Excess body weight and age associated with the carriage of fluoroquinolone and third-generation cephalosporin resistance genes in commensal Escherichia coli from a cohort of urban Vietnamese children

Purpose. Antimicrobial-resistant bacterial infections in low- and middle-income countries (LMICs) are a well-established global health issue. We aimed to assess the prevalence of and epidemiological factors associated with the carriage of ciprofloxacin- and ceftriaxone-resistant Escherichia coli and associated resistance genes in a cohort of 498 healthy children residing in urban Vietnam. Methodology. We cultured rectal swabs onto MacConkey agar supplemented with resistant concentrations of ciprofloxacin and ceftriaxone. Additionally, we screened meta- E. coli populations by conventional PCR to detect plasmid-mediated quinolone resistance (PMQR)- and extended-spectrum b -lactamase (ESBL)-encoding genes. We measured the associations between phenotypic/genotypic resistance and demographic characteristics using logistic regression. Results/Key findings. Ciprofloxacin- and ceftriaxone-resistant E. coli were cultured from the faecal samples of 67.7 % (337/498) and 80.3% (400/498) of children, respectively. The prevalence of any associated resistance marker in the individual samples was 86.7 % (432/498) for PMQR genes and 90.6 % (451/498) for b -lactamase genes. Overweight children were significantly more likely to carry qnr genes than children with lower weight-for-height z-scores [odds ratios (OR): 1.24; 95% confidence interval (CI): 10.5 – 1.48 for each unit increase in weight for height; P =0.01]. Additionally, younger children were significantly more likely to carry ESBL CTX-M genes than older children (OR: 0.97, 95% CI: 0.94 – 0.99 for each additional year, P =0.01). Conclusion. The carriage of genotypic and phenotypic antimicrobial resistance is highly prevalent among E. coli in healthy children in the community in Vietnam. Future investigations on the carriage of antimicrobial resistant organisms in LMICs should focus on the progression of carriage from birth and structure of the microbiome in obesity.


INTRODUCTION
Antimicrobial resistance (AMR) has become an increasingly recognized global health problem. The social complexities associated with AMR are multifaceted; however, a major driver of AMR is thought to be non-prescribed use of antimicrobials in the community, which is particularly common in many low-and middle-income countries (LMICs) [1]. Multiple studies have demonstrated a recent dramatic rise in pathogenic bacteria exhibiting resistance against fluoroquinolones and third-generation cephalosporins [2][3][4], which are two of the most commonly used classes of antimicrobials globally [5]. Consistently, cephalosporins and fluoroquinolones are most consumed classes of antimicrobial classes in Vietnam, with prescribed consumption from hospitals equalling consumption from retail pharmacies [1,6].
Escherichia coli is abundant in the intestine of healthy people and can be used to assess the composition of AMR genes in the human gut [7][8][9]. E. coli are proficient vehicles for AMR gene circulation, and are commonly co-resistant to several clinically important antimicrobials, including cephalosporins, fluoroquinolones and sulfamethoxazole/trimethoprim [10,11]. Therefore, commensal E. coli are sentinel Gram-negative organisms that play a key role in the maintenance and spread of plasmid-associated AMR genes within the Enterobacteriaceae [11,12]. Specifically, E. coli are known to be a major reservoir of plasmids carrying both extended-spectrum b-lactamases (ESBLs) and plasmidmediated quinolone resistance (PMQR) genes [13,14].
The expression of ESBL genes is a key resistance mechanism against later generation beta-lactams; the circulation of genes encoding these enzymes is of critical importance in human health [13]. ESBLs, which can hydrolyze a wide variety of b-lactamases, are presently classified into several subgroups; the most frequently identified include TEM, SHV and CTX-M [15,16]. The bla CTX-M genes in particular have been extremely successful, becoming globally ubiquitous and undergoing extensive diversification (>100 variants have been identified). This important group of ESBL genes is broadly divided into five phylogroups: CTX-M-1, CTX-M-2, CTX-M-8, CTX-M-9 and CTX-M-25. The most dominant global lineage is bla CTX-M-1 , which incorporates bla CTX-M-15 , a particularly successful and persistent variant that is commonly found in Asia [15].
Fluoroquinolone resistance generally occurs as a result of mutations in the chromosomal drug target genes encoding the DNA gyrase (gyrA and gyrB) and the topoisomerase IV (parC and parD) [17]. However, fluoroquinolone resistance can also be induced and enhanced by the PMQR genes. The most common PMQR genes are the qnr family (qnrA, qnrB, qnrC, qnrD and qnrS), which act by altering the DNA gyrase and topoisomerase IV, thus reducing the interaction between drug and target [18,19]. The qnr genes have been found on a broad range of plasmid structures in various pathogens within the Enterobacteriaceae. Notably, PMQR genes frequently co-exist on the same plasmid backbones as ESBL genes, which can be co-transferred during horizontal gene exchange [18,20,21].
Several studies have focused on characterizing the carriage of AMR organisms in hospitalized patients and healthy individuals in the community in LMICs [3,11,22] However, less is known about the prevalence of AMR bacteria and the diversity of associated AMR genes in the commensal microbiota of healthy children living in an urban community. Here, we used a direct plating method on selective media to measure phenotypic and genotypic resistance against fluoroquinolones and third-generation cephalosporins among a cohort of 498 healthy children resident in Ho Chi Minh City (HCMC) [23]. Further, we aimed to identify the epidemiological risk factors associated with the carriage of phenotypic and genotypic resistance in these urban children.

Study population
All rectal swabs were collected from healthy children who were participants in a previously described prospective longitudinal study [24]. In brief, infants were recruited into a birth cohort at Hung Vuong hospital (HVH) in HCMC, the second largest obstetric hospital in south Vietnam, and followed for 12 months for febrile disease through passive surveillance [25]. A subset of these children were followed beyond 12 months and recruited into a continuous cohort for active surveillance for diarrhoeal disease. The children attended scheduled routine follow-up visits at HVH every 6 months. Information on demographics, household water sources, sanitation, health issues in the intervening 6 months between visits and food consumption habits in the 3 days prior to the visit was collected. A rectal and nasopharyngeal swab was collected at the time of the visit. Seven hundred and forty-eight children were enrolled from June 2014 to December 2014 and then followed up to December 2016. To actively monitor diarrhoeal disease episodes, fortnightly SMS messages were sent to the children's parents/ care providers. During diarrhoeal episodes nurses visited their homes to confirm the case, collect samples and provide supportive care. Cases were treated according to local standard of care and were required to attend the Hospital for Tropical Diseases (HTD) in central HCMC for medical care.
Identification of antimicrobial-resistant E. coli From March to August 2016, we collected 498 rectal swabs from healthy children (no diarrhoea or other symptoms of illness) at routine follow-up visits for the purposes of this investigation. At the routine follow-up visits, we recorded information regarding the children's general well-being and their antimicrobial practices in the previous 6 months. All rectal swabs were inoculated onto MacConkey agar to isolate suspected E. coli colonies (e.g. lactose-fermenting and non-lactose-fermenting, circular, elevated, entire margins, smooth texture, pink or colourless, dry and shiny). Meta-E. coli populations (i.e. all E. coli from the same rectal swab) were inoculated into brain heart infusion (BHI) media supplemented with 20 % glycerol and stored at À80 C until being subjected to inoculation onto MacConkey agar supplemented with either ciprofloxacin (4 mg l À1 ) or ceftriaxone (6 mg l À1 ) to select for organisms resistant to these antimicrobials. E. coli ATCC25922 [ciprofloxacin minimum inhibitory concentration (MIC)]=0.008 mg l À1 and ceftriaxone MIC=0.094 mg l À1 ) and clinical E. coli isolate 03_0334 (ciprofloxacin MIC of >256 mg l À1 and ceftriaxone MIC of >256 mg l À1 ) were used as negative and positive culture controls. All suspected E. coli colonies were confirmed using a Bruker biotyper MALDI-TOF bacterial identification system.

Detection of antimicrobial-resistant genes
The meta-E. coli populations in frozen BHI were subjected to nucleic acid extraction and multiplex PCR amplification. Approximately 200 µl of meta-E. coli in frozen BHI was removed and aliquoted into a 96-well plate and centrifuged for 10 min at 4000 r.p.m. The supernatant was removed and discarded. The pellet was resuspended in 200 µl of ultrapure water and agitated gently to ensure mixing. Nucleic acid was extracted by heating the solution to 100 C on a heating block for 10 min, followed by cooling to ambient temperature. The solution was subjected to centrifugation for 10 min at 4000 r.p.m. in a microfuge. The supernatant was removed and used as template DNA in all PCR amplifications. We conducted a series of multiple PCR amplifications aiming to detect PMQR genes (aac(6')-Ib-cr, qepA, qnrA, qnrBand qnrS), ESBL genes (bla CTX-M , bla TEM , bla SHV and bla OXA ) and AmpC lactamase genes (MOX, DHA, EBC, FOX, ACC and CIT). The primer sequences and the predicted amplification sizes are shown in Table S1 (available in the online version of this article). The amplification conditions were as previously described [12,14,26,27]; a positive PCR result was determined by the detection of a PCR amplicon of an appropriate size (using a UV transilluminator and gel documentation system) for each target in the presence of appropriate positive and negative amplification controls.

Statistical analyses
Categorical and continuous variables were described using frequencies (%) and medians (interquartile range), respectively. Bar plots were used to describe the prevalence of ciprofloxacin-and ceftriaxone-resistant E. coli and the various ESBL and PMQR genes. We used univariate and multivariable logistic regression to assess the relationship between the participants' characteristics and the presence of antimicrobial-resistant E. coli and AMR-associated genes. These relationships were described in odds ratios (OR) and their corresponding 95 % confidence intervals (CIs). The significance level for the statistical tests was 0.05. All analyses were performed in R v3.4.0 (R Foundation, Vienna, Austria) and companion packages.

Baseline characteristics
Between March and August 2016, we collected rectal swabs from 498 children without diarrhoea attending a routine health check at the designated healthcare facility. All children were from a single district (eight) in HCMC (the same study area assessed for access to pharmacies in this age group [1]) and enrolled in a longitudinal cohort study of 748 individuals. This population was considered to be generally representative of the child population in HCMC. The baseline characteristics of those sampled are shown in Table 1; 269/498 (54 %) were male, 211/498 (42.4 %) were born by caesarean section, and 64/498 (12.9 %) were preterm deliveries ( Table 1).
The median age of the children at the point of sampling was 46 months. Therefore, at the time of the health check in which they were sampled, the majority of children were attending nursery/school (67.9 %; 338/498). Only a small fraction of the children were underweight, stunted or wasted. However, more than a quarter of the children were overweight (27.5 %; 137/498), with a similar proportion (27.3 %; 136/498) having a minor non-diarrhoeal illness at the time of visit. A substantial proportion (42.8 %; 213/498) of children had received antimicrobials in the 6 months prior to sampling and 26.9 % (134/498) regularly consumed some form of probiotic as part of their diet (Table 1). A third (29.1 %; 145/498) of the children had taken oral vitamin A supplementation, provided via a government public health nutrition programme. The children had consumed a wide variety of animal products in the 3 days prior to sampling, which included pork (89. 6  Ciprofloxacin-and ceftriaxone-resistant E. coli Aiming to assess the prevalence of antimicrobial-resistant E. coli carriage in this population, we independently plated the meta-E. coli populations onto MacConkey agar supplemented with ceftriaxone and ciprofloxacin to isolate resistant organisms. We cultured E. coli colonies that were resistant to ceftriaxone [according to the Clinical and Laboratory Standards Institute (CLSI) guidelines [28]] from 80.3 % (400/498) of the rectal swabs and E. coli colonies that were resistant to ciprofloxacin (according to CLSI guidelines [28]) from 67.7 % (337/498) of the rectal swabs. Resistant E. coli colonies were cultured from at least one of the antimicrobial supplemented plates (i.e. either ciprofloxacin or ceftriaxone) in 84.7 % (422/498) of the samples; no antimicrobial-resistant E. coli colonies were isolated on either plate from the remainder of the samples (76/498; 15.3 %).
Of the 422 samples from which we isolated resistant E. coli colonies on at least 1 plate, 315 produced E. coli colonies on both the ceftriaxone-supplemented media and the ciprofloxacin-supplemented media. We aimed to identify the epidemiological characteristics associated with the detection of fluoroquinolone-and cephalosporin-resistant E. coli in the rectal swabs. However, after investigating multiple factors, including age, sex, antimicrobial usage and other household variables, we found that no factors were significantly associated with their carriage. We suspect that this result either stems from the limited sample size or indicates the ubiquitous nature of antimicrobial-resistant E. coli (Table S2).
Not all antimicrobial-resistant meta-E. coli populations were associated with a corresponding AMR gene marker. For the ciprofloxacin-resistant E. coli, 37/337 (11 %) samples did not generate an amplicon associated with a PMQR gene, which may potentially be explained by chromosomal DNA gyrase mutations. Correspondingly, from the ceftriaxoneresistant E. coli, 20/400 (5 %) samples did not produce an ESBL or AmpC amplicon. Conversely, from the 161 samples with no ciprofloxacin-resistant E. coli, 73 % (118/161) generated an amplicon for at least 1 qnr gene. Of the 98 rectal swabs containing no E. coli exhibiting phenotypic ceftriaxone resistance, a large proportion (72 %; 71/98) generated at least 1 ESBL or AmpC amplicon.

Risk factors for PMQR and b-lactamase gene carriage
We next explored potential epidemiological associations between the participants' characteristics/behaviour and the most common PMQR genes (qnr family: 83.9 %; 418/ 498) ( Table 2) and b-lactamase genes (bla CTX-M : 75.1 %; 374/498) ( Table 3). We found that the detection of qnr genes in meta-E. coli populations was significantly more common in children with higher weight-for-height z-scores (OR: 1.24, 95 % CI: 1.05-1.48, for each unit increase in weight-for-height z-score, P=0.01). However, we did not find any significant association between the age of the children when sampled and the presence of any qnr gene (P>0.05). Conversely, we did not identify a relationship between the weight-for-height z-scores and the presence of bla CTX-M genes (P>0.05). However, bla CTX-M genes were more commonly detected in the commensal E. coli in children of a younger age (OR: 0.97, 95 % CI: 0.94-0.99 for each additional year, P=0.01). Additionally, we investigated potential epidemiological associations between the participants' features and the presence of the most common genes in combination (bla CTX-M and qnr). We found a consistent relationship, in which the co-existence of these genes was significantly higher among younger children (OR: 0.95, 95 % CI: 0.9-1.00, P=0.03) (Table S3).

DISCUSSION
Commensal E. coli are important vehicles for the transmission of AMR determinants [9,11,12]. Here, we investigated both the phenotypic and genotypic characteristics of AMR in a cohort of urban children in Vietnam using direct plating and conventional PCR, an approach that is sensitive, specific, inexpensive and suitable for screening in LMICs [29,30]. Our study outlines several new insights into this potential reservoir of AMR genes and their association with demographic characteristics. We found an exceptionally high prevalence of fluoroquinolone-and third-generation cephalosporin-resistant E. coli and their associated resistance genes in young children in this location. We report that the most commonly identified PMQR genes were within the qnr family (including qnrB and qnrS) and the most commonly detected class of ESBL gene was bla CTX-M . We additionally found that the carriage of qnr genes was associated with higher weight-for-height z-scores and that younger children were more likely to carry bla CTX-M ESBL genes in combination with qnr PMQR genes.
Previous comparable studies in Vietnam demonstrated a low prevalence of quinolone-resistant commensal E. coli [29,31]. These two studies, conducted in northern Vietnam, suggested that even though the overall prevalence of AMR E. coli was low, the isolation of ciprofloxacin-resistant organisms was higher in the urban area than in the rural area. Here we show a trend towards an increasing prevalence of ciprofloxacin-resistant commensal E. coli. This trend may reflect a generic higher degree of resistance to this group of antimicrobials over time, in combination with dramatic increases in antimicrobial use within Vietnam, particularly in urban areas. Using direct plating, we show a higher prevalence of ciprofloxacin-resistant E. coli in the commensal microbiota of children in Vietnam compared to those described in Bolivia and Peru (18 % in 2002 and 33 % in 2005) [32,33]. Additionally, in comparison to a recent study conducted in HCMC among healthy adults [34], we also show a higher prevalence of ciprofloxacin-resistant E. coli (15.5 %, 16/103 E. coli isolates vs 67.7 %, 337/498 rectal swab samples). However, these differences may be derived from the method of resistance detectionthe previous study exploited routine bacterial culture and antimicrobial susceptibility testingor the population sampled.
Here, we additionally detected a substantially higher prevalence of PMQR and ESBL genes than had been observed in previous investigations in human populations. Our data illustrate a much higher prevalence of PMQR genes among the healthy paediatric population (86.7 %; 432/498) than has been described in other countries. For example, in Mexico in 2013 the prevalence of PMQR genes among Enterobacteriaceae isolates from children was 32 % [35]. A global meta-analysis estimated an ESBL positivity rate for Enterobacteriaceae among healthy communities of 14 % in 2016 [36]. The prevalence of ESBL-producing Enteobacteriaceae among the paediatric community varied according to the level of development: the figures were 2.9 % for Sweden [37], 11.6 % for Tanzania [38], 23 % for Laos [39], 24.8 % for Lebanon [40] and 59 % for Bangui (Central African Republic). Potential explanations for this discrepancy may include the use of differing methods for the detection of PMQR and ESBL determinants (i.e. PCR methods) and a temporal difference of 5 years, which may reflect a natural increase in the circulation of organisms and people carrying such AMR genes. Our data indicate that a high proportion of children in urban Vietnam carry multiple AMR genes, implying high exposure to a range of antimicrobials in this local setting [6], which is comparable to the findings of other reports in Southeast Asia [39,[41][42][43].
Notably, we found that the prevalence of the bla CTX-M gene was 75.1 % (374/498), with the majority (62.9 %; 313/498) being bla CTX-M-9 [the next most common was bla TEM (36.9 %; 184/498)]. This distribution is comparable to the patterns observed in a small sample of healthy adults residing in HCMC, where the bla CTX-M class of genes was the most dominant among ESBL-producing E. coli isolates (100 %; 10/10), followed by bla TEM (50 %; 5/10) [34]. We found a relatively consistent prevalence of qnr genes, as previously reported in this setting [12]. We further established that qnrB and qnrS were the most prevalent PMQR genes [67 % (335/498) and 61 % (306/498), respectively]. This prevalence was higher than that found in data from China and some settings in Latin American [35]. However, this disparity may again be associated with a difference in the population sampled.
We found a significant association between children carrying qnr genes and a higher weight-for-height z-score. An explanation for this association could be that overweight children are more likely to be treated with antimicrobials. In this setting, children who are overweight are more commonly from families with a higher socio-economic status, who in turn may be more likely to seek healthcare and, therefore, receive antimicrobials. Furthermore, childhood obesity is known to trigger chronic low-grade inflammation [44], which may lead to their vulnerability to gastrointestinal diseases and diarrhoea [45]. This in turn results in a higher likelihood of antimicrobial consumption prior to hospitalization in cases of diarrhoeal episodes [46]. Quantitative microbiological examination has reported a significant expansion of E. coli, coupled with the contraction of Bifidobacterium, in obese children in China when compared to age-matched controls [47]. As we screened for PMQR genes, which chiefly circulate within the Enterobacteriaceae, we speculate that there is a link between a shift in the gut microbiome during obesity and the selection of genotypic resistance. Fluoroquinolones are amongst the most commonly used antimicrobials worldwide, and their use is particularly common in Vietnam [5,48]. The association of the carriage of bla CTX-M genes and younger children may be induced by reducing antimicrobial usage with age. It is difficult to assess the nature of the selective pressure for specific AMR genes during therapeutic use, but we suggest that selection may reduce as children mature.
Our study has limitations. First, the point prevalence of phenotypic and genotypic resistance could identify non-causal associations with the variables assessed in the children. Second, we investigated associations between the carriage of various AMR genes and a wide range of heterogeneous epidemiological variables among a healthy population, which may explain the limited number of significant epidemiological associations. Furthermore, by screening for the presence of PMQR and b-lactamase genes directly from rectal swabs using conventional PCRs, we were unable to assess the direct association between resistant phenotypes and associated genotypes. Notwithstanding these limitations, the results from this, the first study of its type in Vietnam, will provide an important step towards understanding the role of the microbiome in maintaining AMR genes in the community. We conclude that there is a high prevalence of commensal E. coli in children aged under 5 years in an urban setting within Vietnam exhibiting resistance against fluoroquinolones and third-generation cephalosporins. In addition, we have determined that the prevalence of PMQR-and b-lactamase-encoding genes is extremely high among healthy children. We highlight that high weight-to-height z-scores are a major risk for carrying qnr genes and that younger children are more likely to carry CTX-M genes in combination with qnr. We advocate that future longitudinal studies in Asia assess the maturation of the AMR component of the human microbiome. These data are imperative for understating the ecology of AMR gene carriage and transmission within humans. Our work emphasizes that the role of the commensal bacteria is key for the transmission and maintenance of AMR organisms/genes, especially among children.

Funding information
This work was funded by a Sir Henry Dale Fellowship to S. B., jointly funded by the Wellcome Trust and the Royal Society (100087/Z/12/Z).