Gut carriage of antimicrobial resistance genes in women exposed to small-scale poultry farms in rural Uganda: A feasibility study.

BACKGROUND: Antibiotic use for livestock is presumed to be a contributor to the acquisition of antimicrobial resistance (AMR) genes in humans, yet studies do not capture AMR data before and after livestock introduction.
METHODS: We performed a feasibility study by recruiting a subset of women in a delayed-start randomized controlled trial of small-scale chicken farming to examine the prevalence of clinically-relevant AMR genes. Stool samples were obtained at baseline and one year post-randomization from five intervention women who received chickens at the start of the study, six control women who did not receive chickens until the end of the study, and from chickens provided to the control group at the end of the study. Stool was screened for 87 clinically significant AMR genes using a commercially available qPCR array (Qiagen).
RESULTS: Chickens harbored 23 AMR genes from classes found in humans as well as additional vancomycin and β-lactamase resistance genes. AMR patterns between intervention and control women appeared more similar at baseline than one year post randomization (PERMANOVA R2 = 0.081, p = 0.61 at baseline, R2 = 0.186, p = 0.09 at 12 months) Women in the control group who had direct contact with the chickens sampled in the study had greater similarities in AMR gene patterns to chickens than those in the intervention group who did not have direct contact with chickens sampled (p = 0.01). However, at one year there was a trend towards increased similarity in AMR patterns between humans in both groups and the chickens sampled (p = 0.06).
CONCLUSIONS: Studies designed to evaluate human AMR genes in the setting of animal exposure should account for high baseline AMR rates. Concomitant collection of animal, human, and environmental samples over time is recommended to determine the directionality and source of AMR genes. TRIAL REGISTRATION: ClinicalTrials.gov Identifier NCT02619227.

Introduction

Antimicrobial resistance (AMR) is a global public health crisis. Although estimates vary on the severity of the problem, one report has suggested that by 2050, 10 million deaths a year worldwide will be attributed to antimicrobial resistance [1], with crude estimates of the annual economic costs totaling 55 billion dollars in the United States alone [2]. This problem may be accentuated in resource-limited settings due to a likely higher burden of infectious disease, little to no antimicrobial stewardship, less resources for microbiology testing, possible limited access to antibiotics targeting highly resistant pathogens, insufficient sanitation and hygiene infrastructure for managing human and animal waste. Although prior AMR studies have focused on hospitalized patients and recent administration of antimicrobials to treat infections, an updated view of AMR as a public health problem has highlighted the importance of AMR as a “One Health” problem; that is, viewing human, animal, and environmental health as interconnected and interdependent [3-5]. Antibiotics are widely used in livestock farming to enhance animal health and increase productivity [6], and this practice is thought to be one contributor to the problem of AMR among humans. However, most available studies are cross-sectional and/or focused on single organisms or pathogens [7-9], and these study designs lack the ability to determine causality. More robust study designs are needed to determine the effect size that antimicrobials used in livestock farms has on transmission of AMR genes to humans [10].

Surveillance data in 2005 showed that livestock production in Uganda accounted for about 5% of total Ugandan gross domestic product [11], with an estimated annual production of 70.8 million total livestock including cattle, pigs, sheep and goats, and poultry [12]. Studies of poultry farms in Uganda have identified multiple mechanisms of AMR in Escherichia coli strains isolated from healthy chickens [13, 14], suggesting that poultry farms may serve as a reservoir of AMR genes for humans. Few studies have evaluated how the initiation of chicken farming relate to AMR in humans, partly due to difficulty in obtaining pre-intervention samples for AMR testing. In this study we determined the feasibility of a longitudinal study of AMR gene patterns in a subset of participants enrolled in a delayed intervention RCT of small-scale chicken farms in rural Uganda.

Materials and methods

Study design and study population

We recruited participants from an existing randomized clinical trial (RCT) of small-scale chicken farming (ClinicalTrials.gov Identifier: NCT02619227) [15]. Participants were chosen by convenience sample from the original trial and assessment of AMR gene carriage was added as a post-hoc aim. In the primary waitlist-controlled RCT conducted in 2015, 92 women living in Mbarara, Uganda were recruited and randomized to receive training, raw materials, and broiler hybrid chicks either immediately (intervention group), or after at least a 12-month delay (control group). Chicken coops were constructed to house chicks as part of the study protocol, and study participants were the primary caretakers for the broilers. Chicks, feed, and medications were given to intervention participants as a series of escalating microloans starting from 15, 50, then 100 chicks with loans paid back prior to the receipt of the next batch of chicks (see S1 Fig). Broiler chicks were sourced from a single distributor based in Kampala, Uganda and underwent a standard care protocol by participants during the brooding period which lasted approximately 8 weeks. Under supervision, participants administered vaccines to the chicks against Newcastle, Gumboro, fowl typhoid, and fowl pox. Participants also routinely administered dietary supplements to chicks in their drinking water during the brooding period as part of a protocol to boost growth. This included two oxytetracycline-containing medications; Alamycin chick formula given for the first 2–3 weeks of brooding and Oxiveto given weekly for four weeks. In addition, Coccid (which contains amprolium) was given once weekly for four weeks to prevent coccidiosis. Chicken feed was sourced from a single distributor based in Mbarara, Uganda. Routine surveys were administered as part of the RCT to monitor behaviors such as recent antimicrobial use (in both chickens and humans) and vaccination status in chickens, and data pertinent to this study was extracted from the survey developed for the larger RCT (see Supplement). The primary trial was designed as a series of microloans in the form of chickens. Per study protocol, the time from chick acquisition to slaughter was 8 weeks, although participants were given the option of an additional two weeks to sell their chickens and pay back the loan.

The timing of stool sample collection is depicted in S1 Fig. Stool samples from six chicken coops belonging to the 6 control participants were collected by retrieving fresh chicken stool once at approximately 18 months after randomization, between 20 and 41 days after the control group had received their chickens as part of the delayed-start randomized controlled trial design. Human samples could not be collected at the 18 month timepoint when chicken samples were collected. The 6 control participants were chosen based on participants who had stool samples collected from the chickens, and the 5 intervention participants lived in the same villages as the control participants. Stool sample collection was added post-hoc for this feasibility study as an addition to the original study. At baseline, before chickens were introduced into the intervention households and at 12-month follow up after chicken introduction in the intervention group, we obtained fresh stool samples from participants during research clinic visits. Stool samples were frozen within one hour of collection in generator-backed -80°C freezers in the research laboratories of the Mbarara University of Science and Technology. All samples were subsequently transported on dry ice to Massachusetts General Hospital for further processing. All study procedures were approved by the Research Ethics Committee of Mbarara University of Science and Technology (Protocol #30/11-14) and the Partners Human Research Committee (Protocol #2015P000227/BWH). Consistent with national guidelines, we also received clearance for the study from the Ugandan National Council of Science and Technology (Protocol #HS 1746) and the President’s office.

Sample processing, AMR gene identification and quantification

Microbial DNA was extracted from 100mg of chicken and human stool samples, and from a reagent-only negative control using the PowerSoil DNA extraction kit (Qiagen, Valencia, CA) according to the manufacturer’s instructions. The presence of AMR genes was screened using a commercially available AMR gene identification microbial DNA polymerase chain reaction (PCR) array (Qiagen, Valencia, CA, cat. No. 330261) according to the manufacturer’s instructions. This array targets six major classes of antibiotics (aminoglycoside, β–lactam, erythromycin, fluoroquinolone, macrolide–lincosamide–streptogramin B, tetracycline, and vancomycin) and includes genes with multi-resistance potential. Briefly, 500ng template microbial DNA was mixed with 1275 μl qPCR mastermix (Qiagen) and nuclease-free water was added to reach a final volume of 2550 μl. 25 μl of reaction mix was added to a 96-well PCR plate containing a pre-dispensed mixture of lyophilized primers and probes for each of the 87 AMR genes. qPCR was performed using Applied Biosystems 7500 Fast Real-Time PCR System using thermal cycling conditions of initial denaturation at 95°C for 10 minutes, followed by 40 cycles of denaturation at 95°C for 15 seconds, and annealing at 60°C for 2 minutes. Raw cycle threshold (CT) values were analyzed using the Microbial DNA qPCR Array data analysis template. One replicate per sample was tested. The efficiency of the PCR instrument and the quality of mastermix were determined by measuring the CT for the control sample between 20 and 24. Validity of the control ensured that potential PCR inhibitors in the sample did not interfere with measurements. A no-template and nuclease-free control were also included to evaluate for the presence of laboratory based contaminants.

Data analysis, visualization and statistical analysis

Determination of detection of AMR genes was performed according to the manufacturer’s (Qiagen) guidelines, described here in brief. The presence or absence of each AMR gene was determined as follows: present if ΔCT > 6, not detected if ΔCT <3, and inconclusive if ΔCT was ≥ 3 and ≤6. To visualize the results of AMR gene presence or absence in each sample, we created a heatmap using the ggplot2 R package [16]. In order to visualize global patterns of AMR genes over time in the human samples and difference between the chicken samples, we chose to use the Jaccard dissimilarity index. Briefly, the Jaccard index calculates the proportion of unshared features (here AMR genes) out of the total number of features (here AMR genes) recorded between any two samples, an approach used in other studies of high-dimensional antimicrobial resistance data [17, 18]. To calculate the Jaccard index, we first created a sample by feature matrix denoting the presence or absence of each AMR gene in each sample. Presence/indeterminacy/absence were determined using the ΔCT method described above according to manufacturer recommendations, with the following value assignments; present = 1, indeterminate = 0, absent = 0. Visualization of the dissimilarities in AMR gene patterns was performed using the plot_ordination() function as implemented in the phyloseq R package [19]. To test the hypothesis that AMR gene patterns in human control groups were the same or different at baseline and one year post-randomization, we performed permutational multivariate analysis of variance (PERMANOVA) [20, 21] on the Jaccard index with 10,000 permutations as implemented in the vegan R package [22]. To determine the similarities in AMR gene patterns between human control and intervention groups over time compared to chickens, we computed the distance between the Jaccard index of each sample to the centroid of all chicken samples [23]. In this plot, a shorter distance between data points indicates increased similarity in AMR gene patterns. Measurements were calculated using the dist_between_centroids() function implemented in the usedist R package [24]. For statistical testing, we performed a mixed effects model as implemented in the mgcv R package [25] where the outcome was the calculated distance between each sample and the centroid of the chicken samples, covariates were group membership (intervention vs control) and time (baseline vs follow-up), adjusting for repeated measures in a subject. All statistical analyses were performed in the R programming language [26]. Two-sided p values of < 0.05 were considered statistically significant.

Results

We collected stool from five women in the intervention group and six women in the control group, from 11 separate households in Nyakabare parish, Mbarara district, Uganda. Mbarara is located in a rural area of Uganda approximately 260km southwest of Kampala, the capital city. The local economy is largely dominated by animal husbandry, petty trading, subsistence agriculture, and supplemental migratory work. Food and water insecurity are common [27-29]. In this study, samples were collected between August 11, 2015 and June 8, 2017. The median age of participants was 35 years, and self-reported demographic data are listed in Table 1. All participants were women involved in subsistence farming. At baseline, 10 of the participants reported regular animal contact and a minority reported recent antibiotic use.

Table 1

Baseline characteristics of study participants.

	Control	Intervention
n	6	5
Age, years	40 [34–43]	33 [25–40]
Farming	6 (100%)	5 (100%)
Antibiotic use in prior three months
At 0 months	1 (17%)	0 (0%)
At 12 months	1 (17%)	1 (20%)
Animal contact	5 (83%)	5 (100%)
Village chickensa	2 (33%)	5 (100%)
Cows	2 (33%)	2 (40%)
Goats	4 (67%)	4 (80%)
Pigs	1 (17%)	0 (0%)
Dogs	2 (34%)	2 (40%)
Cats	2 (34%)	2 (40%)

aVillage chickens refer to free-range chickens that do not receive vaccinations or medications, and do not require an enclosure.

AMR genes detected

Stool samples from chickens, and from pre- and post-intervention human control and intervention groups were assayed for AMR genes using a validated quantitative polymerase chain reaction (qPCR) assay. All of the no-template controls and positive PCR controls passed the quality control thresholds determined by the manufacturer (S1 Table). At baseline, the stool of study participants in both control and intervention groups harbored β-lactamase, aminoglycoside, fluoroquinolone, macrolide and tetracycline AMR genes found in the stool (Table 2). Seven new AMR genes were detected after one year in the intervention group, and four of these were present in chickens (SHV, SHV[238G240E], QnrS, QnrB-5 group). Six new AMR genes were detected after one year in the control group, and one of these was present in chickens (CTX-M-1 group). Overall, AMR genes were detected from five classes of antimicrobials in humans, and six classes in chickens.

Table 2

Antimicrobial resistance (AMR) genes in participants detected at baseline and one year post-intervention, and in chickens.

Among study participants, newly detected genes after one year are shown in bold. Baseline grouping includes both intervention and control group participants. AMR gene detection was measured using a qPCR array (Qiagen). Raw cycle threshold (CT) values were used to determine detection of AMR, defined as positive if ΔCT >6, not detected if ΔCT <3 and inconclusive if ΔCT was ≥ 3 and ≤6, as per the manufacturer’s instructions. Raw qPCR data is shown in S1 Table. Gene names are italicized and names of gene classes are not.

Antibiotic classification	Women at baseline (0 months)	Women in intervention group (12 months)	Women in control group (12 months)	Chickens (18 months*)
Aminoglycoside resistance	aadA1	aacC2, aadA1	aacC2, aadA1	aadA1
Class A β-lactamase	CTX-M-1 group, CTX-M-9 group, SHV, SHV(156G), SHV(238G240E)	CTX-M-1 group, SHV, SHV(156G), SHV(238G240E)	CTX-M-1 group, SHV, SHV(156D), SHV(156G), SHV(238G240E)	CTX-M-1 group, SHV, SHV(156G), SHV(238G240E), SHV(238S240E), SHV(238S240K)
Class B β-lactamase	ccrA	ccrA	--	ccrA
Class C β-lactamase	ACT-1 group, ACT 5/7 group, ACC-3, MIR	ACT-1 group, ACT 5/7 group, MIR	ACT-1 group, ACT 5/7 group, CFE-1, LAT, MIR	ACT-1 group, MIR
Class D β-lactamase	--	--	--	OXA-10 group, OXA-58 group
Fluoroquinolone resistance	QnrS, QnrB-1 group, QnrB-5 group	AAC(6)-Ib-cr, QnrS, QnrB-5 group	AAC(6)-Ib-cr, QnrS, QnrB-1 group, QnrB-5 group	QnrS, QnrB-5 group, QnrB-8 group
Macrolide Lincosamide Streptogramin_b	ermB, mefA	ermB, mefA	ermB, mefA	ermA, ermB, ermC, mefA, msrA
Tetracycline efflux pump	tetA, tetB	tetA, tetB	tetA, tetB	tetA, tetB
Vancomycin resistance	--	--	--	vanB, vanC

* Chicken stool was collected 18 months after randomization, but between 20–41 days after chick delivery to the control group.

AMR gene class trends between groups and over time

During the study period there was an overall increase in AMR genes in both the control and intervention groups. The most prevalent AMR genes were tetA and tetB, which confer tetracycline efflux pumps, and these were found in all chickens tested. tetA and tetB were also found in the majority of human participants in the study at baseline and follow-up timepoints, as shown in a heatmap of our overall results (Fig 1 and S1 Table). β-lactamases were also highly prevalent in both humans and chickens, with Class A and C β-lactamase AMR genes found in humans at both baseline and follow-up timepoints, regardless of chicken exposure, and the Class C β-lactamase MIR present in nearly all study participants. However, Class D β-lactamases were found only in chickens. AMR genes in the Class C β-lactamase group, which includes the clinically important ampC β-lactamases responsible for inducible resistance upon exposure to specific antibiotics were particularly dynamic over time, with two AMR genes emerging in the control group after one year that were not seen in other groups (CFE-1 and LAT), and the loss of ACC-3, which was found in the baseline population and not detected upon follow up [30]. Fluoroquinolone and macrolide resistance were widespread over all groups and timepoints. Chicken AMR genes detected included two vancomycin resistance genes that were not found in humans.

Fig 1

Heatmap demonstrating whether antimicrobial resistance (AMR) genes were present, absent, or indeterminate in human and chicken samples at different timepoints.

To address the question of whether AMR gene profiles were different between intervention and control groups at baseline and 12 months after randomization, we used permutational analysis of variance (PERMANOVA) on the Jaccard distance to determine whether the centroids of the intervention and control groups differ at baseline and 12 months after randomization in the control vs. intervention groups. At baseline, there was no difference in AMR resistance patterns between control and intervention groups (PERMANOVA R2 = 0.081, p = 0.61), whereas at 12 months there was a trend towards a difference in AMR patterns between intervention and control groups (PERMANOVA R2 = 0.186, p = 0.09). We used an ordination plot to depict patterns of AMR gene composition across groups and over time (Fig 2).

Fig 2

Ordination plot of the Jaccard dissimilarity index of AMR gene patterns between groups.

The proportion of unshared AMR genes out of the total number of AMR genes detected between any two samples is shown. More similar samples will appear closer together on the plot. The ellipse depicts the 95% confidence ellipse around each sample group. At baseline, there were no statistically significant differences between AMR gene patterns between intervention and control groups (PERMANOVA R2 = 0.081), whereas at 12 months, there was a trend towards different AMR gene patterns (PERMANOVA p = 0.09) between intervention and control groups.

To determine the similarity of AMR gene patterns of human samples compared to the chicken samples, we computed the distance between the Jaccard index of each sample to the centroid of the chicken samples (Fig 3). A shorter distance between data points indicates increased similarity in AMR gene pattern with the chicken samples, while a higher distance indicates decreased similarity in AMR gene pattern with the chicken samples. To identify predictors of similarity between human and chicken AMR patterns, we used mixed effects models where the outcome was the distance between each human sample compared to the centroid of the chicken samples, with predictors being control vs. intervention group and timepoint (baseline vs. 12 months), adjusting for repeated measures in a person. The AMR gene pattern of the control group is more similar to the AMR gene pattern in their chickens than the intervention group to the control group’s chickens (b = 0.128, p = 0.014, intervention vs. control group; Note more positive b indicates less similarity with chicken samples). There was a trend towards increasing similarity of AMR gene patterns between all human groups (control and intervention) and chickens at one year compared to baseline though it did not reach statistical significance (b = -0.067, p = 0.059, 12 month vs baseline). In this latter comparison, the effect size was negative, which is consistent with increased similarity between follow-up AMR gene patterns in both the intervention and control groups compared to chicken AMR gene patterns.

Fig 3

Boxplot of the distance between sample groups and the centroid of the chicken stool samples based on AMR gene pattern.

To demonstrate the comparison of the AMR gene pattern of each human sample to the chicken samples at baseline and follow up, we computed the distance between the Jaccard index of each sample to the centroid of all chicken samples. Here, a shorter distance indicates increased similarity in AMR gene pattern of the human sample in relation to the centroid of the chicken samples gene patterns, whereas a longer distance indicates decreased similarity in AMR gene pattern of that human sample compared to the chicken samples gene patterns. The chicken sample centroid is set at zero. The AMR gene pattern of the chicken samples is more similar to the AMR gene pattern in the control group rather than the intervention group (p = 0.014); note that chicken samples were obtained from the control group. Differences in AMR gene patterns over time did not reach statistical significance (p = 0.059), although at follow-up, the AMR gene patterns in both control and intervention group humans were more similar to AMR gene patterns in chicken samples.

Discussion

In this study, we find that tetracycline-exposed chickens and humans who care for them harbor AMR genes from multiple gene classes. Over one year, AMR gene carriage increased in all study participants although we only tested two time points in humans. Women who did not care for chickens during the 12 months of human sampling (control group) harbored many of the same AMR genes at one year. There were greater AMR gene pattern similarities between chickens and the humans who had direct interaction with the chickens in the study. This latter finding should be interpreted conservatively and with several caveats: first, chicken stool samples were obtained only after human stool sample collection, and were obtained only from the control group who received chickens at the end of the study. It is possible that the control group samples were more similar to their chickens either due to community-wide human to chicken AMR gene transmission, or because the control group and chickens had a common environmental source such as community wells for water.

Shared gut organisms between animals and humans increase when close contact occurs between groups, such as in animal husbandry [31, 32]. These shared environments can result in transmission events, which range from zoonotic infections to the spread of benign commensal microbes, or events that represent potential harm to humans or animals, such as acquisition of AMR genes. Pathogens resistant to antibiotics result in more severe illness and increased mortality in humans compared with infections caused by susceptible bacteria [33]. Consistent with the One Health concept, we found that humans and chickens with direct contact had greater similarities in AMR gene carriage in the gut, although the directionality of transmission could not be determined based on our study design. Stool samples from the control group were obtained prior to chicken introduction, and thus increased similarity in AMR gene patterns with the chickens in this study who had direct contact with the control group could be due to human to chicken transmission, or a common environmental source. Prior studies have demonstrated that the introduction of tetracycline-supplemented feed to chickens led to increased carriage of multi-drug resistant bacteria in the feces of chickens, and after 3 months, a rise also in resistant intestinal bacteria in farm workers caring for these chickens. Bi-directional transfer of AMR genes is possible, likely through mobile genetic elements. Additionally, we observed overall that there are many AMR genes in chickens in rural Uganda.

Tetracycline resistance genes are often found to be widespread among livestock treated with antibiotics, including in Africa [34]. Use of tetracycline in livestock has been associated with increased colony counts of tetracycline-resistant human pathogens in treated animals [35]. While tetracycline was the only antibiotic administered to chickens in this study, a wide range of AMR genes from six different classes were detected in chickens. Many β-lactamase AMR genes with direct links to difficult-to-treat human infections were also detected. For example, the CTX-M-1 Group can confer an extended-spectrum beta lactamase (ESBL) phenotype, and is the most commonly found gene in Escherichia coli in the few surveys of AMR genes that have been conducted in African livestock [34]. CTX-M-1 was detected after one year in our control group and was also present in chickens in this study,. CTX-M-1 was also present in the intervention group both at baseline and follow up. While the directionality of CTX-M-1 transmission between control participants and chicken exposure cannot be evaluated in this study, our results demonstrate that CTX-M-1 is circulating in this population among chickens and humans. The Carbapenemases OXA-10 and OXA-58 Group were also found in chicken stool in our study and were not found in humans, and confer a concerning degree of antimicrobial resistance [36]. Reasons that these AMR genes have not emerged into the human population are unknown, and may be due to a lack of selective pressure (ie chickens and humans not yet exposed to carbapenem antibiotics) at the time of our study. Similarly, the VanB and VanC genes found in chickens are known to confer vancomycin (glycopeptide antibiotic) resistance to Enterococci, a common genus of the colonic flora, resulting in the clinically important vancomycin resistant enterococcus (VRE). Avoparcin, an antibiotic also from the glycopeptide class, was widely used in livestock and poultry in Europe and linked to VRE isolates in animals. This drug was outlawed for use in animals in the European Union in 1997, although VRE isolates have persisted in some poultry populations after use ceased [37, 38]. Although Avoparcin was not known to be administered to the chickens in this study, it is sold in Uganda as a livestock supplement.

In the humans we studied, numerous additional AMR genes were detected in both the intervention and control groups at the one year follow up timepoint compared to baseline. There are several possible explanations for this finding. First, it is possible that the higher amount of AMR gene content at the one year follow up is simply due to ongoing, transient fluctuations in AMR presence that occur over time. Another possibility is that the population may be trending toward increased AMR gene content over relatively short time periods, and that AMR genes in this population are widespread and dynamic. For example, in the fluoroquinolone class, the AAC(6)-lb-cr gene is often found on a multiresistance plasmid with other AMR genes, and this gene was detected in both human groups after follow up and was not found in chickens. This indicates that the increased AMR gene patterns over time seen in this population may also originate from sources unrelated to chicken exposure, such as environmental sources [39]. Over one year, we observed that microbial community profiles in humans were significantly altered with (intervention group) or without chicken exposure (control group). We also note shared AMR genes between humans and chickens. Possible explanations for our findings could be that 1) there is a common source of AMR genes in both chickens and humans, for example environmental sources such as water; 2) the possibility exists that AMR genes may be transmitted from humans to chickens; 3) our data does not allow us to comment on transmission of AMR genes from chickens to humans as chicken stool samples were collected after the human stool samples. However, all chicks were from the same distributor and underwent the same care protocol and thus it is possible that some of the AMR genes acquired by humans over time were from direct or indirect chicken contact.

In this study, we describe point prevalence estimates of AMR genes over two timepoints in humans. Our study has some limitations. This pilot study does not evaluate for the directionality or source of transmission of AMR genes detected in humans and chickens, because chicken stool samples were not collected at the same timepoints, and only two human timepoints were collected. Larger longitudinal studies should include repeated measures within subjects with assessments of correlation within subjects and modeling assessments of associations to account for repeated measures and autocorrelation. This limited our ability to detect random variation from true trends. We also did not assess cross-contact between enrolled participants within and between villages. In this study we did not aim to assess AMR transmitted from other community sources, and focused only on AMR gene content in chickens and humans. Additionally, our sample size was small, and our detection method often identified gene classes, preventing us from commenting on presence of specific genes. Additionally, the number of replicates collected over time in both humans and chickens are unlikely to capture the diversity of timing relative to chicken production cycle and antibiotic use in that cycle. Our qPCR detection of genes was conducted with single replicates due to the cost of the AMR gene arrays. Despite these limitations, our study does highlight the prevalence of circulating AMR genes in people and in chickens in a rural Ugandan population. Our results offer practical design suggestions for future studies evaluating AMR gene transmission in animal husbandry settings. Based on our experience, we would recommend measurement of a wide range of AMR genes at several timepoints, since at baseline a significant number of AMR genes were already present in humans. Sampling from humans, livestock, as well as shared environmental samples (such as water sources) would be required to establish patterns of temporal transmission. A randomized controlled trial design for livestock exposure, as well as molecular evaluation of genetic similarity between bacterial strains harboring AMR genes will be critical for evaluating causality and directionality of transmission.

The World Health Organization Expert Guidelines Development Group tasked with addressing the worldwide crisis of increasing AMR recommend complete restriction of all classes of medically important antibiotics in food-producing animals for growth promotion [40]. Although this was issued as a strong recommendation, evidence to support the recommendation was deemed “low-quality” due to a lack of supportive studies. Here, we describe changes in the AMR gene profile in stool of humans over two timepoints, and highlight the prevalence of AMR genes in both humans and livestock in a rural Ugandan population. In future studies, to confirm the suspected epidemiologic links that may be responsible for the results in this pilot study, genotyping methods to define mobile elements and strain-specific analysis of AMR genes found in humans exposed to antibiotic-treated livestock are needed. A randomized trial design where simultaneous acquisition of human, livestock and environmental samples with a high frequency of sampling may be useful to define susceptibility factors for acquisition of AMR genes.

Study design overview.

(JPG)

Click here for additional data file.

Cycle threshold (CT) values of AMR detection from human and chicken stool samples and controls used in this study.

PPC = Positive PCR Control.

(XLSX)

Click here for additional data file.

(DOCX)

Click here for additional data file.

11 Mar 2020

PONE-D-20-03262

Gut carriage of antimicrobial resistance genes in women exposed to small-scale poultry farms in rural Uganda: a feasibility study

PLOS ONE

Dear Dr. Lai,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

==============================

Please go through the reviewer comments and revise the manuscript indicating how each point raised by the reviewers has been addressed.

==============================

We would appreciate receiving your revised manuscript by Apr 25 2020 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Iddya Karunasagar

Academic Editor

PLOS ONE

Journal Requirements:

1.  When submitting your revision, we need you to address these additional requirements.

Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Additional Editor Comments (if provided):

The reviewers have raised major concerns in the manuscript which needs to be addressed through a major revision addressing all concerns point by point.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Weil et al., provide a manuscript detailing the feasibility of conducting a delayed intervention trial with the objective of measuring the patterns of fecal AMR genes in chicken farmers before and after poultry introduction to a “farm” or household. To my knowledge, the application of a delayed intervention RCT design for a longitudinal study of antimicrobial resistance maker gene prevalence in humans and livestock is novel. This feasibility study identifies many factors that should be considered in design and implementation of such a study. Many of the critical factors identified are omitted in this study, and some additional factors in data analysis are not mentioned. To my knowledge, using a Jaccard dissimilarity index is a novel approach to estimating shared antimicrobial resistance markers in populations or groups.

Major concerns:

1. The objective or hypothesis of the study is unclear. Please revise the sentence at the end of the introduction section to provide a more comprehensive description of study goals or objective(s) or any specific hypothesis.

2. The unbalanced design by not sampling the chickens at the time of introduction in both the intervention and control group, and not sampling humans at the same time as chickens is a major limitation and should motivate a more conservative interpretation of the results. While you recognize this limitation in the discussion, in my opinion it seems biologically unjustified to test associations between the resistance genes found in the human groups and the chickens provided to the control group 6 months after the human samples were collected. As you note in the introduction temporal aspects and history of exposure are critical. Further, the study design seems to assume a single chicken sample, taken 6 months after the human one year samples, is a realistic representation of the chicken samples provided to the intervention group. Previous animal (and human) studies have shown resistance markers change over time and especially in relation to antibiotic exposure (e.g. tetracycline exposure in the brooders) and life cycle or management changes.

3. In my opinion, there are too many places in the manuscript where you imply possible chicken to human transmission, including in the approach to analysis, the presentation of results, the interpretation/discussion, and the conclusions. Is this an objective of the study? (e.g. lines 48, 172, 254) Please refine the description of the study objectives and in doing so justify the approach. Specifically comment on why you did not collect base-line and one year samples from the poultry prior to and after introduction to the households in both groups, and why you did not collect 18 month samples in humans. Please refine the description in the methods describing the duration of contact between humans and poultry at the 18 month time point prior to chicken sample collection.

4. An important aspect of the feasibility analysis would be an analysis of the variability in gene marker occurrence in the control and intervention groups over time and between subjects within groups. Please explore this variability using the data available and describe how such variability would influence future study design (e.g. power calculations to explore effect size estimates). While you have commented on the small sample size limitation in the discussion, you did not specifically comment on the infrequent repeated measures and evaluating variability in outcomes over time and power needed to discriminate between noise (random variation) and true trends over time. In my opinion, 2 time points are insufficient to make inferences regarding trends over time. Longer term surveys of antimicrobial resistance trends demonstrate sample frequency influences interpretation.

5. Because it appears you did not account for within subject repeated measures, nor did you appear to account for clustering (the hierarchical nature of subjects and chickens within households and within villages), in the discussion, please comment on: 1) longitudinal studies that include repeated measures within subjects should include assessments of correlation within subjects and models assessing associations should account for repeated measures and autocorrelation, 2) the potential effect of clustering within households and villages (e.g. humans in the same villages (or connected households within villages) may have greater resistance gene similarity compared to humans from other villages, independent of the chicken exposure treatment). For example, inclusion criteria for this study was the household must have at least 1 child under the age of 5. What if the household also had school-aged children who attend school? This is just one obvious example (of many) that creates a possible contact network between households (not to mention within village shared water sources, which you mentioned, or shared latrines or shared foods or shared food production environments).

6. Building on the previous comment, how do you control for the possibility of community acquired resistance genes in this study? Please comment on the limitations in this study design? Does this preliminary work suggest it would be possible to replicate the study across multiple villages? A challenge with longitudinal infectious disease studies is when intervention and control groups co-mingle and you cannot control for the possible contact and transmission between the intervention and control groups. This issue should be addressed in additional analyses of the data and in the discussion; for example the human baseline resistance gene profiles in the control group (or possibly a pooled baseline of the time zero control and intervention groups) could be used as the centroid to test the similarity between intervention or control subjects at one year or chickens at 18 months compared to controls (or pooled data) at baseline. I would recommend conducting a similarity analysis to first demonstrate that the baselines in the 2 human groups are not different (not sure how one might do tis in the absence of replication), before pooling the baseline data for subsequent analysis. Given that 10 of the 32 markers were found in only chickens and 13 of the 32 markers were found in both chickens and humans, and the majority of the markers 22 were found in humans, I suspect there may be greater similarity in markers found in the one year samples compare to base-line humans than in markers found in chickens compared to base-line humans. And there may be an issue with the uneven frequency of sampling (i.e. lower overall sample numbers in the chickens results in a under-estimate of the true frequency of resistance markers in this population.)

7. Methods – please add more details on the methods of recruitment of the women in this study as a subset of the original RCT. Was this a convenience sample subset of the original eligible subjects? Please add details as to whether stool sample collection was included in the parent RTC plan, or if the sample collection and outcome measures in this feasibility study were a post-hoc additional to the original study.

8. Please add some discussion (at least) and possibly some analysis of the Jaccard dissimilarity index approach to estimating shared antimicrobial resistance markers in populations or groups. Are there other studies that have applied this approach? Are there any studies that have compared using this summary index in a regression model with a more traditional approach using a binary outcome such as presence or absence of a gene marker or class of genes with corrections for multiple comparisons?

9. Are chickens ever free-ranging during the course of the study? How long was it until slaughter for the chickens provided to the intervention group? Were slaughtered chickens sold in the village? How frequently were intervention household chickens purchased by (or shared with) control households? My experience with poultry introduction programs in rural communities which provided coup construction in addition to animals, is that the poultry are often raised extensively, (i.e. allowed to forage outside the coup areas) for significant periods of time during the day or their lifetime, and may be returned to owner’s coups at night. Chickens will then forage in cattle (and other livestock) dung and the broad environment (including human waste and especially any human food waste such as vegetable scraps which may have fecal/manure contaminants and be a source of resistance genes). Thus the chicken diet and potential exposure is not well controlled in many rural systems, and chickens may simply be a sentinel for common environmental exposure risk. Were these variables considered and controlled for in this study? If appropriate this issues should be addressed in the discussion section.

10. In my opinion, the authors over emphasize speculative transmission (based on similarity) of resistance markers between humans and chickens, when alternative transmission dynamics, independent of poultry contact, such as human to human or human to environment to human, can explain the observed data and are not explored in the analysis – if the intervention subjects and the control subjects are considered to belong to the same population (i.e. from the same village) then at base-line almost all of the observed genes present in the human population are also found in either the intervention or control group at 12 months. From table 2 and table S1, it appears 4 (12%) of 32 (did I count correctly?) total genes were found in either the intervention or control groups at 1 year [aaC2, AAC(6)-Ib-cr, CFE-1 and LAT] but not at baseline, and none of these markers were found in chickens. It appears that the majority of genes found in chickens at 18 months were pre-existing in humans at baseline (i.e. before introduction of chickens for this trial), except for 10 genes (31% of all genes), [vanB, vanC, msrA, erm A, ermC, SHV(238S240E), SHV(238S240K), OXA-10 group, and OXA-58 group, QnrB-8 group], which were only found in chickens. There are many examples where a marker was found in one or both groups at one year but was not observed in chickens [AAC(6)-Ib-cr, aacC2, CTX-M-9 Group, SHV(156D), ACC-3, ACT 5/7 group, CFE-1, LAT, QnrB-1 group ] including at least 4 elements found in the control group prior to direct chicken contact. Further there are many examples were a resistance marker existed in baseline samples in one group and emerged in the other group at 12 months [CTX-M-1 Group existed in intervention at base then found in control at 1 year; SHV, SHV(238G240E), ACT 5/7 group, QnrB-5 group, QnrS in controls at base and found in intervention at 1 year]. Overall these findings suggest to me there is more similarity between human samples in the 2 groups at baseline and one year, than there is between human and poultry samples. While you discussed this concepts, with some specific examples in the discussion section. These associations should be explored more thoroughly.

Specific comments:

Line 34 – eliminate “growth promoters” – all antimicrobial use in livestock production likely contributes to emergence and selection of AMR genes in livestock and possibly subsequently to humans either through direct exposure to animals and their housing environments or through indirect exposure through the food chain (i.e. contaminated food products harvested from animals).

Line 34 and line 77 – delete “major” – this is subjective; I think many believe, and some research data would suggest, the major contributor to acquisition of AMR genes in humans is antimicrobial use in humans. See for example the systematic review and meta-analysis by Tang et al. (Lancet Planet Health 2017;

http://dx.doi.org/10.1016/S2542-5196(17)30141-9).

Line 37 – delete “from”

Line 47 – revise order, this is confusing as it does not represent what I interpret is the actual study design and sampling protocol… it might read “After one year of exposure to chickens, seven new AMR genes were detected in the intervention group at one year, while six new AMR genes were detected in the control group at one year and 6 months prior to when this group received chickens.”

Line 49 – revise “Women who had direct contact with the chickens sampled in the study had greater similarities in AMR resistance gene patterns to chickens than those who did not have direct contact with chickens sampled.” This sentence is misleading. In your model using the similarity outcome, at base-line (i.e. before any contact) and at one year (after contact), women with direct contact with chickens during the year (i.e. the intervention group) actually had lower similarity to chicken samples (taken at 18 months) than the control group (who had no direct household chicken contact at either baseline or one year).

Line 54 – revise - there were not chicken samples at both baseline and follow up time-points. The chicken samples are temporally distant from the human samples and it is not clear the length of contact time between humans and chickens before the chicken samples were taken at 18 months .

Line 68 – “The problem is accentuated…” The information in this sentence seems logical, but without data, is speculative – please revise this sentence. Please clarify, is this a hypothesis statement for future studies or can you provide a reference supporting this statement with empirical data?

Line 93 – please end the introduction with a sentence describing the actual objective(s) (or hypothesis?) of the study – was the objective simply to determine the feasibility of conducting a longitudinal study of AMR gene patterns in chicken farmers and their poultry in a delayed intervention RCT? Or was the objective to test a specific hypothesis? From the current sentence, it seems the objective was to determine patterns of AMR genes in stool of chicken farmers before and after poultry introduction, but the methods appear to go beyond this descriptive objective to test associations between resistant elements found in poultry and their caregivers. Yet, as discussed under general concerns, the study design for this test of association is incomplete and the statistical approach ignores the hierarchical nature (clustering) and repeated measures (correlation) in the sample populations.

Line 107 - What was the duration of the brooding period, and more specifically what was the duration of feeding tetracycline-containing dietary supplements? Were these supplements provided in water or in solid feed? What was the target dose of tetracycline as fed per kilogram animal unit?

Line 110 – how are the survey data being used in this study? In particular the records of antimicrobial use for the women study participants and for treatment of all livestock on the farms. Was this a new survey developed only for this feasibility study, or were data extracted from a larger survey developed for the original larger RTC. Please provide a copy of the survey tool in the supplemental data, or provide a link to a permanent archive for the survey tool.

Line 113 – modify figure S1 to indicate the actual number of participants sampled in the 2 arms of the this feasibility study

Line 115 – so at 18 months, when the control group receives their first batch of brooding chicks (which were receiving medicated feed or feed supplements at the time of sampling?) you take chicken fecal samples. This seems to be an apples and oranges comparison group, because you have longitudinal samples from the humans and a single point in time sample from the chicks during a time when the chicks appear to be receiving tetracycline. Also the chicks are introduced to the control households and chicken samples are taken 6 months after the last human samples with no description of the duration of direct contact between chickens and humans in the control group, prior to sampling the chickens. How long were the chicks exposed to the humans and the farm environment before they were sampled? What relevance does the antibiotic resistance gene screening in the poultry at 18 months have toward the study objective? This might be the place in methods section to provide more detail that informs these concerns.

Line 154 – what is meant by environmental contaminants? These samples appear to be controlling for contaminants introduced within the laboratory setting? There would be other more appropriate controls for contaminants introduced during sampling in the farm or household environments (i.e. sham samples collected on farm). Perhaps revise the wording to be more specific.

Line 173 – revise sentence, there are no “baseline” chicken samples

Line 254 – again “both at baseline and follow-up” is modifying chicken samples, when there was only one time point for chicken samples

Line 291 – add something to the effect of, “and people who did not care for chickens during the 12 months of human sampling (control group) harbored many of the same genes at one year.”

Line 294 – revise – this wording is confusing, the humans who had direct contact with chickens at one year in the study when samples were taken was the intervention group, correct? So the control group had no direct contact with chickens at one year, yet their samples were more similar to chicken samples collected at 18 months. Was this expected? What are possible explanations for this finding?

Line 295 – delete or revise “AMR gene patterns in both groups became more similar to AMR…” the use of the word “became” suggests a temporal association that could be perceived to imply chicken gene patterns are the driver of human gene patterns. I suggest, this sentence could also be written as chicken gene patterns at 18 months were more similar to human patterns at one year than to human patterns at base-line. Either delete or revise after analyzing the similarity of one year in both human groups to human base-line. Highlighting this comparison demonstrates the limitation of not having longitudinal chicken samples.

Line 303 – revise, change to “to humans or animals” – there are examples where humans serve as a reservoir of resistant organisms that appear to be transmitted to animals, further antibiotic resistance is also detrimental to animal health

Line 308 – delete or revise – “Additionally, we observed a high rate of AMR genes overall in humans and in chickens in rural Uganda” what is the comparison group – please provide a reference - higher than what other group? Urban humans and chickens in Uganda? Rural humans and chickens in other countries?

Line 366 – please add another limitation; the number of replicates over time (in both humans and chickens) are limited and do not represent a diversity of timing relative to chicken production cycle and antibiotic use in that cycle.

Line 386 – “significant changes” – how do you define “significant” here? This conclusion can be kept if the recommended analysis comparing 1 year human to base-line human demonstrates a significant change in prevalence (frequency) – otherwise change wording to “we identified differences in the 2 time point analyzed” Of course you also identified many (more?) similarities between the 2 time points. And also there were some subjects where markers were found at baseline and not at one year. Highlight the challenges of modeling the dynamics at the level of subjects within treatment group.

Figure 2 – legend – source chicken crtl should be 18 months

In the reference section, organism genus species names should be written in italics

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

23 Apr 2020

Please see uploaded word document with overall changes in the revised manuscript and point by point response.

Submitted filename: R2R.docx

Click here for additional data file.

20 May 2020

Gut carriage of antimicrobial resistance genes in women exposed to small-scale poultry farms in rural Uganda: a feasibility study

PONE-D-20-03262R1

Dear Dr. Lai,

We are pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it complies with all outstanding technical requirements.

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/, click the "Update My Information" link at the top of the page, and update your user information. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

With kind regards,

Iddya Karunasagar

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

All comments addressed satisfactorily

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Drs. Weil, Lai and other study authors,

Thank you for the attention to detail in addressing my suggestions and concerns in the revised manuscript. I have no additional requests for revisions, only minor comments for your consideration. I truly enjoyed reviewing your manuscript and I am glad you believe my comments were constructive. These additional four comments reference the line numbers in the revised document without the tracked changes.

Line 67 – consider adding an additional factor, "insufficient sanitation and hygiene infrastructure for managing human and animal waste."

Line 87 - either here, or later in the discussion, you might consider recognizing one of the classic (seminal?) studies conducted by Stewart Levy

Changes in intestinal flora of farm personnel after introduction of a tetracycline-supplemented feed on a farm. Levy SB FitzGerald GB Macone AB. N Engl J Med 1976 09 Sep;295(11):583-8 PMID:950974

Emergence of antibiotic-resistant bacteria in the intestinal flora of farm inhabitants. Levy SB. J Infect Dis 1978 May;137(5):689-90 PMID:351088

Line 278 – I have always been taught to avoid the concept of a “trend toward a difference” based on a marginal p value, but I suppose this is personal preference and I am not going to be pedantic and push for a change in the language, as I believe it is reasonable in a pilot or feasibility study an association at P<0.1 might be worth recognizing as of interest for future study

Line 298 – typo “uwed”

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

1 Jun 2020

PONE-D-20-03262R1

Gut carriage of antimicrobial resistance genes in women exposed to small-scale poultry farms in rural Uganda: a feasibility study

Dear Dr. Lai:

I am pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

For any other questions or concerns, please email plosone@plos.org.

Thank you for submitting your work to PLOS ONE.

With kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Iddya Karunasagar

Academic Editor

PLOS ONE