Study Design
A cross-sectional study was conducted between August 2017 and May 2018 in seven districts in Zambia that included Lusaka, Chilanga, Chongwe, Kafue, Choma, Kabwe and Kitwe. Choma, Kabwe, Kitwe and Lusaka districts were purposely selected because as provincial headquarters, they are retail destinations for many poultry products from other districts while Chilanga, Chongwe and Kafue districts were included due to their proximity to Lusaka, the capital city. The primary sampling units were the markets (broadly classified as Open markets and Supermarkets) and the secondary sampling units were the raw retail broiler chicken carcasses. An open market was defined as an unrestricted market, not housed in a building, where food products are often sold exposed, while a supermarket was defined as a restricted market, housed in a closed building with modernized facilities (Bumbangi et al., 2016).
Proportion stratified random sampling was employed where Open markets and Supermarkets were the strata. At the time of the study, information collected from the respective City Councils revealed that there were 47 supermarkets and 33 open markets in Lusaka province, five open markets and four supermarkets in Choma, seven open markets and eight supermarkets in Kitwe, nine open markets and six supermarkets in Kabwe, three open markets and two supermarkets in Kafue, three open markets and one supermarket in Chilanga, and one open market and no supermarket in Chongwe. This formed the sampling frame from which a study population was drawn.
Sample size calculation
The sample size for estimation of a single proportion was calculated using Epi tools software (www.epitools.ausvet.com) based on the following assumptions: prevalence of AMR E. coli on raw retail broiler chicken carcasses = 25% (Vindigni et al., 2007); confidence level = 95%; level of precision = 5%. Using the above assumptions, the minimum sample size calculated was 289 and the distribution of sampling sites is as outlined in Table 1.
Sampling from open markets
From the open markets within Lusaka district that were sampled (n=22), two shops trading in raw retail broiler chicken carcasses at each market were selected and from each shop two raw retail broiler chicken carcasses purchased. Furthermore, at the market where open stands (display tables) instead of shops trading in dressed broiler chickens were available, one sample from each stand was collected. For other districts, considering the limited number of markets that traded in raw retail broiler chicken carcasses, all of them were included in the study and at least two samples from each market were collected bringing the total number of samples from open markets to 178 (Table 1).
Sampling from the Supermarket
From each supermarket in all districts, a maximum of four different brands of raw retail broiler chicken carcasses was sampled (one of each brand), though some sampling sites had less than the maximum number. Therefore, the number of broiler carcasses from supermarkets in this study was 154 (Table 1).
Upon purchase, all samples were transported in a cooler box containing ice packs to the laboratory and processed within 8 hours.
Laboratory Analysis
Laboratory isolation included a whole carcass rinse in buffered peptone water (Oxoid) (Figure S1a-c), pre-enrichment of an aliquot of the rinsate and subsequent incubation at 37OC overnight. Processing of pre-enriched broths was undertaken in the Public Health Laboratory, School of Veterinary Medicine. The Carcass Rinse technique outlined by the USDA Food Safety and Inspection Service was used with a few modifications (D’Aoust et al, 1982). Whole chicken carcasses were aseptically placed in sterile 3 litre ziplock bags. 450mL of sterile buffered peptone water (Oxoid UK) was poured into each of the bags containing the carcasses. The open ends of the bags were then twisted to hold the bags closed and the entire carcass was rinsed using a repeated rocking motion 30 times. The bags were then opened and the carcasses removed. The rinsate was then collected in sterile containers by puncturing the “v”-end of the bags and used for further analysis.
The methods proposed by the Food and Drug Administration’s Bacteriological Analytical Manual (U.S. Food and Drug Administration, 2001) were used with a few modifications for the isolation of Salmonella and E. coli. During the carcass rinse technique, 450mL of sterile buffered peptone water was also poured into an empty bag that did not contain a carcass to act as a control. The rinsate was incubated overnight and later streaked onto MacConkey agar plate (Oxoid, UK) to ensure that the batch of bags was sterile and that the organisms isolated were indeed from the chicken carcasses and not the bags used for rinsing. 10µL of the incubated broth was then transferred to MacConkey agar (Oxoid UK) and resulting colonies were gram stained for detection of Gram-negative short rods, which were subsequently sub-cultured onto Eosin Methylene Blue (EMB) agar (Oxoid UK). Colonies that showed a metallic green sheen (Figure S2) were subjected to biochemical tests using Analytical Profile Index (API 20E) (Biomerieux®) for identification for E. coli isolates. 1ml of the incubated pre-enrichment broth was also transferred to Rappaport Vassiliadis (Oxoid UK) and later subcultured on Xylose-Lysine Deoxycholate agar (Oxoid UK). Pink and black colonies on XLD agar (Figure S3) were then Gram-stained and subjected to biochemical tests for identification of Salmonella using API 20E. Further confirmation of the isolates was done using 16S rRNA sequencing (Weisburg et al., 1991).
The Kirby-Bauer disk diffusion technique for AST was used on all confirmed Salmonella and E. coli isolates using a panel of 9 different antibiotics (Kirby-Bauer, 1961). The isolates were prepared by sub-culturing onto Blood agar (Oxoid UK) overnight at 37oC. One or two colonies were then suspended in 4mL of 0.9% sodium chloride solution and their turbidity compared to that of a 0.5 McFarland’s turbidity standard. An inoculum of the suspension was then spread on two Müeller Hinton agar (4 ml thickness) plates (Oxoid UK) until the entire surfaces of the plates were covered. 5 different antibiotic wafers from the 9 chosen for the study were then placed on the surface of each of 2 plates using the applicator (Oxoid). Two plates were used for each isolate to accommodate the 9 antibiotics. The antibiotics that were used were amoxicillin-clavulanic acid (30µg), ampicillin (10µg), cefotaxime (30µg), chloramphenicol (30µg), ciprofloxacin (5µg), imipenem (10µg), nalidixic acid (30µg), tetracycline (30µg) and trimethoprim-sulfamethoxazole (25µg). The list of antibiotics was prioritized based on the most frequently used in the poultry industry in Zambia and also based on the priority list by the WHO and OIE of critical antibiotics (Oie, 2007; WHO, 2017). The plates were then incubated at 37oC for 24hrs and the diameters of the zones of inhibition entered and analyzed in WHONET 2018 software. An isolate was considered to be completely resistant to an antimicrobial when it had no zone of inhibition around the antimicrobial disc after the incubation (Kirby-Bauer, 1961). The interpretation of AST results was based on the Clinical and laboratory standards institute (CLSI) guidelines 2018 (Replaces et al., 2018)
Isolates which showed resistance to tetracyclines, sulphonamides and beta-lactam antibiotics were then forwarded for molecular analysis that involved extraction of DNA and checking for the presence of target resistance genes. The process of DNA extraction involved the suspension of a few bacterial colonies in 100µL of nuclease-free water and heating of the vials at 80OC for 10 minutes. The suspension was then centrifuged at 60000G with a temperature of 4°C for 3 minutes. Multiplex polymerase chain reaction (PCR) was performed to check the presence of resistant genes of interest according to the method described by Adesiji et al. 2014. The mastermix volumes and PCR reaction were as outlined in Table S1 and Table S2 (Supplementary Material). The target genes were selected based on the antimicrobial susceptibility results. The 3 target genes were tetA (for tetracycline resistance), sul1 (for sulfonamide resistance) and blaCTX-M (for beta-lactam resistance) (Table S3; Figures S4). Every batch of samples was processed along with a positive and negative control using E. coli 25922 (ATCC) and Salmonella typhimurium 14028 (ATCC).