Isolation and molecular characterization of multidrug-resistant Escherichia coli from chicken meat

Antibiotic-resistant Escherichia coli (E. coli) are common in retail poultry products. In this study, we aimed to isolate and characterize multidrug resistant (MDR) E. coli in raw chicken meat samples collected from poultry shops in Sylhet division, Bangladesh, as well as to determine correlation between resistance phenotype and genotype. A total of 600 chicken meat swabs (divided equally between broiler and layer farms, n = 300 each) were collected and the isolates identified as E. coli (n = 381) were selected. Disc diffusion antimicrobial susceptibility assay showed resistance of these isolates to ampicillin, erythromycin, tetracycline, streptomycin, trimethoprim-sulfamethoxazole, chloramphenicol, and gentamicin. Polymerase chain reaction (PCR) identified several antibiotic resistance genes (ARGs) in our isolates. Among these ARGs, the prevalence of tetA (for tetracycline) was the highest (72.58%) in broiler chicken isolates, followed by sul1 (for sulfonamide; 44.16%), aadA1 (for streptomycin; 33.50%), ereA (for erythromycin; 27.41%), aac-3-IV (for gentamicin; 25.38%), and the two genes cmlA (24.87%) and catA1 (8.63%) for chloramphenicol. On the other hand, the respective prevalence in layer chicken isolates were 82.06%, 47.83%, 35.87%, 35.33%, 23.91%, 19.02%, and 5.43%. Furthermore, 49.23% of the isolates from broiler chicken were MDR, with the presence of multiple antibiotic resistance genes, including 3 (40.11%) and 4 (9.13%) genes. On the other hand, 51.09% of layer chicken E. coli isolates were MDR, with 3, 4 or 5 ARGs detected in 36.41%, 14.13%, and 0.54% of the isolates, respectively. We also found that 12.8% of broiler chicken E. coli isolates and 7.61% of layer chicken isolates carried genes coding for extended-spectrum SHV beta-lactamases. Lastly, we report the presence of the AmpC beta-lactamase producing gene (CITM) in 4.56% and 3.26% of broiler and layer chicken E. coli isolates, respectively. We found significant correlations between most of the antimicrobial resistant phenotypes and genotypes observed among the investigated E. coli isolates. Our findings highlight the need for the prudent use of antimicrobials in chickens to minimize the development of antibiotic-resistant bacterial strains.


Isolation and identification of E. coli.
A total of 600 swabs were collected randomly from broiler (n = 300) and layer (n = 300) chicken meat samples, derived from 100 different retail poultry shops at Sylhet division of Bangladesh. Gram staining, growth characteristics on culture media (including nutrient broth, nutrient agar, MacConkey's agar, Eosin Methylene Blue agar; all from Merck, Germany), and results of biochemical tests (including sugar fermentation, indole, methyl red (MR), Voges-Proskauer (VP), and citrate utilization tests) were used for identification and isolation of E. coli, as previously described 19 . Molecular confirmation of the isolates was performed using PCR targeting the 16S rRNA, using a primer set specific for E. coli, as previously described 20 . The isolates identified as E. coli (n = 381; 197 from broiler and 184 from layer chickens) were selected for further investigation.
Antimicrobial susceptibility testing. The susceptibilities of the 381 chicken meat-derived E. coli isolates to a panel of commonly used antibiotics were determined using the Kirby-Bauer method on Mueller-Hinton agar plates (Merck, Germany) according to the guidelines and breakpoints of the Clinical and Laboratory Standard Institute 21 . The antimicrobial discs used, which were all obtained from Oxoid (UK), included: trimethoprimsulphamethoxazole (23.75 µg), chloramphenicol (30 µg), erythromycin (15 µg), gentamicin (10 µg), tetracycline (30 µg), streptomycin (10 µg), and ampicillin (10 µg). E. coli ATCC 25,922 (American Type Culture collection, Manassas, VA, USA) was used as a control strain. Test results were only validated when the diameters of the inhibition zones of the E. coli ATCC 25922 control strain were within the performance ranges. Resistant and intermediate resistant isolates were considered as non-susceptible as previously described 22 . E. coli was defined as multidrug resistant isolate when it was found non-susceptible to at least one agent in three or more different classes of antimicrobial agents, excluding the broad-spectrum penicillins without a β-lactamase inhibitor 22 .
Extraction of bacterial genomic DNA. All E. coli isolates (n = 381) were cultured overnight in nutrient broth at 37 °C and then bacterial genomic DNA was extracted using the Phenol-Chloroform Isoamyl Alcohol (PCI) method, as described previously 23 . The average concentration and purity of the extracted DNA were determined using the Nano Drop™ 2000c spectrophotometer (ThermoScientific, USA).

Scientific Reports
| (2020) 10:21999 | https://doi.org/10.1038/s41598-020-78367-2 www.nature.com/scientificreports/ PCR amplification and detection of antibiotic resistant genes. All the tested E. coli isolates (n = 381) were PCR-screened for the presence of seven non-beta-lactam and six beta-lactam antibiotic resistant genes (ARGs) using a combination of two uniplex and two multiplex assays. The resistant genes for tetracycline (tetA) and streptomycin (aadA1) were amplified individually using set 1 and 2 primers, respectively (Table 1). Set 3 and 4 primers were used to detect some other ARGs (sul1, catA1, cmlA, ereA, aac-3-IV, blaSHV, and CITM) and four types of ESBL genes (blaTEM, blaCTX-M, blaCTX-M-1, and blaCTX-M-2), respectively (Table 1). All PCR amplifications were conducted in a thermal cycler (Gene Atlas, Japan) using the conditions listed below. The basic setup of the uniplex PCR amplification for set 1 consisted of an initial denaturation step at 95 °C for 15  Statistical analysis. The antibiotic resistance data are expressed as percentages or frequency of the E. coli isolates. A two-way analysis of variance (ANOVA) without replication was used to determine the significant differences in the levels of resistance prevalence among the selected antibiotics, between broiler and layer chickens, Table 1. List of primers used in the current study for uniplex (sets 1 and 2) and multiplex (sets 3 and 4) PCR assay formats for the amplification of beta-lactam and non-beta-lactam ARGs in the investigated E. coli isolates. F forward primer, R reverse primer.  www.nature.com/scientificreports/ as well as among the four districts under study. A P value of < 0.05 was considered to be statistically significant. These statistical analyses were carried out using the GraphPad Prism (version 6; GraphPad Software Inc.; USA).

Results
Prevalence of E. coli in broiler and layer chicken meat swabs. We collected a total of 600 chicken meat swab samples (75 from broiler and 75 from layer chickens from each of the four districts of Sylhet division; Sylhet, Moulavibazar, Sunamganj, and Habiganj). Out of the 600 samples, 381 E. coli isolates (63.5%) (197 from broiler and 184 from layer chicken) were identified using staining, cultural, and biochemical tests ( Table 2). There was no statistically significant difference in the prevalence of E. coli between samples from broiler and layer chickens (P = 0.06), nor among the four districts (P = 0.37).
Prevalence of antimicrobial resistance. All E. coli isolates (n = 381) were tested for resistance to seven different antimicrobial agents by disc diffusion method. As shown in Table 3, 286 (75.06%) of the isolates were MDR. Resistance to ampicillin, erythromycin, and tetracycline were the most prevalent in the isolates (98.95%, 89.5%, and 85.3%, respectively). No significant difference in resistance patterns was observed in isolates from broiler and layer chicken meat (P = 0.18).
Prevalence of non-beta-lactam ARGs. The overall prevalence of non-beta-lactam ARGs among the investigated E. coli isolates in relation to type of hens (broiler or layer) in Sylhet division are given in Table 4. The most prevalent gene was that of tetA (for tetracycline resistance; harboured by 77.17% of the isolates), which was followed by sul1 (for sulphonamide resistance; 45.94%), aadA1 (for streptomycin resistance; 34.65%), ereA (for erythromycin resistance; 31.23%), aac-3-IV (for gentamicin resistance; 24.67%), and the two genes cmlA (22.05%) and catA1 (7.09%) for chloramphenicol resistance. Additionally, there was a significant difference in the prevalence of the seven non-beta-lactam ARGs among the E. coli isolates (P = 0.0001) but there was no significant difference in the prevalence of each antibiotic resistance gene in broiler versus layer chickens (P = 0.42). As mentioned previously, the investigated E. coli isolates were from broiler and layer chickens that have been collected from four districts within Sylhet division. The prevalence of E. coli isolates from broiler or layer chicken harbouring non-beta-lactam resistance genes in relation to these districts is shown in Table 5. There was no significant difference in the prevalence of sul1, catA1, ereA, aac-3-IV, tetA and aadA1genes between broiler and layer chickens, nor between the four districts (P > 0.05). On the other hand, the prevalence of cmlA gene was significantly higher in broiler chicken than in layer chicken. Moreover, the prevalence of cmlA gene was significantly different among the four districts under study (P < 0.05). Table 2. Prevalence of E. coli in broiler or layer chicken meat swab specimens, collected from various retail shops at Sylhet division. a P values were calculated using a two-way analysis of variance (ANOVA) without replication. P values > 0.05 were considered to be statistically nonsignificant. # Variance between broiler and layer chickens. ## Variance among the four districts under study.  The analysis was performed against five antimicrobial categories (representative antimicrobials tested in this analysis and the respective genes involved are shown in brackets): aminoglycosides (gentamicin-aac-3-IV, streptomycin-aadA1), tetracyclines (tetracycline-tetA), phenicols (choloramphenicol-cmlA and catA1), macrolides (erythromycin-ereA) and folate pathway inhibitors (sulfonamide/trimethoprim-sul1). The analysis showed 26 resistance profiles (Fig. 1), the most frequent among which (n = 16) correlated with isolates from layer chickens and harboured the resistance genes for sulfonamide, erythromycin and tetracycline. A total of 191 (50.13%) out of 381 E. coli isolates from broiler and layer chickens of Sylhet division carried more than 3 ARGs in their genomes. The majority of the E. coli isolates harboured resistance genes to three classes of antibiotics (146 isolates; 38.32%), while 44 isolates (11.55%) and 1 isolate (0.26%) possessed four and five antibiotic resistance genes, respectively (Table 6). Within the triple-antibiotic resistant E. coli isolates, 79 isolates were from broiler chickens, while 67 isolates were from layer chickens. Out of the 44 isolates that carried four antibiotic resistance genes, 18 (9.13%) were from broiler chickens and 26 (14.13%) were from layer chickens. However, in contrast to the one E. coli layer chicken isolate (0.54%), none of the broiler chicken isolates carried five antibiotic resistance genes. There was a significant difference in the prevalence of MDR genes (P = 0.01); however, there was no significant difference in their prevalence between broiler and layer chickens (P = 0.83).
Next, we aimed to determine the prevalence of E. coli isolates (from broiler or layer chicken) harbouring MDR genes (non-beta lactam antibiotics) in relation to the four districts of Sylhet division under study (Table 7). Within the three categories of MDR isolates (those having 3, 4, or 5 MDR genes), all the prevalence differences between the four districts were found to be non-significant (P > 0.05). Similarly, there were no significant Table 4. Overall prevalence of non-beta-lactam ARGs among the investigated E. coli (n = 381) isolates in relation to type of hens (broiler or layer) in Sylhet division of Bangladesh. a P values were calculated using a two-way analysis of variance (ANOVA) without replication. P values < 0.05 are highlighted in bold. # Variance between broiler and layer chickens. ## Variance among the antibiotic resistant genes.      Statistically significant differences between the prevalence of both types of beta-lactamase coding genes (ESBL and AmpC) were observed (P = 0.001) but there were no significant differences between broiler and layer chickens (P = 0.24). The data regarding the prevalence of beta-lactam (ESBL and AmpC) ARGs among the investigated isolates (from broiler or layer chickens) in relation to the four districts of Sylhet division are shown in Table 9. There was a statistically significant higher prevalence of isolates possessing the SHV (ESBL) gene in broiler chicken than in layer chicken (P = 0.01), while there were no significant differences among the four districts (P = 0.07). In the case of the CITM gene (AmpC), there was no significant differences in the prevalence of isolates possessing this gene between broiler and layer chickens, nor between the four districts (P > 0.05).

Correlation between antimicrobial resistant phenotypes and genotypes.
In this study, significant correlations (r 2 > 0 and P < 0.05) were found between most of the antimicrobial resistant (AMR) phenotypes and genotypes observed among the investigated E. coli (n = 381) isolates (Table 10). Comparatively stronger correlation was found between Gentamycin and aac-3-IV among the E. coli strains isolated from layer chicken (r 2 = 0.791 and P < 0.001). On the other hand, no significant correlation was observed between chloramphenicol AMR phenotype and catA1 gene among isolates from broiler chicken (r 2 = 0.018 and P = 0.067). Similarly, in the case of layer chicken, no significant correlation was observed between erythromycin AMR phenotype and ereA gene among the E. coli isolates (r 2 = 0.001 and P = 0.672).

Discussion
Chicken meat is a potential source of multi-drug resistant ESBL-producing E. coli strains, which are responsible for serious human health concerns worldwide 32 . In this molecular study, we isolated E. coli from chicken meat and examined the existence of ARGs. The high prevalence of antibiotic resistant E. coli isolates in our findings indicates that the raw chicken meat from retail poultry shops could be contaminated with antimicrobial-resistant E. coli. This is alarming for developing countries like Bangladesh, where retail poultry shops hardly maintain proper hygienic condition during processing of chicken meat.
In the current study, the overall prevalence of E. coli in chicken meat was 63.5% whereas, 65.67% of broiler and 61.33% of layer meat swabs tested positive for E. coli. Other research groups detected high frequency of E. coli in poultry meat [33][34][35] . In contrast, Ranjbar et al. 36 , Moawad et al. 37 and Younis et al. 38 showed lower prevalence of E. coli in raw chicken meat. Another study in India found that 78% of broiler chicken meat specimens from retail shops were contaminated with E. coli 39 . Jakaria et al. 40 reported that in Bangladesh, the prevalence rates of E. coli in layer, broiler, and indigenous chicken were 78.67%, 82% and 70%, respectively.
In this study, the disc diffusion method showed relatively higher frequency of antibiotic resistance and multidrug resistance among the investigated E. coli isolates than the genotypic analysis. This may be due to the possible protective role of the tested genes against multiple (often related) antimicrobial drugs that are not structurally or mechanistically related 41 . We found that more than 80% of the tested E. coli isolates were resistant to the common Table 8. Overall prevalence of beta-lactam (ESBL and AmpC) antibiotic resistant genes among the investigated E. coli (n = 381) isolates in relation to type of hens (broiler or layer) in Sylhet division of Bangladesh. a P values calculated using a two-way analysis of variance (ANOVA) without replication. P values < 0.05 are highlighted in bold. # Variance between broiler and layer chickens. ## Variance among the beta-lactam antibiotic resistant genes.

Selected ESBL genes
AmpC gene
In this study, among the seven tested non-beta-lactam antibiotic resistance genes, the prevalence of tetracycline (tetA), sulphonamides (sul1), and streptomycin (aadA1) resistance genes were the highest (72.58%, 44.67% and 33.50%, respectively), followed by erythromycin (ereA) (31.23%) and the two chloramphenicol resistant genes cmlA (22.05%) and catA1(7.09%). Our findings indicate that chicken meat E. coli could be a reservoir of resistance genes, which may later become transferred to other common pathogens.
In a similar study, the prevalence rates of tetA, aadA1 and sul1ARGs in E. coli isolated from Vietnam were found to be 81%, 81%, and 27.1%, respectively 24 , while their respective rates in Portugal were 41.1%, 70.6% and 23.5% 45 . Additionally, Moawad et al. 37 reported that E. coli isolates from poultry meat were resistant to tetracycline (80.9%), streptomycin (61.9%) and trimethoprim/sulphamethoxazole (61.9%), which is higher than the prevalence rates reported in our present study. In Bangladesh, 37-100% of the poultry-derived E. coli strains were found resistant to chloramphenicol, tetracycline, streptomycin, erythromycin and penicillin, as reported by Rahman et al. and Islam et al. 46,47 . The lowest prevalence rate was observed with aac-3-IV (gentamicin resistance gene; 24.67%) which may be attributed to the very low absorption rate of gentamycin in poultry 48 .
We showed that 75.06% of our E. coli isolates were resistant to at least three antibiotics. In a study conducted in Iran, a high prevalence rate (64.91%) of MDR strains among E. coli isolates from commercial chicken meat has been reported 49 . Studies have demonstrated even higher prevalence rates of MDR E. coli in broiler (94%) and layer (60%) chicken in India 50 and in Nepal (80.0%) 51 . Such high prevalence of MDR isolates may be due to misuse of antibiotics, which may ultimately replace the drug sensitive microorganisms in an antibiotic saturated environment 52 .
In our study, a total of 53 (13.91%) beta-lactam antibiotic resistant E. coli isolates were identified from broiler (n = 33; 16.75%) and layer (n = 20; 10.87%) chicken meat. Among the 5 types of bla genes (TEM, CTX-M, CTX-M-1, CTX-M-2 and SHV) tested by multiplex PCR in the current study, only the bla SHV gene was detected in our E. coli isolates. In a study conducted in Netherlands, bla CTX-M-1 (58.1%) has been found to be the most common gene in chicken meat, followed by bla TEM-52 (14%) and bla SHV-12 (14%) 53 . In another study conducted in Egypt, TEM, CTX-M and SHV genes have been detected in 57.55%, 46.23% and 23.58% of the isolates, respectively 54 . Highly prevalent ESBL-producing E. coli strains in meat samples from broiler (87%) and layer (42%) chickens have been previously reported in India 50 , Nepal (36.9%) 51 and Vietnam (37%) 55 . In a previous study, the AmpC gene has been found to be mostly plasmid-associated whereas the chromosomal AmpC was found in a small percentage of E. coli 56 . In Vietnam and Italy, 84.2% 24 and 11.2% 57 , respectively, of the chicken meat isolates carried plasmid-associated AmpC genes, which is higher than the prevalence among the isolates of the current study. Over expression of ESBL and AmpC beta-lactamases in gram-negative bacteria may reduce therapeutic options for treatment of their infections by providing resistance to most beta-lactam antibiotics 58 . Table 10. Correlation between the AMR phenotypes and genotypes among the investigated E. coli (n = 381) isolates in relation to the type of hens (broiler or layer). a Chl Chloramphenicol, Ery Erythromycin, Gen Gentamicin, Tet Tetracycline, Str Streptomycin. Please note that the correlation between trimethoprimsulfonamide and sul1 was not included in this table because sul1 is responsible for resistance to sulfamethoxazole only not to the trimethoprim-sulfonamide combination. b n-Pr: number of strains expressing phenotype resistant to the indicated antimicrobial agent. c n-Gp: number of strains carrying the indicated resistance gene. d P+/G−: number of phenotypically resistance strains (P+) with no resistance genes (G−) for the antimicrobial identified. e P−/G+: number of phenotypically susceptible strains (P−) with one or more resistance genes (G+) for antimicrobials. f r square = 1 indicates positive correlation, r square = 0 indicates no correlation. The highest correlation obtained is highlighted in bold. g P values < 0.05 are highlighted in bold. www.nature.com/scientificreports/ The strong correlation between the phenotypes and genotypes of AMR in bacteria may indicate that the resistance to these antibiotics is mainly attributed to the presence of certain AMR genes in their genome. Based on this, a number of strong correlations between the phenotypes and genotypes of AMR in E. coli were found in our study, such as between gentamicin and aac-3-IV, tetracycline and tetA, and streptomycin and aadA1, indicating that the resistance to some antimicrobials may be mediated, at least partially, by a single gene. This finding is more or less similar to the results of previous studies 59, 60 .

Characteristics of strains
On the other hand, interestingly, we found that some strains possessed resistance phenotypes but did not have the corresponding ARGs and vice versa. This finding is similar to the results reported by Rosengren et al. 60 . A possible explanation is that resistance phenotypes can be expressed upon the stimulation of many different genetic factors, and that each factor may present a unique epidemiological character 61,62 . Also, this may be due to the co-selection pressure of one antimicrobial class on another. It is known that the use of a particular antimicrobial agent can select for resistance not only to its own, but also act as potential co-selection marker for other antimicrobials agents. It means the use of single antimicrobial agent can lead to the selection and co-selection of multiple resistance phenotypes and ARGs 63 . It is reported that the high prevalence of MDR isolates as well as their persistence is governed by co-selection processes, even in the absence of antibiotic selection pressure 64 .
In the same context, in the current study we found comparatively large number of phenotypic erythromycin resistance E. coli strains than those carrying the ereA gene. This might be the result of the carriage of erythromycin resistance genes other than that included in our study. Thus, further detailed investigation is necessary to unveil the exact mechanism of AMR in E. coli isolates of broiler and layer chickens produced in Bangladesh.
There is very limited data on antibiotic use in chicken production in Bangladesh. In addition to the AMR genes that could be detected given the available resources, there may be other AMR genes that can be revealed in future studies. Further studies on the roles of TEM, SHV, and CTX genes will be important.

Conclusion
Our study aimed primarily to characterize the antibiotic resistant E. coli isolates from commercial broiler and layer chicken meat samples in greater Sylhet division of Bangladesh. The correlations between the phenotypic and the genotypic susceptibility were also explored. Overall, the prevalence of E. coli was 63.5% (381/600) in these samples, from which 75.06% (286/381) of the isolates were MDR and 50.13% (191/381) contained 3 to 5 MDR genes. The isolates from raw chicken meat were highly resistant to ampicillin, erythromycin and tetracycline and 13.91% (53/381) of the isolates contained beta-lactamase (ESBL and AmpC) producing genes. This situation is alarming for Bangladesh, where facilities for health care, surveillance for antibiotics medication, and facilities to detect MDR and ESBL genes are underdeveloped. Our results highlight the need to develop novel antibiotic with potent activity against MDR and ESBL-producing bacteria. At the same time, promoting the rational use of antibiotics in livestock, as well as adopting safe food handling and proper cooking practices are crucial to reduce or eliminate the risk from pathogenic antibiotic resistance bacteria originating from raw foods. www.nature.com/scientificreports/