Impact of Elements Containing Glycopeptide Resistance Genes on Expression of Virulence in Enterococcus faecalis Peritonitis: a Pilot Study with Rats

Abstract

The relationship between virulence and chromosomal elements containing glycopeptide resistance genes was experimentally assessed for two transconjugant strains of Enterococcus faecalis (VanA and VanB phenotypes) and compared to that for a susceptible wild-type strain. Microbiologic and inflammatory effects were assessed in a polymicrobial rat model of peritonitis. Mean peritoneal enterococcus concentrations ± standard deviations at day 1 were 2.1 ± 1.9, 1.3 ± 1.1, and 1.7 ± 2.0 log₁₀ CFU/ml for susceptible, VanA, and VanB strains, respectively (P < 0.05). At day 3 also there were lower concentrations of glycopeptide-resistant enterococcal strains in peritoneal fluid (3.2 ± 3.4, 1.8 ± 1.8, and 2.1 ± 2.4 log₁₀ CFU/ml for susceptible, VanA, and VanB strains, respectively [P < 0.05]). Transconjugant glycopeptide-resistant strains were associated with increased peritoneal cell counts at the different evaluation times of the experiment (P < 0.001). Plasma α1-acid glycoprotein concentrations were lower in the presence of the susceptible strain (667 ± 189 mg/liter) than in the presence of the VanA or VanB strain (1,193 ± 419 or 1,210 ± 404 mg/liter, respectively [P < 0.05]), while concentrations of tumor necrosis factor alpha and interleukin-6 in peritoneal fluid remained similar for the strains. These results suggest a trend toward variation of virulence of transconjugant strains compared to the wild-type strain in this peritonitis model.

Introduction of a plasmid and/or of a large chromosomal element changes features of the results from animal models (13). Plasmids are known to both increase and decrease the virulence of bacteria in animals. However, the effects of large genetic elements have been minimally evaluated. This point is of particular interest for enterococci, because in real life, glycopeptide resistance genes are often present in such elements (19).

Expression of the vanA or vanB gene and other genes involved in their regulation results in the synthesis of abnormal peptidoglycan precursors. The influence of glycopeptide resistance on the virulence of enterococci remains unclear due to lack of experimental studies and disparate clinical findings (10, 14, 15). Impaired virulence leading to a limited pathogenic potential of these bacteria could be expected due to the genetic load of multidrug resistance, as observed for other resistant gram-positive organisms (26). As previously reported for intraabdominal infection, enterococci often express their pathogenicity essentially as components of polymicrobial inocula, likely reflecting the activity of synergistic virulence mechanisms (17, 18, 20). In the rat model of peritonitis, cell wall components of Enterococus faecalis such as peptidoglycan or lipoteichoic acid might influence the pathogenicity of this bacterium (18), while virulence factors such as cytolysin-bacteriocin aggregation substance and gelatinase do not seem to play a major role (8).

As an initial approach to clarifying relationships between the inclusion of large elements containing vancomycin resistance genes and changes in virulence, this pilot study was designed to evaluate the role of genetic elements mediating glycopeptide resistance in the expression of virulence in E. faecalis. Because the pathogenicity of E. faecalis in the rat peritonitis model is minimal when it is inoculated alone and enhanced in polymicrobial models of infection (8), we addressed this issue in a rat model of polymicrobial peritonitis by focusing on the microbiologic and inflammatory effects of the enterococcal strains.

(This study was presented in part at the 40th International Conference for Antimicrobial Agents and Chemotherapy, Toronto, Ontario, Canada, September 2000.)

MATERIALS AND METHODS

Microorganisms.

Three strains of E. faecalis were obtained from the Institut Pasteur of Paris (P. Courvalin). E. faecalis JH2-2 (vancomycin and teicoplanin susceptible [Vm^s and Te^s, respectively]) is susceptible to β-lactam antibiotics and intrinsically resistant to low levels of aminoglycosides (12). E. faecalis BM4316 harbors a 50-kb plasmid genetic element conferring VanA-type resistance and was obtained by conjugal transfer of vancomycin and teicoplanin resistance from clinical isolate Enterococcus faecium HM1074 to JH2-2 (3). Expression of resistance was assessed by testing of susceptibility to vancomycin and teicoplanin. The agar dilution method was used to determine the MICs of vancomycin and teicoplanin in Mueller-Hinton II agar. E. faecalis BM4316 was highly resistant to vancomycin (MIC, 256 μg/ml) and teicoplanin (MIC, 32 μg/ml). E. faecalis BM4275 harbors a 250-kb chromosomal genetic element conferring VanB-type resistance and was obtained by conjugal transfer of vancomycin resistance from clinical isolate E. faecium BM4120 (Vm^r Te^s) to JH2-2 (13, 23). This strain was resistant to vancomycin (MIC, 64 μg/ml) and susceptible to teicoplanin (MIC, 0.5 μg/ml). The Bacteroides fragilis (AIP5-86) and Escherichia coli (CB1496) strains used in this study have been described previously (17).

The model of polymicrobial peritonitis previously described involves implantation of a gelatin capsule containing barium sulfate along with E. faecalis (10⁹ CFU/ml), B. fragilis (10⁹ CFU/ml), and E. coli (10⁸ CFU/ml) (8, 17). A group of animals underwent a sham procedure consisting of implantation of a gelatin capsule containing barium sulfate, semisolid agar medium, and sterile broth.

Determination of growth curve of E. faecalis strains.

To assess the consequences of movements of large genetic elements in the three strains, the growth curves for the three strains were determined. The growth of VanA- and VanB-type resistant transconjuguant strains of E. faecalis was compared to that of wild-type JH2-2 by using optical density and bacterial colony counts. Overnight cultures of each strain were diluted 100-fold with brain heart infusion broth and incubated at 37°C with shaking. Six to eight different tubes were used for each point of the in vitro growth curve. Triplicate measurements were made for each tube. Each hour, 100-μl aliquots of each culture and 100 μl of 10⁻⁴ to 10⁻⁸ dilutions of each sample were plated onto brain heart infusion agar plates. The generation time was defined as the time to reach twice the bacterial count and was graphically determined during the logarithmic phase of growth. For each strain, the slope of the logarithmic phase of growth was used to determine the specific growth rate, which was defined as ln2 divided by the generation time (11).

Animals.

Male Sprague-Dawley rats (Janvier, Le Genest St Isle, France) weighing 250 to 300 g and housed 4 per cage had access to chow and water ad libitum throughout the experiment. All of these experiments were performed according to current European regulations.

Preparation of microorganisms.

B. fragilis was grown at 37°C under anaerobic conditions and diluted anaerobically in prereduced thioglycolate broth. E. coli and the three strains of E. faecalis were grown anaerobically in brain heart infusion broth at 37°C. The final inoculum was prepared when the bacteria were in the log phase of growth. The inocula were adjusted spectrophotometrically, and bacteria were diluted to obtain the number of microorganisms required for bacterial challenge. Purity was assessed by Gram staining and microscopic examination, and counts were validated for each strain by taking samples for enumeration immediately before mixing.

Intraabdominal infection in rats.

Semisolid agar medium was prepared by adding 2% (wt/vol) agar to the diluted broth cultures along with 10% (wt/vol) barium sulfate. Aliquots (0.5 ml) of the final product were placed in double gelatin capsules for peritoneal implantation. Each E. faecalis strain was studied using 20 animals. Each strain of E. faecalis (10⁹ CFU/ml) was inoculated with E. coli (10⁸ CFU/ml) and B. fragilis (10⁹ CFU/ml).

Implantation of inoculum.

Rats were anesthetized with an intramuscular injection of ketamine (30 mg/kg of body weight; Parke-Davis, Courbevoie, France), and the gelatin capsule was inserted into the pelvic cavity via a midline abdominal incision. The wound was closed at the musculoperitoneal and skin layers by using interrupted nylon sutures.

Assessment of spontaneous outcome.

After implantation of the inoculum, the animals were returned to separate cages and then observed and weighed daily until sacrifice. In each group, four animals were sacrificed at 6 h after inoculation for cytokine measurements and the remaining animals were sacrificed at 24 h (n = 8) or on day 3 (n = 8). The following pathological criteria were studied: clinical criteria (mortality and change in body weight), microbiologic criteria (positive blood cultures and bacterial counts in peritoneal fluid), and inflammatory response (peritoneal phagocyte, tumor necrosis factor alpha [TNF-α], and interleukin-6 [IL-6] concentrations and serum α1-acid glycoprotein concentrations).

Sacrifice.

Animals were killed with isoflurane (Abbott Laboratories, Rungis, France). Blood samples were obtained by aseptic percutaneous transthoracic cardiac puncture for qualitative blood culture on day 1 (at 24 h after inoculation) and day 3 and determination of serum α1-acid glycoprotein concentrations on day 3. After intraperitoneal injection of 10 ml of cold phosphate-buffered saline, a midline laparotomy was performed, and peritoneal fluid samples were recovered from all regions of the peritoneal cavity for bacterial and cell counts (days 1 and 3). Peritoneal fluid cytokine concentrations were specifically evaluated 6 h after bacterial challenge. A dilution factor taking into account the fluid present in the peritoneal cavity prior to the injection of 10 ml of phosphate-buffered saline was applied to all calculations according to a previously described technique (8, 18). Blood cultures were inoculated immediately after collection by using aerobic and anaerobic blood culture bottles (BIO AER and BIO ANAER; Bio-Rad, Sanofi Diagnostics Pasteur, Marnes-la-Coquette, France). The bottles were incubated immediately in accordance with the manufacturer's recommendations. They were analyzed daily until day 5 for identification. Serial dilutions of the peritoneal fluid were performed, and 0.1 ml of each dilution was spread onto agar plates for colony counts. The limit of detection for each microbiologic test was <1 log₁₀ CFU/ml. For values below this limit, results were expressed as ≤1 log₁₀ CFU/ml and were treated as 1 log₁₀ CFU/ml in statistical analysis. Plates were incubated under appropriate conditions (aerobic or anaerobic) for 2 to 5 days. The selective medium used for detection of B. fragilis was Columbia agar base (Sanofi Diagnostics Pasteur) with 5% sheep blood containing 2 μg of ampicillin/ml and 4 μg of pefloxacin/ml. The selective media used for aerobic culture were Drigalski agar and bile-esculin-azide agar (both from Sanofi Diagnostics Pasteur). The stability of plasmids was assessed by antibiotic susceptibility testing. The disk agar diffusion method was used with disks containing 30 mg of vancomycin and 30 mg of teicoplanin (Bio-Rad) (25).

Peritoneal cell counts.

The total cell count (polymorphonuclear neutrophils [PMN], macrophages, and lymphocytes) was determined on an aliquot of the original peritoneal fluid by using a Malassez counting chamber.

Cytokine assay.

Four 1-ml samples of the original peritoneal fluid collected from all regions of the peritoneal cavity were centrifuged (at 300 × g for 15 min at 4°C), divided into 200-μl aliquots, and stored at −80°C until assay. Samples were assayed in triplicate. TNF-α activity was measured by an enzyme-linked immunosorbent assay (ELISA) using a recombinant rat anti-TNF-α monoclonal antibody (R & D Systems, Abington, United Kingdom). The sensitivity of this test, which was defined as the lowest concentration of the standard which showed an absorbance greater than the mean absorbance of the 0-pg/ml sample ± 2 standard deviations (SD), was 12.5 pg/ml. IL-6 activity was measured by the same ELISA method using a recombinant rat anti-IL-6 monoclonal antibody (R & D Systems). The limit of detection of IL-6 in peritoneal fluid was 12.5 pg/ml.

α1-Acid glycoprotein assay.

Blood samples were obtained by cardiac puncture on day 3 of the experiment and transferred to sterile glass tubes. After coagulation, the serum was collected and centrifuged at 500 × g for 5 min at 4°C. The supernatant was divided into four aliquots and stored at −80°C until the assay. Protein immunoblotting was performed to detect α1-acid glycoprotein in serum. Each sample was applied to a nitrocellulose membrane (pore size, 0.45 μm; Trans-Blot; Bio-Rad Laboratories, Richmond, Calif.). Nitrocellulose membranes were incubated with immune serum containing a rabbit anti-rat α1-acid glycoprotein antibody (5). After a wash, the substrate reaction was performed using a peroxidase-linked rabbit anti-IgG secondary antibody. Each intensity of grey could be associated with a defined α1-acid glycoprotein concentration. Comparison of the different samples with this reference permitted determination of each concentration. Values below 300 mg/liter were considered normal.

Statistical analysis.

Results are expressed as means ± SD or as proportions. Continuous parameters were compared by Kruskal-Wallis and Mann-Whitney tests. Chi-square tests were used for quantitative data. A P value of <0.05 was considered significant.

RESULTS

Determination of growth curves of glycopeptide-resistant transconjugant strains.

Bacterial growth curves of susceptible, VanA, and VanB strains are shown in Fig. 1. The bacterial concentration for the susceptible strain after growth for 24 h was (3.5 ± 3.0) × 10⁹ CFU/ml, while the concentration for the VanA strain was (2.5 ± 1.6) × 10⁹ CFU/ml and that for the VanB strain was (1.6 ± 1.4) × 10⁹ CFU/ml (not significant). The generation times were 0.23 h (13.8 min) for the susceptible strain, 0.17 h (10.2 min) for the VanA strain, and 0.33 h (19.8 min) for the VanB strain. The specific growth rates were 3.01 for the susceptible strain, 4.07 for the VanA strain, and 2.1 for the VanB strain.

FIG. 1.

In vitro growth curves of susceptible (JH2-2) (solid circles), VanA (shaded squares), and VanB (open triangles) strains of E. faecalis. Bacterial titers are expressed as optical density readings (means ± SD) and were measured every hour until stationary phase was reached.

Effect on survival in the rat peritonitis model.

No unexpected deaths were observed in the three groups through the study period.

Effect of E. faecalis strains on body weight.

The most severe weight loss occurred on day 1 for all groups. Progressive weight recovery was observed in control and VanB-infected animals, while persistent weight loss was noticed in VanA-infected animals. No difference in weight variations in comparison with baseline was observed among the groups (susceptible, VanA, and VanB) on day 1 (−7, −7, and −6%, respectively) or on day 3 (−2, −3, and −1%, respectively). Lower weight loss occurred in sham-treated animals than in the other groups both on day 1 (−4%) and on day 3 (+3%) (P = 0.003).

Effect of E. faecalis strains on blood cultures in the rat model.

All animals had positive blood cultures on day 1. More enterococcal bacteremias occurred on day 1 than on day 3. Sixteen of 20 animals (70%) developed positive E. faecalis bacteremia at day 1, compared to only 5 of 20 animals (21%) at day 3. However, no difference in the frequency of bacteremia was noted among the strains of E. faecalis (data not shown). No variation in the frequency of E. coli and B. fragilis bacteremia was observed between day 1 and day 3 (data not shown).

Effect of E. faecalis strains on peritoneal cultures in the rat model.

Bacterial titers of the E. faecalis strains in the peritoneal cavity between days 1 and 3 are shown in Table 1. Concentrations of the susceptible E. faecalis strain in the peritoneal cavity on days 1 and 3 were significantly higher than those of the VanA and VanB strains (P = 0.001). All bacteria isolated from peritoneal cultures retained their resistance phenotypes. No new teicoplanin- or vancomycin-resistant mutant bacteria were isolated in the peritoneal cultures after animal experiments. At day 1, bacterial titers of E. coli were 2.4 ± 2.7 log₁₀ CFU/ml in coinfection with the susceptible strain of E. faecalis, 2.2 ± 2.5 log₁₀ CFU/ml with the VanA strain, and 2.6 ± 2.9 log₁₀ CFU/ml with the VanB strain. At day 3, bacterial titers of E. coli were 3.3 ± 3.7, 3.0 ± 3.0, and 2.7 ± 2.9 log₁₀ CFU/ml with the susceptible, VanA, and VanB strains, respectively. At day 1, bacterial titers of B. fragilis were 3.0 ± 3.4 log₁₀ CFU/ml in coinfection with the susceptible strain of E. faecalis, 3.6 ± 3.8 log₁₀ CFU/ml with the VanA strain, and 1.0 ± 0.1 log₁₀ CFU/ml with the VanB strain. At day 3, bacterial titers of B. fragilis were 5.3 ± 5.7, 3.7 ±3.9, and 5.4 ± 5.8 log₁₀ CFU/ml with the susceptible, VanA, and VanB strains, respectively.

TABLE 1.

Bacterial titers, peritoneal cell counts, levels of IL-6 and TNF-α in peritoneal fluid, and α1-acid glycoprotein dosage in serum

Group	Mean bacterial titer (log10 CFU/ml) ± SD		Mean peritoneal cell count (106/ml) ± SD		Mean peritoneal IL-6 level (ng/ml) ± SD	Mean peritoneal TNF-α level (ng/ml) ± SD	Mean α1-acid glycoprotein dosage in serum (mg/liter) ± SD
	Day 1	Day 3	Day 1	Day 3
Sham	NAa	NA	2 ± 2	4.5 ± 1.8	1,018 ± 226	16 ± 3.2	NDb
Susceptible	2.1 ± 1.9c	3.2 ± 3.4c	4.6 ± 2.8d,e	7.6 ± 2.8e,f	1,294 ± 204f	46.7 ± 14.7g	667 ± 189
VanA	1.3 ± 1.1	1.8 ± 1.8	35.9 ± 9.9	73.6 ± 43.9h	1,564 ± 248	47.3 ± 21.4	1,193 ± 419
VanB	1.7 ± 2.0	2.1 ± 2.4	53.2 ± 25.1	48.9 ± 26.2	1,431 ± 91	34.7 ± 9.1	1,210 ± 404

^aNA, not applicable.

^bND, not done.

^cP < 0.05 for susceptible versus glycopeptide-resistant strains. For all P values given, data were compared by the Kruskal-Wallis and Mann-Whitney tests.

^dP = 0.05 for the sham group versus the susceptible group.

^eP < 0.001 for susceptible versus glycopeptide-resistant strains.

^fP = 0.02 for the sham group versus the susceptible group.

^gP = 0.0007 for the sham group versus the susceptible group.

^hP < 0.05 for the VanA strain at day 1 versus day 3.

Effect of E. faecalis strains on peritoneal cell counts in rats.

Increased peritoneal cell counts were observed on days 1 and 3 for animals infected with glycopeptide-resistant E. faecalis strains compared to those infected with the susceptible strain (Table 1). Sham-treated animals had significantly lower peritoneal cell counts than animals infected with any of the strains (P < 0.0001).

Effect of E. faecalis strains on peritoneal cytokine levels.

Peritoneal TNF-α concentrations and peritoneal IL-6 concentrations are shown in Table 1. TNF-α concentrations were significantly higher in infected rats than in sham-infected rats but were not different in rats infected with different enterococcal strains.

Effect of E. faecalis strains on serum α1-acid glycoprotein concentrations.

Animals inoculated with resistant enterococcal strains had higher serum α1-acid glycoprotein concentrations than animals inoculated with the susceptible strain (Table 1).

DISCUSSION

We observed decreased concentrations of the two transconjugant E. faecalis strains in peritoneal fluid, combined with an increase in local and systemic inflammatory responses with these organisms, relative to those with the wild-type susceptible strain. We deliberately studied a pathophysiological model without any treatment considerations. While vanA and vanB gene clusters can both be induced by glycopeptides (2), we did not attempt antibiotic induction in our experimental model. We cannot exclude an effect on virulence due to the additional genetic information contained on the elements used to introduce the VanA and VanB resistance phenotypes (23). In our attempt to clarify the relationships between virulence and insertion of genetic elements, we deliberately chose to remain as close to real life as possible. Therefore, we used genetic material commonly reported in clinical isolates rather than constructing strains containing the vanA or vanB resistance gene on a plasmid vector lacking additional genes. It is noteworthy that very few papers have addressed this issue. In addition, our data confirm that acquisition of vanA and vanB gene clusters did not have a major effect on in vitro growth curves and that the glycopeptide resistance phenotype remained stable even after animal passage, indicating no gross effect on bacterial reproduction.

Pathogenicity is a complex procedure resulting from host-microbial interactions (7). As stated by Finley et al. (8a), the definition can be divided into the ability of microbes to persist and grow in the host milieu (the definition initially chosen in the present study) and their ability to induce disease. The results of this study are disparate with regard to these two definitions, as we found increased levels of the susceptible strain in vivo but no overall effect on mortality and weight loss. Thus, no hypothesis regarding overall pathogenesis in relation to glycopeptide resistance in enterococci can be supported or rejected based on our results. However, our results are in accord with previous reports concerning the effects of antibiotic resistance on the virulence of both gram-negative and gram-positive microorganisms (1). Although it has long been thought that antibiotic resistance genes and accessory elements would engender a cost in the fitness of bacteria, the actual evidence for this being the case is, at best, modest. The cost of resistance in the virulence of enterococci has never been assessed, and therefore the discrepancies in our results might be explained by such mechanisms. Indeed, the cost in fitness could induce decreased concentrations of bacteria in the peritoneum, but the increased virulence of the pathogen could compensate for the growth defect (produced by antibiotic resistance). A bacterial strain could easily accumulate compensatory mutations which restored the original pathogenic potential either totally or in part, allowing it to manifest such traits as induction of inflammatory responses (1, 16). On the other hand, the growth rate assessed in vitro did not demonstrate any major differences among the strains, suggesting that the cost of resistance is not of major importance in these strains, especially with VanA.

Despite decreased peritoneal counts of glycopeptide-resistant enterococcal strains relative to those of the susceptible strain on days 1 and 3 of the study, similar frequencies of bacteremia were observed. However, because blood cultures were only qualitatively analyzed, we cannot exclude a possible decrease in concentrations of bacteria in the blood. A possible decrease in bacterial growth could explain these differences. However, we did not observe any significant difference in the in vitro bacterial growth of glycopeptide-resistant strains in the present study.

Increased inflammatory responses and increased peritoneal fluid leukocyte counts were observed in animals infected with the glycopeptide-resistant strains. Two of the three markers of inflammation assessed in our study, peritoneal cell counts and α1-acid glycoprotein, reflect this systemic and local inflammatory response. α1-Acid glycoprotein is one of the major acute-phase proteins in rats and various other species. In vitro and in vivo studies suggest that α1-acid glycoprotein exerts natural antiinflammatory and immunomodulating effects, especially when its concentration increases rapidly, induced by acute-phase mediators (9). In contrast, no differences in proinflammatory cytokine levels in peritoneal fluid evaluated at 6 h after bacterial challenge were observed among the various enterococcal strains. In a previous investigation with this rat peritonitis model, Montravers et al. did not report any increase in peritoneal IL-6 and TNF-α concentrations beyond 12 h after inoculation (18). Compared to the usual kinetics of these mediators reported by Walley et al., this 6-hour interval appears to be appropriate for evaluating the peak of cytokine release (28). In the present study, no within-strain differences in peritoneal cytokine concentrations were observed despite increased peritoneal cell counts in animals that had been given transconjugant microorganisms. Several mechanisms could be involved in such features. For instance, lipoteichoic acid alone (4), as well as heat-inactivated enterococci (18), is able to stimulate production of proinflammatory cytokines, such as IL-6 and TNF-α, involved in PMN recruitment. However, TNF-α may not be the only trigger of PMN migration into the peritoneal cavity. Recently, it has been shown that intraperitoneal inoculation of E. faecalis alone failed to induce a detectable systemic and local TNF-α response (22).

Structural differences in muropeptide production between susceptible and glycopeptide-resistant strains could be involved in the different patterns of inflammatory response to infection, as certain experimental data involving various bacteria suggest a major role for peptidoglycan in the pathophysiology of host reactions to infection (6, 27). In vitro, the Streptococcus pyogenes peptidoglycan enhances the production by leukocytes of reactive oxygen species involved in intracellular microbial killing (24). Similarly, muramyl dipeptides, injected intraperitoneally into guinea pigs, increase the number of cells and the proportion of PMN entering the peritoneal cavity (21).

In conclusion, our findings suggest some variations in E. faecalis virulence in cases of genotypic inclusion in a rat polymicrobial model of peritonitis. Slightly decreased bacterial proliferation and increased peritoneal and systemic inflammatory responses were observed for two transconjugant E. faecalis strains. The relevance to clinical practice of the complex relationships between genotypic inclusion and virulence remains to be assessed. Additional investigations are required in order to clarify the role of glycopeptide resistance genes and of additional genetic material in these mechanisms.