• Open Access

Antimicrobial Resistance Impacts Clinical Outcome of Granulomatous Colitis in Boxer Dogs


  • The abstract was given as an oral presentation at the Forum of the ACVIM in Montreal, 2009.

Corresponding author: Kenneth W. Simpson, Department of Clinical Sciences, College of Veterinary Medicine, VMC 2001, Cornell University, Ithaca, NY 14853-9655; e-mail: kws5@cornell.edu.


Background: Escherichia coli have recently been identified within the colonic mucosa of Boxer dogs with granulomatous colitis (GC). Eradication of invasive E. coli is associated with clinical and histological remission.

Objectives: To determine antimicrobial susceptibility profiles of E. coli strains from GC and healthy dogs, and the association of antimicrobial resistance with clinical outcome.

Animals: Fourteen Boxer dogs with GC and 17 healthy pet dogs.

Methods: Prospective study: E. coli was cultured from GC biopsies and rectal mucosal swabs of healthy dogs. Individual strains were selected by phylogroup and overall genotype, determined by triplex- and random amplified polymorphic DNA-polymerase chain reaction respectively. Antimicrobial susceptibility was determined by broth microdilution minimal inhibitory concentration.

Results: Culture yielded 23 E. coli strains from GC (1–3/dog, median 2) and 34 strains from healthy (1–3/dog, median 2). E. coli phylogroups were similar (P= .18) in GC (5A, 7B1, 5B2, 6D) and healthy (2A, 10B1, 15B2, 7D). Resistance to ampicillin, amoxicillin-clavulanate, cefoxitin, tetracycline, trimethoprim-sulfa (TMS), ciprofloxacin, and chloramphenicol was greater (P < .05) in GC (21–64%) than healthy (0–24%). Enrofloxacin resistant E. coli were isolated from 6/14 GC versus 0/17 healthy (P= .004). Of the enrofloxacin resistant cases, 4/6 were also resistant to macrophage-penetrating antimicrobials such as chloramphenicol, rifampicin, and TMS. Enrofloxacin treatment before definitive diagnosis was associated with antimicrobial resistance (P < .01) and poor clinical outcome (P < .01).

Conclusions and Clinical Importance: Antimicrobial resistance is common among GC-associated E. coli and impacts clinical response. Antimicrobial therapy should be guided by mucosal culture and antimicrobial susceptibility testing rather than empirical wisdom.


adherent invasive E. coli


Crohn's disease


fluorescence in situ hybridization


granulomatous colitis IBD inflammatory bowel disease


inflammatory bowel disease


minimal inhibitory concentration


polymerase chain reaction

Granulomatous colitis (GC), also known as histiocytic ulcerative colitis, is typically seen in young Boxer dogs <4 years of age, and was first described in 1965.1 Affected dogs usually have signs of colitis including hematochezia, diarrhea, mucus, tenesmus, and increased frequency of defecation that is often accompanied by ill-thrift, weight loss, anemia, and hypoalbuminemia. Diagnosis is based on colonic histology, and typified by large numbers of periodic acid-Schiff-positive macrophages, accompanied by lymphocytes, plasma cells, eosinophils, epithelial ulceration, and loss of goblet cells.1,2,3

Recent studies have documented clinical responses to enrofloxacin,4–7 and an association between GC and intramucosal Escherichia coli in affected Boxer dogs.5,7 Bacterial localization by fluorescence in situ hybridization (FISH) with an E. coli probe typically reveals multifocal clusters of mucosally invasive bacteria within the mucosa and intracellularly within macrophages.5,7 Given the susceptibility of Boxers it appears likely that an underlying host defect in mucosal immunity enables opportunistic invasion by resident E. coli.5 A role for the resident enteric microflora, particularly the Enterobacteriaceae, is increasingly recognized in the pathogenesis of inflammatory bowel diseases (IBD), and increased numbers of mucosa-associated E. coli and serological responses against E. coli are observed in people with Crohn's disease (CD).8–15 The E. coli strains isolated from Boxer dogs display a similar pathotype in cultured intestinal epithelial cells and macrophages to a novel group of E. coli strains, termed adherent and invasive E. coli (AIEC), that are associated with ileal inflammation in CD patients.5,8–10,16

Traditionally, empirical treatment of GC with a variety of antimicrobials and immunosuppressive agents was met with poor success (reviewed in Hostutler and colleagues6,7). We suspect that the positive clinical responses reported in GC affected Boxer dogs treated with enrofloxacin or chloramphenicol1,4–7 reflects the susceptibility of E. coli to these antimicrobials, and their ability to accumulate in E. coli colonized macrophages.17 This is supported by the correlation between clinical response to enrofloxacin and eradication of intramucosal and intracellular E. coli.5,7

To optimize antimicrobial selection against GC-associated E. coli we sought to determine the antimicrobial susceptibility profiles of E. coli strains isolated from GC and healthy dogs, and the effect of antimicrobial resistance on clinical outcome.


Animals and Sampling

Formalin-fixed, paraffin-embedded colonic mucosal biopsies from 14 Boxer dogs (8 male, 6 female; median age 12 months, range 4–60 months, “GC” group) with histologically confirmed GC were evaluated for intracellular E. coli by FISH analysis with eubacterial (EUB338-6FAM) and E. coli-specific probes (E. coli-Cy3), targeting 16S ribosomal DNA as described previously.5 Biopsy samples for microbial culture were collected into Luria-Bertani broth with 20% glycerol on ice and stored at −80°C until processing. Clinical and outcome data were obtained from referring veterinarians and owners where possible. Affected dogs were classified as “complete responders” if their clinical signs had completely resolved with antimicrobial treatment; “partial responders” if their clinical signs had reduced in frequency or severity, and “nonresponders” if treatment had failed to elicit any improvement in clinical signs. Rectal mucosal swabs from 17 clinically healthy pet dogs (group “H”) of various breeds (6 male, 11 female; median age 60 months, range 6–132 months) were inserted into BBL Port-a-cul transporter tubesa and processed immediately. The rationale for sampling E. coli was principally based on consideration of the localization of E. coli in biopsies evaluated by previous studies.5,7 In healthy dogs E. coli are not invasive, and are restricted to superficial mucus and mucus containing glands, in contrast to dogs with GC where the invasive E. coli are located intramucosally and within macrophages. Hence we considered that collection of E. coli by biopsy in GC and swab in H would provide a representative sampling of mucosa-associated strains in each population. A secondary consideration that impacted our choice of methodology was the lack of justification for performing endoscopic biopsy on healthy client-owned dogs. Because the affected dogs were sampled as part of their diagnostic evaluation, approval by the Institutional Animal Care and Use Committee was not required. Healthy dogs were sampled with informed client consent as an adjunct to fecal parasitology screening at routine health checks.

Isolation and Characterization of  E. coli

For microbial culture, biopsies were aseptically transferred from collection vials into a disposable tissue grinder and homogenized. The tissue homogenate or swabs were used to inoculate trypticase soy agar with 5% sheep blood and Gram-negative broth. Media were incubated at 37°C for 18–24 hours in 6% CO2 and visually inspected for bacteria before being subcultured onto Levine EMB agar and MacConkey MUG agar containing the chromogenic indicator methylumbelliferyl-β-d-glucuronidase, and incubated at 37°C for 18–24 hours. Aerobic bacterial colonies were screened by Gram stain, catalase, and oxidase reaction, and were identified with the computer controlled, automatic, Sensititre System.b Conventional biochemical reactions by standard identification strategies were used as needed to supplement the Sensititre identification system, and confirmed by polymerase chain reaction (PCR) targeting the E. coli housekeeping gene uidA (β-glucuronidase).10 Ten to 15 individual E. coli colonies per sample were selected at random and screened by triplex PCR for major E. coli phylogenetic groups (A, B1, B2, and D),18,19 and genotyped by random amplified polymorphic DNA (RAPD)-PCRs with primer 1283, followed by additional primers (1254 and 1290) on strains with a similar banding pattern. E. coli strains that differed in genotype were archived at −80°C, and fresh nonpassaged bacteria were used for subsequent investigations.5

Antimicrobial Susceptibility Testing

The minimum inhibitory concentrations (MIC) of E. coli to selected antimicrobials were determined by the microdilution method and interpreted according to Clinical Laboratory Standards Institute (CLSI) interpretive criteria20 (formerly NCCLS). The panel of antimicrobials tested included amikacin, amoxicillin-clavulanic acid (amox-clav), ampicillin, cefazolin, cefoxitin, cefpodoxime, chloramphenicol, ciprofloxacin, clindamycin, enrofloxacin, erythromycin, gentamicin, imipenem, marbofloxacin, orbifloxacin, streptomycin, tetracycline, and trimethoprim-sulfa (TMS). For enrofloxacin resistant strains, additional susceptibilities to macrophage-penetrating antimicrobials florfenicol, clarithromycin, and rifampin were determined.17 Breakpoints were based on CLSI interpretive standards20 and where CLSI interpretive standards were unavailable, breakpoints from the National Antimicrobial Resistant Monitoring System (NARMS) 2004 Annual Report were used.21E. coli strain ATCC 25922 was used as a quality control standard.

Statistical Analysis

A Fisher's exact test was used to compare phylogroups, the number of antimicrobial resistant strains between GC and H, and clinical outcome with (a) susceptibility of E. coli strains to enrofloxacin and (b) prior treatment with enrofloxacin. Significance was set at P < .05.


E. coli were observed in all GC (14/14), within the colonic mucosa and as multifocal clusters of bacteria within macrophages (Fig 1). Characterization of E. coli isolates by Triplex—and RAPD-PCR (Fig 2) yielded 23 E. coli strains from GC (1–3 per dog, median 2) and 34 from H (1–3 per dog, median 2) (P > .05). The distribution of E. coli phylogroups was similar (P= .18) in E. coli isolated from GC (5A, 7B1, 5B2, 6D) and H (2A, 10B1, 15B2, 7D). In the GC group, 3 dogs (Fig 2, lanes 5, 8, and 17) harbored a phylogroup B1 E. coli with identical banding patterns for RAPD 1283; subsequent analysis with RAPD primers 1254 and 1290 showed identical banding patterns in 2 of these 3 strains (Fig 2, lanes 8 and 17). In the H group, 2 dogs with E. coli in phylogroup A, and 4 with E. coli in phylogroup B2, had strains that were identical with all 3 RAPD primers. When comparing the GC and H groups, 1 strain from GC and H in phylogroup B1 (Fig 2, lane 26 GC, lane 28 H), and 1 from phylogroup B2 (Fig 2, lane 27 GC, lane 29 H) were identical on RAPD 1283, 1254, and 1290.

Figure 1.

 Fluorescence in situ hydridization of granulomatous colitis using Escherichia coli-Cy3 (red) and eubacterial (EUB338-6FAM, green) probes. Multifocal clusters of E. coli (orange) can be seen within the mucosa and within cells (× 60). Inset: periodic acid-Schiff-positive macrophages (× 40).

Figure 2.

 Genetic diversity of 23 granulomatous colitis (GC) associated Escherichia coli evaluated by random amplified polymorphic DNA-polymerase chain reaction (RAPD-PCR) with primer 1283 (lanes 2–24), showing identical banding in lanes 5, 8, and 17 of 3 strains from 3 dogs (isolates 5 and 17 were also identical with RAPD primers 1254 and 1290). Lanes 26–29: RAPD-PCR primer 1254 on 2 isolates from GC and H groups with identical RAPD patterns on 1283 (GC lanes 26 and 28; H lanes 27 and 29). Lane 1: mastermix/primers/H20 control. Lane 25: 100 bp plus DNA ladder.

A significantly greater number of GC-associated E. coli strains were resistant to ampicillin (P < .05, GC n = 11, H n = 5), amox-clav (P < .05, GC n = 8, H n = 3), cefoxitin (P < .001, GC N = 7, H = 0), tetracycline (P < .05, GC n = 11, H n = 6), TMS (P < .01, GC n = 11, H n = 2), ciprofloxacin (P < .001, GC n = 8, H n = 0), and chloramphenicol (P < .05, GC n = 4, H n = 0), when compared with strains from H (Table 1). Enrofloxacin resistant E. coli were isolated from 6/14 GC versus 0/17 healthy (P= .004). The susceptibilities of these 8 resistant strains to other fluoroquinolones and macrophage-penetrating antimicrobials with efficacy against E. coli17 are shown in Table 2. Where resistance to enrofloxacin occurred, it was present in all E. coli strains isolated from an individual, and was associated with resistance to all other fluoroquinolones tested (ciprofloxacin and marbofloxacin 8/8 strains, orbifloxacin 3/3 strains). Only 4 of 8 enrofloxacin resistant strains (in 2 individuals) were susceptible to chloramphenicol; 2 of 8 strains were susceptible to tetracycline or TMS; 1 of 8 was susceptible to clarithromycin; and none were susceptible to rifampin.

Table 1.   Prevalence of antimicrobial resistance in Escherichia coli strains isolated from 14 granulomatous colitis (GC) affected boxer dogs and 17 healthy dogs.
AntimicrobialResistant Strains (%)Resistant Individuals (%)
  1. Fisher's Exact, GC/histiocytic ulcerative colitis versus H :
    *P < .05 , ** < 0.01 , ***P < .001.

Amoxicillin-clavulanate8/23 (35)*3/34 (9)8/14 (57)*2/17 (12)
Ampicillin11/23 (48)*5/34 (15)9/14 (64)**3/17 (18)
Cefoxitin7/23 (30)***0 (0)7/14 (50)***0 (0)
Tetracycline11/23 (48)*6/34 (18)9/14 (64)*4/17 (24)
Trimethoprim-sulfa10/23 (43)**2/34 (6)8/14 (57)**1/17 (6)
Ciprofloxacin8/23 (35)***0 (0)6/14 (43)**0 (0)
Gentamicin3/23 (13)6/34 (18)2/14 (14)6/17 (35)
Chloramphenicol4/23 (17)*0 (0)3/14 (21)0 (0)
Table 2.   Susceptibility of granulomatous colitis-associated enrofloxacin resistant Escherichia coli to additional fluoroquinolones and macrophage-penetrating antimicrobials in 6 dogs that did not respond or responded only partially to treatment.
Dog ID24561011
  1. E, euthanized due to disease; I, intermediate susceptibility; P, partial response; R, resistant; S, susceptible.

E. coli strains12111211

Most of the GC (11/14) had received empirical treatment with a number of antimicrobials (range 1–4, median 1), including metronidazole (11/14), enrofloxacin (8/14), tylosin (7/14), TMS (4/14), amoxicillin (3/14), ampicillin (1/14), and doxycycline (1/14) before colon biopsy. Empirical treatment with enrofloxacin was associated with the isolation of a resistant E. coli (P < .01). Resistance to the antimicrobials given before biopsy was also documented in 3/4 dogs treated with TMS, 4/4 treated with ampicillin or amoxicillin, and 1/1 treated with doxycycline. There was no relationship of antimicrobial resistance to E. coli phylogroup (A, B1, B2, or D).

The time from a biopsy confirmed diagnosis of GC to last follow-up in the surviving dogs (n = 10) was 27.5 months (median, range 18–78). Six GC were classified as complete responders, 4 were partial responders, and 4 were nonresponders that had been euthanized because of disease. In the 6 complete responders, all E. coli strains were susceptible to enrofloxacin, and only 2/6 had received enrofloxacin before definitive diagnosis. In contrast, the 4 nonresponders harbored E. coli strains that were resistant to enrofloxacin and had received treatment with enrofloxacin before definitive diagnosis. Additional antimicrobial therapy in these dogs with various combinations of amikacin, neomycin enemas, erythromycin, cephalexin, and amox-clav was unsuccessful. In the 4 partial responders, 2 were colonized with enrofloxacin susceptible E. coli and 2 with enrofloxacin resistant E. coli. Both of the partial responders with enrofloxacin resistant E. coli had received empirical treatment with enrofloxacin before biopsy. The owners of the partial responders reported that enrofloxacin was associated with a reduction in severity and frequency of clinical signs, and had employed intermittent 2–4 week treatments with enrofloxacin (3 dogs) or continuous enrofloxacin (1 dog). To summarize, 6/6 complete responders harbored enrofloxacin susceptible E. coli, whereas 2/4 partial responders and 4/4 (euthanized) nonresponders harbored enrofloxacin resistant strains. Clinical response (complete versus partial responders + euthanized) was negatively impacted by resistance of mucosal E. coli to enrofloxacin (P < .01), and treatment with enrofloxacin before definitive diagnosis (P < .01).


The present study was undertaken to inform the treatment of GC by determining the antimicrobial susceptibility profiles of E. coli associated with GC. We found that over 50% of GC harbored mucosal E. coli that were resistant to one or more nonfluoroquinolone antimicrobial, and resistance to fluoroquinolones (including enrofloxacin) was present in an alarming 43%. These observations suggest that antimicrobial resistance in GC-associated E. coli may have been selected for as a consequence of the growing use of fluoroquinolones in human and veterinary medicine, though we did not uncover any relationship of resistance to specific E. coli phylogroups.22–25 It is also possible that invasive E. coli in GC affected Boxers have acquired resistance through prior treatment, as supported by the correlation between enrofloxacin resistance and empirical therapy. The majority of E. coli strains isolated from GC and H were diverse in overall genotype with minimal overlap between H and GC. However, it is interesting to note that 2 GC dogs from different geographical locations harbored phylogroup B1 strains with identical RAPD-PCR patterns with 3 different RAPD primers. The absence of clustering by phylogroup and overall genotype suggests that no single clonal group is responsible for mucosal invasion in GC. In-depth analysis of genotype (virulence and phylogeny) and phenotype of GC and H E. coli is underway to try to identify the microbial characteristics associated with invasion of, and persistence within, the colonic mucosa. Four independent studies have documented dramatic clinical responses of Boxer dogs with GC to antimicrobial regimens containing enrofloxacin,4–7 and clinical and histological remission is associated with the eradication of invasive E. coli, typically within colonic macrophages.5,7 These findings have led to GC in Boxers being considered a breed-specific defect in mucosal immunity that enables opportunistic invasion by resident E. coli,5 and enrofloxacin being regarded as the treatment of choice, with a positive clinical response reported in 19/20 dogs.4–7 The lack of clinical response to enrofloxacin in the only nonresponder reported to date was associated with persistence of intramucosal E. coli, the isolation of enrofloxacin resistant E. coli, lack of response to alternative antimicrobials, and eventually euthanasia.7

Enrofloxacin resistance was associated with clinical outcome, and complete remission of disease occurred only in the dogs that harbored enrofloxacin susceptible E. coli, with partial response or euthanasia resulting in all 6 GC with resistant strains. Because of the retrospective nature of obtaining follow-up and clinical details in this study, we were not able to determine precisely the dose, frequency, and duration of enrofloxacin therapy before colonic biopsy of GC. In a prior study we found that enrofloxacin at 7 mg/kg PO for at least 6 weeks was associated with a good clinical response, and we speculate that short-term empirical treatment might result in clinical improvement (often within 1–2 weeks7) that leads to cessation of therapy before complete eradication of E. coli and the emergence of resistant strains. The potential adverse effects of enrofloxacin on cartilage development in skeletally immature Boxer dogs could additionally influence practitioners to avoid treatment beyond clinical improvement. Prospective evaluation of the impact of antimicrobial therapy on E. coli isolated from GC affected Boxers is required to clarify these issues.

Given the high frequency of antimicrobial resistance in GC, we sought to identify alternative antimicrobials with efficacy against GC associated E. coli. The lack of clinical response of dogs that were unresponsive to enrofloxacin to antimicrobials with efficacy against their E. coli strains in vitro (eg amikacin, amox-clav, neomycin) suggests that factors such as drug distribution impact the eradication of mucosally invasive E. coli. Support for this is provided by a recent study of CD-associated E. coli (AIEC), where the ability of an antimicrobial to kill E. coli within macrophages was determined by a combination of antimicrobial susceptibility and the ability to accumulate intracellularly.17 For example, ampicillin was effective against CD-associated E. coli in culture, but ineffective against bacteria within cultured macrophages.17 The antimicrobials with best efficacy for intracellular killing were ciprofloxacin, 99.5%; rifampin, 85.1%; tetracycline, 62.8%; clarithromycin, 62.1%; sulfamethoxazole, 61.3%; TMS, 56.3%; and azithromycin, 41.0%.17 In the present study, the most effective antimicrobials against GC associated E. coli on the basis of MIC testing were amikacin (23/23 strains and 14/14 dogs susceptible) and chloramphenicol (19/23 strains and 11/14 dogs susceptible). The aminoglycosides are ineffective against facultative intracellular pathogens such as Shigella spp., Salmonella spp. during their intracellular growth phase, because of their inability to effectively penetrate mammalian cells.26 Amikacin is therefore unlikely to be an effective agent for elimination of intracellular E. coli in GC. Chloramphenicol is small and lipid soluble, penetrates effectively into most tissues, and may therefore be an effective alternative in fluoroquinolone resistant cases. Chloramphenicol has also been associated with positive treatment responses in 6/8 GC affected Boxer dogs.1 However, in the present study 50% of the fluoroquinolone resistant E. coli strains were also resistant to chloramphenicol, and only 2/6 GC with enrofloxacin resistant E. coli harbored chloramphenicol susceptible strains. The potential adverse effects of chloramphenicol (aplastic anemia) have also tended to restrict its use in dogs. Florfenicol is a newer derivative of chloramphenicol that has not been associated with aplastic anemia, is almost completely absorbed after oral administration in dogs, and is not subject to the action of acetyltransferase, an enzyme used by bacteria to develop resistance to chloramphenicol.27 In the present study we observed intermediate susceptibility of GC associated E. coli to florfenicol that were susceptible to chloramphenicol, but it is unclear if this is a consequence of the breakpoints used for interpretation of MIC for florfenicol, which have not been determined for canine isolates (the same is true for chloramphenicol).

Interestingly, in contrast to CD-associated E. coli,17 isolates from GC were frequently resistant to rifampin, clarithromycin, TMS, and tetracycline, which severely limits the selection of alternative antimicrobials in enrofloxacin resistant GC. We speculate however that the in vivo activity and potential for synergism in bacterial killing may be underestimated by in vitro testing of individual antimicrobials. In this respect, synergistic responses to combination therapy with multiple antimicrobials, eg ciprofloxacin, tetracycline, and trimethoprim have been reported in CD-associated E. coli.17 This multidrug approach may also be helpful in enrofloxacin resistant GC but remains to be evaluated.

In conclusion, the findings of this study indicate that antimicrobial resistance is common in E. coli associated with GC. Resistance to enrofloxacin is associated with a poor clinical outcome, and empirical antimicrobial treatment before definitive diagnosis. To optimize outcome it would seem preferable to perform colonic biopsy to achieve a definitive diagnosis (histopathology and FISH), and isolation and susceptibility testing of colonic mucosal E. coli to optimize treatment before administration of antimicrobials.


aBD Diagnostics, Franklin Lakes, NJ

bTREK Diagnostic Systems, Cleveland, OH


Francis Davis, Research Support Specialist, Simpson Laboratory, Cornell University, and submitting veterinarians. The work was not supported by a specific grant.