b-lactam Antibiotics and Gastrointestinal Colonization with Vancomycin-Resistant Enterococci

MAJOR ARTICLE b-lactam Antibiotics and Gastrointestinal Colonization with Vancomycin-Resistant Enterococci Louis B. Rice, 1,2,3 Rebecca Hutton-Thomas, 1 Viera Lakticova, 2 Marion S. Helfand, 1 and Curtis J. Donskey 1 1 Medical and Research Services, Louis Stokes Cleveland VA Medical Center, 2 The Cleveland Veterans Affairs Research and Education Foundation, and 3 Department of Medicine, Case Western Reserve University, Cleveland, Ohio We studied the effect of different subcutaneously administered b-lactam antibiotics on the establishment of gastrointestinal colonization by vancomycin-resistant Enterococcus faecium C68 in a mouse model. Aztreonam, cefazolin, cefepime, and, to a lesser extent, ceftazidime, which neither have significant antienterococcal activity nor are secreted into human bile at high concentrations, did not promote significant vancomycin-resistant enterococci (VRE) colonization. Piperacillin-tazobactam, which has antienterococcal activity and is secreted in human bile at high concentrations, inhibited colonization after limited exposure to the inoculum but was associated with progressively increased VRE colony counts in stool samples after repeated exposure to the VRE inoculum. Ceftriaxone and cefotetan, which lack antienterococcal activity but are secreted into human bile at high concentrations, were associated with rapid and high-level colonization. These data suggest that the risk of VRE colonization varies during exposure to different b-lactam antimicrobial agents and that the risk is related to biliary concentration and antienterococcal activity of the specific b-lactam. Despite never having been described before the late 1980s, enterococci expressing resistance to the glycopeptide antibiotic vancomycin have spread widely and have grown in clinical importance over the past decade [1, 2]. Data from the Center for Disease Control s National Nosocomial Infection Surveillance System indicate that 25% of enterococcal strains isolated from documented infections occurring in patients from intensive care units are now resistant to vancomycin [3]. Moreover, the vast majority of vancomycin-resistant enterococci (VRE) are Enterococcus faecium that also express high levels of resistance to ampicillin [1, 2]. As a result, our 2 most reliable antienterococcal drugs are Received 6 May 2003; accepted 7 September 2003; electronically published 1 March 2004. Financial support: Elan. Potential conflicts of interest: L.B.R. is a consultant for Elan, Wyeth-Ayerst, Merck, Pharmacia, Intermune, Basilea, Essential Therapies, Cubist, Shire, and Genome Therapeutics; belongs to the speakers bureaus of Elan, Wyeth-Ayerst, and Merck; and has received grant support from Elan, Wyeth-Ayerst, andintermune. C.J.D. is a consultant for and has received grant support from Ortho-McNeil and IPSAT Therapies. Reprints or correspondence: Dr. Louis B. Rice, Medical Service 111(W), Louis Stokes Cleveland VA Medical Center, 10701 East Blvd., Cleveland, OH 44106 (louis.rice@med.va.gov). The Journal of Infectious Diseases 2004; 189:1113 8 This article is in the public domain, and no copyright is claimed. 0022-1899/2004/18906-0020 inactive against these strains. This inactivity results in the use of less-effective regimens or, more recently, in the use of newer antienterococcal agents, such as linezolid or quinupristin-dalfopristin. Although both of these newer agents have proved to be effective in treating VRE infections, baseline rates of reduced susceptibility to quinupristin-dalfopristin have been relatively high, and emergence of resistance to linezolid during treatment is increasingly being documented [4, 5]. Substantial data now support the concept that gastrointestinal colonization with VRE commonly precedes VRE infection [6]. VRE infections, in turn, occur most commonly in patients with immunocompromising conditions who have been treated with multiple antimicrobial agents during prolonged hospitalizations [5, 7]. Although, initially, there was debate about whether VRE were truly pathogenic in humans, 2 recent studies have demonstrated a clear association between infection with VRE and mortality, compared with patients infected with vancomycin-susceptible enterococci [5, 7]. Moreover, at least 2 potential virulence determinants, enterococcal surface protein [8] and hyaluronidase [9], have now been described to occur in VRE and may increase their pathogenicity. Understanding the factors that promote colonization and infection with VRE has, therefore, assumed considerable importance. b-lactam Antibiotics and VRE Colonization JID 2004:189 (15 March) 1113

Receipt of antimicrobial treatment has emerged as a critical factor for promoting gastrointestinal VRE colonization in hospitalized patients. Although vancomycin was initially presumed to be the primary risk factor for acquisition of VRE, careful clinical studies soon identified a variety of antimicrobial exposures as risk factors for VRE colonization. The most prominent among these has been exposure to cephalosporins (in particular third-generation cephalosporins) and antimicrobial agents with potent activity against anaerobic bacteria [10, 11]. In an animal study designed to investigate these associations, Donskey et al. [12] used a mouse model of VRE colonization to show that potent activity against anaerobic bacteria was a common characteristic of antibiotics associated with persistence of high-level VRE colonization in mouse feces. Contrary to the initial hypotheses of these studies, the fourth-generation cephalosporin cefepime did not promote persistence of VRE. The third-generation cephalosporin ceftriaxone promoted persistence, to some extent, in these studies, but not to the degree provided by those agents with very potent activity against anaerobic bacteria. Piperacillin-tazobactam, which has not been frequently associated with VRE colonization or infection in clinical studies, was found to promote persistence of VRE colonization in these studies, presumably because of its high levels of gastrointestinal secretion and potency against anaerobic bacteria. Donskey et al. [13] extended these studies to humans, documenting a strong tendency for increased quantities of VRE to be present in stool samples from colonized patients exposed to potent antianaerobic regimens. In an effort to further understand the association between exposure to antimicrobials and VRE colonization, Donskey et al. [14] established a second model of VRE colonization, in which the ability to promote the establishment of colonization (rather than persistence in animals already colonized) was tested. In these studies, they found that exposure to piperacillintazobactam was protective against establishment of VRE colonization, whereas exposure to either ceftriaxone or ticarcillinclavulanic acid promoted rapid establishment of high-level VRE colonization. They concluded that the modest activity of piperacillin-tazobactam against ampicillin-resistant E. faecium (MIC, 500 1000 mg/ml), combined with substantial biliary secretion of piperacillin (1000 2000 mg/ml in bile in human studies [15]) was sufficient to suppress growth of VRE in the upper gastrointestinal tract and to forestall high-level colonization. Ceftriaxone, on the other hand, was so devoid of activity against ampicillin-resistant E. faecium (MIC, 110,000 mg/ml [16]) that even the very high concentrations of this antibiotic that are achievable in human bile ( 5000 mg/ml [17]) were insufficient to inhibit growth in the upper gastrointestinal tract, thereby favoring establishment of colonic VRE colonization. Ticarcillinclavulanic acid also promoted establishment of VRE colonization in these studies, presumably because of its very poor activity against ampicillin-resistant E. faecium (MIC, 10,000 mg/ml [16]), combined with its potent activity against anaerobic bacteria. Taken together, these data suggest that the likelihood that a particular b-lactam antibiotic will promote initial VRE colonization will depend on several factors, including the intrinsic activity against VRE and the potency of the agent against anaerobic bacteria. Ceftriaxone and piperacillin-tazobactam are both significantly concentrated in the bile. Because of this concentration, it is conceivable that ceftriaxone is particularly likely to select for VRE colonization and that other cephalosporins without this characteristic but with weak antianaerobic activity may be less likely to select for colonization. To further delineate the associations between exposures to b-lactam antibiotics and VRE colonization, we tested a variety of b-lactam antibiotics in the mouse establishment model. MATERIALS AND METHODS Bacterial strain. We used E. faecium strain C68, an ampicillinand vancomycin-resistant (VanB-type) clinical strain. This strain has been extensively described in the past and is the strain used in our previous animal studies [12, 14]. Microbiologic techniques. Broth macrodilution MICs were performed in brain-heart infusion broth. Overnight cultures grown without antibiotic selection were diluted to achieve an inoculum of 10 5 cfu/ml. Cultures were incubated overnight at 37 C, and the MIC was determined to be that in the tube containing the lowest concentration of antibiotic for which there was no detectable turbidity after 24 h of incubation. Standard 2-fold dilutions of antibiotics were used. Mouse model. The mouse model of VRE colonization was essentially the same as that described in our previous study [14], with a few modifications. In brief, 25 30-g female CF1 mice (Harlan Sprague-Dawley) were administered subcutaneous (sc) injections of antibiotics 2 times per day for 2 days before gavage of organisms and for the duration of the experiment. These antibiotics were chosen on the basis of differing concentrations in the biliary tract and differing activities against anaerobic bacteria. Except for piperacillin-tazobactam, none exhibited significant in vitro activity against ampicillin-resistant E. faecium. The characteristics of the different test antibiotics are shown in table 1. The dose of antibiotics chosen reflected the usual daily dose for a 70-kg human, adjusted for the weight of the mouse, and was divided into 2 doses administered 12 h apart. The choice of administration frequency does not reflect either human or mouse pharmacokinetics of the antibiotics. It was chosen to make antimicrobial administration more practical and to minimize repeated injections for the mice over the 9 days of the study. Total daily doses of each antibiotic administered to the mice were the following: aztreonam (3 mg/ 1114 JID 2004:189 (15 March) Rice et al.

Table 1. Characteristics of antimicrobial agents. Antibiotic Class/generation Activity against ampicillin-resistant Enterococcus faecium Biliary concentration (multiple of serum concentration) Antianaerobic activity Aztreonam Monobactam None 1 None Cefazolin Cephalosporin/first None 3 Weak Cefepime Cephalosporin/fourth None 2 Weak Cefotetan Cephalosporin/second, cephamycin None 20 30 Potent Ceftazidime Cephalosporin/third, antipseudomonal None 0.2 Weak Ceftriaxone Cephalosporin/third None 10 30 Modest Piperacillin-tazobactam Ureidopenicillin-inhibitor combination Modest 5 10 Potent NOTE. These data are derived from previous studies [15, 17 20]. day), cefazolin (3 mg/day), cefepime (3 mg/day), cefotetan (2 mg/day), ceftazidime (3 mg/day), ceftriaxone (1 mg/day), and piperacillin-tazobactam (6.75 mg/day). Control mice were administered a volume of saline equivalent to that used in the administration of antibiotics. After 2 days of sc administration of antibiotics, mice were administered an inoculum of 2 4 10 2 cfu of E. faecium C68, derived by saline dilutions of a frozen stock of known inoculum, by use of a stainless steel feeding tube (Perfektum; Popper & Sons). Fresh stool samples were obtained from all mice on days 1, 3, and 6 after gavage. Stool samples were weighed, homogenized in sterile saline, and serially diluted. Dilutions were plated onto Enterococcosel agar containing vancomycin (6 mg/ml). Colonies that hydrolyzed esculin in the agar were counted after 48 h of incubation. When no VRE were detectable on initial screening, a larger aliquot (100 ml) was plated, to increase detection of small numbers of VRE (lower limit of detection, 2.0 log 10 cfu/g); if no VRE were detected from these samples, a number equal to the lower limit of detection was assigned. In a deviation from the previous protocol, in the first 2 series of experiments, the bedding in the cages was not changed on a regular basis. This increased the risk that coprophagia (the act of mice ingesting their own stool) would lead to ingestion of continuous inocula of VRE. To determine whether this difference in protocol had an effect on our results, we performed additional experiments in which ceftazidime, piperacillin-tazobactam, ceftriaxone, or saline was administered in doses equivalent to those in the earlier experiments, but with daily changes in the cage bedding. Results of these experiments are reported separately. In this set of experiments, stool samples were obtained on days 1, 3, 6, and 9. These experiments were reviewed and approved by the Louis Stokes Cleveland VA Medical Center Animal Care and Use Committee. Statistical analysis. Statistical analyses were performed by use of STATA software (version 5.0; StataCorp). Log 10 colonyforming units were determined across the 3 data-collection time points for individual mice. These were then examined by use of a 1-way analysis of variance, by antibiotic treatment. Overall differences and pair-wise differences were examined, with P values adjusted for multiple comparisons by use of the Scheffe correction. RESULTS Establishing VRE colonization. Log 10 colony-forming units per gram of stool for the initial experiments are shown in figure 1. Consistent with previous experiments, ceftriaxone promoted colonization at high levels very quickly, with extensive colonization demonstrable by day 1 after inoculation of organisms. The numbers of VRE colony-forming units found in the stool of ceftriaxone-treated mice decreased over time, which also is consistent with previous experiments. We interpret this decrease as reflecting the relatively modest activity of ceftriaxone against anaerobic bacteria minimizing the persistence of high levels of VRE colonization, as observed in previous experiments using the VRE persistence model [12]. Cefotetan also promoted colonization at high levels, and these levels persisted over time. It is noteworthy that cefotetan, like ceftriaxone, is highly concentrated in the bile and, as such, may be particularly likely to promote the establishment of colonization. Moreover, its potent activity against anaerobic bacteria will promote persistence of high levels of colonization. It is similar in its antibacterial activity to ticarcillin-clavulanic acid, which also promotes establishment and persistence of VRE colonization in this and other models [12, 14]. In contrast to the results for ceftriaxone and cefotetan, there was no tendency to promote colonization on the part of aztreonam, cefazolin, or cefepime. Administration of these antibiotics was statistically equivalent to administration of saline. As detailed in table 1, these antimicrobial agents are neither concentrated in the bile nor potent in their activity against anaerobic bacteria. In this setting, their lack of activity against ampicillin-resistant E. faecium may well become irrelevant, since they are present in concentrations too low to provide a significant selective advantage for enterococcal proliferation. Results obtained in mice treated with either ceftazidime or b-lactam Antibiotics and VRE Colonization JID 2004:189 (15 March) 1115

Figure 1. Bar graph demonstrating log 10 cfu of Enterococcus faecium C68 per gram of feces, 1, 3, and 6 days after inoculation of 200 cfu of E. faecium C68 into the stomach of mice pretreated with the indicated antimicrobial agents. Results represent the 2 experiments in which cage bedding was not changed on a daily basis. Lines extending above the bars represent the SD for the log 10 cfu per gram of feces. Means of the 3 determinations were examined by use of a 1-way analysis of variance, by antibiotic treatment. Overall differences and pair-wise differences were examined, with P values adjusted for multiple comparisons by use of the Scheffe correction. *Difference is significant, vs. all other groups except ceftriaxone. Difference is significant, vs. all other groups except cefotetan and piperacillin-tazobactam. Difference is significant, vs. normal saline, aztreonam, and cefazolin. piperacillin-tazobactam were more complex. In the first experimental group treated with ceftazidime, it appeared that this antibiotic significantly promoted colonization with VRE. The second experiment, however, suggested only minimal promotion of colonization (data not shown). For the piperacillin-tazobactam treated mice, results on day 1 were consistent with those of previous experiments, showing no significant promotion of VRE colonization. However, results obtained on days 3 and 6 of these experiments were at odds with those of the previous experiments, in that they indicated an association of administration of piperacillin-tazobactam with increasing quantities of fecal VRE over time. Review of the current and previous protocols revealed 2 potential explanations for this difference. First, the doses of piperacillin-tazobactam administered to the mice in the present study were slightly lower than those administered in previous studies (6.75 vs. 8 mg/day). Second, in contrast to the investigators in previous experiments [14], we did not regularly change cage bedding. As such, it was possible that, in the mice, continued ingestion of colonized feces could have resulted in a persistent exposure to increasing inocula of VRE. Preliminary experiments with a small number of mice suggested that daily bedding changes would prevent the association of piperacillin-tazobactam with colonization over time (data not shown). We therefore performed a third set of experiments, with ceftazidime (8 mice), piperacillin-tazobactam (8 mice), ceftriaxone (2 mice), or saline (2 mice), in which we made certain that bedding was changed on a daily basis, in an attempt to minimize the contribution of coprophagia to our results. The results of these experiments are presented in figure 2. VRE colonization was virtually undetectable in all 8 piperacillin-tazobactam treated mice. Overall, levels of colonization for piperacillin-tazobactam treated mice were comparable to those for mice administered saline, which is similar to the result of our previous study [14]. Ceftazidimetreated mice exhibited variable levels of colonization. Overall, these levels of colonization were statistically greater than those for piperacillin-tazobactam treated mice and less than those for ceftriaxone-treated mice but were not different from those for mice that were administered saline alone. These data are consistent with a contribution of coprophagia to the discrepancies between the results of piperacillin-tazobactam treatment in these experiments and those that we reported previously. DISCUSSION The data presented here suggest that one strategy to avoid promoting gastrointestinal VRE colonization is to select antimicrobial agents with minimal biliary excretion. Previously published data suggest that neither aztreonam, cefazolin, cefepime, nor ceftazidime is significantly concentrated in human bile after normal dosing [18 20]. Also, these agents do not have potent activity against anaerobic bacteria. If these data on mice can be applied to humans, it would, therefore, seem that aztreonam, cefazolin, and cefepime could probably be used for the infections for which they are indicated without significant concern about promoting gastrointestinal VRE colonization. These data are particularly reassuring for cefazolin, given the frequency with which this antibiotic is used for surgical pro- 1116 JID 2004:189 (15 March) Rice et al.

Figure 2. Bar graph demonstrating log 10 cfu of Enterococcus faecium C68 per gram of feces, 1, 3, 6, and 9 days after inoculation of 200 cfu of E. faecium C68 into the stomach of mice pretreated with the indicated antimicrobial agents. The nos. of mice in each group were as follows: saline, 2; ceftazidime, 8; piperacillin-tazobactam, 8; and ceftriaxone, 2. Results represent the single experiment in which cage bedding was changed daily. Lines extending above the bars represent the SD for the log 10 cfu per gram of feces. P!.05, ceftriaxone vs. normal saline, ceftazidime, and piperacillin-tazobactam. P!.05, ceftazidime vs. piperacillintazobactam. phylaxis, and for cefepime, which is used frequently in immunocompromised patients. Our results indicate that the effect of ceftazidime in our model is less predictable. Of note, ceftazidime has been specifically implicated in human gastrointestinal VRE colonization in at least 1 study [21]. Among antimicrobial agents that achieve significant gastrointestinal concentrations, our results suggest that piperacillintazobactam exerts an inhibitory effect against VRE colonization. However, this inhibition is not absolute, since persistent exposure to VRE inocula in mice treated with piperacillintazobactam results in high-level colonization with VRE over time. It should be acknowledged that our choice of dosing 2 times per day could put piperacillin-tazobactam at a disadvantage in the face of persistent and possibly increasing exposures to VRE. In humans, piperacillin-tazobactam is usually administered 4 or 6 times per day, which could increase its protective effect against VRE colonization by yielding biliary concentrations that are more consistently elevated than would be likely with dosing 2 times per day in mice. We are not aware of data from clinical studies linking exposure to piperacillintazobactam with increased risk of VRE colonization or infection, despite the fact that its potency against anaerobic bacteria is significant. The lack of clinical data associating piperacillintazobactam with VRE supports the hypothesis that the activity of this combination in the upper gastrointestinal tract is usually enough to inhibit high-level VRE colonization in humans. Our data serve to emphasize the importance of infection control measures for control of VRE in the hospital, since the quantities of VRE in the stool samples of mice subjected to repeated inocula were higher than observed when repeated inocula were not present, even in mice that were administered normal saline. The repeated observation that high-level VRE colonization cannot be established to a significant extent in the absence of antibiotic administration should also be emphasized and used to promote judicious and parsimonious use of antibiotics in all settings. Finally, our data suggest that all cephalosporins are not alike in their tendency to promote VRE colonization and that the tendency to promote colonization can be predicted on the basis of their intrinsic characteristics (biliary excretion, activity against enterococci, and anaerobic bacteria). When assessing the risk of colonization or infection with VRE, future clinical studies should try, as much as possible, to distinguish between exposure to different types of cephalosporins. References 1. Sahm DF, Marsilio MK, Piazza G. Antimicrobial resistance in key bloodstream bacterial isolates: electronic surveillance with The Surveillance Network Database USA. Clin Infect Dis 1999; 29:259 63. 2. Murdoch DR, Mirrett S, Harrell LJ, Monahan JS, Reller LB. Sequential emergence of antibiotic resistance in enterococcal bloodstream isolates over 25 years. Antimicrob Agents Chemother 2002; 46:3676 8. 3. Centers for Disease Control and Prevention. National Nosocomial Infection Surveillance System. Available at: http://www.cdc.gov/ncidod/ hip/surveill/nnis.htm. Accessed 19 February 2004. 4. Marshall SH, Donskey CJ, Hutton-Thomas R, Salata RA, Rice LB. Gene dosage and linezolid resistance in Enterococcus faecium and Enterococcus faecalis. Antimicrob Agents Chemother 2002; 46:3334 6. 5. Vergis EN, Hayden MK, Chow JW, et al. Determinants of vancomycin resistance and mortality rates in enterococcal bacteremia: a prospective multicenter study. Ann Intern Med 2001; 135:484 92. 6. Wells CL, Juni BA, Cameron SB, et al. Stool carriage, clinical isolation, and mortality during an outbreak of vancomycin-resistant enterococci in hospitalized medical and/or surgical patients. Clin Infect Dis 1995; 21:45 50. 7. Lodise TP, McKinnon PS, Tam VH, Rybak MJ. Clinical outcomes for patients with bacteremia caused by vancomycin-resistant enterococcus in a level 1 trauma center. Clin Infect Dis 2002; 34:922 9. 8. Baldassarri L, Bertuccini L, Ammendolia MG, Gherardi G, Creti R. Variant esp gene in vancomycin-sensitive Enterococcus faecium. Lancet 2001; 357:1802. 9. Rice LB, Carias L, Rudin S, et al. A potential virulence gene, hyl Efm, predominates in Enterococcus faecium of clinical origin. J Infect Dis 2003; 187:508 12. 10. Bonten MJM, Slaughter S, Ambergen AW, et al. The role of colonization pressure in the spread of vancomycin-resistant enterococci. Arch Intern Med 1998; 158:1127 32. 11. Puzniak LA, Mayfield J, Leet T, Kollef M, Mundy LM. Acquisition of vancomycin-resistant enterococci during scheduled antimicrobial rotation in an intensive care unit. Clin Infect Dis 2001; 33:151 7. 12. Donskey CJ, Hanrahan JA, Hutton RA, Rice LB. Effect of parenteral antibiotic administration on persistence of vancomycin-resistant Enterococcus faecium in the mouse gastrointestinal tract. J Infect Dis 1999; 180:384 90. 13. Donskey CJ, Chowdhry TK, Hecker MT, et al. Effect of antibiotic therapy on the density of vancomycin-resistant enterococci in the stool of colonized patients. N Engl J Med 2000; 343:1925 32. 14. Donskey CJ, Hanrahan JA, Hutton RA, Rice LB. Effect of parenteral antibiotic administration on establishment of colonization with vancomycin-resistant Enterococcus faecium in the mouse gastrointestinal tract. J Infect Dis 2000; 181:1830 3. 15. Taylor EW, Poxon V, Alexander-Williams J, Jackson D. Biliary excretion of piperacillin. J Int Med Res 1983; 11:28 31. b-lactam Antibiotics and VRE Colonization JID 2004:189 (15 March) 1117

16. Donskey CJ, Schreiber JR, Jacobs MR, et al. A polyclonal outbreak of predominantly VanB vancomycin resistant enterococci in northeast Ohio. Clin Infect Dis 1999; 29:573 9. 17. Hayton WL, Schandlik R, Stoeckel K. Biliary excretion and pharmacokinetics of ceftriaxone after cholecystectomy. Eur J Clin Pharmacol 1986; 30:445 51. 18. Martinez OV, Levi JU, Devlin RG. Biliary excretion of aztreonam in patients with biliary tract disease. Antimicrob Agents Chemother 1984; 25:358 61. 19. McLeish AR, Strachan CJ, Powis SJ, Wise R, Bevan PG. The influence of biliary disease on the excretion of cefazolin in human bile. Surgery 1977; 81:426 30. 20. Okamoto MP, Gill MA, Nakahiro RK, et al. Tissue concentrations of cefepime in acute cholecystitis patients. Ther Drug Monit 1992; 14:220 5. 21. Bradley SJ, Wilson AL, Allen MC, Sher HA, Goldstone AH, Scott GM. The control of hyperendemic glycopeptide-resistant Enterococcus spp. on a haematology unit by changing antibiotic usage. J Antimicrob Chemother 1999; 43:261 6. 1118 JID 2004:189 (15 March) Rice et al.