|Home | About | Journals | Submit | Contact Us | Français|
The emergence of plasmid-mediated multidrug resistance (MDR) among enteric bacteria presents a serious challenge to the treatment of bacterial infections in humans and animals. Recent studies suggest that avian Escherichia coli commonly possess the ability to resist multiple antimicrobial agents, and might serve as reservoirs of MDR for human extraintestinal pathogenic Escherichia coli (ExPEC) and commensal E. coli populations. We determined antimicrobial susceptibility profiles for 2202 human and avian E. coli isolates, then sought for associations among resistance profile, plasmid content, virulence factor profile, and phylogenetic group. Avian-source isolates harbored greater proportions of MDR than their human counterparts, and avian ExPEC had higher proportions of MDR than did avian commensal E. coli. MDR was significantly associated with possession of the IncA/C, IncP1-α, IncF, and IncI1 plasmid types. Overall, inferred virulence potential did not correlate with drug susceptibility phenotype. However, certain virulence genes were positively associated with MDR, including ireA, ibeA, fyuA, cvaC, iss, iutA, iha, and afa. According to the total dataset, isolates segregated significantly according to host species and clinical status, thus suggesting that avian and human ExPEC and commensal E. coli represent four distinct populations with limited overlap. These findings suggest that in extraintestinal E. coli, MDR is most commonly associated with plasmids, and that these plasmids are frequently found among avian-source E. coli from poultry production systems.
Extraintestinal pathogenic Escherichia coli (ExPEC) have received considerable attention because of their complex nature and ability to cause a variety of important extraintestinal diseases in humans and animals (Johnson and Russo, 2002). Several subpathotypes of ExPEC have also been described, based on host source, specific disease syndrome, and virulence genotype. These include uropathogenic E. coli (UPEC) causing urinary tract infection (UTI), neonatal meningitis-associated E. coli (NMEC) causing meningitis of the newborn, and avian pathogenic E. coli (APEC) causing colibacillosis in poultry (Kaper, 2005). These diseases are costly to the human health care system and poultry industries, and cause considerable morbidity and mortality. Thus, the control of these diseases is an important area of focus.
It has been shown that ExPEC commonly possess large, transmissible plasmids encoding multidrug resistance (MDR) (Johnson and Nolan, 2009). By comparison, less is known about the prevalence of such plasmids in commensal E. coli. Further, the scope of horizontal gene transfer in relation to the dissemination of MDR in E. coli in the fecal and vaginal flora of healthy humans and animals is not known. Since ExPEC that cause clinical disease are thought to emerge from the fecal microbiota of healthy hosts, it is plausible that some commensal intestinal E. coli could also harbor large, transmissible plasmids conferring a multidrug-resistant phenotype. Notably, subsets of avian E. coli may represent a zoonotic threat via the consumption of contaminated poultry meat (Ewers et al., 2007; Johnson et al., 2007a, 2008, 2009a; Smith et al., 2007b). However, it is unclear whether these subsets also represent a threat with regard to the dissemination of MDR to human bacterial populations, which would be more likely if MDR in avian strains is encoded on mobile genetic elements.
To address these important knowledge gaps, we assessed 2202 previously characterized, disease-associated ExPEC and commensal E. coli from healthy human and avian hosts for their antimicrobial susceptibilities. Our goal was to combine these data with existing information to determine the associations among antimicrobial susceptibility, plasmid content, and virulence potential, in relation to host species and clinical origin.
The 2202 study isolates originated from a variety of sources, isolated between 1990 and 2005 (Table 1) (Obata-Yasuoka et al., 2002; Johnson et al., 2002a, 2002b, 2008; Rodriguez-Siek et al., 2005a, 2005b). All these isolates have been previously characterized for their virulence gene content and phylogenetic group membership, and some have been previously characterized for plasmid content (Johnson et al., 2007c). All isolates were stored frozen at −80°C in Brain Heart Infusion Broth (Difco Laboratories) with 10% glycerol and had undergone limited passage since their initial isolation in an attempt to ensure their genetic stability.
All E. coli isolates were examined for their antimicrobial susceptibilities by using the National Antimicrobial Resistance Monitoring System panels CMV5CNCD (some APEC isolates) and CMV1AGNF (remaining isolates) by Trek Diagnostics according to Food and Drug Administration, United States Department of Agriculture, and Clinical Laboratory Standards Institute recommendations (Clinical and Laboratory Standards Institute, 2010). A 96-well microtiter plate was used to test the susceptibility of strains to the following 14 antimicrobials (drug name abbreviation; breakpoint used): amikacin (AMI; ≥64μg/mL), amoxicillin/clavulanic acid (AUG; ≥32/16μg/mL), ampicillin (AMP; ≥32μg/mL), cefoxitin (FOX; ≥32μg/mL), ceftiofur (TIO; ≥8μg/mL), ceftriaxone (AXO; ≥4μg/mL), chloramphenicol (CHL; ≥32μg/mL), ciprofloxacin (CIP; ≥4μg/mL), gentamicin (GEN; ≥16μg/mL), kanamycin (KAN; ≥64μg/mL), nalidixic acid (NAL; ≥32μg/mL), streptomycin (STR; ≥64μg/mL), trimethoprim/sulfamethoxazole (SXT; ≥4/76μg/mL), and tetracycline (TET; ≥16μg/mL). Inoculation of panels was carried out according to the manufacturer's instructions. CLSI-specified control strains of E. coli, Staphylococcus aureus, Enterococcus faecalis, and Pseudomonas aeruginosa were used to validate each batch of plates. Strains displaying resistance to ≥3 classes of antimicrobial agents tested were defined as exhibiting MDR.
Isolates were also examined for the presence of plasmid replicon types by using multiplex polymerase chain reaction (PCR), as previously described (Carattoli et al., 2005; Johnson et al., 2007c). Additionally, selected isolates were examined for the presence of class 1 integron-associated genes using primers designed in this study (Table 2). PCR was performed as previously described (38). Amplicons were visualized on 2% TAE agarose gels alongside appropriate size standards (Minnesota Molecular, Inc.). Reactions were performed twice, and, if a discrepancy was identified, they were repeated again.
For all isolates, multiplex PCR-based genotyping for 32 ExPEC-associated virulence factor-encoding genes (VFs) was performed as previously described (Johnson and Stell, 2000; Rodriguez-Siek et al., 2005b). Some of these data are previously described (Rodriguez-Siek et al., 2005a, 2005b; Johnson et al., 2007a, 2008). Determination of major E. coli phylogenetic group (A, B1, B2, and D) was done according to the interpretive approach described by Clermont et al. (2000).
Comparisons of proportions were tested by using Fisher's exact test (two-tailed) or Chi-squared distributions (Snedecor and Cochran, 1989; Westfall, 1999) using SAS. Hierarchical two-way clustering, which clusters data based on overall traits on both the X and Y axis, was performed on the raw MIC values and visualized by using JMP for a graphical display of all characters used, in the context of the groups obtained from the cluster analysis (Johnson et al., 2008). Overall similarity relationships among the individual isolates with regard to VF profiles and phylogenetic group were assessed by using principal coordinates analysis (PCoA), a multivariate technique related to correspondence analysis enabling plotting of the major patterns within a dataset (Peakall and Smouse, 2006). By means of Genalex6 (Peakall and Smouse, 2006), PCoA was applied to the entire dataset. Each axis in PCoA represents a unique weighted composite of all the individual variables in the dataset. Individual isolates were assigned values on each axis on the basis of study variables and each variable's weighting factor on the particular axis. These values (for pairwise combinations of the first three axes) were plotted as a series of Cartesian grids, to show the distribution of the individual isolates (and their respective source groups) in two-dimensional space. They were also used in multivariate analysis of variance (MANOVA) to determine whether the comparison groups differed significantly according to the first three PCoA axes. If the initial multivariate ANOVA identified a significant overall difference, then univariate ANOVA was used to test pairwise comparisons of individual groups according to each PCoA axis, with use of a Bonferroni correction for multiple post-hoc comparisons as appropriate.
The 2202 total E. coli isolates were examined for 67 traits, including susceptibility to 14 antimicrobial agents, possession of 17 plasmid types, possession of 32 ExPEC virulence genes, and E. coli phylogenetic group membership. The goal of this work was to identify associations between MDR, plasmid replicon content, and virulence genotype. Compared with avian commensal E. coli (n=422), APEC isolates (n=909) exhibited a significantly greater prevalence of resistance (p<0.05) to AMP, GEN, KAN, STR, SXT, and TET (Table 3), and a significantly higher prevalence of the IncB/O, IncP1-α, IncFIIA, IncFIB, IncN, and IncHI2 replicons. Similarly, compared with human fecal E. coli (n=156), UPEC (n=559) exhibited a significantly greater prevalence of resistance to AMP, CHL, KAN, STR, SXT, and TET, and NMEC (n=70) had a significantly higher prevalence of resistance to STR, and a significantly higher prevalence of the IncB/O, IncP1-α, and IncFIB plasmid replicons. Likewise, compared with the human vaginal E. coli, NMEC had a significantly higher possession of the IncB/O, IncP1-α, and IncFIB plasmid replicons. When analyzed by phylogenetic group, the group B2 isolates had a lower prevalence of antimicrobial resistance, whereas the group A and B1 isolates tended to have a higher prevalence of resistance (Table 3). The B2 isolates also had a lower prevalence of some plasmid replicon types, including IncP1-α and IncI1 (Table 3).
Overall, 37.4% of isolates were susceptible to all antimicrobial agents tested, with most of the pan-susceptible isolates belonging to the UPEC group (Fig. 1). Among the remaining 62.6% of isolates, 20 distinct resistance profiles shared by 15 or more isolates were identified (Fig. 2). Among these MDR isolates, two profiles were identified with ≥8 resistances: AMP-AUG-CHL-FOX-GEN-STR-TET-TIO (n=41) and AMP-AUG-CHL-FOX-GEN-KAN-STR-TET-TIO (n=34); these occurred only among avian-source isolates. Overall, MDR (≥3 resistances) was most prevalent among APEC (34.9%) and AFEC (31.3%) isolates, and was less prevalent among UPEC (19.5%), NMEC (11.4%), and human commensal isolates (10.3%).
Comparisons of drug-resistant and drug-susceptible isolates according to plasmid replicon content showed that several plasmid types occurred in a significantly higher proportion of resistant isolates (p<0.05) than of their susceptible counterparts (Table 4). Chi-squared distributions were also used to identify significant associations between plasmid replicon type and resistance phenotype (Table 5). Using both approaches, several replicon types were strongly associated with MDR. Replicons associated with the greatest numbers of resistance markers included IncA/C (resistance to AUG, AMP, AXO, CHL, CIP, FOX, GEN, KAN, STR, SXT, TET, and TIO), IncP1-α (GEN, KAN, NAL, STR, and TET), IncI1 (AMP, AUG, AXO, GEN, KAN, NAL, STR, TET, and TIO), and IncFIB (AUG, AXO, GEN, KAN, STR, and TIO).
Among human isolates, significant associations of VF presence with individual resistance phenotype included those of kII, pap, ibeA, fyuA, iutA, traT, iha, and afa individually with AMP resistance; and of ibeA, bmaE, iutA, gafD, and afa individually with TET resistance (Supplementary Table S1; Supplementary Data are available online at www.liebertonline.com/fpd). Among avian isolates, significant associations included those of cvaC, iss, iutA, and traT individually with resistance to FOX, GEN, KAN, STR, and TIO, individually.
Similar analyses were performed to identify associations of plasmid replicon type with VFs (Supplementary Table S1). Among human isolates, highly significant associations included those of the IncFIB plasmid type with kI, kII, malPAI, ibeA, fyuA, cvaC, iss, iutA, traT, and fliC individually, and of the IncB/O plasmid type with cvaC and iss individually. Among avian isolates, significant associations included those of the IncFIB plasmid type with kI, ireA, ibeA, fyuA, cvaC, iss, iutA, and traT individually; of the IncN plasmid type with pap, ompT, ireA, fyuA, cvaC, iss, and iutA individually; and of several other plasmid types (i.e., IncB/O, IncA/C, IncP, IncFIIA, and IncI1) with cvaC, iss, and iutA individually.
Due to the previously established association of E. coli MDR with class 1 integrons, a subset of 1244 isolates were also examined for class 1 integron genes (intI1, sulI, and qacEΔ1), and merA and tetA, which are components of Tn21 and Tn10, respectively (Liebert et al., 1999). The three class 1 integron genes were jointly present in 344 (27.7%) of the isolates examined, presumptively defining the presence of class 1 integrons (Table 6). Of the presumptive class 1 integron-containing isolates, 55.2% contained merA and 66.6% contained tetA, values significantly greater than for the remaining isolates (2.2 and 13.0%, respectively; p<0.001). Plasmid replicon types significantly associated with class 1 integron positivity included IncA/C, IncP1-α, IncFIB, and IncI1. The only pathotype significantly associated with class 1 integron positivity was APEC, which comprised 89% of the isolates positive for class 1 integron genes (vs. 59.7% of class 1 integron-negative isolates; p<0.0001).
For an integrated analysis, PCoA was used to collapse the entire dataset into a small number of derived variables (i.e., principal coordinates or axes). The PCoA showed a separation of isolates based on source of isolation (Fig. 3), in which avian-source isolates were usually separated from human-source isolates. The overall ANOVA was highly significant (p<0.01), although only 18% of the variance in the dataset was explained by between-population differences, whereas 82% was explained by within-population differences. In pairwise comparisons between individual source groups according to their values on individual PCoA axes using MANOVA, all groups were significantly different (p<0.05) from one another on one or more axes (Supplementary Table S2).
The goal of this study was to assess and compare avian and human ExPEC and commensal E. coli for their antimicrobial susceptibility, and to identify correlations between antimicrobial resistance phenotype, plasmid replicon possession, virulence factor possession, and E. coli phylogenetic group membership. Multiple studies have examined the antimicrobial susceptibilities of ExPEC and commensal E. coli, including E. coli isolated from commercial poultry, the commercial farm environment, retail poultry meat, human UTI, and the fecal flora of healthy humans and animals (Schroeder et al., 2003; Yang et al., 2004; Johnson et al., 2005a, 2005b; Zhao et al., 2005; Miles et al., 2006; Diarrassouba et al., 2007; Wallmann et al., 2007; Khaitsa et al., 2008; Ozawa et al., 2008; Ahmed et al., 2009; Bonnet et al., 2009; Cook et al., 2009). However, fewer studies have explored the correlations between antimicrobial resistance and genetic traits, with contrasting results regarding the correlations between antimicrobial resistance phenotype, phylogenetic distribution, and virulence gene content (Johnson et al., 2009b).
Our results showed that MDR and the plasmids and mobile elements encoding for MDR are widespread among avian-source E. coli, irrespective of the clinical status of their host of isolation, while being only sporadically found among human-source E. coli isolates. When we compared VFs and resistance phenotype, certain highly significant correlations were observed. These included correlations between CHL resistance and possession of iha and pap; AMP resistance and possession of afa, iha, and iutA; and SXT resistance and possession of afa and iutA. Precedent exists for the association between VF and MDR. For example, it has been previously shown that MDR is common in bovine isolates carrying the afimbrial AfaE-VIII adhesin (Girardeau et al., 2003). In addition to these positive associations, though, a greater number of individual VFs were negatively associated with resistance phenotypes, including genes such as kI, kII, papACEFG, ompT, fyuA, sfa, fliC, and cdtB, and resistances including FOX, GEN, STR, TET, and TIO. The nature of these negative associations remains unclear.
A notable finding was the strong positive association of avian-source isolates harboring APEC-associated VFs (cvaC, iss, iutA, and traT), and a number of plasmid types (IncB/O, IncA/C, IncP1-α, IncFIIA, IncFIB, IncI1, and IncN), with certain resistance phenotypes (FOX, GEN, KAN, STR, TET, and TIO). These APEC VFs typically are encoded on IncFIB/IncFIIA plasmids known as ColV plasmids (Johnson et al., 2006a, 2006b). Although some ColV plasmids have been shown to contain resistance modules, they seem to be rare among ColV plasmids and typically involve Tn10-like elements encoding a limited number of resistances (Mellata et al., 2009; Fricke et al., 2009a). However, APEC strains also commonly possess ColV virulence plasmids with co-transferring R plasmids (Johnson et al., 2005c, 2006c, 2007c). The results of our replicon typing suggest that co-transferring ColV and MDR-encoding plasmids are widely prevalent among APEC isolates. Seemingly, then, the presence of an F plasmid in an avian E. coli strain might enhance its ability to acquire and disseminate other MDR-encoding plasmids, such as IncA/C, IncI1, and IncP1-α plasmids. Certainly, though, the complexities of the poultry production environment could also drive the selection of multidrug-resistant APEC, as multiple selective pressures exist (Singer and Hofacre, 2006). The mechanisms driving the emergence of co-transferring ColV plasmids and MDR-encoding plasmids need to be further investigated.
Regarding the commonality (or lack thereof) between human and avian-source E. coli, definitive conclusions are limited by our inclusion of isolates that differ temporally, geographically, and by source of isolation. Nonetheless, our results, as exemplified by the PCoA, suggest that although some overlap exists between isolates from poultry and humans, overall relatively few human-source isolates resemble the pool of avian-source isolates. Certainly, the human ExPEC isolates generally lacked MDR, whereas the APEC isolates had a high occurrence of MDR, and the two groups have previously been shown to differ in their VF content (Johnson et al., 2004, 2005b, 2007a, 2008, 2009a). However, within-group variation was extensive, and some of the multidrug-resistant avian-source isolates fell within the human-source clusters, and vice versa. This is supportive of previous work suggesting that certain subsets of ExPEC are capable of zoonotic transfer (Johnson et al., 2007a, 2008). Thus, although the potential for zoonotic transmission of multidrug resistant avian-source clones to humans probably does exist (Johnson et al., 2007b; Manges et al., 2007; Price et al., 2007; Jakobsen et al., 2010), the actual frequency of such transmission relative to the entire human and avian ExPEC populations might be relatively low.
The main MDR-associated plasmid types in this study were IncA/C, IncP1-α, IncF, and IncI1. IncA/C plasmids have received extensive recent attention because of their emergence in human clinical and production animal settings, broad host range, and ability to encode for extended-spectrum β-lactamases (ESBLs) (Welch et al., 2007; Fricke et al., 2009b; Call et al., 2010; Suzuki et al., 2010; Veldman et al., 2010). The IncA/C plasmids identified were exclusive to avian-source isolates and were associated with resistance to 12 or more of the antibiotics tested. IncI1 and IncF plasmids have also been associated with ESBL genes and MDR (Garcia-Fernandez et al., 2008; Marcade et al., 2009; Woodford et al., 2009; Smet et al., 2010; Sampei et al., 2010), and IncP1-α plasmids include the “Birmingham” plasmids known for their broad host range, promiscuity, and carriage of resistance genes (Thomas and Smith, 1987). Although MDR-encoding IncI1, IncF, and IncP1-α plasmids were identified among human-source E. coli isolates, their prevalence was low compared with that of avian-source isolates. This again suggests that the transfer of these elements, or of E. coli clones possessing these plasmids, from poultry to humans might be rare. However, these isolates were derived from multiple geographical sources, and these populations are rapidly changing with regard to plasmid possession and resulting MDR phenotypes. This underscores the need for continued monitoring for these mobile genetic elements.
Although disagreement exists regarding to what extent multidrug-resistant, poultry-associated strains have emerged and are persisting due to antimicrobial usage in the poultry production environment (Smith et al., 2007a), it is evident that MDR is now widespread in E. coli of poultry origin and is associated with conjugative plasmids. It should be acknowledged that there is bias in the dataset analyzed here with regard to geographical and temporal origins, thus limiting our ability to draw conclusions with regard to antimicrobial resistance phenotypes between the groups analyzed. However, this did not hamper our observations regarding the strong correlations between resistance phenotype and plasmid content. It is essential that future efforts address the risk of transfer of such plasmids from food animal bacterial populations to humans, and the underlying biological mechanisms enabling the dissemination and persistence of these plasmids among bacterial populations.
This project was funded, in part, by The Alliance for the Prudent Use of Antibiotics (APUA) NIH grant no. U24 AI 50139 (LKN and TJJ), the National Science Foundation grant no. EFO062666 (LKN and TJJ), and the Office of Research and Development, Medical Research Service, Department of Veterans Affairs (JRJ). The authors wish to thank Sandra Cloud-Rosenberger and the E. coli Reference Center for providing some of the strains used in this study, and Jessica Danzeisen for performing portions of the experimental procedures. This work was carried out in part by using computing resources at the University of Minnesota Supercomputing Institute.
No competing financial interests exist.