Antimicrobial Resistance of Escherichia coli Isolates from Livestock and the Environment in Extensive Smallholder Livestock Production Systems in Ethiopia

The objective of this study was to characterize the distribution of antimicrobial resistance (AMR) of Escherichia coli (E. coli) isolated from livestock feces and soil in smallholder livestock systems. A cross-sectional study was carried out sampling 77 randomly selected households in four districts representing two agroecologies and production systems. E. coli was isolated and the susceptibility to 15 antimicrobials was assessed. Of 462 E. coli isolates tested, resistance to at least one antimicrobial was detected in 52% (43.7–60.8) of isolates from cattle fecal samples, 34% (95% CI, 26.2–41.8) from sheep samples, 58% (95% CI, 47.9–68.2) from goat samples and 53% (95% CI, 43.2–62.4) from soil samples. AMR patterns for E. coli from livestock and soil showed some similarities, with the highest prevalence of resistance detected against streptomycin (33%), followed by amoxycillin/clavulanate (23%) and tetracycline (8%). The odds of detecting E. coli resistance to ≥2 antimicrobials in livestock fecal samples were nearly three times (Odd Ratio—OR: 2.9; 95% CI, 1.72–5.17; p = 0.000) higher in lowland pastoral than in highland mixed crop–livestock production systems. These findings provide insights into the status of resistance in livestock and soil, and associated risk factors in low-resource settings in Ethiopia.


Introduction
Antimicrobial resistance (AMR) has been recognized as one of the most significant threats to the health of people and food-producing animals. The report from the World Health Organization (WHO) on AMR indicates that resistance of common bacteria has reached alarming levels in many parts of the world. For example, the resistance of Escherichia coli (E. coli) and Klebsiella spp. to last-resort third-generation cephalosporins and carbapenems antibiotics has reached up to 54% [1,2]. Some reports estimated that the economic loss due to AMR will increase dramatically, causing trillion-dollar losses by the mid-21st century [3]. In line with this, the 2030 Agenda for Sustainable Development Goals emphasized the need to address growing antimicrobial resistance [4,5].
understood. However, this understanding is essential to develop effective strategies to reduce the emergence and spread of resistance, which is a global priority. This paper characterized the distribution of AMR of E. coli isolated from livestock feces and soil in a low-resource, extensive smallholder livestock production system.
The Biolog system (Omni Log ID) laboratory reader identified 42 (9.1%) of the 462 isolates as E. coli O157:H7. Table 1 presents the occurrence of E. coli O157:H7 by sample types and species of animals examined. The higher occurrence of E. coli O157:H7 was found among isolates from goats (20.2%), cattle (11.4%) and soil samples (6.8%). Conversely, only two isolates (1.4%) were characterized as E. coli O157:H7 from sheep samples. The proportion of resistance to specific antibiotic classes among E. coli isolates from cattle, sheep, goat and soil was generally low ( Table 2). Among 440 isolates, 23.2% showed resistance to amoxycillin/clavulanate, with 25.8% in cattle, 11.2% in sheep, 37.0% in goat and 24.8% in soil samples. Among 451 isolates tested for tetracycline, overall, 8% were resistant, with 3.9% resistance in cattle samples, 5.8% in sheep, 13.6% in goat and 11.3% in soil isolates. A higher level of resistance against streptomycin was found in the 449 isolates tested, with 33.2% of all samples testing positive, with 37.3% positive in cattle samples, 16.8% in sheep samples, 49.4% in goat samples and 36.4% in soil samples. A much smaller proportion of isolates were resistant against the other antibiotics tested (Table 2). Each superscript letter (*) denotes a resistance phenotype percentage in different host types: the same letter or no letter means the percentages do not differ significantly from each other at the 0.05 level (ANOVA).

Antimicrobial Resistance among E. coli O157:H7 Isolates
Resistance among E. coli O157:H7 isolates was generally low regardless of the source of isolation. Among the 42 isolates, 5 (17.2%) were resistant to streptomycin and 3 (10.3%) were resistant to amoxycillin/clavulanate. Resistance to other antibiotics tested was rare, with resistance to chloramphenicol (3.5%) and cefotaxime (3.5%) detected at low levels. All isolates were susceptible to cefoxitin (second-generation cephalosporin).

Risk Factors for the Occurrence of Antimicrobial Resistance Phenotypes
Univariable analyses showed that six variables were potentially associated (p-value < 0.25) with the occurrence of more than two or more antimicrobial resistance phenotypes in E. coli isolated from livestock, and seven variables for E. coli isolated from soil. Table 4 shows risk factors associated with the occurrence of ≥2 antimicrobial resistance phenotypes among E. coli isolated from livestock using multivariable logistic regression analysis. The variables 'manure management' and 'what do you do with dead animals?' were both highly correlated with agroecology and not used in the model. Agroecology (pastoral system) and lack of access to professional animal health services were associated with E. coli resistance for ≥2 antibiotics in the models. The odds of detecting E. coli resistance to ≥2 antimicrobials in livestock fecal samples were nearly three times higher in lowland pastoral production systems than in highland mixed crop-livestock production systems (OR: 2.9; 95% CI, 1.72-5.17; p = 0.000). Predictor variables for the occurrence of resistance to ≥2 antibiotics among E. coli isolate in soil samples are shown in Table 5. Manure management was retained in this model. Households that either used manure for fuel (incl. biogas) or sold manure for cash had a reduction of 85% (OR: 0.15; 95% CI, 0.03-0.75; p = 0.03) and had higher odds of having ≥2 antimicrobial resistance phenotypes in soil compared with those who left manure either on the farm or in the open air or discarded manure into the environment.

Discussion
We found a similar prevalence of resistant E. coli in livestock fecal samples and soil. Goat fecal samples had the highest prevalence of resistance (58%) to at least one antimicrobial in E. coli, followed by soil samples (53%). However, it is difficult to pinpoint the origin of the antimicrobial resistance observed. The lowest proportion of resistance was observed in isolates from sheep fecal samples. Despite the absence of drug stewardship in the study area, the resistance level for individual antibiotics tested was generally low, with a higher level of resistance against 'older' drugs such as streptomycin, followed by amoxycillin/clavulanate and tetracycline. This is expected, as resistance in E. coli mainly occurs against drugs that have been commonly used for farm animal treatments and/or prophylaxis for a long time [16,[40][41][42]. Resistance against 'newer' drug classes (Cephalosporins, Quinolones, Chloramphenicol, Nitrofuran) was lower. This is in agreement with a recent systematic review and meta-analysis which noted that drug resistance in various samples, including animal-sourced foods, was against older drugs such as ampicillin, amoxicillin, streptomycin and tetracycline [35].
AMR occurrence in E. coli isolated from food-producing animals has been reported in different countries, but due to differences in sampling strategies, isolation methods and methods of AMR phenotype determination comparisons between studies is difficult. A study in Kenya showed that E. coli resistance to aminoglycosides, sulfonamides, tetracyclines, trimethoprim and penicillin was high in both humans and livestock, while resistance to cephalosporins and fluoroquinolones was low [43]. There is also evidence from community settings within countries in sub-Saharan Africa and in South Asia where E. coli resistances to 'older' antimicrobials was common, with 65% of isolates resistant to ampicillin, 67% to trimethoprim, 66% to trimethoprim/sulphamethoxazole, 56% to tetracycline and 43% resistant to streptomycin [21]. In a study in the USA, out of 746 E. coli isolates recovered from animal sources, 71.1% were resistant to tetracycline, 59% to streptomycin, 57.7% to sulfonamide and 34.1% to ampicillin [24].
Multi-drug-resistant pathogens have emerged worldwide [16]. In our study, the prevalence of MDR in E. coli was 26.7% and the most common co-resistant phenotype observed was to amoxycillin/clavulanate and streptomycin (18.8%). A relatively larger proportion of MDR E. coli isolates was recovered from animals in a US study [24]. They found that concurrent resistance to tetracycline and streptomycin was the most common coresistance phenotype (30%), followed by resistance to tetracycline and sulfonamide (29%).
In our study, 42 (9%) E. coli isolates from livestock feces and soil were E. coli O157:H7, with higher proportions among isolates from goats and cattle. Hunduma (2018) also reported comparable prevalence of 4.7% E. coli O157 in both milk and feces samples in cows from a similar setting [44]. It is commonly cited that cattle are the primary reservoir of E. coli O157:H7 [45], with small ruminants also implicated [46][47][48]. The resistance level among E. coli O157:H7 isolates was generally low regardless of the source of isolation. Resistance to streptomycin (17%) and amoxycillin/clavulanate (10.3%) were the most common resistance profiles seen in these isolates. However, Hunduma found a higher level of E. coli O157 resistance to streptomycin (65%), tetracycline (59%) and Trimethoprim (24%) [44]. These drugs are still commonly administered in humans [49] and animals [41,42] in Ethiopia.
Bekele et al. reported E. coli O157:H7 isolates from raw meat in Addis Ababa that were resistant to different antibiotics including streptomycin (33%) and tetracycline (5%) [36]. With this study, we report for the first time AMR in E. coli O157:H7 isolated from soil, and thus confirm that soil contaminated via feces can act as a source of drug-resistant microbial pathogens including E. coli O157:H7. Greater attention should be paid to prevent E. coli O157:H7 contamination of the human food chains, given its health impact. Many studies have shown that the survival of E. coli O157:H7 in soil can lead to contamination of drinking water, fruits and fresh vegetables and constitutes a major public health threat [9,[50][51][52][53][54]. Furthermore, E. coli O157:H7 can cause severe hemorrhagic colitis and hemolytic uremia in humans [55].
Our results also found higher odds for resistance in livestock from the lowland pastoral production system. This may be due to higher infection pressure and probability of recirculation of resistant isolates in the lowland agroecology and pastoral production system. It is possible that warm temperatures offer more potential for bacteria to multiply, with greater transference of antimicrobial resistance genes. Warmer temperatures are also associated with higher insect populations, which can play a role in disseminating resistant bacteria [56]. It could, however, also reflect the fact that improper use of antibiotics, mostly without a proper diagnosis, is more common in these production systems [42]. Such information is important to target AMR management practices.
Poor management, including the management and disposal of manure (i.e., leaving manure either on the farm, or in the open-air or discarding manure into the environment) was also strongly associated with detecting a higher level of E. coli resistance to more than two or more antimicrobial resistance phenotypes in soil samples. Similarly, Muloi et al. (2019) found that keeping manure inside the household compound was also significantly associated with AMR carriage in humans [43]. Animal manure has been implicated as a reservoir of AMR bacteria and AMR determinants [57,58]. Transmission of antibioticresistant E. coli and resistance genes may also occur through environments contaminated with feces, especially in developing countries [57,59].
Low levels of resistance are often overlooked, but can play an important role in the expansion of resistance [60]. In these extensive smallholder and pastoral settings, there is little to no testing of drug susceptibility during treatment of cases for both humans and animals. Hence, minimizing resistance is crucial. There is a need to maintain an overview of drug susceptibility through an AMR surveillance system that monitors resistance patterns and trends.

Study Area
The study was conducted in two agroecological zones and production systems: (i) a highland mixed crop-livestock production system (Menz Mama and Menz Gera district) and (ii) a pastoral system (Yabello and Eleweya districts). Details of the characteristics of the study area were published elsewhere [42]. Briefly, highland agroecology with a mixed crop-livestock system is typical for areas above 2200 m above sea level (masl) in which livestock husbandry depends on rain fed cropping. In lowland agroecology, pastoral livestock production is widespread, with the community mainly dependent on livestock and livestock products.
In recent publications, differences were reported between locations and production systems in terms of characteristics of antimicrobial usage including: (1) access to antimicrobials, (2) types of antimicrobials used and (3) when they were used. Livestock producers in mid/lowland pastoral systems appeared to use antibiotics more frequently than their counterparts in highland and lowland mixed crop-livestock systems [42].

Study Design and Sample Size Determination
A cross-sectional study was conducted with 77 households selected from extensive smallholder livestock systems in four districts. A total of 539 samples, which included 462 livestock fecal samples (cattle = 152, sheep = 180 and goats = 130) and 77 soil samples, were collected.
For fecal samples, the number of animals to be sampled in the study was estimated by the formula [61]; where z = 1.96, Pexp (the expected prevalence) = 0.11 and d = 0.05 (the desired level of precision). Based on the result of a systematic review and meta-analysis, the overall pooled prevalence of E. coli expected was 15% in all samples [62]. The required sample size was, therefore, n = 231. To account for herd level clustering, the target sample size was adjusted using an intra-cluster correlation coefficient of 0.2 with an average of 6 animals sampled per herd. Accordingly, the design effect (D) of the study was calculated as 1.4 according to: where m was cluster size (i.e., 6), ρ was 0.2 and the calculated sample size was adjusted by multiplying by D. Therefore, the new sample size was 462 animals from 77 households. Hence, 77 soil samples were collected from the homestead and barn areas of 77 households. Household data were previously collected. In each household, details of household demographics, farm characteristics, manure management, feed types, animal health constraints, disease prevention, animal health services, antimicrobial use and animal product consumption were collected. Information on the selection of agroecological zones, districts and villages and random household selection was described in [42].
Because this study does not focus on the number of isolates per animal, we restricted the number of isolates to one or zero per animal.

Sample Collection and Pre-Enrichment Procedure
Fecal samples were taken from the rectum using a gloved hand and a sterile 50 milliliter (mL) capacity Falcon tube.
Soil samples were collected from either the homestead or the barn area of ruminants, from 2-5 cm beneath the surface. Approximately 10 g of soil, free of obvious fecal contamination, was collected into sterile vials. If the surface of the area was not flat, samples were collected from the lowest as well as middle and highest points, and mixed.
The fecal and soil samples were refrigerated and transported for laboratory analysis at either Yabello Pastoral and Dryland Agricultural Research Center (for samples collected from lowland pastoral areas) or the International Livestock Research Institute, Addis Ababa (for samples from highland agroecology) within 4-6 h of collection.
Immediately upon arrival at the lab, a sample suspension was prepared using 1 g of the sample in 9 mL of phosphate-buffered solution (5%). Samples were pre-enriched in buffered peptone water and incubated at 370 • C for 24 h.

Isolation and Identification of E. coli
A loop full of pre-enriched cultures was taken and inoculated on MacConkey agar and then incubated at 37 • C for 24 h. Typical colonies on MacConkey agar (pink, due to their ability to ferment lactose) were Gram-stained and observed for their staining and morphological characteristics, then transferred to eosin-methylene-blue (EMB) agar and incubated for 24 h at 37 • C. The colonies with metallic sheen on EMB agar, which is a typical characteristic of E. coli, were then considered as E. coli-positive and transferred to nutrient agar to be used for additional confirmatory biochemical tests (IMViC tests) and for further identification for Biolog tests.
Presumptive pure E. coli isolates were further analyzed using the Biolog system (OmniLog ID system, Hayward, CA, USA) following the standard procedures of the manufacturer. Briefly, the purified cultures of E. coli were inoculated on BUG (Biolog Universal Growth) agar medium; inoculums were prepared at a specified cell density using inoculating fluid A (IF A); the Biolog microplate GEN III was inoculated with the inoculums; the plate was incubated at 330 • C for 22 h into the Ominilog apparatus; the reaction pattern was entered; and results were obtained from the apparatus. The system also further identified E. coli O157:H7 serotype.
The purified cultures of E. coli were then stored in glycerol added to TSB broth at −200 • C for further biotyping and other studies.

Antimicrobial Susceptibility Test
The antimicrobial susceptibility test on the isolates was performed according to the National Committee for Clinical Laboratory Standards [63] using the Kibry-Bauer disk diffusion test method on Muller-Hinton agar (Oxoid CM0337 Basingstoke, England).
From each isolate, four to five confirmed colonies grown on nutrient agar were transferred to a test tube of 5 mL Tryptone Soya Broth (TSB) (Oxoid). The broth culture was incubated at 37 • C for 18 h until growth reached the 0.5 McFarland turbidity standard.
Mueller-Hinton agar plates were readied according to the manufacturer's guidelines and held at room temperature for 30 min to allow drying. A sterile cotton swab was dipped into the suspension and then swabbed uniformly in three directions over the surface of Muller-Hinton agar plate (Oxoid Ltd, Hampshire, UK). After the plates dried, antibiotic disks were placed using an automatic Oxoid antimicrobial disk dispenser onto the inoculated plates and incubated at 37 • C for 24 h. After incubation for 24 h, the diameter of the zones of inhibition was measured and compared with zone size interpretation guidelines described by the Clinical Laboratory Standard Institute [64] for the family Enterobacteriaceae and determined as sensitive, intermediate or resistant.
The isolated E. coli were tested for sensitivity to the most commonly used antimicrobials. The zone of inhibition was interpreted based on the Performance Standards for Antimicrobial Susceptibility Testing; Sixteenth Informational Supplement, as detailed in Table 6.

Data Analysis
Overall and host-and sample-specific (livestock and soil) distribution of resistance phenotypes were determined by dividing the number of isolates with specific resistance phenotypes by the total number of isolates examined. Difference in proportions of samples in a group with resistance to one or more antibiotics was determined using Chi-squared tests and one-way analysis of variance (ANOVA; soil vs. different livestock groups). Bonferroni's multiple-comparison test was performed post hoc for pairwise comparisons between groups, and p-values < 0.05 were considered significant.
Host-and sample-specific occurrence of E. coli O157:H7 were determined by dividing the number of E. coli O157:H7 isolated in each host by the total number of samples examined. The proportions of antimicrobial resistance were also determined by dividing the number of E. coli O157:H7 isolates with specific resistance phenotypes by the total number of isolates examined.
From the household data, eight variables were used as potential predictor variables for the occurrence of AMR. The variables were agroecology (highland mixed croplivestock production system vs. pastoralist system); species mix (keep <3 livestock species vs. keep ≥3 livestock species); manure management (leave on farm or open air, discard into environment vs. use as fertilizer vs. use for fuel (incl biogas), sale for cash); isolation of sick animals (yes vs. no); allow mix of animals on treatment (yes vs. no); what do you do with dead animals (leave as it is vs. give to the dog vs. bury vs. human consumption); access to professional animal health services/treatments (yes vs. no); access to regular animal health services/vaccination and deworming (yes vs. no). A subset of the data that only included isolates tested for all antimicrobials (n = 288 for livestock and n = 87 for soil) was used for analysis of risk factors. To identify risk factors for having resistance to ≥2 antibiotics, odds ratios (ORs) were calculated using univariable logistic regression, followed by multivariable logistic regression.
Separate models were developed for livestock and soil, and a forward selection method with a significance level of 5% was used to select a suitable model. Models were adjusted for the cluster effect using robust standard error estimation for cluster sampling. When selecting important variables to the model, the Wald statistic associated with the variable was used instead of the likelihood ratio test (LRT) statistic [65].
Collinearity was assessed pair-wise via calculation of Spearman rank correlations between predictors. Variables with a p-value < 0.25 were considered for multivariable analysis, provided that there was no collinearity (r < |0.7|) between them. Potential confounders were considered in every model as variables that, if present, changed the coefficient for one or more significant variables by an amount that was important to report, taken as 10% [66]. All variables with a p-value ≤ 0.05 were retained in the final model. There were no biologically plausible interactions between the main effects expected and tested. Data were analyzed using Stata software version 16 (College Station, TX, USA).

Conclusions and Recommendations
These findings provide insights into the status of resistance in livestock and soil, and associated risk factors in low resource settings in Ethiopia. AMR patterns for E. coli from livestock and soil showed similarities. The resistance level for individual antibiotics tested was generally low, with the highest prevalence of resistance detected against streptomycin (33%), followed by amoxycillin/clavulanate (23%) and tetracycline (8%). The most common co-resistant phenotype observed was against amoxycillin/clavulanate and streptomycin. The findings showed that soil contaminated via feces can act as a source of drug-resistant microbial pathogens including E. coli O157:H7. The results also revealed that agroecology and production systems and manure management were strongly associated with occurrence of resistance. Additional molecular analysis of resistance genes is now planned to further investigate evidence of overlapping patterns of transferable resistance genes between livestock and soil. This would provide more definitive data on how resistance gene clusters have evolved and the context in which genes are maintained in the absence of known selection pressures.
In conclusion, we recommend that animal health and biosecurity practices, such as manure management, are radically improved, and that an integrated AMR surveillance system is established. Funding: This research was conducted under CRP livestock and continued under the CGIAR Initiative Sustainable Animal Productivity for Livelihoods, Nutrition and Gender (SAPLING). CGIAR research is supported by contributions from the CGIAR Trust Fund. CGIAR is a global research partnership for a food-secure future dedicated to transforming food, land and water systems in a climate crisis. The German Academic Exchange Service (DAAD) supported the project through a ILRI-DAAD Doctoral Fellowship for the first author. The funders played no role in the design or conclusion of the study and any opinions, findings, conclusions and recommendations expressed here are those of the authors alone. Informed Consent Statement: Not applicable. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

Data Availability Statement:
The datasets generated for this study are available on request to the corresponding author.

Acknowledgments:
The authors thank the Animal Health Institute (AHI) and Yabello Regional Veterinary Laboratory in Ethiopia and their technical staff, who were involved in field data collection and laboratory analysis. The farmers and pastoralists who participated in the study are greatly appreciated.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.