Antibiotic-Resistant Escherichia coli and Salmonella from the Feces of Food Animals in the East Province of Rwanda

Simple Summary A paucity of information on antimicrobial resistance in animals in Rwanda prompted us to conduct this study, the objective of which was to estimate the prevalence of antibiotic resistance among Escherichia coli and Salmonella from the feces of cattle, goats, pigs, and poultry. We found that resistance to tetracycline, ampicillin, and streptomycin were the most frequent among non-type-specific E. coli isolates. Resistance to chloramphenicol, quinolone-based antibiotics, amoxicillin–clavulanic acid, and azithromycin were also observed among these bacterial isolates, but with lower percentages. Most of resistant E. coli, including multidrug-resistant strains, were isolated from poultry fecal samples. More than 30% of samples were positive for E. coli resistant to third-generation cephalosporins or quinolone-based antibiotics. All isolated Salmonella were pan-susceptible. These results give an insight into the status of antibiotic resistance in food animals in Rwanda, as well as a call for further research. Also, the findings indicated a need for antibiotic stewardship and antimicrobial resistance surveillance initiatives. Abstract In Rwanda, information on antibiotic resistance in food animals is scarce. This study was conducted to detect and phenotypically characterize antibiotic-resistant Escherichia coli and Salmonella in feces of cattle, goats, pigs, and poultry in the East province of Rwanda. We isolated non-type-specific (NTS) E. coli and Salmonella using plain culture media. In addition, we used MacConkey agar media supplemented with cefotaxime at 1.0 μg/mL and ciprofloxacin at 0.5 μg/mL to increase the probability of detecting E. coli with low susceptibility to third-generation cephalosporins and quinolones, respectively. Antibiotic susceptibility testing was performed using the disk diffusion test. Among 540 NTS E. coli isolates, resistance to tetracycline was the most frequently observed (35.6%), followed by resistance to ampicillin (19.6%) and streptomycin (16.5%). Percentages of NTS E. coli resistant to all three antibiotics and percentages of multidrug-resistant strains were higher in isolates from poultry. All isolated Salmonella were susceptible to all antibiotics. The sample-level prevalence for resistance to third-generation cephalosporins was estimated at 35.6% with all third-generation cephalosporin-resistant E. coli, expressing an extended-spectrum beta-lactamase phenotype. The sample-level prevalence for quinolone resistance was estimated at 48.3%. These results provided a baseline for future research and the development of integrated surveillance initiatives.


Introduction
The factors influencing the emergence, propagation, and spread of bacteria resistant to antibiotics are complex and not fully understood [1]. Despite this uncertainty, antimicrobial family-run production system without any specialization. There is still a lower number of specialized animal farms with intensive production of milk, meat, or eggs [28].
Livestock production contributes 4% to the total gross domestic product, and is one of the fastest-growing subsectors in Rwanda. Between 2012 and 2016, its growth was estimated at 8.3% [28]. This high performance is attributed to governmental efforts to improve and modernize the animal production sector through various programs, such as programs of intensification, the Girinka program (one cow per family), and small stock development.
Despite governmental efforts to improve animal production to contribute to food security and income generation, animal diseases represent an important constraint to the sector. Various infectious diseases, including bacterial diseases, prevail in food animals in Rwanda [29][30][31][32][33][34][35][36]. Some of the prevalent bacterial diseases include anthrax, bovine tuberculosis, brucellosis (Brucella abortus, Brucella melitensis), swine erysipelas, contagious bovine pleuropneumonia, contagious caprine pleuropneumonia, fowl typhoid, heartwater, and ovine epididymitis (Brucella. Ovis) [29]. High levels of morbidity and mortality in livestock are considered as major drivers of antibiotic use in animals. In addition, poor diagnostics and uncontrolled access to antibiotics can lead to antibiotics misuse [37,38]. Unfortunately, data on types and quantities of antibiotics used in food animals are missing.
The paucity of information on the occurrence and the magnitude of AMR in food animals makes it difficult to objectively design contextual strategies to prevent and control AMR in food animals in Rwanda.
The end-goal of this study was to address gaps in the information on AMR in food animals in Rwanda. Specifically, the study's objective was to provide initial estimates of isolate-and sample-level prevalence of indicator E. coli and pathogenic Salmonella resistant to antibiotics among cattle, goats, pigs, and poultry feces in the East province of Rwanda.

Fecal Sample Collection
From September to November 2019, a convenience sample size of 180 feces was collected from cattle, goats, pigs, and poultry in farms located in the East province of Rwanda ( Figure 1). In general, visited farms were diverse in terms of animal population and production management. All cattle farms had between 15 and 30 animals, goat farms had between 15 and 50 animals, and pig farms had between 10 and 35 animals. All cattle, goat, and pig farms were non-specialized, intensive-production farms. They can be categorized as semi-intensive farms [38]. All poultry farms were large in terms of chicken population and were commercial farms with either broiler or layer production.
The northeast region of Rwanda has an established livestock farming culture and higher livestock density than other regions in Rwanda [27]. For each animal species, 15 farms, identified and recruited via snowball sampling [39], were visited. At each farm, three freshly voided feces were randomly collected in clean cups using clean spoons. In total, 45 fecal samples were collected per animal species. Samples were kept on ice in a cooler box and transported by car to the Rwanda Agriculture and Animal Resources Development Board's microbiology laboratory in Kigali for processing.
The collection of fecal samples was done without any interaction with animals. Voided feces were collected from the floor. Therefore, ethical approval was not required.

Bacteria Isolation
The protocol used in this study was adapted from a protocol developed and fieldtested on dairy faecal samples in Texas [40]. Briefly, 10 gr of each faecal sample was enriched in 90 mL of buffered peptone water (Hardy Diagnostics, Santa Maria, CA, USA) overnight. The obtained non-selective enrichment was used for bacterial isolation. Ten µL of the non-selective enrichment was streaked on plain (without antibiotics) MacConkey agar (Hardy Diagnostics, Santa Maria, CA, USA) for the isolation of non-type-specific (NTS) E. coli. Moreover, to increase the probability of isolating third-generation cephalosporinresistant (3GCr) E. coli and quinolone-resistant (Qr) E. coli, a screening step was performed, using MacConkey agar containing cefotaxime (Acros Organics, Fair Lawn, NJ, USA) at 1.0 µg/mL and MacConkey agar containing ciprofloxacin (Acros Organics, Fair Lawn, NJ, USA) at 0.5 µg/mL, respectively [40,41]. In addition, 1 mL of the non-selective enrichment was transferred into 9 mL of Rappaport-Vassiliadis Salmonella (Becton Dickinson, Franklin Lakes, NJ, USA) and 9 mL of Tetrathionate (Becton Dickinson, Franklin Lakes, NJ, USA) broths for Salmonella-selective enrichments. Thereafter, 10 µL of each of the selective enrichments was streaked on plain brilliant green sulfa agar (Becton Dickinson, Franklin Lakes, NJ, USA) and plain xylose lysine tergitol4 agar (Becton Dickinson, Franklin Lakes, NJ, USA) for the isolation of Salmonella. Based on morphological appearance, three typical E. coli colonies grown on plain MacConkey agar and one colony from each of the MacConkey agar containing antibiotics were selected for confirmation as E. coli using the indole spot test (Hardy Diagnostics, Santa Maria, CA, USA). For each sample, two colonies with morphology typical to Salmonella were selected and confirmed as Salmonella by detection of the invA gene using loop-mediated isothermal amplification (LAMP)m as described by Hara-kudo [42]. The LAMP was performed using the WarmStart Colorimetric LAMP 2X Master Mix (New England Biolabs, Ipswich, MA, USA) following the manufacturer guidelines. The primers used were developed by Hara-kudo [42], and they are listed in Table S1 (Supplementary Materials).
Furthermore, the combination disk test [43] was used to detect bacterial isolates, producing extended-spectrum beta-lactamases (ESBLs) among 3GCr isolates. E. coli ATCC 25,922 was used for quality control purposes.

Statistical Analysis
Statistical analyses were performed using R software version 3.0.0 (R Core Team, Vienna, Austria). The Fisher's exact test of independence was used for comparison of the prevalence of resistant bacteria among food animal species, with a p-value threshold of 0.05. The Wilson score method was used to calculate 95% confidence intervals (95% CI).

Results
Five hundred and forty NTS E. coli were isolated from plain MacConkey agar, with 135 isolates arising from each animal species. The screening process resulted in 69 samples with presumptive 3GCr E. coli and 160 samples with presumptive Qr E. coli. Four Salmonella isolates were recovered from only two pig fecal samples (n = 2 of 180 samples; 1.1%).

Antibiotic Susceptibility
At the isolate level, among the 540 NTS E. coli, the prevalence of isolates resistant to tetracycline was the most observed (Table 1). The isolate-level prevalence of NTS E. coli resistant to ampicillin, quinolones, streptomycin, and tetracycline differed statistically (p < 0.05) among animal species, and were higher among isolates from poultry samples than from all other food animal species (Table 2).  In total, 71 of the 540 NTS E. coli (13.1%) were of a multidrug resistance phenotype. The prevalence of MDR NTS E. coli was statistically different among food animal species (p < 0.05), with isolates from poultry samples (34.8%) being the highest vis-à-vis the prevalence of MDR NTS E. coli among isolates from the other three animal species (i.e., 3.7% in cattle, 5.9% in goats, and 8.1% in pigs) (Figure 2). The four Salmonella isolates were all pan-susceptible to the 12 antibiotics tested. Among 69 E. coli isolates screened on MacConkey agar containing cefotaxime, 64 were confirmed to be resistant to ceftriaxone. The sample-level prevalence for third-generation cephalosporin resistance among all food animal species was estimated at 35.6% (95% CI: 28.9-42.8%). The recovery of confirmed 3GCr E. coli was significantly different among food animal species (p < 0.05), with pigs having the highest (57.8%) in comparison to samples from cattle, goats, and poultry ( Figure 3). Confirmed 3GCr E. coli (n = 64) exhibited resistance mostly to ampicillin, followed by tetracycline (Table 3).
From 160 E. coli isolates screened on MacConkey agar with ciprofloxacin, 87 were confirmed as resistant to quinolone-based antibiotics (nalidixic acid or ciprofloxacin). The sample-level prevalence for quinolone resistance was estimated at 48.3% (95% CI: 41.1-55.6%). The recovery of confirmed Qr E. coli differed significantly among the food animal species (p < 0.05), with samples from poultry (73.3%) and goats (55.6%) higher in comparison to samples from cattle and pigs ( Figure 4).
The proportion of Qr E. coli resistant to tetracycline was high (Table 4).

E. coli Producing ESBLs and Patterns of Resistance to Quinolones
In total, 72 isolates were confirmed as 3GCr E. coli. Among them, 64 were isolated from MacConkey agar with cefotaxime, while eight were isolated in MacConkey agar with ciprofloxacin. Most 3GCr E. coli had a phenotypic resistance to cefazolin and sulfisoxazole on the second panel of disks. It was found that all of the isolated 3GCr E. coli were phenotypically ESBL producers, except for one isolate that remains unclassified (Table 5).
In total, 141 E. coli isolates were resistant to quinolone-based antibiotics. Among them, 87 were isolated from MacConkey agar with ciprofloxacin, 19 were isolated from MacConkey agar with cefotaxime, and 35 were isolated from plain MacConkey agar. Among the 141 Qr E. coli, 24.8% were susceptible to ciprofloxacin but resistant to nalidixic acid, 19.9% were intermediate to ciprofloxacin and resistant to nalidixic acid, while 55.3% were resistant to ciprofloxacin and nalidixic acid concurrently.

Discussion
This study revealed that resistance to tetracycline was the most prevalent among NTS E. coli isolated from food animals, followed by resistance to ampicillin and streptomycin. This is similar to what was reported in some countries of East Africa, such as Uganda, Tanzania, and Kenya, where resistance to tetracycline, ampicillin, and streptomycin were ranked among the most predominant in food animals [7,8,46]. Resistance to oxytetracycline was the most common resistance among E. coli from chicken in Thailand, Indonesia, and Vietnam [47]. Our results are not that different from the global trend of antimicrobial resistance in animals in low-and middle-income countries. Van Boeckel and collaborators reported that resistance to tetracycline, sulfonamide, and penicillin were the most frequently observed in animals in these countries [48]. We also noted that most of the resistant NTS E. coli, specifically those resistant to tetracycline and multidrug-resistant strains, were isolated from poultry fecal samples. Even if the most common antibiotic used in food animals in Rwanda is not documented, situations reported in adjacent countries may be comparable for farmers in Rwanda. It is reported that tetracycline is the most widely used antibiotic in food animals in Tanzania and Kenya [49,50]. This is also true for other African countries, where it has been reported that tetracycline, aminoglycoside, and penicillin groups are the most commonly used antibiotics in animals [5]. In the majority of low-and middle-income countries, tetracycline, sulfonamide, and penicillin antibiotic groups have been reported as the most commonly used [48]. In Rwanda, oxytetracycline and peni-streptomycin were identified as the main antibiotics recognized or known by farmers [38]. In the same study, poultry farmers were predicted to have a moderate level of antibiotic use in their animals, where they could use antibiotics on regular basis for disease prevention or growth promotion [38]. If tetracycline is heavily used in food animals, specifically in poultry, in the study area, this would help to explain the high level of resistance to tetracycline observed among NTS E. coli isolated from poultry. We recommend further studies to establish the types and amounts of antibiotics used in food-producing animals to address this hypothesis.
In our study, the recovery of Salmonella was low. The detection and isolation of Salmonella can be challenging. It is recommended to combine various Salmonella selective media for efficient detection [51]. In this study, two Salmonella-selective broths (i.e., Rappaport-Vassiliadis Salmonella and Tetrathionate) were used to selectively grow Salmonella in samples before isolation on two Salmonella-selective agar-based media. A study has demonstrated that Rappaport-Vassiliadis and Tetrathionate are effective media to isolate Salmonella [52]. On the other hand, another study demonstrated a relatively low ability of Rappaport-Vassiliadis medium to recover Salmonella when used alone [53]. The low recovery of Salmonella in our study could also indicate a low prevalence of Salmonella in food animals. In some countries, it is documented that Salmonella prevalence in animals varies with seasons [54,55]. Unfortunately, due to the lack of studies involving the isolation of Salmonella in food animals in Rwanda, it remains difficult to explain the low recovery rate observed in this study. With such a low recovery rate of only four strains of pan-susceptible Salmonella, it remains impossible to provide an overview of antibiotic resistance among Salmonella in food animals in Rwanda.
It was found that 35.6% of samples were positive for 3GCr E. coli, even though in many East African countries and the majority of low-and middle-income countries, third-generation cephalosporin antibiotics are not widely used in food animals [49,50]. Direct selection pressure seems unlikely; therefore, exploration as to the likely co-selection pressures [56,57] is needed. Further characterization of 3GC resistance genes would be needed to confirm this, along with other resistance genes that might be co-located on mobilizable genetic elements. In low-and middle-income countries, the rate of bacteria resistant to third-and fourth-generation cephalosporins in animals was reported to be moderate, ranging between 10% and 40%. Comparable to results reported in Thailand [58], the recovery rate of 3GCr E. coli was higher in pig fecal samples. Similar to results reported in Tanzania [7], all 3GCr E. coli isolated in this study presented a phenotype of ESBL production. Additionally, it was noted that most of 3GCr E. coli were resistant to cefotaxime rather than ceftazidime. This may suggest that bla CTX-M genes are the most prevalent in E. coli from food animals in the study area. Previous studies [59] have reported that bla CTX-M genes generally confer lower resistance to ceftazidime than to cefotaxime, which is borne out in our study (Table 5).
Furthermore, our recovery of Qr E. coli (48.3%) was within the estimated range for quinolone resistance in animals in low-and middle-income countries (20% to 60%) [48]. The phenotype of quinolone resistance in E. coli can help in the prediction of the resistance mechanism involved [60,61]. Based on previously reported observations [60,61], we predicted that most of Qr E. coli isolated in this study had mutated gyrA and parC genes. This prediction was based on the fact that the majority of Qr E. coli isolated was resistant to all quinolone-based antibiotics tested (nalidixic acid and ciprofloxacin). In addition, we predicted that Qr E. coli classified as susceptible or intermediate to ciprofloxacin, but with a resistance to nalidixic acid, had a mutation in the gyrA gene or parC gene, or else had the qnr gene. In Vietnam, a study demonstrated that 30 of 33 E. coli resistant to nalidixic acid but intermediate to ciprofloxacin had a qnr gene [62].
Several E. coli isolates from the feces of food animals in the present study were resistant to various antibiotics, including critically important antibiotics for human and veterinary medicine [63,64]. Antibiotic-resistant bacteria or genes of resistance in food animals can be transmitted to humans directly through contact with animals, or indirectly through the consumption of animal products or a contaminated environment [65]. The results of this study indicate a need to use a one health approach to control the challenge of antimicrobial resistance in Rwanda.
The absence of data on AMR in food animals in Rwanda needs to be addressed urgently. Consequently, the present results are made public to provide preliminary information on antibiotic resistance in food animals in Rwanda. Nevertheless, the next step of this study will be the exploration of genes and genetic determinants of antibiotic resistance among the isolated bacteria.

Conclusions
Overall, this study provides an overview of the distribution of AMR among food animals in the East province of Rwanda. This early insight feeds a call for more research that can cover the whole country and entire food chains. These findings can also serve as a basis of design upon which an integrated AMR surveillance system in Rwanda can be developed.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/ani11041013/s1, Table S1: List of primers used in the Loop-mediated isothermal amplification to detect invA gene of Salmonella.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.