Prevalence of plasmid-mediated AmpC beta-lactamases in Enterobacteria isolated from urban and rural folks in Uganda

Background: AmpC beta-lactamase-producing bacteria are associated with increased resistance to third-generation cephalosporins. Here, we describe plasmid-mediated AmpC beta-lactamase-producing enterobacteria isolated from urban and rural dwellers in Uganda. Methods: Stool and urine from 1,448 individuals attending outpatient clinics in Kampala and two rural districts in central Uganda were processed for isolation of Escherichia coli and Klebsiella. Following antibiotic susceptibility testing, cefoxitin resistant isolates, and amoxicillin/clavulanate resistant but cefoxitin susceptible isolates, were tested for AmpC beta-lactamase production using the cefoxitin-cloxacillin double-disc synergy test. Carriage of plasmid-mediated AmpC beta-lactamase-encoding genes (pAmpC) and extended spectrum beta-lactamase (ESBL) encoding genes was determined by PCR. Results: Nine hundred and thirty E. coli and 55 Klebsiella were recovered from the cultured samples, yielding 985 isolates investigated (one per participant). One hundred and twenty-nine isolates (13.1%, 129/985) were AmpC beta-lactamase producers, of which 111 were molecularly characterized for pAmpC and ESBL gene carriage. pAmpC genes were detected in 60% (67/111) of the AmpC beta-lactamase producers; pAmpC genes were also detected in 18 AmpC beta-lactamase non-producers and in 13 isolates with reduced susceptibility to third-generation cephalosporins, yielding a total of 98 isolates that carried pAmpC genes. Overall, the prevalence of pAmpC genes in cefoxitin resistant and/or amoxicillin/clavulanate resistant E. coli and Klebsiella was 59% (93/157) and 26.1% (5/23), respectively. The overall prevalence of pAmpC-positive enterobacteria was 10% (98/985); 16.4% (45/274) in Kampala, 6.2% (25/406) Kayunga, and 9.2% (28/305) Mpigi. Ciprofloxacin use was associated with carriage of pAmpC-positive bacteria while residing in a rural district was associated with protection from carriage of pAmpC-positive bacteria. Conclusion: pAmpC beta-lactamase producing enterobacteria are prevalent in urban and rural dwellers in Uganda; therefore, cefoxitn should be considered during routine susceptibility testing in this setting.


Introduction
Enterobacteriaceae is a family of Gram-negative bacteria that inhabit the mammalian gut and includes the leading causes of community-and hospital-acquired infections 1 . Enterobacteriaceae have become increasingly resistant to antibiotics, especially beta (β)-lactam agents, the mainstay of treatment for infections caused by them. One of the main mechanisms underlying resistance to β-lactam antibiotics among the enterobacteriaceae are the AmpC β-lactamases. These enzymes are clinically relevant as they confer resistance to most β-lactam antibiotics, except the fourth-generation cephalosporins and carbapenems [2][3][4] .
AmpC β-lactamases are chromosomally encoded in most species of the enterobacteriaceae, particularly Citrobacter freundii, Enterobacter, Morganella morganii, Hafnia alvei, Aeromonas and Serratia spp. 3 . However, Escherichia coli and other enterobacteria, notably Proteus mirabilis, Salmonella and Klebsiella spp., can acquire plasmid-encoded AmpC β-lactamases (pAmpC), which are highly transferable between species. Note that Proteus mirabilis, Salmonella and Klebsiella spp. lack chromosomallyencoded AmpC enzymes, while a chromosomally-encoded AmpC β-lactamase occurs in E. coli but it is expressed at low basal levels due to presence of a weak promoter and attenuator, which makes E. coli susceptible to cephamycins (e.g. cefoxitin, cefotetan) 2,5,6 . Acquisition of pAmpC β-lactamases by species like E. coli and Klebsiella pneumoniae (K. pneumoniae) is worrying as it enables an efficient spread of extended resistance in bacteria and ultimately their spread in the community 3 . Furthermore, the widespread use of cephamycins and β-lactamase inhibitor combinations (e.g. clavulanic acid/amoxicillin and tazobactam/piperacillin) has contributed to selection of pAmpC β-lactamase producing strains worldwide 7,8 .
Once they have colonized the gut, pAmpC β-lactamase producing strains may initiate an infection at various anatomical sites 9 . In low-income countries where such infections are empirically treated with third-generation cephalosporins, failure to detect AmpC β-lactamase-related resistance may lead to treatment failure. Moreover, the cut-off / break points for the disc diffusion / minimum inhibitory concentrations (MICs) may not detect AmpC β-lactamase production and resistance to third-generation cephalosporins. Carbapenems are currently the only effective drugs against infections caused by AmpC β-lactamase producing bacteria, since these bacteria tend to be multidrug resistant (MDR) 10 . However, carbapenamase-producing enterobacteria are now common in low-income countries including Uganda 11 .
Although pAmpC β-lactamase producing bacteria have been reported throughout the world 3,7,12,13 , there is little information about them in Uganda, especially their frequency among enterobacteria. Therefore, the aim of this study was to estimate the prevalence of pAmpC β-lactamase producing bacteria among Enterobacteriaceae isolated from individuals attending outpatient clinics in Kampala city and two rural districts in central Uganda. We show that pAmpC β-lactamase producing bacteria are prevalent in enterobacteriaceae isolated from urban and rural dwellers in Uganda, implying that ceftriaxone, an antibiotic commonly used to treat systemic infections in Uganda, could be associated with treatment failure.

Study setting
This cross-sectional study was conducted on flora from stool and urine of clients attending outpatients clinics in Kampala and two rural districts (Kayunga and Mpigi) in central Uganda 14 . The study sites were purposively selected with an assumption that urban areas are associated with high bacterial carriage and exposure to antibiotics compared to rural areas 15 . Kampala (urban) and Mpigi districts were assumed to have a wet-tropical climate, while Kayunga has a wet-dry tropical climate 16 .

Sample size estimation and sampling
The sample size for each study subsite was proportional to the contribution of the facility to the total outpatient clinic attendance in the months of April, May and June of 2006. All individuals attending the clinics were eligible to participate. By then, there was no data on antibiotic resistance among E. coli and/or K. pneumoniae isolates in well-defined community infections in Uganda. As such, the sample size was estimated based on an observed prevalence of 19.6% for K. pneumoniae carriage in clinical samples (urine, etc.) at Makerere University's Clinical Microbiology Laboratory (unpublished observations). In each of the three districts, multistage sampling was done based on the average clinic attendance for the district. Thirty clusters of 16-20 participants were selected from each district using probability proportion to size sampling. Two busy days of a week were purposively chosen to visit a selected health care facility. When the number of participants exceeded 20, systematic sampling was done. A standardized interviewer-administered questionnaire was used to collect clinical and demographic data. Participants were instructed to provide stool or urine (if unable to provide stool) in a sterile screw-cap container. Samples from the rural districts were stored at 4°C for up to 24 hours prior to transportation, while those from Kampala were immediately transported to the laboratory at Makerere University College of Health Sciences for culturing.
Culturing and identification of E. coli and K. pneumoniae The procedure for culturing and isolate identification was described previously 14 . Briefly, samples were streaked on MacConkey agar medium on the third/fourth quadrant, and incubated at 37°C for 18-24 hours in ambient air. In case of stool, samples were first emulsified in sterile normal saline before inoculation onto MacConkey agar plates. Lactose fermenting isolates with colony morphology suggestive of E. coli and Klebsiella spp. were subjected to oxidase testing and when negative, they were cultured for 18-24 hours on triple sugar and iron (TSI) agar, Simmons citrate agar, urea and Sulphide Indole Motility (SIM) medium for identification. Inconclusive isolates were confirmed as E. coli or Klebsiella by using the API 20E system (BioMerieux Marcy 1'Etoile, France).
Testing for AmpC β-lactamase production Typically, AmpC β-lactamases confer resistance to cephamycins (e.g. cefoxitin), a characteristic widely used to distinguish them from the extended-spectrum β-lactamases (ESBLs), and to functionally screen for AmpC β-lactamase producing isolates 3,4,18-21 . As such, all cefoxitin resistant isolates in this study were screened by the cefoxitin/cloxacillin double-disc synergy test (CC-DDST) to detect AmpC β-lactamase production as previously described 22 . As AmpC β-lactamases are associated with clavulanate resistance 17,23 , isolates with reduced susceptibility to amoxicillin/clavulanate were also tested for AmpC β-lactamase production. Susceptibility to cefoxitin and to third-generation cephalosporins was determined based on the CLSI guidelines (2007) 24 . Furthermore, isolates that were positive on the CC-DDST (i.e. AmpC β-lactamase producers) were re-tested with E-test strips containing cefotetan and cefotetan/ cloxacillin (CN/CNI, AB BIODISC, Solna, Sweden). E-test screening was considered positive for AmpC β-lactamase production when MIC ratio for cefotetan / cefotetan/cloxacillin was ≥8 22 . Testing for ESBL production was carried out as described previously 25 , on isolates with reduced zone diameters to third-generation cephalosporins. Briefly, isolates with zone diameters of 22 mm, 25 mm, 27 mm and 27 mm for discs of ceftazidime, ceftriaxone, cefotaxime and aztreonam respectively, were considered suspect for ESBL-production, which was subsequently confirmed by the double disc synergy test (DDST). Detection of ESBLs was The DDST for detection of ESBLs was performed by using amoxicillin/clavulanate disc in the center, and discs of ceftazidime, ceftriaxone, aztreonam, and cefotaxime placed 15-20 mm center-to-center from the amoxicillin-clavulanate disc. Extension of the zone of inhibition towards the clavulanate disc was indicative of ESBL-production.
Screening for ESBL and pAmpC β-lactamase genes All isolates testing positive for AmpC β-lactamase production on the CC-DDST were tested by polymerase chain reaction (PCR) for pAmpC gene carriage. Further, as pAmpC β-lactamase genes have also been detected in isolates with reduced susceptibility to third-generation cephalosporins, we tested isolates with inhibition zone diameters of ≤27 mm, ≤25 mm and ≤22 mm for cefotaxime, ceftriaxone and ceftazidime, respectively, for pAmpC gene carriage 17,23 . In-house multiplex PCRs targeting AmpC β-lactamase genes bla CIT , bla DHA , bla MOX , bla FOX , bla EBC , bla ACC and bla CMY-2 were performed using published primers, Thermo-Fisher Taq DNA polymerase and PCR master-mixes, and conditions 6,26 . Amplification was performed in a 3Prime Mid-size thermocycler (Techne, UK) and the expected amplicon sizes were successfully generated. PCR amplification of ESBL genes bla CTX-M , bla TEM and bla SHV was performed with Taq DNA polymerase (Thermo-Fisher Inc.) using published primers 27 . Amplicons were sequenced (ACGT, Wheeling, IL, USA) by the chain termination method (Sanger sequencing) and sequences confirmed through BLAST-searching at NCBI. Phylogenetic group typing of E. coli was done according to the method of Clermont et al, in which PCR of a combination of two genes (chuA & yjaA) and an anonymous DNA fragment are used to classify strains 28 .

Data analysis
The data were double entered for validation using EPIDATA software version 3.1, cleaned and exported to STATA (v14) for analysis. Data were compared across the districts using descriptive statistics, frequencies and bivariate analyses (cross-tabulations). Associations between outcome variables, i.e. isolates with ESBL/pAmpC genes, and categorical independent variables, i.e. socio-demographics, use of antibiotics, history of hospital admission and medical procedures three months prior to visits, were tested using Pearson's Chi-square. A significant level was set at p<0.05. Similarly, odds ratios (ORs) between the categorical independent variables and outcome variables were determined. Variables with p<0.2 at bivariate analysis were entered into multivariate logistic regression models with backward elimination. Independent variables used were gender (male vs. female), health center level, health sub-district and district, history of admission, history of medical procedures and antibiotic use recalled by client and from health record during the previous three months. To control for the effect of clustering, regression with robust standard errors was used.

Ethical statement
The study protocol and consent procedure were reviewed and approved by the Research Ethics Committee and the Higher Degrees committee of Makerere University Medical School (IRB #-2006-009) and the Uganda National Council for Science and Technology (HS246). All adult participants and guardians gave written informed consent before participation. The consent process included storage and use of the collected stool and urine samples for further studies. We obtained assent from participants below the age of 18 years in addition to informed consent from their parent/guardians/caregivers.  Table S2 (see Extended data 29 ). None of the urine samples grew bacteria at ≥10 4 colony forming units (CFU) per milliliter implying that there was no infection-related growth. Overall, 37% (535/1,448) of the participants visited outpatient clinics for general conditions and bacteria grew in 68% (363/535) of these participants. Of the 731 participants who presented with infectious conditions, 67% (488/731) had bacterial growth in their samples. Of the 122 participants who visited HIV/AIDS clinics for routine checks, 72% (88/122) had growth (Tables S1 and S2, Extended data 29 ). Furthermore, 1,093 participants reported to have taken antibiotics three months prior to the visit, of whom 69% (755/1,093) had growth. Of the 125 participants who reported to have been previously admitted to hospitals, 62% (78/125) had growth. Of the 130 participants who reported to have undergone medical procedures three months prior to the clinic visit, 55% (71/130) had growth (Tables S1 and S2, Extended data 29 ).
Factors associated with carriage of pAmpC β-lactamase producing bacteria There was significant association between health center level, district of residence, use of ciprofloxacin and sample type with carriage of pAmpC β-lactamase producing bacteria, Table 3. After adjusting for each of them, the district of residence remained an independent risk factor for carriage of pAmpC β-lactamase producing bacteria, Table 3. Overall, we found that residing in a rural district (Kayunga/Mpigi) was associated with low carriage of pAmpC β-lactamase producing bacteria (aOR 0.23 (95% CI:0.11, 0.47) and aOR 0.49 (95% CI:0.25, 0.99), respectively). Similarly, participants who were 45 years and above carried less pAmpC positive bacteria (aOR 0.17; 95% CI:0.05, 0.62). Ciprofloxacin use was an independent risk factor for carriage of pAmpC positive bacteria (aOR 2.61; 95% CI:1.28, 5.32), Table 3.

Discussion
AmpC β-lactamases are clinically important in that community-acquired infections arising from strains producing these enzymes may not respond to empiric treatment with common antibiotics. The cefoxitin/cloxacillin double-disc synergy screening of cefoxitin and amoxicillin/clavulanate resistant isolates for AmpC β-lactamase production is a simple and efficient method of quickly detecting these resistance mechanisms in isolates 21,22 . Using this approach, the prevalence of AmpC β-lactamase producing bacteria in this study (13.2%), and the prevalence of pAmpC β-lactamase gene carriage (26-59%) was high but comparable to the rate of 36.5% at Mbarara Regional Referral Hospital in South-western Uganda 20 . The prevalence in our study is higher than rates from other East African settings 31-33 ; however, the study populations were varied, making direct comparison difficult. Furthermore, in this study, co-carriage rates of ESBL-and pAmpC genes was high, which is a cause for concern as individuals carrying strains producing these enzymes could be reservoirs of spread for MDR bacteria 34,35 . A significant number of isolates that were susceptible to third-generation cephalosporins but cefoxitin resistant carried pAmpC genes, a discordance that has been reported before 3,36 . The overall carriage rate (10%) for pAmpC genes in enterobacteriaceae in this study reflects extensive use of antibiotics, as found in Libya where a fecal carriage rate of 6.7% for pAmpC β-lactamase producing bacteria in the community was reported 37 . Importantly, ceftriaxone has been used in Uganda for the last two decades, mainly in empirical treatment of systemic bacterial infections and currently its prescription among in-patients is higher than that of other antibiotics 38 . Such extensive use of ceftriaxone could be a driving force behind the high pAmpC gene carriage rates. Coexistence between bla EBC and bla CIT genes in isolates has been reported before in Africa and Asia 36,39 . bla CIT reported in this study comprised of the bla CMY-2 and bla CMY-4 genes. In Africa, bla CMY-4 was first reported in North Africa. The detection of bla DHA and bla EBC genes is of concern as bla ACT-1 , a prototype gene for the bla EBC and bla DHA genes, is linked to a functional ampR regulator and is inducible 40,41 . In Seoul, Korea, four of the five deaths from bloodstream infections due to bla DHA producing K. pneumoniae were associated with treatment with extended spectrum cephalosporins 42 .
In this study, about half of the isolates exhibiting reduced susceptibility to third-generation cephalosporins carried pAmpC β-lactamase genes, which contrasted findings from Northern Europe where 100% of isolates exhibiting reduced susceptibility to third-generation cephalosporins carried pAmpC genes 23 . Furthermore, 73% of cefoxitin resistant and pAmpC positive isolates were susceptible to third-generation cephalosporins, in contrast with 26% (10/38) reported from Switzerland, for pAmpC This project contains the following underlying data: -Participants_recruited_from_outpatient_clinics.xls (participant data underlying Table S1) -Participants_whose_samples_had_growth.xls (participant data underlying Table S2) -Susceptibility profile of PampC-isolates-09092020-1.xls (antibiotic susceptibility data underlying Table S3) - producing isolates that were susceptible to third-generation cephalosporins 43 . Overall, findings in this study suggest that antibiotic susceptibility testing of enterobacteria in Uganda may yield false results for third-generation cephalosporins e.g. ceftriaxone, cefotaxime and ceftazidime. Given that bacterial isolates are not routinely tested for AmpC β-lactamase production, region specific protocols guided by surveillance data are necessary.
In E. coli, phylogenetic group analysis has been used to differentiate virulent/extra-intestinal strains, which predominantly belong to phylogenetic groups B2 and D, from commensal strains that belong to groups A and B1 28 . In this study, the predominance of groups B2 and D (n=60) compared to groups A and B1 (n=33) in the community is cause for concern as they are associated with pathogenicity, implying that strains with potential to cause extra-intestinal disease are prevalent, supporting the notion that occurrence of pAmpC β-lactamase producing strains in the community is of public health concern 44,45 . One limitation in this study was that we were not able to genotype isolates with internationally acceptable procedures like the multilocus sequence typing (MLST) to determine the sequence types.

Conclusions
AmpC β-lactamase production and pAmpC β-lactamase encoding genes are prevalent among E. coli and K. pneumoniae isolates from urban and rural dwellers in Uganda. As pAmpC genes are easily transferrable between species and have been associated with outbreaks of community-and hospital-acquired infections, pAmpC beta-lactamase producing bacteria may represent a threat in low-income settings. There is need for testing for cefoxitin resistance during routine antibiotic susceptibility testing, especially among isolates that are resistant to amoxicillin/clavulanate, as well as isolates that are susceptible to third-generation cephalosporins.

Department of Medical Microbiology, University Malaya, Kuala Lumpur, Malaysia
The study presented a detailed analysis of the prevalence of AmpC beta-lactamase producers, pAmpC and ESBL genes in the two most common enteric bacterial species, E. coli and K. pneumoniae. The authors observed a relatively higher prevalence of AmpC producers in urban areas than in rural areas. Detailed molecular characterization was performed to identify the pAmpC and ESBL genotypes, and a great diversity of the beta-lactamase genes was observed among the strains. Moreover, the common carriage of beta-lactamase-producing and potentially virulent E. coli strains in the population indeed poses a great threat to public health, not only causing infections in susceptible hosts but also serves as a reservoir for spreading antimicrobial resistance genes among the pathogenic bacteria. This study adds knowledge to the molecular epidemiology of beta-lactamase-producing enteric bacteria, which is one of the main concerns pertaining to the rising trend of antimicrobial-resistant organisms. However, the authors should make the following changes: The sampling design and procedures, bacterial isolation, antimicrobial susceptibility testing, and phenotypic detection of AmpC and ESBL producers have already been reported in detail in the previous article published by the authors 1 . Please remove the related sections in Methods and Results to avoid duplication of data. I suggest the authors include a summary of the previous findings in the last paragraph of the Introduction since this study is considered an extension of the previously reported work.
1. Figure 2 clearly depicts the distribution of the different bla genotypes according to geographical regions. This is very interesting as there seems to be a predominance of specific bla genes in the different districts. Could the authors elaborate more on this finding? Statistical analysis could be done to determine if the correlation between the 2.

Response to comments:
Thank you for your insightful comments. We are using these to improve our manuscript. Please find below, our response to each of the raised concerns. Kindly note that we are awaiting an additional reviewer's report to upload the revised manuscript hence, the changes to the manuscript in light of your suggestions will be seen when we upload the revised manuscript.

Comment 1:
The sampling design and procedures, bacterial isolation, antimicrobial susceptibility testing, and phenotypic detection of AmpC and ESBL producers have already been reported in detail in the previous article published by the authors 1 . Please remove the related sections in Methods and Results to avoid duplication of data. I suggest the authors include a summary of the previous findings in the last paragraph of the Introduction since this study is considered an extension of the previously reported work.

Response:
We agree. We have revised the Introduction accordingly. Figure 2 clearly depicts the distribution of the different bla genotypes according to geographical regions. This is very interesting as there seems to be a predominance of specific bla genes in the different districts. Could the authors elaborate more on this finding? Statistical analysis could be done to determine if the correlation between the genotypes and the geographical regions is significant or not. Also, the figure legends should be standardized to ease viewing.

Response:
Again, thanks for your appreciation of our analysis of the bla genotypes according to geographical distribution. We repeated the statistical analysis as advised; however, the results showed no statistically significant correlation between the genotypes and the geographical regions i.e., none of the enzymes was significantly in greater proportion compared to others across the different districts (p-value = 0.208). The figure legends are now standardized and easily viewable.

Comment 3:
Ciprofloxacin use was found to be a risk factor for the carriage of pAmpC-positive strains. It would be of interest if the authors could discuss the use of fluoroquinolones in the region.

Response:
As advised, in our revised manuscript we have discussed the use of fluoroquinolones in Uganda and generally the East African region.

Minor comments:
Please standardize the nomenclature of K. pneumoniae throughout the text.