration appears to be a more active mechanism than efflux to decrease the cefoxitin and ertapenem activity. As a result, a noticeable level of resistance to cefoxitin with a cross-resistance to chloramphenicol and nalidixic acid indicates the presence of both an overexpression of the efflux system and an altered porin expression in K. pneumoniae as suggested in our previous study (27).

The notion that resistant isolates are less virulent than those that are susceptible is widely debated, notably for E. coli (14, 20). In the present study, the transformation of the antibiotic-susceptible strain KPBj1 Rev with a plasmid encoding the ESBL CTX-M-15 (resistance to diverse β-lactam molecules) and other antibiotic markers (data not shown) rendered this strain MDR without, however, decreasing its virulence potential. A similar result was obtained from strain KPBj1 E+ ESBL, which remained as virulent as the original strain KPBj1 E+, in spite of an extension of its resistance spectrum. Thus, contrary to Sahly et al., we did not observe an increase in the virulence potential of strains when they acquired an ESBL (33). We observed, as previously described, that the introduction of CTX-M-15 into strains KPBj1 E+ and KPBj1 Rev with porin alteration renders them not only resistant to cefotaxime and ceftazidime but also to ertapenem (10). We did not measure the impact of these β-lactam resistances on the virulence potential of the strains with altered porin expression because they had already been shown significantly less virulent in the absence of CTX-M-15. All these experiments emphasize that the key point, with regard to the balance between virulence and MDR, is certainly the type of mechanism(s) involved in this resistance. In other words, an interaction between virulence and resistance could be expected when the mechanism of resistance is related to systems which play a role in antibiotic influx and efflux.

In conclusion, by using the C. elegans model, the present study allowed us to show that an overexpression of the AcrAB efflux system is linked to an increased virulence potential of K. pneumoniae, whereas porin alteration decreases it. This study clearly showed that chloramphenicol and quinolone cross-resistance is related to efflux system overexpression and not to porin alteration. Of the two systems, efflux and influx, it is the influx system that has the most impact on bacterial susceptibility to β-lactams. Although the bacteria with altered porin are significantly less virulent, we think that porin alteration is a threatening mechanism because it participates in increasing the level and in extending the spectrum of the β-lactam resistance in clinical isolates that also produce β-lactamases, such as the very common and globally found CTX-M-15 enzyme (21).