Antimicrobials use and resistance on integrated poultry-fish farming systems in the Ayeyarwady Delta of Myanmar

Antimicrobials are used to support livestock health and productivity, but might pose a risk for the development of antimicrobial resistance; in particular, when multiple livestock species are raised together in production systems. On integrated chicken-fish farms, chickens are raised over fish ponds and poultry faeces is excreted into the ponds. We investigated antimicrobial usage and the antimicrobial susceptibility of Escherichia coli cultured from poultry faeces on 301 integrated farms in Ayeyarwady Delta of Myanmar. Antimicrobials were used by 92.4% of farmers for chickens, but they were not applied to fish. The most common antimicrobials used were Octamix (amoxicillin and colistin sulfate) on 28.4%, enrofloxacin on 21.0% and amoxicillin on 16% of farms. Overall, 83.1% (152/183) of the E. coli were resistant to at least one antimicrobial. The highest level of resistance was to amoxicillin (54.6%), tetracycline (39.9%), sulfamethoxazole/trimethoprim (35.5%) and enrofloxacin (34.4%). Multidrug resistance was identified in 42.4% of isolates. In general, we found similar levels of antimicrobial resistance in non-users of antimicrobials as in users of antimicrobials for more commonly applied antimicrobials. Overall, antimicrobial resistance was lower in chickens on these integrated farms in Myanmar, compared to poultry farms in other countries of South East and East Asia.

Antimicrobial use in integrated poultry production. On broiler farms no antimicrobials were used on 10.8% (N = 24) farms, 33.2% (N = 74) used one antimicrobial and 56.0% (N = 125) used two or more antimicrobials. On layer farms no antimicrobials were used on 4.2% (N = 3) farms, 11.1% (N = 8) used one antimicrobial and 84.7% (N = 61) used two or more antimicrobials. Antimicrobials were used by all farms raising both broilers and layers (N = 6). Antimicrobials were more frequently used on layer compared to broiler farms (p < 0.001). All antimicrobials were administered in water. The most common antimicrobials used were Octamix (amoxicillin and colistin sulfate) in 42.2% and 44.4% of broiler and layer flocks, respectively; enrofloxacin in 27.4% and 50.0%, and amoxicillin in 18.4% and 43.1%. Amoxicillin, enrofloxacin and tylosin were used more frequently (P < 0.05) in layer flocks compared to broilers flocks (Table 1), while some antimicrobials (neomycin, doxycycline, oxytetraclycine, lincomycin) were only used in broiler, but not in layer flocks.
Ocatmix, amoxicillin and tylosin were the antimicrobials most commonly used for treatments, while also being used prophylactically by other farmers. Streptomycin and sulfadiazine were predominately used for treatments only. Of all the farmers that provided octamix (42.7%, N = 126), about 43.7% (N = 55) administered it daily (Supplementary Table S1 online). In 99.3% of cases antimicrobials were obtained by integrated chicken-fish farmers from feed shops, and only on 0.7% of cases antimicrobials were supplied to farmers through veterinarians.
Antimicrobial susceptibility testing. Escherichia coli was cultured from the faecal samples in 183 of 301 (60.8%) farms. Overall, 83.1% (152/183) of the E. coli were resistant to at least one antimicrobial ( Table 2). The highest level of resistance was to ampicillin (54.6%), tetracycline (39.9%), sulfamethoxazole/trimethoprim (35.5%) and enrofloxacin (34.4%). Lower levels of resistance were identified to gentamicin (13.1%), cephalothin (10.2%), neomycin (6.6%), colistin (6.0%), amoxicillin clavulanic acid (3.8%) and cefoxitin (2.2%). No isolates were resistant to chloramphenicol. Multidrug resistance was identified in 65 isolates (42.4%). Many antimicrobial resistance patterns were detected (Supplementary Table S2 online). www.nature.com/scientificreports/ Relationship between antimicrobial use and antimicrobial resistance. For 172 integrated chickenfish farms the relationship between antimicrobial usage and antimicrobial resistance was explored. Focusing on antimicrobials with less than 95% susceptibility, we explored the relationship between usage and non-usage of antimicrobials in poultry on the integrated chicken-fish farms and the relevant resistance patterns for these antimicrobials. For more commonly applied antimicrobials (usage on more than 5% of farms), a similar resistance pattern was observed for users and non-users of this specific antimicrobial ( Fig. 1), for example, amoxicillin resistance was similar for non-users and users of Octamix. For less commonly applied antimicrobials (usage in less than 5% of farms), more variability in antimicrobial resistance was observed between users and non-users. About 16.4% of broiler and 10.2% of layer on the integrated chicken-fish farms showed no antimicrobial resistance, while resistance to one group or one antimicrobial category 6 was 26.2% and 12.2% and for two or more antimicrobial categories was 57.4% and 77.6% for integrated broiler and layer farms, respectively (Fig. 2). Resistance to multiple antimicrobial categories was more prevalent on integrated layer compared to broiler farms (p = 0.048).
However, the correlation of the number of antimicrobials used on integrated chicken-fish farms did not show a linear relationship with the number of antimicrobial resistance groups found on these farms, thus a sequential larger number of antimicrobials used on integrated chicken-fish farms was not necessarily associated with a proportional increased resistance to the number of antimicrobial categories ( Supplementary Fig. S1 online).

Discussion
Antimicrobial resistance was found on integrated chicken-fish farms in Myanmar, but in general, we found lower levels of antimicrobial resistance in E. coli from healthy chickens compared with other countries in Southeast Asia 1,7-9 and China 10,11 . In this study, the highest level of resistance was detected to older generation antimicrobials; ampicillin (54.6%), tetracycline (39.9%), and sulfamethoxazole/trimethoprim (35.5%).
Of most concern is the identification of resistance to the highest priority critically important antimicrobials, fluoroquinolones (e.g. enrofloxacin) (34.4%) and colistin (6%). These drugs are recognized as critically important for human therapy 12 and the World Health Organisation recommends these drugs should not be used for treatment of food-producing animals" 13 . Fluoroquinolone resistance in this study was lower than that detected in faecal E. coli from chickens in Vietnam (59.6-70.2%) 8 and Bangladesh (83-100%) 9 and similar to a study from Indonesia (35.9%) 7 and across studies from Southeast Asia > 20 to 50% 1 . Resistance to colistin in Southeast Asia in previous studies was reported as > 20 to 50% 1 .
Resistance to critically important third generation cephalosporins was low, cefoxitin (2.2%) similar to previous studies in Southeast Asia where levels < 10% were reported 1,8 . No third generation cephalosporins were reported as administered to chickens in this study. The macrolide, tylosin was also administered commonly to poultry in this study. Macrolides are generally administered to prevent or treat chronic respiratory disease (CRD) associated with Mycoplasma gallisepticum. Macrolides are also identified as highest priority critically important to human health 12 . Macrolides are generally not used to treat colibacillosis in poultry and no clinical breakpoints are available, therefore the faecal E. coli were not screened for susceptibility to macrolides. Future studies could involve culturing a gram-positive indicator organism e.g. Enterococcus spp. and screening to detect macrolide resistance.
A recent study on salmonella from poultry meat, from the Yangon Region Myanmar identified higher levels of resistance to sulfamethoxazole/trimethoprim (70.3%), tetracycline (54.3%), chloramphenicol (29.7%), amoxicillin clavulanic acid (17.4%) and multidrug resistance (52.2%) and lower levels of resistance to ampicillin (47.1%), ciprofloxacin (9.4%) and gentamicin (8%) than the current study 14 . Other studies in Myanmar have focused on MDR bacteria, extended spectrum beta-lactamase and carbapenemase producing Enterobacteriaceae from human clinical isolates [15][16][17] , or isolates cultured from the environmental 16 or foods 18 . www.nature.com/scientificreports/ In our study we did not test for carbapenems, however, we had no isolates resistant to ceftiofur, and only four isolates resistant to cefoxitin, three of which were MDR and MDR was identified in 42.4% of isolates.
Interestingly, we were able to show that for frequently used antimicrobials, similar antimicrobial resistance patterns were observed, even on those farms that did not use these antimicrobials. This suggests that a possible spread of antimicrobial resistance genes between farms for commonly used antimicrobials, including to farms that did not use specific antimicrobials. Our field data support and highlight the concerning relationship that antimicrobial resistance is a global public health concern as it did not stop at farm boundaries involving even farms that abstain of using specific antimicrobials.
However, our study also was not able to identify a clear linear correlation between antimicrobial use and resistance. A study investigating development of antimicrobial resistance in bacteria at integrated pig-fish farms in Vietnam described an increase in prevalence of naladixic acid and enrofloxacin resistance in faecal E.coli with enrofloxacin in feed, however the association between development of resistance and provision of feed containing enrofloxacin was not clear 19 . Tetracycline resistance was high prior to the study, and no association with provision of tetracycline in feed and increase in tetracycline resistance was identified 19 . However, the antimicrobial relationship between usage and resistance and spread of antimicrobial resistance genes is complex and is difficult to describe with field data, therefore modelling approaches and molecular studies are needed to illustrate the development of antimicrobial resistance over time.
We only used E. coli as an indicator pathogen for chicken faeces and did not sample the water in the ponds. Two types of indicator organisms for antimicrobial resistance in aquatic environments were used in another studies: Acinetobacter spp. and Enterococcus spp. 4 . In future studies E. coli and Enterococcus spp. could be used as indicator organisms in poultry and pigs while Acinetobacter spp. and Enterococcus spp. could be used for detection of resistance in the aquatic environment.
Overall, antimicrobial resistance was lower in broilers and layers on the studied integrated chicken-fish farms in Myanmar, compared to poultry farms in other countries of South East and East Asia. Although all famers used antimicrobials for poultry, none of the farmers used them for fish. Thus, chicken-fish farming in the study region of Myanmar seems to be a well-placed with respect to antimicrobial resistance, but adequate training of farmers is required to reduce the development and dissemination of antimicrobial resistance.

Materials and methods
Study design. A cross-sectional study was used to interview integrated chicken-fish farmers in the Ayeyarwady Delta (AD) of Myanmar. As no 'lists' or directories of integrated chicken-fish farmers existed, integrated chicken-fish farmers had to be located through participatory approaches. Thus, fishery officers from the Myanmar Livestock and Fisheries Department and veterinary officers from the Myanmar Livestock, Breeding and Veterinary Department 'identified' integrated chicken-fish farmers in selected townships by discussing with individual chicken or fish farmers where integrated chicken-fish farms were located. Integrated chicken-fish farmers were then visited and invited to participate in the interviews. Additional chicken-fish farmers were identified by snowball sampling, i.e. interviewed chicken-fish farmers provided information on other integrated chicken-fish farmers in the same township.
As the proportion of farms where E. coli is resistant to at least one antimicrobial was unknown, we assumed a prevalence of resistance 0.5, which would generate the largest sample size. A sample size of 302 was required to estimate a confidence interval up to 0.10 wide if the observed proportion is between 0.3 and 0.7 (Confidence level = 95%, Acceptable difference = 0.05, Assumed proportion = 0.5, Size of population = 1200). Sample size calculations were conducted in WinPepi version 11.65 20 .
Questionnaire and data collection. A questionnaire was developed in English and digitized using the mobile data collection platform CommCare (Dimagi, Inc, https ://www.commc arehq .org/home/). CommCare is an open source platform that allows the construction of specific mobile apps tailored for collection data. Digital tablets and smartphones were used to record data on the administration of antimicrobials (type, amount, type of administration, source), to collect spatial information on the farm location and to take photographs of labels of antimicrobials and supplements used by the farmers. Questionnaire data were collected by two veterinarians (who are co-authors of this paper) that were enrolled in master's program at the University of Veterinary Science, Yezin, Myanmar at the time of the interviews.
Human Ethics Approval was obtained for this survey from the UQ Research and Innovation Ethical Review Committee (#2016001804). All research was performed in accordance with relevant guidelines/regulations of the Australian National Health and Medical Research Council Act. Informed consent was obtained from all  Antimicrobial susceptibility testing. Antimicrobial susceptibility testing was performed by disc diffusion as per Clinical and Laboratory Standards Institute guidelines (CLSI) for 12 antimicrobials of veterinary and/ or human health importance: amoxicillin/clavulanic acid (30 μg); ampicillin (10 μg); cefoxitin (30 μg); ceftiofur (30 μg); cephalothin (30 μg); chloramphenicol (30 μg); colisitn (10 μg); enrofloxacin (5 μg); gentamicin (10 μg); neomycin (30 μg); sulfamethoxazole/trimethoprim (1.25/23.75 μg) and tetracycline (30 μg) 21 between March 2017 and January 2018. All antimicrobial discs were obtained from Oxoid. The quality control organism used was E. coli ATCC 25922. Only enrofloxacin has a veterinary specific breakpoint for poultry 21 , therefore, human breakpoints were used for other antimicrobials 22 . For neomycin breakpoint information was obtained directly from the manufacturer (Zoetis, West Ryde, NSW, 2114, Australia). Isolates were considered multidrug resistant (MDR) if they acquired non-susceptibility to at least one agent in three or more antimicrobial categories 6 .
Data analysis. Frequency and contingency tables and descriptive statistics were produced focusing on the usage of antimicrobials. The relationship between antimicrobial usage and resistance was visualized using a range of graphical displays and compared between integrated layer and broiler farms using the Fisher's Exact test. Data analysis was conducted using Stata 14

Data availability
The dataset analysed in this study is available from the https ://doi.org/ 10.6084/m9.figshare.11971995.