Severe Sepsis Facilitates Intestinal Colonization by Extended-Spectrum-β-Lactamase-Producing Klebsiella pneumoniae and Transfer of the SHV-18 Resistance Gene to Escherichia coli during Antimicrobial Treatment

Abstract

Infections caused by multidrug-resistant pathogens are frequent and life threatening in critically ill patients. To investigate whether severe sepsis affects gut colonization by resistant pathogens and genetic exchange between opportunistic pathogens, we tested the intestinal-colonization ability of an extended-spectrum beta-lactamase-producing Klebsiella pneumoniae strain carrying the SHV-18 resistance gene and the transfer ability of the resistance gene to endogenous Escherichia coli under ceftriaxone treatment in rats with burn injury only or severe sepsis induced by burns plus endotoxin exposure. Without ceftriaxone treatment, the K. pneumoniae strain colonized the intestine in both septic and burned rats for a short time, with clearance occurring earlier in burn-only rats but never in sham burn rats. In both burned and septic rats, the colonization level of the challenge strain dropped at the beginning and then later increased during ceftriaxone treatment, after which it declined gradually. This pattern coincided with the change in resistance of K. pneumoniae to ceftriaxone during and after ceftriaxone treatment. Compared with burn-only injury, severe sepsis had a more significant effect on the change in antimicrobial resistance to ceftriaxone. Only in septic rats was the resistance gene successfully transferred from the challenge strain to endogenous E. coli during ceftriaxone treatment; the gene persisted for at least 4 weeks after ceftriaxone treatment. We concluded that severe sepsis can facilitate intestinal colonization by an exogenous resistant pathogen and the transfer of the resistance gene to a potential endogenous pathogen during antimicrobial treatment.

INTRODUCTION

The emergence of multidrug-resistant (MDR) bacteria is a life-threatening event in critically ill patients in the intensive care unit (ICU), and the emergence of MDR bacteria may lead to increases in mortality, cost, and the length of the hospital stay (1, 2). The gastrointestinal (GI) tract is recognized as an important reservoir of antibiotic-resistant bacteria, including Enterobacteriaceae spp., Pseudomonas aeruginosa, Acinetobacter baumannii (3), Enterococcus spp., and Klebsiella pneumoniae (4). The selective pressure exerted by antibiotics plays a particularly important role in colonization by and spread of MDR bacteria. Antibiotics may inhibit colonization by susceptible pathogens but may select antimicrobial-resistant strains and give these resistant strains an advantage in the colonization of the GI tract (3, 5–8). The GI tract is also an important reservoir involved in the transmission of antibiotic resistance genes (9). Several studies have demonstrated that antibiotics facilitate the transfer of antibiotic resistance genes between different species of bacteria in the GI tract (8–10). However, all of those studies were performed in healthy animals, which may not recapitulate the complicated conditions associated with critical illnesses, specifically severe sepsis. It is unclear whether sepsis affects the colonization pattern of possible exogenous pathogens that carry resistance genes under selective antibiotic pressure and whether sepsis affects the transferability of antimicrobial resistance genes to potentially pathogenic endogenous bacteria. One clinical study reported that the intestinal microbiota changed significantly after a sudden insult even before the administration of antibiotics (11). Another clinical study showed that the frequency of antimicrobial resistance in Escherichia coli was significantly higher in ICU patients during hospitalization, but not in general ward patients (12). Apart from the more frequent use of broad-spectrum antibiotics in ICU patients, resulting in the development of resistance, it is possible that severe sepsis itself may play an additional role in changes in antimicrobial resistance in ICU patients.

In this study, we used a rat model of two-hit sepsis to determine whether sepsis affects intestinal colonization by an exogenous opportunistic pathogen carrying an antibiotic resistance gene under selective antibiotic pressure and to assess the transferability of this antibiotic resistance gene to endogenous bacteria in the GI tract.

We hypothesized that two-hit sepsis caused by burns and endotoxin exposure would facilitate intestinal colonization by K. pneumoniae ATCC 700603, a strain carrying the bla_SHV-18 gene, and the transfer of the SHV-18 resistance gene to endogenous E. coli in the GI tract during ceftriaxone treatment.

MATERIALS AND METHODS

This study was approved by the Medical Ethics Committee of The Second Military Medical University and was in compliance with The Guide for the Care and Use of Laboratory Animals prepared by the Institute of Laboratory Animal Resources and published by the National Institutes of Health.

Preparation of the K. pneumoniae challenge strain culture.

K. pneumoniae strain ATCC 700603, which produces the extended-spectrum beta-lactamase (ESBL) SHV-18 (13), was selected as the challenge strain and was purchased from the Shanghai Center for Clinical Laboratory, Shanghai, China. The strain was verified based on the phenotypic characteristics of ESBL and PCR detection of the bla_SHV-18 gene (14). This strain is intermediate resistant to ceftriaxone (13). The bacterium was diluted to yield a 1 × 10⁸-CFU/ml suspension in broth culture on the day of gastric inoculation.

Preparation of the burn and two-hit sepsis model.

Forty male Sprague-Dawley rats weighing 220 to 250 g were obtained in two batches at an interval of 1 month from the Shanghai Laboratory Animal Research Center. They were individually caged and fed commercial rodent chow in our animal facility for 1 week to allow acclimatization. The animals were then fasted with free access to drinking water for 12 h before the experiment. The two-hit sepsis model was prepared as previously described (15, 16). Briefly, the animals were anesthetized with 1% pentobarbitone at 40 mg/ml after they had been weighed and their dorsa had been shaved. The rats were fixed in place and immersed in boiling water for 15 s to produce a 30% full-thickness dorsal burn injury. The animals were then resuscitated with saline at a dose of 100 ml/kg of body weight intraperitoneally. For the next 24 h and the remainder of the experiment, the rats were individually caged, and the cages were handled in a sterile manner. All rats that received only the burn injury consumed food and defecated normally after regaining consciousness from anesthesia and survived uneventfully. The burn-only injury may therefore represent a less severe systemic inflammatory response and stress. The animals randomly assigned to the sepsis model were intraperitoneally administered endotoxin at a dose of 10 mg/kg (E. coli O111B4 lipopolysaccharide [LPS]; Sigma Company, St. Louis, MO, USA) (17, 18) for the second hit 1 day after the initial burn injury to further increase the plasma catecholamine level (19) and substantially amplify the inflammatory response (20). The intensified inflammatory response was indicated by inactivity and anorexia of the rats within at least 24 h after surgery. The two-hit sepsis model is widely considered to reflect the clinical scenario of most ICU patients (16, 21–22). The animals randomly assigned to the burn model were given an identical volume of saline intraperitoneally.

Study protocol and grouping.

Forty rats were randomly assigned into five groups of eight animals each. The groups were defined by the experimental conditions (burn injury only or burn injury followed by endotoxin challenge) and the treatment (ceftriaxone or saline), and a sham burn group was used as a negative control (Fig. 1). (i) In the sepsis-treat group, animals received endotoxin 24 h after burn injury. Orogastric gavage with 1 ml of the K. pneumoniae challenge strain (1 × 10⁸ CFU/ml) was performed 2 h later, followed by 9 days of treatment with ceftriaxone at a dose of 60 mg/kg/day by intraperitoneal injection at 12-h intervals. The dose of ceftriaxone was based on the daily dose recommended for human adults. (ii) In the sepsis-control group, the animals were subjected to procedures similar to those for the sepsis-treat group, but an identical volume of saline was injected intraperitoneally instead of ceftriaxone. (iii) In the burn-treat group, the rats were subjected to procedures similar to those for the sepsis-treat group, but an identical volume of saline was injected intraperitoneally instead of endotoxin. (iv) In the burn-control group, neither endotoxin nor ceftriaxone was given to the animals in the group. Instead, identical volumes of saline were injected intraperitoneally in place of endotoxin and ceftriaxone. (v) In the sham-control group, the animals were subjected to a sham burn procedure. No burn injury was given to the animals in this group after shaving the dorsa, and neither endotoxin nor ceftriaxone was given to the animals in the group. To assess intestinal colonization, fresh stool samples were collected at baseline (before model setup, denoted BL); 1 day after K. pneumoniae inoculation (denoted I1); 1, 3, 6, and 9 days after ceftriaxone or saline treatment (denoted T1, T3, T6, and T9, respectively); and at 2, 7, 14, 21, and 28 days after the end of ceftriaxone treatment (denoted W2, W7, W14, W21, and W28, respectively).

FIG 1.

Experimental protocol and grouping. SD, Sprague Dawley; ETX, endotoxin; I.P., intraperitoneal; K.P., K. pneumoniae. Time zero on the timeline represents the day of orogastric inoculation with K. pneumoniae, and the symbols indicate the preset sampling times. d, days.

Quantification of stool organisms.

Fresh stool specimens were collected from all of the animals on preset dates using established sterile procedures. After collection and weighing, the samples were diluted 10 times with saline to produce stock suspensions. Appropriate 10-fold serial dilutions of stock suspensions were performed. At each level of dilution, three aliquots of 100 μl of diluted suspension were plated separately onto three MacConkey culture dishes. After overnight culture in a CO₂ incubator at 37°C, Enterobacteriaceae were selected for further cultivation based on Gram staining and colony morphology. A Vitek 2 compact identification system (bioMérieux, Marcy l'Etoile, France) was used for bacterial identification and susceptibility testing (control strains included E. coli ATCC 25922 and P. aeruginosa ATCC 27853). K. pneumoniae and E. coli were counted separately. The dilution level that yielded 20 to 100 colonies on each plate was taken as the valid dilution for counting. The density of bacteria was expressed as log₁₀ CFU/g stool and was quantified using the following formula: bacterial count (CFU/g) = [(total number of colonies in three dishes × dilution level)/weight of specimens]/3. The lower limit of detection was 2.56 log₁₀ CFU/g stool.

ESBL phenotype confirmation and ceftriaxone susceptibility testing.

We randomly selected zones that contained 10 to 20 colonies of E. coli or K. pneumoniae on the plate at the valid dilution level and purified all of the target colonies in these zones. Following the National Committee for Clinical Laboratory Standards (NCCLS) methods (14), we confirmed the ESBL phenotypes of the clones using the double-disk synergy test. A 5-mm increase in the inhibition zone diameter for cefotaxime (30 μg) or ceftazidime (30 μg) tested in combination with clavulanic acid (10 μg) relative to the diameter when the antibiotic was tested alone was considered indicative of ESBL production (i.e., the presence of a clavulanic acid effect). The quality control strains were E. coli ATCC 25922 (negative control) and K. pneumoniae ATCC 700603 (positive control).

Detection and confirmation of the antibiotic resistance gene bla_SHV-18.

Plasmids were extracted from purified and amplified colonies of E. coli and K. pneumoniae with the Qiagen Plasmid Mini Kit (Qiagen, Germany) following the instructions of the manufacturer. PCR was used to determine if bla_SHV-18 was present. The oligonucleotide primers designed to amplify the gene were 5′-AGAATAGCGCTGAGGTCTG-3′ (forward) and 5′-AGCGCGAGAAGCATCCTG-3′ (reverse). The amplification was performed in a mixture of 50 μl, including 0.5 μl of the upstream primer (25 pmol/μl), 0.5 μl of the downstream primer (25 pmol/μl), 25 μl of 2× KOD FX buffer (containing Mg²⁺), 1 μl of deoxynucleoside triphosphates (dNTPs) (2 mmol each), 1 μl of KOD FX enzyme, 2 μl of template DNA, and 11 μl of double-distilled H₂O (ddH₂O). The reaction was performed as follows: 94°C for 5 min; 35 cycles of 98°C for 10 s, 62°C for 30 s, and 68°C for 1 min 20 s; and 72°C for 5 min. The purified PCR product was confirmed to be a 1,369-bp gene segment by agarose gel electrophoresis analysis in 2% agarose using 1× Tris-acetate-EDTA buffer and a 250-bp ladder (TaKaRa, Japan) size standard. The purified PCR product extracted and amplified from K. pneumoniae ATCC 700603 was used as a positive control, which is indicative of SHV-18 gene positivity (13). The purified PCR product was also sent to Shanghai Sangni Bio-tech Company for two-way sequencing (the primer sequences were the same as those of the PCR primers), and the resulting sequence was compared with the nucleotide sequence of bla_SHV-18 (accession number AF132290) downloaded from GenBank (http://www.ncbi.nlm.nih.gov/nuccore/AF132290).

Statistical analysis.

All data were analyzed with SPSS 18.0 statistical software (SPSS Inc., Chicago, IL, USA). The continuous variables were expressed as the means ± standard deviations, and differences among the study groups were compared by analysis of variance (ANOVA). The least-squares method was used to identify differences between pairs of groups. The categorical variables were compared by the chi-square test and Fisher's direct probability method. Statistical significance was defined as a P value of <0.05. Because no challenge strain was detected at any time in the sham-control group, the group was not included in the statistical analysis.

RESULTS

Effects of burn-only injury and two-hit sepsis on intestinal colonization by exogenous ESBL-producing K. pneumoniae and on the growth of endogenous E. coli under or not under selective pressure from ceftriaxone treatment.

The challenge strain was not detected in any group at baseline, before the burn injury, or in any of the sham-control animals at any preset time point (Fig. 2). Twenty-four hours after injury and inoculation, K. pneumoniae readily colonized all 16 septic rats (including the sepsis-treat and sepsis-control groups) at 8.54 ± 0.96 log₁₀ CFU/g stool and all 16 burn-only rats (including the burn-treat and burn-control groups) at 5.36 ± 0.84 log₁₀ CFU/g stool (P < 0.001 compared to the level in septic rats). In the absence of ceftriaxone treatment, the density of K. pneumoniae in both the sepsis-control and burn-control rats declined quickly to an undetectable level within 4 days after inoculation and remained undetectable thereafter throughout the study period (P < 0.001 for both compared with levels 1 day after K. pneumoniae inoculation). The clearance of K. pneumoniae from the GI tract in septic rats lagged behind that in burn-only rats (P < 0.001, comparing the K. pneumoniae level 2 days after inoculation between the sepsis control group and the burn control group) (Fig. 2). With respect to the density of endogenous E. coli, there were no significant differences at baseline among all groups. In the sepsis-control animals, the E. coli density was ∼6 log₁₀ CFU/g stool at baseline and significantly increased to 7.8 log₁₀ CFU/g stool 1 day after injury (P < 0.001 versus baseline). The E. coli density then gradually declined to baseline level within 9 days. The E. coli density in the burn-control animals exhibited almost no change (Fig. 2). These results indicate that both burn-only injury and severe sepsis may facilitate transient colonization of exogenous K. pneumonie, and severe sepsis may facilitate transient overgrowth of commensal E. coli.

FIG 2.

Densities of K. pneumoniae and E. coli in the stool during and after ceftriaxone treatment in burn-only or severe-sepsis rats. CRO, ceftriaxone; SLN, saline; BL, baseline. T1, T3, T6, and T9 represent 1, 3, 6, and 9 days after continuous CRO treatment, respectively. W2, W7, W14, W21, and W28 represent 2, 7, 14, 21, and 28 days after the end of CRO treatment. The error bars represent the standard deviations. The dashed lines represent the detection limit (2.56 log₁₀ CFU/g stool). xx, P < 0.01, comparing the sepsis-control group with the burn-control group; ††, P < 0.01, comparing the sepsis-treat group with the burn-treat group.

In the burn-treat group, the density of K. pneumoniae rapidly declined to an undetectable level 1 day after ceftriaxone treatment (2 days after inoculation) and remained undetectable at 3 and 7 days after treatment. On day 9 after treatment, K. pneumoniae was detected in two of eight rats, but its level was not statistically different from that in burn control animals. After the withdrawal of ceftriaxone, the density of K. pneumoniae rapidly increased to 8.35 ± 0.73 log₁₀ CFU/g stool 2 days later (P < 0.001 compared with the density on day 9 after treatment) but gradually declined thereafter to 3.32 ± 1.06 log₁₀ CFU/g stool by the end of the study 4 weeks later (P < 0.001 compared with the density on day 2 after the withdrawal of ceftriaxone) (Fig. 2).

In the sepsis-treat group, the density of K. pneumoniae reached its lowest detectable level of 4.28 ± 1.39 log₁₀ CFU/g stool 1 day after ceftriaxone treatment. This density was significantly higher than that in the burn-treat group (<2.56 log₁₀ CFU/g stool; P = 0.004) but was lower than that in the sepsis-control group (6.09 ± 1.56 log₁₀ CFU/g stool; P = 0.003). By the end of 9 days of ceftriaxone treatment, the density of K. pneumoniae had gradually increased to 6.40 ± 2.80 log₁₀ CFU/g stool, a density significantly higher than that in the burn-treat group (2.96 ± 0.46 log₁₀ CFU/g stool; P < 0.001) at the same time. After the withdrawal of ceftriaxone treatment, the density of K. pneumoniae in the stool gradually declined to 3.33 ± 0.91 log₁₀ CFU/g stool in a manner similar to that in the burn-treat group. At 2 weeks after the end of treatment and thereafter, there was no difference in the density of K. pneumoniae between the burn-treat group and the sepsis-treat group (Fig. 2). These results indicate that selective pressure from ceftriaxone greatly facilitates the colonization of ESBL-producing K. pneumoniae and that removal of this pressure should thus facilitate the clearance of exogenous K. pneumoniae in the current animal model.

In the burn-treat group, the density of E. coli rapidly declined to an undetectable level 1 day after ceftriaxone treatment and remained undetectable during the 9-day treatment period. Then, the density of E. coli increased to a peak level of 8.47 ± 0.51 log₁₀ CFU/g stool 1 week after the end of ceftriaxone treatment, after which the density gradually returned to the baseline level over the following 3 weeks (Fig. 2).

In the sepsis-treat group, the density of E. coli in the stool also decreased to an undetectable level soon after ceftriaxone treatment. However, unlike in the burn-treat group, E. coli was detected in four of eight rats on day 9 of ceftriaxone treatment, and the density reached a level of 5.07 ± 2.70 log₁₀ CFU/g stool (P = 0.293 compared with baseline level). After the end of ceftriaxone treatment, the density gradually returned to the baseline level within the following 4 weeks, but there was no increase to a peak level, as observed for the burn-treat group (Fig. 2). These results indicate that endogenous E. coli responded well to ceftriaxone treatment in the burn-only rats, but not in the septic rats, in the later stage of treatment.

Change in the ceftriaxone resistance of exogenous K. pneumoniae and endogenous E. coli during and after ceftriaxone treatment.

As the challenge strain, ESBL-producing K. pneumoniae successfully colonized all of the rats 24 h after either burn-only injury or burn-plus-endotoxin injury. All K. pneumoniae isolates had intermediate resistance to ceftriaxone before treatment (Fig. 3).

FIG 3.

Change in the levels of ceftriaxone resistance of K. pneumoniae isolates during and after ceftriaxone treatment in burn-only and severe-sepsis rats. CRO-I, intermediate resistant to ceftriaxone; CRO-R, resistant to ceftriaxone. xx, P < 0.01, comparing the number of rats harboring both intermediate resistant and resistant isolates of K. pneumoniae between the burn-treat group and the sepsis-treat group. †, P < 0.05, comparing the number of rats harboring resistant isolates of K. pneumoniae between the burn-treat group and the sepsis-treat group.

In the burn-treat group, no isolates of K. pneumoniae were detected in the stool until day 9 of ceftriaxone treatment. At this time, K. pneumoniae was detected in five rats, four of which carried ceftriaxone-resistant isolates. The number of rats harboring ceftriaxone-resistant K. pneumoniae reached a peak of five on day 2 after the end of ceftriaxone treatment (P = 0.031 compared with day 1 after K. pneumoniae inoculation) and then gradually decreased to zero rats after 4 weeks (Fig. 3). In addition, the total number of rats harboring both ceftriaxone-resistant and ceftriaxone-intermediate K. pneumoniae also declined gradually, nearly coinciding with the change in the number of rats harboring ceftriaxone-resistant K. pneumoniae (Fig. 3).

In the sepsis-treat group, resistant isolates of K. pneumoniae were identified starting on day 3 of ceftriaxone treatment, 6 days earlier than in the burn-treat group. After 6 days of ceftriaxone treatment, 5 rats in the sepsis-treat group carried ceftriaxone-resistant K. pneumoniae, but no rats in the burn-treat group carry ceftriaxone-resistant K. pneumoniae (P = 0.031). Furthermore, after 9 days of ceftriaxone treatment, all 8 rats carried ceftriaxone-resistant K. pneumoniae (P < 0.001 compared with day 1 after K. pneumoniae inoculation). After the end of ceftriaxone treatment, both the number of rats harboring ceftriaxone-resistant K. pneumoniae and the number of rats harboring K. pneumoniae of any type declined gradually. However, one septic rat still carried ceftriaxone-resistant K. pneumoniae until the end of the observation period (Fig. 3). These results indicate that selective pressure from ceftriaxone may increase the resistance of ESBL-producing K. pneunoniae. Severe sepsis might intensify the change in antimicrobial resistance.

Before ceftriaxone treatment, all of the E. coli isolates in all groups were sensitive to ceftriaxone. In the burn-control group and the sepsis-control group, endogenous E. coli was detected in all the animals without ceftriaxone administration at all time points throughout the study period. In the burn-treat group, no E. coli isolates were detected during the 9 days of ceftriaxone treatment (Fig. 4). After the end of ceftriaxone treatment, E. coli was soon detected and temporarily overgrew to a peak level that was 2 log₁₀ units higher than baseline in this group (Fig. 2). No ESBL-producing E. coli isolates that had intermediate resistance to ceftriaxone were detected in the burn-treat group (Fig. 4). In the sepsis-treat group, E. coli was detected again on day 9 of ceftriaxone treatment after at least 7 days of absence in the stool. The E. coli isolates detected in four rats were all ESBL producers (Fig. 4). After the end of ceftriaxone treatment, E. coli isolates with intermediate resistance to ceftriaxone were detected in the same four rats and in one other rat at the subsequent time points until the end of the study (Fig. 4). Therefore, a total of 5 rats were detected to carry ceftriaxone-intermediate E. coli in the sepsis-treat group from 2 weeks after the end of ceftriaxone treatment to the end of the observation period (P = 0.031 for both compared with those at the same time in the burn-treat group and compared with the baseline level).

FIG 4.

Change in the levels of ceftriaxone resistance of isolated E. coli during and after ceftriaxone treatment in burn-only and two-hit sepsis rats. CRO-S, sensitive to ceftriaxone. †, P < 0.05, comparing the number of rats harboring intermediate resistant isolates of E. coli between the burn-treat group and the sepsis-treat group.

Detection and confirmation of the SHV-18 antimicrobial resistance gene in E. coli and K. pneumoniae isolates.

All E. coli isolates at baseline were confirmed not to contain the SHV-18 gene, as verified by PCR. All of the E. coli isolates with intermediate resistance to ceftriaxone carried the SHV-18 gene, as verified by PCR. These results indicate that the SHV-18 resistance gene was successfully transferred from ESBL-producing K. pneumoniae to endogenous E. coli during ceftriaxone treatment and that E. coli isolates that received the resistance gene persisted in the gut for at least 4 weeks after ceftriaxone treatment. All K. pneumoniae isolates and ceftriaxone-intermediate E. coli isolates were confirmed to contain a gene segment of 1,369 bp by agarose gel electrophoresis, which indicated that the isolates were positive for the SHV-18 gene (13). The gene sequences of four randomly selected PCR products that were determined to be SHV-18 gene positive by agarose gel electrophoresis analysis were further confirmed to be identical to the original sequence of the SHV-18 gene.

DISCUSSION

In this study, we established a rat model of severe sepsis using a burn injury and endotoxin challenge to investigate the effect of sepsis on intestinal colonization by ESBL-producing K. pneumoniae and on the transfer of the SHV-18 resistance gene to endogenous E. coli under selective pressure from ceftriaxone. Our main findings were as follows. Endogenous E. coli in the GI tract overgrew transiently in rats with burn injury plus ceftriaxone-induced sepsis. Exogenous ESBL-producing K. pneumoniae successfully colonized the GI tract in both burned and septic rats but was cleared in several days without ceftriaxone treatment. In both burned and septic rats, the intestinal-colonization level of ESBL-producing K. pneumoniae decreased at the beginning and then increased during ceftriaxone treatment. After ceftriaxone treatment, the colonization level decreased gradually. This change was coincident with the change in the resistance of K. pneumoniae to ceftriaxone during and after ceftriaxone treatment. In comparison with burned rats, septic rats experienced an earlier appearance of the increase in antimicrobial resistance to ceftriaxone during ceftriaxone treatment and a later recovery after the end of ceftriaxone treatment. Only in septic rats was the ESBL-producing SHV-18 gene successfully transferred from exogenous K. pneumoniae to endogenous E. coli during ceftriaxone treatment; the gene persisted for at least 4 weeks after the end of ceftriaxone treatment. Correspondingly, the resistance of E. coli to ceftriaxone increased from sensitive to intermediate. In burn-only rats, neither the change in E. coli resistance to ceftriaxone nor the transfer of the SHV-18 gene was observed during ceftriaxone treatment.

Wang et al. (23) showed that burn injury led to a significant decrease in the number of bifidobacteria but significant increases in E. coli and fungi in the rat gut mucosa, accompanied by elevations of the levels of plasma endotoxin and interleukin 6 (IL-6). We observed gut overgrowth of E. coli in septic rats, but not in burn-only rats. This result may be explained by the different severities of the insults between the two studies. It is unclear how sepsis induced the overgrowth of endogenous E. coli. One possible explanation is that stress may have induced a transient but significant release of catecholamine, which has been reported to enhance the adherence of E. coli to the cecum and colon (24, 25), to stimulate the growth of intestinal commensal E. coli (26), and to enhance the virulence of enterohemorrhagic E. coli (27). Endotoxin may have an additional effect on the adherence of E. coli to the surface of the intestinal mucosa (28).

Our data indicated that exogenous K. pneumoniae successfully colonized the burn-only rats and the septic rats, but not the rats in the sham group. Our data suggest that the host gut is more vulnerable to possible exogenous pathogens under conditions of stress and sepsis. Several studies may help explain this statement. Shimizu et al. (29) showed that the gut flora and environment are significantly altered in patients with severe systemic inflammatory response syndrome (SIRS). A dramatic change in the gut flora may occur immediately after severe and sudden insults, even before the administration of antibiotics (11). Bailey et al. (30) showed that the alteration of the intestinal microbiota by stressor exposure led to increased colonization by an exogenous pathogen. The disruption of the normal intestinal microbiota has long been known to be one of the risk factors for decreased resistance to colonization (31). Immunosuppression caused by endotoxin challenge may also have facilitated the intestinal colonization by exogenous K. pneumoniae observed in this study. For example, Wang et al. (23) observed endotoxemia and decreased secretory IgA (sIgA) levels in the mucus of the small intestine after burn injury. Davis et al. (32) suggested that sepsis-induced immunosuppression increases susceptibility to Candida infection. However, the immunosuppression following the initial septic insult would resolve over several days without further insult (33). The return of the growth level of endogenous E. coli to baseline without ceftriaxone treatment in this study indicates that the normal microbiota was restored. This restoration of the normal microbiota may help explain why the level of colonization by exogenous K. pneumoniae quickly decreased to below the detection limit.

This study revealed several interesting features of intestinal colonization by and resistance dynamics of endogenous E. coli and exogenous ESBL-producing K. pneumoniae during and after selective antibiotic pressure in burn-only rats and septic rats. The K. pneumoniae strain ATCC 700603 that we used in this study produces ESBL and is intermediate to ceftriaxone (13). Endogenous E. coli is susceptible to ceftriaxone. After receiving the SHV-18 resistance gene, E. coli developed intermediate resistance to ceftriaxone, as demonstrated by Rasheed et al. (13).

In both the burn-only rats and the septic rats, ceftriaxone treatment resulted in an immediate and significant decrease in K. pneumoniae colonization at first. Colonization by E. coli exhibited a very similar response to ceftriaxone treatment. This initial decrease reflects the suppressive effect of ceftriaxone on ESBL-producing K. pneumoniae and E. coli. Ceftriaxone has already been demonstrated to have a profound suppressive effect on the intestinal microbiota (34). Our data are also in accordance with the clinical study performed by Filius et al. showing that the prevalence of intestinal colonization by Gram-negative bacteria was lowest during hospital stays both in the ICU and in the general ward population (12); the colonization rates were inversely related to the measured levels of antibiotic use (12). Furthermore, Filius' study showed that the suppression of the normal intestinal flora was more pronounced in the ICU population than in the general ward population, manifesting as a decreased frequency of E. coli and increased frequencies of Klebsiella spp. and P. aeruginosa. Our study also demonstrated that colonization by ESBL-producing K. pneumoniae in septic rats was not suppressed as much as in burn-only rats, resulting in a higher K. pneumoniae-to-E. coli ratio in the gut. Sepsis-induced immunosuppression and microbiota disturbances may reduce the effect of ceftriaxone on exogenous K. pneumoniae.

Our study further demonstrated that prolonged ceftriaxone treatment finally led to the loss of the suppressive effect of ceftriaxone on exogenous ESBL-producing K. pneumoniae in both septic and burn-only rats and on endogenous E. coli in septic rats. After the discontinuation of ceftriaxone treatment, the colonization levels of both K. pneumoniae and E. coli gradually returned to baseline levels within 4 weeks. The dynamic changes in intestinal colonization by K. pneumoniae and E. coli closely coincided with the dynamic changes in their levels of resistance to ceftriaxone. Increased K. pneumoniae resistance to ceftriaxone occurred earlier during ceftriaxone treatment but subsided later in septic rats than in burn-only rats after the discontinuation of ceftriaxone. It seems that sepsis has a more pronounced effect on the changes in bacterial resistance induced by ceftriaxone treatment. Clinical studies have observed a higher prevalence of antimicrobial resistance among ICU patients than among non-ICU patients (35). The higher level of antibiotic use in ICU patients may be one of the reasons for this finding. However, it is not well documented whether antibiotic use affects resistance at the individual level during antibiotic treatment and whether sepsis plays an additional role in the process. Our data clearly demonstrate that there was a resistance change at the individual level during ceftriaxone treatment and that sepsis had an additional effect during this process. To our knowledge, the underlying mechanisms have not been reported in the current literature. Further studies are needed in the future. Also, to our knowledge, no other recent study has measured dynamic intestinal colonization and antibiotic resistance simultaneously at the individual level during and after antibiotic treatment.

The intestinal tract is an important reservoir for bacteria carrying resistance genes and is a source for the transmission of resistance genes between bacteria of the same and different species (9, 36–39). Inappropriate antibiotic use facilitates the conjugative transfer of resistance genes (8–10, 40) by increasing intestinal colonization by resistant bacteria, which serve as donors of the resistance gene. In this study, we detected increased E. coli resistance to ceftriaxone in 5 of the 8 septic rats but in none of the burn-only rats. The E. coli isolates with increased resistance to ceftriaxone were all confirmed to harbor the ESBL-producing SHV-18 gene by PCR sequencing, suggesting the successful transfer of the SHV-18 gene from the exogenous K. pneumoniae to the endogenous E. coli. No evidence was found in the current literature showing that sepsis has a direct impact on the transferability of resistance genes. The most plausible explanation may be that in comparison with burn-only injury, two-hit sepsis has a more pronounced effect on intestinal colonization by ESBL-producing K. pneumoniae, therefore yielding a sufficient donor pool for the resistance gene to accomplish the transfer. Another possible explanation is that sepsis induced by endotoxin significantly increases plasma catecholamine concentrations (19). Catecholamine at physiological concentrations can enhance the efficiency of horizontal gene transfer between enteric bacteria (41).

Our study also showed that after the discontinuation of ceftriaxone treatment, resistant E. coli harboring SHV-18 could persist in the gut for at least 4 weeks. Our findings are consistent with those from previous animal (9) and clinical (12) studies. Persistent intestinal colonization by endogenous flora carrying resistance genes would increase the likelihood of the transmission of resistance to the community (42) and of future infection by resistant pathogens (43, 44). Clinicians caring for patients who have had recent ICU stays should consider previous colonization by resistant pathogens and their antibiotic resistance spectra.

This study has several limitations. First, we used a single burn injury and a single dose of endotoxin to produce the sepsis model. The inflammatory response may occur early and subside early. This pattern may differ from the typical clinical sepsis process, in which the causative agents may persist for longer periods. Second, we measured colonization by only E. coli rather than the whole intestinal microbiota, which includes anaerobic microflora. Previous studies have suggested that ceftriaxone treatment may also significantly decrease the intestinal density of anaerobes (34, 45), and intestinal anaerobic microflora are more likely to be responsible for colonization resistance (5, 6, 31). Third, the intrinsic physiological differences between rats and humans may result in disparate microbiological and pharmacological responses, thus limiting the external validity of the conclusions of this study.

In summary, severe sepsis induced by burn injury and endotoxin exposure facilitates intestinal colonization by exogenous ESBL-producing K. pneumoniae and the transfer of the SHV-18 resistance gene to endogenous E. coli during ceftriaxone treatment. E. coli isolates that have received the resistance gene persist in the gut for at least 4 weeks after the discontinuation of ceftriaxone treatment.