Abstract
Background Escherichia coli is commonly used as an indicator for antimicrobial resistance (AMR) in food, animal, environment, and human surveillance systems. Our study aimed to characterize AMR in E. coli isolated from retail meat purchased from grocery stores in North Carolina, USA as part of the National Antimicrobial Resistance Monitoring System (NARMS).
Methods Retail chicken (breast, n=96; giblets, n=24), turkey (n=96), and pork (n=96) products were purchased monthly from different counties in North Carolina during 2022. Label claims on packages regarding antibiotic use were recorded at collection. E. coli was isolated from meat samples using culture-based methods and isolates were characterized for antimicrobial resistance using whole genome sequencing. Multi-locus sequence typing, phylogroups, and a single nucleotide polymorphism (SNP)-based maximum-likelihood phylogenic tree were generated. Data were analyzed statistically to determine differences between antibiotic use claims and meat type.
Results Of 312 retail meat samples, 138 (44.2%) were positive for E. coli, with turkey (78/138; 56.5%) demonstrating the highest prevalence. Prevalence was lower in chicken (41/138; 29.7%) and pork (19/138;13.8%). Quality sequence data was available from 84.8% (117/138) of the E. coli isolates, which included 72 (61.5%) from turkey, 27 (23.1%) from chicken breast, and 18 (15.4%) from pork. Genes associated with AMR were detected in 77.8% (91/117) of the isolates and 35.9% (42/117) were defined as MDR (≥3 distinct classes of antimicrobials). Commonly observed AMR genes included tetB (35%), tetA (24.8%), aph(3’’)-lb (24.8%), and blaTEM-1 (20.5%), the majority of which originated from turkey isolates. Antibiotics use claims had no statistical effect on MDR E. coli isolates from the different meat types (X2=2.21, p =0.33). MDR was observed in isolates from meat products with labels indicating “no claims” (n=29; 69%), “no antibiotics ever” (n=9; 21.4%), and “organic” (n=4; 9.5%). Thirty-four different replicon types were observed. AMR genes were carried on plasmids in 17 E. coli isolates, of which 15 (88.2%) were from turkey and two (11.8%) from chicken. Known sequence types (STs) were described for 81 E. coli isolates, with ST117 (8.5%), ST297 (5.1%), and ST58 (3.4%) being the most prevalent across retail meat types. The most prevalent phylogroups were B1 (29.1%) and A (28.2%). Five clonal patterns were detected among isolates.
Conclusions E. coli prevalence and the presence of AMR and MDR were highest in turkey retail meat. The lack of an association between MDR E. coli in retail meat and antibiotic use claim, including those with no indication of antimicrobial use, suggests that additional research is required to understand the origin of resistance. The presence of ST117, an emerging human pathogen, warrants further surveillance. The isolates were distinctly diverse suggesting an instability in population dynamics.
Introduction
Surveillance of antimicrobial resistance (AMR) along the one-heath continuum, including in the food chain, is necessary to reduce the impact of resistant bacteria on health and prioritize policies and areas of intervention. The value and importance of one-health surveillance systems have been well described [1]. Escherichia coli, a gram-negative bacteria common to the human and animal gastrointestinal tract, is often used as an indicator organism in AMR surveillance systems [2]. Studies have shown that food products are potential reservoirs of pathogenic E. coli for humans [3, 4]. Most foodborne outbreaks in humans caused by E. coli have been associated with the consumption of contaminated food products of animal origin or contaminated with animal feces [5]; additionally, the role of these organisms in transmission of AMR to naturally-occurring human strains remains unknown and a concern. E. coli have flexible fitness mechanisms, making them easily adaptable to environmental conditions [6]. They can persist on surfaces for long periods of time and can be isolated from ready-to-eat food products [7].
To track AMR in food products of animal origin in the United States, the National Antimicrobial Resistance Monitoring System (NARMS) was established in 1996. The NARMS program which is operated by the Centers for Disease Control and Prevention (CDC), Food and Drug Administration (FDA) and U.S. Department of Agriculture (USDA) routinely conducts surveillance of foodborne pathogens in humans, retail meat products and food animals respectively. E. coli is one of the indicator bacteria identified for AMR surveillance in NARMS [8]. The NARMS program contributes to the promotion and protection of public health by disseminating knowledge regarding new bacterial resistance, the distinction between resistant and susceptible illnesses, and the effects of treatments meant to stop the spread of resistance [9]. Since its inception, the NARMS program has been solely responsible for continued tracking of antimicrobial resistance among human and animal related populations through the food supply.
The use of antimicrobials in any setting has potential implications on the development and maintenance of AMR. In the United States, retail food products including meat have been marketed with antibiotic use claims since the mid-2000s [10]. There are public concerns that the use of antimicrobials in food producing animals may affect the efficacy of similar drugs in human medicine especially through selection pressure of resistant bacteria and their ability to be transferred to humans through the food chain [11, 12]. The scientific evidence to support the risk of food products in disseminating AMR bacteria, including E. coli, is less clear. Studies have shown that resistance genes in bacteria such as E. coli can be transmitted at the human-animal-environment interface [13, 14]. Increasingly, modern molecular methods such as whole-genome sequencing (WGS) have been used to detect and trace the presence of resistance genes, mobile genetic elements and plasmids in E. coli, including from retail meat [15]. The NARMS surveillance program uses a WGS-based method for evaluating AMR in E. coli isolates from retail meat products.
Historically, the most prevalent AMR genes detected in E. coli from retail meat from the NARMS surveillance program include genes associated with erythromycin (mph(A)), tetracycline (tet(A), tet(B) and tet(C)), sulfonamide (sul1 and sul2) and plasmid mediated quinolone (qnr; gyr or par mutations) resistance [16–18]. Recent analysis of the NARMS Genome Trackr database showed that common phylogroups detected in E. coli recovered from retail meat were A, B1, B2, C, Clade I, D, E, F, and G [18]. Previous NARMS studies have reported E. coli prevalence of 47.5% in all retail meat products, with higher prevalence (90.7%) reported in turkey products [17, 19]. While one study reports more than 50% of the isolates demonstrated multidrug resistance (MDR), which showed turkey meat-derived E. coli isolates with the highest resistance other studies have reported a lower MDR prevalence of 14.3% [17, 19].
Here, we provide updated prevalence estimates and characterization of the common AMR mechanisms and phylogeny characteristics of E. coli isolated from retail meat with different antimicrobial use claims including chicken breasts, pork chops and ground turkey purchased from retail grocery stores in North Carolina, USA.
Materials and methods
Retail meat sampling
Between January and December 2022, 312 meat products comprised of retail chicken cuts (breast; bone-in/skin-on; n = 96), ground turkey (n=96), pork chop (n=96) and chicken giblets (liver, gizzard, or heart; n=24) were purchased from grocery supermarkets across five municipalities and eight counties in North Carolina in accordance with NARMS project protocol [20]. Monthly, 26 samples were randomly selected and purchased from grocery stores using zip codes that geographically represented each location based on US FDA assignment. At each sampling, eight chicken breast, eight pork chop, eight ground turkey and one package each of chicken heart, liver or gizzards were purchased.
As per the NARMS protocol for 2022, the demographic information collected for each purchased meat sample included antibiotic use claims apparent on product label, store name, store location, brand name, sell-by-date, purchase date, lab processing date and season of retail meat purchase (winter, spring, summer, and fall). Winter is considered December, January, and February; spring is March through May; summer is June through August; and fall is September through November. The samples were kept on ice in an isothermal container, during transportation from the grocery stores to the laboratory. On arrival at the lab, samples were refrigerated at 4°C and processed for detection of multiple bacteria, including E. coli.
Bacterial culture
We followed the provided NARMS 2022 retail meat surveillance laboratory protocol; briefly, this began with 50g of each sample aseptically cut into a stomacher bag where 250 ml of Buffer Peptone Water (Thermo Fisher Scientific, Waltham, MA) was added, and the sample was homogenized on a shaker at 200rpm for 15 mins. After mixing, the sample was placed into a sterile, plastic container (Fisherbrand™) with 50 ml of double strength (2X) MacConkey broth (Thermo Fisher Scientific, Waltham, MA), mixed well and incubated at 35°C for 24h. Following incubation, a 10ul loopful of enrichment broth was plated onto MacConkey agar (BD BBL™, Sparks, MD) and incubated at 35°C for 24h. One colony demonstrating typical E. coli phenotypic morphology (pink, round) was picked from each plate and streaked for isolation onto blood agar plates (Thermo Fisher Scientific, Lenexa, KS) which were incubated at 35°C for 18-24h. Indole (BD BBL™, Sparks, MD) and oxidase (BD BBL™, Sparks, MD) quick tests were performed on suspected E. coli colonies; indole-positive, oxidase-negative colonies were confirmed to be E. coli by matrix-assisted laser desorption/ionization-time of flight (MALDI TOF; BioMerieux). All confirmed E. coli isolates were saved in cryovials containing Brucella broth (Thermo Fisher Scientific, Lenexa, KS) with 15% glycerol and subsequently stored at -80 °C for sequencing analysis.
DNA extraction and Whole Genome Sequencing
DNA was extracted from each E. coli isolate using a modified version of the Qiagen DNeasy PowerLyzer microbial kit (Qiagen, Hilden, Germany). Using a 10μl loop, four passes of cells were added to 300μl of PowerBead solution in a 1.5 ml microcentrifuge tube. Next, 50μl of SL solution was added to the PowerBead tubes which were immediately vortexed for 5 seconds to re-suspend. The bacterial suspension (300μl) was transferred to the PowerBead tubes which were placed in a bead mill (speed=6; time=1min) to homogenize. The PowerBead tubes were centrifuged (30secs at 17,000xg), and the supernatant was transferred to new 1.5 ml microcentrifuge tube (Qiagen, Hilden, Germany). Afterwards 100μl of IRS solution was added to the supernatant and vortexed briefly, followed by a 5 min incubation at 4°C. After incubation the tubes were centrifuged, and the supernatant again transferred to a new 1.5 ml microcentrifuge tube each containing 900μl of SB solution. The tubes were vortexed and 650μl of supernatant transferred to the spin column for spin washing (3 washes) and final elution. The Nanodrop 2000 Spectrophotometer was used to conduct a quality check for all the DNA samples. The ratio of absorbance at 260 nm and 280 nm is used to assess the purity of DNA. A ratio of ∼1.8 is generally accepted as “pure” for DNA. Subsequently the DNA concentration was quantified using Qubit 4.0 Fluorometer (ThermoFisher Scientific, Waltham, MA). DNA libraries of each sample were prepared for whole genome sequencing (WGS) using a Nextera XT kit (Illumina, San Diego, CA). Briefly, 0.3 ng/μl of DNA from each E. coli isolate was pooled together and sequenced on an Illumina MiSeq platform (Illumina, San Diego, CA) using 2 x 250 or 2 x 300 paired-end approach. The raw paired end reads from the sequencer were demultiplexed and submitted to the NCBI database where they were assigned an accession number [21].
Assembly and assessment of genes
E. coli genome sequences were assembled de novo using SPAdes version 3.15.4 via a web-based genome assembly service provided by the Bacterial and Viral Bioinformatics Resource Center accessed online at https://www.bv-brc.org/app/Assembly2. The in silico analysis of acquired resistance genes and replicon typing for each E. coli isolate was conducted using the Mobile Element Finder tool (database version 1.0.2, 2020-06-09) accessed online via the Center for Genomic Epidemiology (CGE) website (https://cge.food.dtu.dk/services/MobileElementFinder/) [22]. The Mobile Element Finder tool interfaced with the ResFinder 4.1 tool to match individual genes for each E. coli isolate to an annotated resistance gene using a 90–100% identity, 60% minimum length, and 90% threshold [22]. PlasmidFinder 2.1 tool (database version 2023-01-18) set at a 95% minimum identity and 60% coverage was used for replicon typing. We defined molecular MDR as the presence of resistance genes conferring AMR to three or more distinct classes of antimicrobials in the database.
Multi-locus sequence typing (MLST) of E. coli isolates
In silico prediction of MLST was performed by submitting the assembled genome of an isolate to the E. coli PubMLST database (https:// pubmed.ncbi.nlm.nih.gov/30345391/). The MLST 2.0 (2022-11-14) tool on CGE website analyzed the contigs using previously described schemes by Achtman to assign sequence types (STs) based on allelic variations amongst seven housekeeping genes (adk, fumC, gyrB, icd, mdh, purA, and recA) [23]. E. coli isolates with identical sequences at all seven loci were assigned STs however those without perfect matches were usually identified as novel or unknown.
Determination of E. coli phylogroups and phylogeny
The phylogenetic classification of the E. coli genomes was conducted using in silico ClermonTyping 1.4.1 tool as previously described [24]. The ClermonTyper web interface is freely accessible at http://clermontyping.iame-research.center/. The Nextflow workflow was used to map the E. coli whole genome fastq pair end reads to a reference genome as well as call variants which were used to generate a maximum likelihood phylogenetic tree (https://gitlab.com/cgps/ghru/pipelines/snp_phylogeny). The clonal relationship between isolates was estimated using a pairwise single nucleotide polymorphism (SNP) analysis. The SNP analysis was used to determine how closely related the isolates were using ‘snp-dists’, a command line bioinformatics tool for transforming multiple DNA sequence alignment into a distance matrix (https://github.com/tseemann/snp-dists). Isolates that were less than 30 SNPs apart were related. The maximum likelihood phylogenetic tree was visualized using the interactive Tree of Life tool – iTOL version 6 (http://itol.embl.de/itol.cgi).
Data analyses
Data were analyzed using R version 4.3 statistical software. The data were summarized by calculating frequencies and proportions. Pearson’s chi-squared test and Fisher’s exact test were used to compare MDR prevalence across retail meat types and significance was determined at a p-value < 0.05. The accession numbers for 117 paired end reads for the NARMS E. coli isolates collected in North Carolina by the NCSU CVM Thakur Molecular Epidemiology Laboratory for 2022 have been uploaded onto the National Center for Biotechnology Information (NCBI) database https://www.ncbi.nlm.nih.gov under bio projects accession number PRJNA292663. The additional data file for this study contains a list of accession numbers for individual Sequence Read Archive (SRA) for the E. coli isolates.
Results
From January through December 2022, a total of 312 meat samples were purchased from 48 grocery stores in North Carolina, representing 8 counties. Of these samples, 69 (22.1%) were labeled with “no-antibiotic-ever”, 39 (12.5%) were labeled as organic, and 204 (65.3%) were identified as having no antibiotic use claim present. Of the 312 samples, 138 were positive for E. coli with an overall prevalence of 44.2%. Of these, 78 (56.5%) E. coli isolates were recovered from ground turkey, followed by chicken breast (n=27, 19.6%), chicken giblets (n=14, 10.1%) and pork chops (n=19, 13.8%). Regardless of meat type, the distribution of E. coli isolates across meat types was 15.2% (n=21) in organic products, 18.8% (n=26) in “no-antibiotic-ever” products and 65.9% (n=91) in products with “no claims”. The observed prevalence difference was not statistically significant across label claims (X2=2.67; p=0.26). The prevalence of E. coli isolates differed by month of sampling with the highest prevalence observed in May (17/26; 65.3%), followed by February and November (14/26; 53.8% each) while the least prevalence was observed in December (4/26; 15.3%) as shown in Figure 1. This observed difference between months was statistically significant (X2=19.29; p=0.05). However, there was no significant association between E. coli prevalence and season of retail meat purchase from 138 positive isolates (X2=4.73; p=0.19). The highest prevalence (41/78; 52.6%) was observed in Spring and the lowest prevalence (29/78; 37.2%) was observed in Summer.
The highest E. coli occurrence was observed in May and the lowest in December 2022.
Of all E. coli isolates, 84.8% (117/138) had good quality sequence data available for further characterization. Of these, 72 (61.5%) were recovered from ground turkey, 27 (23.1%) from chicken breast and 18 (15.4%) from pork chops. AMR genes were detected in 77.8% (91/117) of isolates, and 35.9% (42/117) were defined as MDR. The distribution of MDR E. coli isolates across retail meat type was 83.3% (35/42) for ground turkey, 9.5% (4/42) from chicken breast and 7.1% (3/42) from pork chop, respectively. Sequencing analysis of AMR genes showed that the most prevalent resistance genes detected in the E. coli isolates from retail meats belonged to aminoglycosides (91.5%; 107/117), tetracyclines (59.8%; 70/117), beta-lactamases (31.6%; 37/117), folate pathway antagonists (25.2%; 29/117) and quinolones (11.1%; 13/117) as shown in Table 1.
Within each retail meat type, we assessed the impact of label antibiotic use disclosure as displayed on the packaging on MDR. The different label types did not statistically impact the resistance of MDR E. coli isolates (p = 0.33). MDR in turkey E. coli isolates was significantly different when compared to all other retail meat types regardless of the label claim (p < 0.01) (Table 2).
The MLST analysis showed that the 117 E. coli isolates belonged to 81 known sequence types (STs) with the most prevalent being ST117 (8.5%;10/117), ST297 (5.1%; 6/117), ST58 (n=4; 3.4%), and three isolates each (2.6%) for ST10, ST126, ST602, and ST1079. Numerous other STs were identified among the E. coli isolates with one or two isolates represented; ST131 and ST371 were not detected. The most prevalent MLST types detected in the isolates are shown in Figure 2.
Each bar represents the various E. coli sequence types for isolates obtained from the different retail meat sources.
The 117 isolates in this study belonged to nine different phylogroups with the most prevalent phylogroup representing B1 (n=34; 29.1%) followed by phylogroup A (n=33; 28.2%), phylogroup B2 (n=16; 13.7%) and phylogroup G (n=11; 9.4%). Isolates with phylogroup B1 were recovered from ground turkey (19/34; 55.9%), chicken breast (8/34; 23.5%) and pork chop (7/34; 20.6%) (Figure 3). Most isolates assigned ST297 (n=4) and ST58 (n=2) belonged to phylogroup A, while half of the isolates assigned ST117 (n=5) belonged to phylogroup B1.
Thirty-four different replicon types were observed among E. coli isolates from retail meat. The most prevalent replicon types were IncFIB(AP001918) (n=73, 62.4%), Col(MG828) (n=50, 42.4%) and IncFIC(FII) (n=43, 36.4%) as shown in Table 3. Thirteen replicon types were common to isolates from all retail meat sources: IncFIB(AP001918), IncFII, IncFIC(FII), IncFII(29), Col(MG828), Col156, ColpVC, Col440I, IncX1, IncFII(pCoo), IncFIA, IncI1, and p0111. IncFIB(AP001918) was the most prevalent in all meat types: ground turkey (n = 51), chicken breast (n=18) and pork chop (n= 3) followed by Col(MG828) detected in ground turkey (n = 35), chicken breast (n=12) and pork chop (n= 3).
Overall, 117 isolates were used to construct a SNP-based maximum likelihood phylogenetic tree (Figure 4). The E. coli isolates from retail meat were very diverse and only isolates from the same meat type were clonally related. More than half (57.3%) of the isolates were clustered into two main phylogroups and seven different STs. The SNPs matrix based on the core genome of 117 E. coli strains showed five clonal relationships with a pairwise SNP difference of below 30 (Table 4). E. coli isolates with a clonal relationship were clustered together based on phylogroups and STs as shown on the phylogenetic tree in Figure 4.
SNP-based maximum likelihood phylogeny of E. coli isolates visualized in interactive Tree of Life tool (iTOL). The tree was rooted in a reference E. coli strain K-12 MG1655. Clustering of isolates was found to be following the core genome and SNP-based phylogenies as clustering of isolates with the same sequence types and phylogroups was consistent. Shown for each isolate is the meat type, season, phylogroups, sequence types, AMR genes (brown) and plasmids (purple).
Seventeen isolates (14.5%) carried AMR genes on plasmid replicons identified on the same assembly scaffold, out of which 88.2% (15/17) were recovered from ground turkey and 11.8% (2/15) from chicken breast. The most prevalent replicon types were IncHI2 and IncHI2A harboring aminoglycoside resistance gene aph(3”)-Ib in eight isolates (47.1%) and in combination with aph(6)-Id in three of the eight isolates recovered from ground turkey. Other AMR genes carried on plasmids include blaTEM-1A on IncI-1; qacE + sul1 on IncFIC (FII); qnrB19 on Col440I amongst others (Table 5).
Discussion
In this study, we assessed the current prevalence and describe the characteristics of clonal types and AMR in E. coli isolates from retail meat, including chicken breast, ground turkey and pork chops purchased from retail grocery stores in North Carolina, USA. Using labeling claims, we were able to assess the role of antibiotic use on the occurrence and characteristics of AMR. We found the presence of AMR genes against different antimicrobial drug classes (aminoglycosides, tetracyclines, beta-lactamases, folate pathway antagonists, quinolones, and macrolides) in E. coli isolates from chicken, turkey and pork sources, but the prevalence differed by the various retail meat types. Our study results are consistent with historical genotypic prevalence reported in previous NARMS surveillance program [16–18].
In the present study, the highest prevalence of E. coli in the retail meat types was observed in May (spring). Our results show that there was no statistically significant association between the season of retail meat purchase and the detection of E. coli, although E. coli prevalence did differ by the month of sampling. The role of season on the presence of E. coli in retail meats is unclear as reports of another NARMS study in California found a prevalence of 36.7% in spring when compared to winter [17], while a similar study done in Canada found no definite seasonal trend [25]. The interplay between the effect of season and the prevalence of E. coli in retail meat is not fully understood [25].
Across all isolates from all retail meat sources, we found that there was a high proportion of genotypic AMR (77.8%) among E. coli and a moderate occurrence of MDR E. coli (35.9%); this prevalence is consistent with previous data collected from retail meat in the United States over the last decade [9, 17, 19, 26, 27]. The prevalence of resistance was highest in E. coli from retail ground turkey when compared to other meat types. This also is consistent with findings from similar studies in Arizona which reported a high resistance prevalence of E. coli (90.7%) in retail turkey [19] and California reporting a prevalence of 70.4% [17]. This suggests there are differences in the management and production of turkey which may select for antimicrobial resistance or allow its persistence. It is important to note that producers in the United States are not required to make public reports on antimicrobial use data, hence it is difficult to link on-farm use of antimicrobials to development of AMR. Although between 2013 and 2017, the largest US turkey production corporations significantly reduced their overall use of antimicrobials attributed to the full implementation of FDA GFI #213, and improved antimicrobial stewardship amongst other factors [28].
The prevalence of genotypic resistance for ampicillin detected in E. coli from all meat types (31.6%; 37/117) was higher than the national average (11.6%; 54/466) for 2022; however, the prevalence of genotypic resistance for tetracycline was not different in E. coli from ground turkey (68.1%; 49/72) when compared to the national average of 62.5% (619/990). This was lower for chicken (25.9%; 7/27) and pork (22.2%; 4/18) when compared to the national average of 31.5% (147/466) and 50.4% (192/381) respectively [9]. In addition, to the ampicillin and tetracycline resistance genes, our results also show sulfonamide resistance genes and plasmid mediated quinolone resistance genes detected in retail meat. This finding aligns with past reports of the NARMS surveillance program [16–18].
Our data indicated that retail ground turkey may serve as an important source of MDR E. coli with the potential for transmission to humans if proper food handling practices are not used. However, there was no significant association between MDR E. coli, and the different antibiotics-use claims on the meat package labels. High resistance rates observed among E. coli isolates from turkey with “no antibiotic-use claims” may indicate that they were raised using conventional methods, however, this cannot be confirmed. This is consistent with the reports of a similar study in Arizona, USA [19]. In conventional turkey production, while there may be a greater potential for antimicrobial use when compared to “no antibiotics ever” and “organic” production, we are not able to definitively determine the use.
The E. coli isolates from retail meat samples displayed diverse STs and phylogroups suggesting the possibility that the isolates had evolved overtime from different E. coli clones. The most prevalent known STs among the isolates were ST117 (recovered from chicken and turkey), ST297 and ST58 (both recovered from chicken breast). ST10 was also detected in isolates from turkey and pork. A study conducted in California, USA in 2018 identified ST117 and ST10 in E. coli isolates from retail meat samples and humans [29]. Our results are consistent with reports of another related study that detected ST117 in E. coli recovered from commercial chicken and turkey [30] but at variance with others [30, 31] that reported ST58 and ST10 were dominant in turkey clinical E. coli isolates. The most dominant phylogroups in the present study were phylogroups A, B1 and B2 which was detected in isolates from all retail meat types and consistent with the literature [18, 32, 33].
The different replicon types along with the numerous E. coli STs in the present study highlight the diversity and complex nature of these indicator organisms. The most prevalent replicon types observed were IncFIB(AP001918), Col(MG828), IncFIC(FII), Col156, Incl1, p0111, IncFII, IncHI2, IncHI2A, Col440I and IncFIA. Seventeen isolates carried AMR genes on plasmid replicon identified on the same assembly scaffold. Replicon typing showed IncHI2 and IncHI2A harbored AMR gene aph(3”)-Ib + aph(6)-Id in some isolates recovered from ground turkey. Interestingly, the IncFIC(FII) plasmid was observed to harbor the qacE and sul1 AMR gene on the assembly scaffold originating from one E. coli ST117 strain, recovered from retail chicken meat in the present study. This is of concern as ST117 is a known pathogenic E. coli lineage that has been implicated in both human and animal diseases [29, 34]. Other AMR genes carried on plasmid contigs in our study include blaTEM-1A on IncI-1; qacE + sul1 on IncFIC (FII); qnrB19 on Col440I amongst others. Other studies have reported E. coli isolates from food animals carrying the plasmid mediated quinolone resistance gene qnrB on a Col440I replicon carrying plasmid contig supporting our study results [35, 36].
Studies have reported a high diversity between E. coli isolates recovered from retail meat types as evident by the wide variety of STs observed [29, 37]. The E. coli isolated from retail meat in the present study were genetically diverse hence we did not detect any clonal relationship between isolates from the different meat types. However, our results show that isolates from the same meat type i.e., ground turkey were found to be closely related within 20 SNPs.
This study is not without limitations, particularly the small number of E. coli isolates characterized by sequencing may not fully encompass the diversity of E. coli strains recovered from retail meats, although it is a representative random sampling. Another limitation of retail level surveillance used in this study is the inconsistent or limited availability of metadata for each sample. This impacts the ability to assess factors along the food production chain from farm-to-fork that can potentially contribute to AMR and MDR in E. coli recovered from retail meat products. Nevertheless, meat type, label claim, packaging and season only accounted for a small portion of the variability in AMR genetic determinants among the E. coli isolates in this investigation, which were genotypically heterogeneous with regards to AMR.
Conclusions
In conclusion, we detected MDR E. coli from retail meat types in North Carolina. Our results showed that > 40% of the retail meats purchased from grocery stores were contaminated with E. coli, and of these the vast majority were resistant to aminoglycosides (aph(3”)-Ib & aph(6)-Id), tetracycline (tetA & tetB) or ampicillin (blaTEM-1). The resistance prevalence was highest among E. coli isolates from turkey for most AMR genes detected. Our data indicated that ground turkey may serve as an important source of MDR E. coli. The isolates were diverse as only ten showed clonal relationships with a pairwise difference of < 20 SNPs.
Ten of the E. coli isolated from retail meat had ST117 which is an emerging sequence type implicated as a human pathogen. It is important to emphasize that the bacteria isolated from retail meat samples in the present study were generic E. coli hence the risk of impacting negatively on human health is quite low. However, we cannot rule out the possibility of horizontal gene transfer between E. coli strains because our results show that plasmids, which are mobile genetic elements harbored some AMR genes which should be a source of concern for human health. Therefore, surveillance of these indicator bacteria strains should continue to serve as a warning for preventing the spread of AMR along the food chain. To further understand the transmission dynamics, additional studies are required especially due to the persistence of these acquired AMR genes and plasmid replicon types in different E. coli STs.
Authors’ contributions
MJ, ST and PC made substantial contributions to conception and design. EH collected data; EH, MA and CG performed laboratory analysis; EH and LH generated the whole genome sequence profile for all the isolates. MA performed bioinformatic analysis, interpreted the data and wrote the first draft of the manuscript. MJ, ST and PC revised article critically for important intellectual content. All authors read and approved the final manuscript.
Acknowledgements
The authors acknowledge the funding received from FDA National Antimicrobial Resistance Monitoring System – ST Grant number 1U01FD007145-01 (URL: https://www.cdc.gov/narms/index.html) and GenomeTrakr program – Grant number 1U18FD00678801 for whole-genome sequencing of the isolates.
References
