Transfer of Plasmid-Mediated CTX-M-9 from Salmonella enterica Serotype Virchow to Enterobacteriaceae in Human Flora-Associated Rats Treated with Cefixime

Abstract

Food animals are a potential source of CTX-M resistance genes for humans. We evaluated the transfer of the bla_CTX-M-9 gene from an animal strain of Salmonella enterica serotype Virchow to Enterobacteriaceae of the human intestinal flora by using human flora-associated (HFA) rats with and without cefixime treatment. In the absence of antibiotic, no transconjugant enterobacteria were found in the feces of HFA rats. However, the transfer rate was high if Escherichia coli J5 recipient strains were coinoculated orally with Salmonella. S. enterica serotype Virchow persisted in the rat fecal flora both during and after treatment with therapeutic doses of cefixime. The drug did not increase the transfer rate, and E. coli J5 transconjugants were eliminated from the flora before the end of cefixime treatment. No cefixime was recovered in the rat feces. In the presence of recipient strains, the bla_CTX-M-9 resistance gene was transferred from a strain of animal origin to the human intestinal flora, although transconjugant colonization was transient. Antibiotic use enhanced the persistence of donor strains, increasing the resistance gene pool and the risk of its spread.

CTX-M-type extended-spectrum β-lactamases (ESBLs) are enzymes responsible for catalyzing the hydrolysis of monobactam and extended-spectrum cephalosporins. Over the last 10 years, these β-lactamases have been identified around the world, and they now have the highest prevalence of any type of ESBL (9, 10, 14, 15, 28, 32). This dramatic increase in their frequency may be attributed to the rapid dissemination of resistant bacterial clones and to their genetic structure, with the resistance genes carried on plasmids and transposons (14, 43). The consumption of contaminated food is thought to be the principal mode of spread of ESBL-producing resistant strains to the general population (31, 38). Several epidemiological analyses have investigated possible clonal relationships between resistant animal and human strains (4, 22, 25, 30, 45). Experimental studies have generated conflicting findings concerning the transfer of resistance genes in the human intestinal tract. Bonner et al. (6) suggested that bacteria from livestock cannot persist in humans and therefore do not constitute a threat. Prescott et al. (37) agreed that conditions in the human gut do not favor the transfer of resistance genes but pointed out that continuous exposure to an antibiotic would lead to the selection of organisms best able to colonize the gut (i.e., resistant strains). However, plasmid-mediated antibiotic resistance transfer may occur in the ileum (5), and Schjørring et al. (40) recently highlighted the effect of antimicrobial treatment on horizontal gene transfer from exogenous bacteria to a susceptible strain present in the mouse intestinal flora.

From 2002 to 2003, the dual emergence in France of CTX-M-9-producing multiresistant strains of Salmonella enterica serotype Virchow from poultry sources, and to a lesser extent from humans, was reported (45). Salmonella are mostly described as food-borne pathogens leading to therapeutic difficulties, especially in children, for whom extended-spectrum cephalosporins are the treatment of choice (20). Moreover, dissemination in humans of CTX-M-9 and CTX-M-14 from food has already been highlighted in 2001 in Spain (7, 36). Since CTX-M-9-producing human isolates were still rare in France in 2003, we investigated the probability of the spread of this plasmid-borne bla_CTX-M-9 gene from one strain of S. enterica serotype Virchow isolated from poultry to enterobacteria of the normal human intestinal flora in a human flora-associated (HFA) rat model. We also assessed the effect of cefixime on the rate of resistance transfer and the persistence of resistant strains.

MATERIALS AND METHODS

Donor and recipient strains.

We studied the transfer of the plasmid-borne bla_CTX-M-9 gene from a donor strain isolated from chickens, S. enterica serovar Virchow strain 3464b (45). This strain was resistant to amoxicillin-clavulanic acid, cefixime, cefotaxime, ceftazidime, ceftiofur, cefuroxime, nalidixic acid, streptomycin, sulfamethoxazole, tetracycline, trimethoprim, and trimethoprim-sulfamethoxazole and has a rifampin MIC of 4 μg/ml. Escherichia coli J5 was used as the recipient strain (45). This strain is susceptible to cefixime and has a rifampin MIC of higher than 512 μg/ml.

Human fecal flora.

We collected human fecal flora from two donors selected from seven healthy volunteers living in Brittany, France. None had received pharmacological treatment for at least 3 months before the study, and all ate traditional European meals. We checked for the absence of cefixime- and rifampin-resistant enterobacteria and Salmonella in each donor.

Media.

Fresh human stool or HFA rat fecal samples were mixed with tryptone glucose yeast extract (AES, Bruz, France) (1%, wt/vol) under anaerobic conditions. Appropriate 10-fold serial dilutions of the supernatant were plated on selective agar media for colony counts. The compositions of the fecal aerobic and anaerobic flora were determined as described by Perrin-Guyomard et al. (34). The total numbers of enterobacteria in the fecal flora of humans and rats were determined on Drigalski agar (Fischer-Bioblock Scientific, Illkirch, France) and compared with the counts obtained on Drigalski agar supplemented with 4 μg/ml cefixime (Sigma-Aldrich, Saint-Quentin Fallavier, France) to estimate the percentage of bacteria resistant to the antibiotic. The donor S. enterica serotype Virchow and recipient E. coli J5 strains were cultured aerobically in brain heart infusion (AES, Bruz, France). S. enterica serotype Virchow bacteria were counted on Brilliant green agar (Difco, BD Biosciences, Le Pont de Claix, France) supplemented with 4 μg/ml cefixime. E. coli J5 bacteria were counted on Drigalski agar supplemented with 250 μg/ml rifampin with the addition of 4 μg/ml cefixime for E. coli J5 transconjugants. All Enterobacteriaceae, donor, and recipient strains were incubated at 37°C for 24 h to 48 h under aerobic conditions. The detection limit for bacterial counts was 2 log₁₀ CFU/g of feces. The transfer rate was defined as the number of transconjugants divided by the number of donor colonies.

Test substances.

Cefixime (Oroken, 100 mg; Sanofi-Aventis, France) was administered by gavage in a dosage regimen of 4 mg/kg twice daily for 8 days, corresponding to the highest dose of cefixime that can be administered in France (41). The antibiotic was reconstituted and stored according to the manufacturer's instructions by adding sterile water to obtain a 20-g/liter stock solution and storing it at room temperature below 25°C for 1 week. The cefixime-clavulanic acid mixture (Promochem, Molsheim, France) was administered by gavage at a dose of 4 mg/kg twice daily for 8 days. The ratio of cefixime to clavulanic acid (12.5%, wt/wt) used was consistent with that commonly used in human treatment with a combination of a β-lactam and clavulanic acid (44).

Animals.

All procedures for animal experiments were performed under license, with approval from the institutional review board, in accordance with French national legislation. Germ-free female and male consanguineous C3H rats (mean age, 3 weeks) from Charles River Laboratories, Arbresle, France, were transferred into a sterile Trexler-type plastic film isolator (Esi Flufrance, Massy, France) on arrival at our facilities. Each group of rats was kept in a different isolator, and each rat was housed individually in a cage, isolated from the litter by a floor grid. The germ-free status of animals was checked immediately after they were received and during the acclimatization period by testing fecal samples for the growth of aerobic and anaerobic bacteria and yeasts. Rats were provided with ad libitum access to a commercial diet sterilized by gamma irradiation and were supplied with sterile water.

Experimental design.

HFA rats were assigned to five groups, each containing three females and two males. These groups were named A, B, C, D, and E. After 9 days of acclimatization, all germ-free rats were inoculated intragastrically with Bacteroides fragilis ATCC 25285 (14 log₁₀ CFU/animal) to reduce oxygen and substrate concentrations in the gut. Two days later, all of the rats were inoculated intragastrically with 1 ml of a suspension of mixed human feces. After 16 days, all of the rats were inoculated intragastrically with 8 log₁₀ CFU of S. enterica serotype Virchow (day −1).

In addition, 8 log₁₀ CFU of the recipient E. coli J5 strain was inoculated intragastrically into the rats in groups C, D, and E 2 h before the donor strain S. enterica serotype Virchow. The next day (day 0), each group received antibiotic or water by gavage as follows: group A, sterile water; group B, cefixime; group C, sterile water; group D, cefixime; group E, cefixime-clavulanic acid (Table 1). Feces were collected from individual rats by provoking defecation one time before cefixime administration (day 0), five times per week during treatment, and six times after the end of treatment. Each experiment lasted 2 months.

TABLE 1.

Overview of the experimental design

Group	Donor strain	Recipient strain	Treatment	Cefixime dose (mg/kg)
A	S. enterica serotype Virchow		Water
B	S. enterica serotype Virchow		Cefixime	8
C	S. enterica serotype Virchow	E. coli J5	Water
D	S. enterica serotype Virchow	E. coli J5	Cefixime	8
E	S. enterica serotype Virchow	E. coli J5	Cefixime-clavulanic acid	8

Verification of transconjugants.

On day 0, five isolates growing on Drigalski agar containing cefixime and rifampin was arbitrarily isolated from each rat sample for groups C, D, and E. Only five bacterial colonies representing the group were identified as E. coli by PCR amplification of the uidA gene, as described by Bej et al. (2, 3). DNA was prepared from samples by adding the InstaGene Matrix kit (Bio-Rad, Marnes la Coquette, France) to the bacterial suspension according to the manufacturer's instructions. The transconjugants were then characterized by pulsed-field gel electrophoresis (PFGE). PFGE analysis was performed with BlnI (Amersham Biosciences, Orsay, France) digestion in a CHEF-DRIII system (Bio-Rad, Marnes la Coquette, France). The running conditions were 6 V/cm at 14°C for 24 h with pulse times ramped from 10 s to 60 s.

PCR detection of the bla_CTX-M-9 gene.

A gene conferring resistance to cefixime (bla_CTX-M-9) was detected by PCR in all transconjugants. Plasmid DNA was isolated with the QIAprep Spin Midiprep kit (Qiagen, Hilden, Germany). PCR assays using primers targeting the bla_CTX-M-9 gene were carried out as described by Weill et al. (45), except that annealing was carried out for 30 s at 60°C in each cycle. The donor and recipient strains were included as positive and negative controls.

Antimicrobial drug susceptibility testing.

The donor and recipient strains and 15 transconjugants were tested for susceptibility to the antimicrobial agents amoxicillin-clavulanic acid, cefotaxime, cefoxitin, ceftazidime, ceftiofur, cefuroxime, nalidixic acid, streptomycin, sulfamethoxazole, tetracycline, trimethoprim, and trimethoprim-sulfamethoxazole by using a commercially prepared, dry panel (Trek Diagnostic Systems Ltd., East Grinstead, West Sussex, United Kingdom). Etest strips on Müller-Hinton agar (AB Biodisk, Solna, Sweden) were used to determine the MIC of cefixime. The E. coli ATCC 25922 strain was used for quality control (QC). Isolates were defined as susceptible or resistant in accordance with clinical breakpoints proposed by CLSI (12) or epidemiological cutoff values presented by EUCAST (www.eucast.org) for cefotaxime, ceftazidime, ceftiofur, cefoxitin, and cefuroxime.

Determination of cefixime levels in fecal samples by the LC-MS method.

Cefixime concentrations in fecal samples were quantified by the liquid chromatography-mass spectrometry (LC-MS) method as previously described (29). A validation study was performed in which QC rat feces samples containing known amounts of cefixime were compared with standard curves. The interassay coefficients of variation were 10.69% and 2.67% for QC samples containing 2 and 50 mg/kg, respectively, of the drug. The intra-assay coefficients of variation were 10.69% and 1.36% for QC samples containing 2 and 50 mg/kg, respectively. The lower limit of quantification was 2.18 mg/kg.

Statistical analysis.

Statistical analysis was performed with Systat 12 software (Systat Software Inc., CA). The experimental unit was the rat. The effects of treatment with cefixime and cefixime-clavulanic acid were assessed by analysis of variance with interaction. The transfer rate was calculated by dividing the number of transconjugants by the number of donor colonies for each rat. The geometric mean was calculated to represent the ratio of transconjugants to donors.

RESULTS

Implantation of the human fecal flora in a rat model.

HFA rats were used to study conjugal transfer in the human gastrointestinal environment, with a complex microbiota providing the colonization barrier. Most human anaerobic and aerobic bacteria were transferred and persisted in the intestine of HFA rats after inoculation. The mean composition of the rat fecal flora was similar to that of the human fecal flora, with up to about 15 log₁₀ CFU/g of feces for the total anaerobic flora, 14 log₁₀ CFU/g for the Bacteroides fragilis group, 10 log₁₀ CFU/g for lactobacilli, 11 log₁₀ CFU/g for bifidobacteria, 10 log₁₀ CFU/g for clostridia, 11 log₁₀ CFU/g for the total aerobic flora, 10 log₁₀ CFU/g for enterococci, and 7 log₁₀ CFU/g for Enterobacteriaceae. Before the transfer experiment, the fecal flora of HFA rats was devoid of Salmonella and indigenous cefixime-resistant members of the family Enterobacteriaceae.

Transfer of the bla_CTX-M-9 gene to the intestines of HFA rats in the absence of selective pressure (groups A and C).

The Enterobacteriaceae populations in groups A and C remained stable at ∼7 log₁₀ CFU/g throughout the experiment.

In group A, the level of the donor S. enterica serotype Virchow reached 6 log₁₀ CFU/g of feces on day 0 and decreased significantly to ∼3 log₁₀ CFU/g of feces after 2 days (P < 0.001) (Fig. 1A). Eight days later, S. enterica serotype Virchow was undetectable, although it was found to have persisted in one-fifth of the HFA rats at the end of the experiment. No transconjugants of endogenous Enterobacteriaceae were detected in the feces of any of the rats during the experiment. In group C, counts of the E. coli J5 recipient strain and the S. enterica serotype Virchow donor strain were ∼4 log₁₀ CFU/g and 5.5 log₁₀ CFU/g of feces, respectively, on day 0 (Fig. 1C). E. coli J5 transconjugants appeared rapidly after the introduction of the donor strain into all of the rats (data not shown), and their level reached ∼3 log₁₀ CFU/g of feces on day 0. The transfer rate was estimated at about 7 (± 1.5) × 10⁻¹ transconjugants per donor and was much higher than the in vitro rate of 5.9 (±5.7) × 10⁻⁸ (19). On day 2, a significant decrease in the counts of the donor, recipient, and transconjugant strains was observed. As reported above, S. enterica serotype Virchow was found to have persisted in one-fifth of rats at the end of the experiment.

FIG. 1.

Bacterial counts in the feces of HFA rats inoculated with S. enterica serotype Virchow (A and B) and E. coli J5 (C, D, and E). Animals were treated with sterile water (A and C), cefixime (B and D), or both cefixime and clavulanic acid (E). Symbols: •, total Enterobacteriaceae; □, S. enterica serotype Virchow 3464b; ▴, E. coli J5; ▵, E. coli J5 transconjugants. The bar under the x axis represents the time of treatment. The values shown are mean results, and error bars represent standard deviations.

Effect of antibiotic exposure on bla_CTX-M-9 gene transfer (groups B, D, and E).

In groups B, D, and E, the mean Enterobacteriaceae counts were significantly reduced, by ∼2 log₁₀, by treatment with cefixime or cefixime-clavulanic acid (P < 0.001) (Fig. 1B, D, and E). After treatment, Enterobacteriaceae counts returned to the initial level of about 7 log₁₀ CFU/g of feces in all of the groups of rats.

In group B, no Enterobacteriaceae transconjugants were detected in fecal samples from rats treated with cefixime (Fig. 1B). The Salmonella counts in feces from all of the rats fell from about 6 log₁₀ CFU/g of feces after inoculation to ∼2 log₁₀ CFU/g at the end of the period of drug treatment. Two weeks after the end of treatment, donor strains continued to be detected in the feces of three of the five rats and bacterial counts were significantly higher than those of the rats in the control group (P < 0.03) (group A). In groups D and E, E. coli J5 and S. enterica serotype Virchow counts were ∼5 log₁₀ CFU/g of feces on day 0 (Fig. 1D and E). At the same time point, E. coli J5 transconjugant levels reached about 3 log₁₀ CFU/g of feces. Transfer rates in both groups were similar to that in the control group, at ∼7 × 10⁻¹ transconjugants per donor. Two days after the beginning of cefixime or cefixime-clavulanic acid treatment, the recipient and transconjugant populations rapidly decreased in size, becoming undetectable. Salmonella counts also decreased after drug administration in the feces of all of the rats, falling below the detection threshold in the animals in group E, whereas the donor strain level was maintained at ∼3 log₁₀ CFU/g of feces in the animals in group D. Two weeks after of the end of cefixime treatment, Salmonella were still present at countable levels in the feces of three of the five rats in this group.

Analysis of transconjugants.

The 15 transconjugants isolated from the rats in groups C, D, and E were identified as E. coli by PCR. All had PFGE profiles identical to that of the E. coli J5 strain used for inoculation (data not shown). The presence of the bla_CTX-M-9 gene was confirmed by PCR in all of the isolates. The β-lactam antibiotic susceptibility of the transconjugants was higher than that observed in the parental strain (Table 2). Cotransfer of other genes conferring resistance to nalidixic acid, streptomycin, tetracycline, trimethoprim, sulfamethoxazole, or trimethoprim-sulfamethoxazole was not observed.

TABLE 2.

Antibiotic susceptibilities of the strains used in this study

Antibiotic(s)	E. coli ATCC 25922	Donor S. enterica serotype Virchow 3464b	Recipient E. coli J5	E. coli J5 transconjugantsa
				Control	Cefixime	Cefixime-clavulanic acid
Amoxicillin-clavulanic acid	4b	8	4	16	8-32	8-16
Ampicillin	4	128	4	16	16-32	16
Cefixime	0.125	8	0.125	8	8	8
Cefotaxime	0.03	>8	0.03	0.5	0.5	0.5
Cefoxitin	4	16	4	8-32	16	16-128
Ceftazidime	0.125	8	0.125	1-4	1	1
Ceftiofur	0.25	>8	0.25	1-8	1	1
Cefuroxime	2	>32	2	16-32	8-16	16-32
Nalidixic acid	4	256	4	4	4	4
Rifampin	4	4	>512	>512	>512	>512
Streptomycin	2	128	2	2	2	2
Sulfamethoxazole	8	>512	8	4-8	4-8	4-8
Tetracycline	1	64	1	2-4	2	4
Trimethoprim	0.5	>64	0.5	0.5	0.5-2	1
Trimethoprim-sulfamethoxazole	<1	>16	<1	1	1	1

^aFive of 15 E. coli J5 transconjugants isolated from animals treated with water (control), cefixime, or cefixime-clavulanic acid were tested to determine the MICs of antibiotics.

^bMICs are expressed in μg/ml.

Determination of the cefixime fraction in fecal samples.

Despite a quantification limit of 2.18 mg/kg, cefixime was not detected in fecal samples from any of the treated HFA rats.

DISCUSSION

In our HFA rat model, we detected no bla_CTX-M-9 gene transfer from S. enterica serotype Virchow to endogenous Enterobacteriaceae. However, the addition of a recipient strain to the normal flora of HFA rats, together with the donor strain of Salmonella, rapidly led to the appearance of transconjugants containing the bla_CTX-M-9 gene in rat feces. The administration of cefixime at therapeutic concentrations did not increase the transfer of the bla_CTX-M-9 gene between S. enterica serotype Virchow and E. coli J5, and the number of transconjugants was not found to be higher when the selective pressure was removed.

The lack of gene transfer from an exogenous strain to the indigenous flora can be related to the results of Bourgeois-Nicolaos et al. (8), who found no transfer of the vanA gene from E. faecium to E. faecalis in a similar HFA model. Colonization resistance of the indigenous flora toward exogenous Salmonella has rapidly decreased the number of donor strain bacteria, whereas for transfer to occur, large numbers of both the donor and recipient strains must be present simultaneously (27). The presence of the plasmid-free strain would almost certainly have inhibited the establishment of the plasmid-bearing strain in the control group (16). Even in the presence of an exogenous recipient strain, the transfer was transient, like that which Lester et al. (27) also observed from an E. faecium strain of animal origin to an E. faecium isolate of human origin. This likely reflects an ecological disadvantage of the transconjugants relative to the recipient strain. Such disadvantages have already been observed by Johnsen et al. (24), who, using competition experiments, demonstrated that E. faecium strains with newly acquired resistance are less fit than their susceptible parental strains.

Cefixime treatment did not enhance the transfer rate, whereas Duval-Iflah et al. (17) observed the establishment of transconjugants in the dominant population and the replacement of the parental recipient strain during ampicillin administration in HFA mice, even after the period of drug administration had ended. These discrepancies may result from differences in the plasmid incompatibility group and/or the adaptability to gastrointestinal conditions of the recipient strains used in our in vivo transfer experiment and that of Duval-Iflah et al. (13, 17, 21).

Cefixime treatment nevertheless had an effect on the persistence of Salmonella in treated rats in both trials. Indeed, at the end of the experiment, CTX-M-resistant strains were found to have persisted in a larger proportion of treated HFA rats (three of five) than of rats in the control group (one of five). In the control group, the persistence of Salmonella may be attributed to the individual variability of animals. Previous studies on the selection and persistence of antibiotic-resistant Salmonella have indicated that antibiotic treatment increases the likelihood of the strain being maintained in the digestive tract (1, 33-35). As native Enterobacteriaceae may prevent the implantation of exogenous Enterobacteriaceae, the persistence of the CTX-M-producing Salmonella strain in our study may result from “substitution colonization” and proliferation of the strain in the digestive tract of rats in which the Enterobacteriaceae population had been disturbed by antibiotic treatment (23).

We detected no cefixime in HFA rat feces during treatment. However, the >2-log₁₀ decrease in Enterobacteriaceae counts during cefixime treatment in our study demonstrated that the antibiotic was present in the digestive tract of our rats. According to the pharmacokinetic study of cefixime in rats, the bioavailability of cefixime was estimated at 30% (data not shown) and comparable to the rat and human data previously described (18, 39), so the part excreted in the gastrointestinal tract may be considered approximately 70%. Based on these observations, we hypothesized that cefixime was hydrolyzed during transit by the β-lactamases-producing strains of anaerobic endogenous flora as previously described (11, 26, 42). We evaluated the impact of these β-lactamases by treating animals with cefixime together with clavulanic acid. However, cefixime was no longer recovered in the feces of rats also treated with clavulanic acid. The only effect of adding clavulanic acid was an elimination rate of Salmonella counts similar to that in the control group, with no regrowth when antibiotic treatment was ended. Cefixime seems to be hydrolyzed by the endogenous β-lactamases of the flora even in the presence of clavulanic acid. The combination of both drugs appeared, nevertheless, to be more inhibitory toward Salmonella than cefixime alone.

In summary, the lack of plasmid transfer between S. enterica serotype Virchow and the Enterobacteriaceae of HFA rats suggests that the probability of dissemination of extended-spectrum β-lactamases such as CTX-M-9 from animals to humans is very low. However, cephalosporin treatment contributes to the acquisition and overgrowth of antimicrobial-resistant pathogens in the gut, including multidrug-resistant S. enterica serotype Virchow, jeopardizing antibiotic treatment and constituting an important reservoir of resistance genes in the human digestive tract.