Antimicrobial resistance and genetic diversity in ceftazidime non-susceptible bacterial pathogens from ready-to-eat street foods in three Taiwanese cities

Bacterial contamination of ready-to-eat (RTE) street foods is a major concern worldwide. Dissemination of antibiotic resistant pathogens from food is an emerging public-health threat. To investigate the prevalence of antibiotic resistance genes and ceftazidime resistance-associated efflux pumps in foodborne pathogens, 270 RTE street foods samples were collected in three densely populated Taiwanese cities. Among 70 ceftazidime non-susceptible isolates, 21 Stenotrophomonas maltophilia, 12 Pseudomonas spp., 22 Acinetobacter spp., and 15 Enterobacteriaceae isolates were identified. Phylogenetic analyses revealed high levels of genetic diversity between all of the different strains. Multi-drug resistance was observed in 86.4% (19/22) of Acinetobacter spp., 100% (12/12) of Pseudomonas spp., 71.4% (15/21) of S. maltophilia, and 93.3% (14/15) of Enterobacteriaceae. Of 70 ceftazidime non-susceptible isolates, 13 contained ESBLs or plasmid-mediated ampC genes and 23 contained ceftazidime resistance-associated efflux pumps, with Acinetobacter spp. identified as predominant isolate (69.6%; 16/23). AdeIJK pump RNA expression in Acinetobacter isolates was 1.9- to 2-fold higher in active efflux strains. Nine clinically resistant genes were detected: catIII and cmlA (chloramphenicol); aacC1, aacC2, aacC3, and aacC4 (gentamicin); tet(A), tet(C), and tet(D) (tetracycline). The scope and abundance of multidrug-resistant bacteria described in this report underscores the need for ongoing and/or expanded RTE monitoring and control measures.

literature contains several studies that assessed foodborne pathogens in RTE food in other countries. For example, among 154 foodborne Staphylococcus aureus isolates analyzed in Turkey, 39 (25.3%) were found to have multi-drug resistance 14 . In one study of vegetables collected in Switzerland, 78.3% (47/60) of multidrug-resistant extended-spectrum β-lactamase (ESBL)-producing Enterobacteriaceae isolates were identified 15 . ESBLs are enzymes that hydrolyze penicillin as well as aztreonam and first-, second-, and third-generation cephalosporins 16 . Although, third-generation cephalosporins are broad-spectrum antimicrobial agents against bacterial infection and useful for different type of clinical situations 17 , the problem of ESBL-producing bacteria has been identified in medical systems and communities worldwide 16,18 . A primary reason for this rapid spread is that ESBL-producing bacteria can be transmitted by contaminated food or water 9,19,20 . In Taiwan, at least two research teams have focused on microbiological quality in RTE foods, but no efforts have been made to determine antibiotic resistance levels and associated resistance mechanisms in these foods 10,12 . Bacterial isolates containing ESBLs have been detected in hospital patients and livestock, though transmission mechanisms remain unidentified [21][22][23][24] . To determine the potential contribution of food to the dissemination of antibiotic resistant pathogens, we collected and analyzed 270 RTE street food samples in the densely populated cities of Kaohsiung, Taichung and Taipei, including spring rolls, cold noodles and fruit platters. One of our primary motivations was to determine the prevalence and mechanisms of antibiotic resistance, specifically the contribution of efflux pumps to ceftazidime resistance.

Antimicrobial susceptibility testing. As shown in
Among the 21 S. maltophilia isolates, we noted high resistance levels (71.4-100%) against 3 antibiotics (Fig. 3D). The ticarcillin-clavulanic acid resistance level for S. maltophilia was 66.7%; only 1 isolate was found to be levofloxacin-resistant.
For this study, multidrug resistance was defined as resistance to a minimum of 3 antibiotics. Our data indicate the presence of multidrug resistance in 86.4% (19/22) of Acinetobacter spp., 100% (12/12) of Pseudomonas spp., 71.4% (15/21) of S. maltophilia, and 93.3% (14/15) of Enterobacteriaceae isolates (Fig. 3E). These high percentages indicate that RTE street foods in Taiwan may be a significant reservoir for antibiotic resistant pathogens.

Investigation of ceftazidime non-susceptible bacterial isolates.
Phylogenetic typing identified 20 pulsotypes in 22 Acinetobacter spp. isolates. Two isolates sharing the same pulsotype, TP36 and TP47, were identified in two fruit platter samples collected in Taipei ( Supplementary Fig. S1). PFGE analysis of 12 Pseudomonas spp., 21 S. maltophilia, and 15 Enterobacteriaceae isolates indicate that they were non-clonal ( Supplementary  Figs S2-S4). The high levels of genetic diversity and multi-drug resistance identified among these bacterial isolates suggest that they developed antibiotic resistance mechanisms independently in the three cities where samples were collected.
Ceftazidime resistance mechanism. Among the 70 ceftazidime non-susceptible isolates that were the focus of this study, 13 contained ESBLs or plasmid-mediated ampC genes (Table 1). bla PER-1 was found in 5 Acinetobacter spp. isolates and bla CTX-M-9 was detected in E. coli (n = 1), S. maltophilia (n = 1), and A. baumannii (n = 1). This suggests that bla CTX-M-9 is horizontally transferred among various bacterial species.
To clarify the contribution of efflux pumps to ceftazidime resistance, we used the efflux pump inhibitor CCCP to analyze the 70 ceftazidime non-susceptible isolates. Bacterial isolates containing ceftazidime resistance-associated efflux pumps were identified as any CCCP-inoculated strain exhibiting a minimum 4-fold decrease in minimum inhibitory concentration (MIC). Our data indicate that 72.7% (16/22) of the Acinetobacter spp. and 25% (3/12) of the Pseudomonas spp. isolates contained efflux pumps conferring resistance to ceftazidime (Fig. 4). A ceftazidime-induced effect on efflux activity was observed in only 3 S. maltophilia isolates (14.3%) and 1 Enterobacteriaceae isolate (6.7%). These findings suggest a significant prevalence of ceftazidime resistance-associated efflux pumps in Acinetobacter spp.

RNA expression of two efflux pumps in Acinetobacter spp. Previous studies have determined a
link between the active efflux pumps adeDE and adeIJK and resistance to third-generation cephalosporins in Acinetobacter spp. 25,26 We analyzed expression of adeDE and adeIJK in four representative isolates exhibiting efflux pump activity (strains KA19, KA82, TP36, TP68) relative to a control strain that did not (strain TC24). Our data indicate that adeE expression was not significantly different from that of the control strain (Fig. 5A), while adeI expression was 1.9-to 2.2-fold (p < 0.05) higher than in the control strain (Fig. 5B). These results suggest that ceftazidime resistance in Acinetobacter spp. may be attributed to adeI expression levels.

Discussion
The literature contains studies of the microbiological quality of various RTE foods in countries around the world. Oliveira et al. tested 162 minimally processed leafy vegetable samples in Brazil and reported that 53.1% contained E. coli, 3.7% Listeria spp., and 1.2% Salmonella spp. 13 . In Argentina, the foodborne pathogens Bacillus cereus, Clostridium perfringens, S. aureus and Salmonella spp. were found in 101 RTE cooked food samples collected from the central kitchen facility of a school district 27 . The researchers focused on B. cereus (found in 63.4% of their samples) and reported bacterial counts of <4 log CFU/g. In a Portuguese study, Campos et al. described poor microbiological quality in a number of RTE food types and food products prepared by vendors working out of trailers: 100% tested positive for Enterobacteriaceae and coliforms, 20% for E. coli (4 hamburgers, 4 other trailer foods), and 20% for L. monocytogenes (2 hamburgers/2 hotdogs, 3 other trailer foods) 7 . In an Italian study of whole vegetables and RTE salads, Salmonella spp., L. monocytogenes, E. coli O157:H7, hepatitis A, and noroviruses were detected in 964 samples 28 . Of these, only 2 and 3 tested positive for Salmonella spp. and L. monocytogenes, respectively.    supermarkets, and 9.5% from warehouse stores 12 . They also identified E. coli in 5-7% of all food samples obtained from supermarkets and warehouse stores. Neither one of these studies looked at antimicrobial susceptibility patterns or resistance mechanisms.
In the present study, the highest percentages of ceftazidime-resistant bacteria were isolated in RTE street food samples of cold noodles (43.3%), followed by fruit platters (17.8%) and spring rolls (16.7%) (Fig. 2B). Results from efforts to determine antimicrobial resistance patterns indicate unexpectedly high levels of multi-drug resistance (71.4-100%) (Fig. 3E). Low resistance was determined for levofloxacin ( Fig. 3A-D), indicating that it should be considered for treating severe cases of food poisoning. While meropenem resistance levels were high in Enterobacteriaceae  Table S2), carbapenemases were not detected in these isolates. Carbapenemases such as bla KPC-2 and bla IMP-8 tend to be detected in hospitalized patients, with low potential for widespread dissemination to communities or via RTE foods 29,30 .
Researchers studying different types of salads in Portugal and Switzerland have reported that RTE food pathogens share the same ESBL genes as clinical isolates collected from hospital patients, suggesting the potential for commensal bacteria to act as antibiotic-resistant gene reservoirs, especially in hospital settings 9,20 . Further, ESBL-producing Enterobacteriaceae isolates have been reported in beef burger, seafood, and raw vegetable samples 15,31,32 . Based on these reports, ESBL gene sources in human food chains include both RTE and non-RTE food products. In the present study we identified various ESBLs and plasmid-mediated ampC genes commonly found among bacteria isolates collected in Taiwanese hospitals [21][22][23] . These genes were especially abundant in ceftazidime non-susceptible non-fermenting gram-negative bacilli, mostly Acinetobacter spp. (Table 1). The data confirm the presence of antibiotic-resistant gene reservoirs in RTE street foods sold in Taiwan's three largest cities, and that stepped-up surveillance efforts are required to monitor their movement.
Efflux pump activity has been associated with ceftazidime resistance 25,26 . Quantitative RT-PCR results indicate that the AdeIJK efflux pump contributed to the ceftazidime resistance that we observed in Acinetobacter spp. (Fig. 5). Specifically, among the 22 ceftazidime non-susceptible Acinetobacter spp. isolates we examined, 3 contained ESBL genes, 5 contained ESBL genes plus efflux pump activity, 11 efflux pump activity only, and 3 unknown resistance mechanisms. To our knowledge, this is the first report to demonstrate a link between the AdeIJK efflux pump and ceftazidime resistance in foodborne pathogens. One previous study reported that efflux pumps contributed to 94.4% of ceftazidime-resistant (17/18) and 72.7% of chloramphenicol-resistant (16/22) Burkholderia cepacia complex isolates 21 . Combined, these reports and data from the present study indicate that efflux pump activity plays an important role in antibiotic resistance mechanisms. The low ceftazidime resistance levels that we observed in S. maltophilia and Enterobacteriaceae isolates suggest that the ceftazidime resistance noted in this study is attributable to an unknown ESBL gene or different efflux pump type. We will attempt to confirm this finding and identify the gene and pump types in a future study.
Our results confirm the capacity of ready-to-eat street foods found in Taiwan, especially cold noodles, to act as antimicrobial-resistant bacteria reservoirs. Additional research efforts are required to quantify the antimicrobial-resistant bacteria that were isolated and analyzed in the present study. The potential for the horizontal transfer of resistance genes and the transmission of multidrug-resistant bacteria warrant further monitoring and control measures throughout Taiwan, but especially in densely populated urban areas. Unfortunately, one study concluded that special training in hygienic practices in Taiwan did not result in a sufficient increase in knowledge or effective practices among food vendors 33 . Clearly a renewed effort needs to be made to establish strict education standards for hygienic practices among street food vendors and food preparers, with local health officials monitoring such effort to ensure RTE food product safety.  Microbiological analysis. Ten grams of each sample were homogenized in a stomacher blender, enriched using brain-heart infusion broth (BD, East Rutherford, NJ), and incubated overnight at 37 °C. Individual loopfuls of bacterial suspension were added to eosin-methylene blue (EMB) agar containing ceftazidime (8 μg/ml) and incubated overnight at 37 °C. When different colony morphologies were observed in the same food sample, all colonies were subcultured on EMB agar containing 8 μg/ml ceftazidime. Bacterial identification was performed using 16S rRNA gene sequencing as described in a previous study 30 . ESBL production was confirmed by tests recommended by the Clinical and Laboratory Standards Institute, using cefotaxime (30 mg) and ceftazidime (30 mg) disks alone and in combination with clavulanic acid (10 mg) (Becton, Dickinson, MD) 34 . Positive phenotypes were defined as growth-inhibitory zones from either cefotaxime or ceftazidime disks plus clavulanic acid ≥5 mm of growth-inhibitory zones from either cefotaxime or ceftazidime disks.
Antimicrobial susceptibility. These tests were performed using agar dilution methods according to Clinical and Laboratory Standards Institute guidelines 34 . Minimum inhibitory concentration (MIC) was defined as the lowest concentration of an antibiotic preventing bacterial growth after 16 to 20 h of incubation at 37 °C. We tested 11 antimicrobial agents: cefotaxime, ceftazidime, chloramphenicol, colistin, gentamicin, levofloxacin, meropenem, ticarcillin, ticarcillin-clavulanic acid, tetracycline, and trimethoprim-sulphamethoxazole.  37 ; tetracycline resistance genes (tet(A), tet(B), tet(C), tet(D), tet(E) and tet(G)); and gentamicin resistance genes (aacC1, aacC2, aacC3, and aacC4) 38 were detected by PCR, using primer sets listed in previous reports (Supplementary Table S1). Positive controls for these resistance genes were included in all PCR analyses.

Pulsed-field gel electrophoresis (PFGE
Efflux pump activity assays. To evaluate the contributions of efflux pump activity to ceftazidime resistance among the isolates, MIC resistance patterns were determined via Mueller-Hinton agar dilution with and without 12.5 μM of the efflux pump inhibitor carbonyl cyanide-m-chlorophenyl hydrazone (CCCP; Sigma-Aldrich, St. Louis, MO) 39 . Bacterial isolates containing ceftazidime resistance-associated efflux pumps were identified as any CCCP-inoculated strain exhibiting a minimum 4-fold decrease in MIC.
Quantitative RNA expression for efflux pumps. RNA extraction and cDNA synthesis were performed as described in our previous study 21 . PCR reactions took place in buffer containing 1X FastStart Universal SYBR Green Master (Roche), 300 nM primers, and 2 μl cDNA in an ABI7000 machine following manufacturer instructions (96-well plates). AdeIJK and AdeDE efflux pumps have previously been described as contributing to ceftazidime resistance in Acinetobacter spp. 26,40 . Quantitative RNA expression of adeI and adeE was analyzed using the primers listed in Supplementary Table S1 40 . Relative fold changes in the transcript levels of indicated genes were normalized to the 16S rDNA gene (internal control) and calculated using the 2 −ΔΔCT method. Non-parametric Mann-Whitney U tests were used to examine differences in quantitative RNA expressions of adeI and adeE between test (ceftazidime resistance-associated efflux pumps) and control strains (no efflux pump activity). Statistical significance was established as p < 0.05.