Background

AMR is a global public health threat associated with morbidity, rehospitalization, longer hospital stays, and mortality [1, 2]. In sub-Saharan Africa (SSA), antibiotic-resistant pathogens are major drivers of morbidity and mortality in children under five years of age, fuelled by inappropriate antibiotic use and poor sanitation [3] leading to the selection and spread of antibiotic-resistant bacteria [4]. Estimates by the 2019 report on the Global Burden of Bacterial AMR identified SSA as the highest contributor to the global AMR burden [5].

Commensal resident gut bacteria such as E. coli and Klebsiella pneumoniae (K. pneumoniae) play a critical role in AMR as they act as reservoirs for the carriage of AMR determinants [6, 7]. These enteric bacteria can become pathogenic or may transfer AMR genes to other pathogenic Enterobacterales, such as Salmonella and Shigella [8]. The acquisition of AMR and virulence factors by commensal E. coli and Klebsiella spp is mediated by mobile genetic elements, such as plasmids and transposons, via horizontal gene transfer [9]. Antibiotic resistance in commensal enteric bacteria, such as K. pneumoniae and E. coli, has been reported in 85–90% of World Health Organization (WHO) member state regions [10].

Fluoroquinolones such as ciprofloxacin (CIP) are effective broad-spectrum antibiotics used for the treatment of bacterial infections making them a recommended choice of therapy for enteric infections such as salmonellosis and shigellosis [11]. The emergence of fluoroquinolone resistance has reduced therapeutic options, especially for Enterobacterales infections [12]. Fluoroquinolone resistance is mediated by two mechanisms: chromosomal mutations in DNA gyrase and topoisomerase IV enzymes and plasmid-mediated quinolone resistance (PMQR). Mutations in fluoroquinolone binding sites during DNA replication mediate high-level fluoroquinolone resistance[13]. Mechanisms of PMQR genes include protection of DNA gyrase and topoisomerase IV from quinolone activity mediated by qnr genes: qnrA, qnrB, qnrC, qnrD, and qnrVC [14]. The aminoglycoside-modifying enzyme encoded by aac(6’)-Ib-cr is involved in the acetylation of fluoroquinolones leading to reduced susceptibility to CIP and norfloxacin [14]. The final mechanism is enhanced efflux pump activity mediated by quinolone efflux pump (qepA) and oqxAB associated with reduced susceptibility to fluoroquinolone and increased ESBLs [15].

There is a paucity of data available on fluoroquinolone resistance and PMQR determinants in commensal bacteria, especially in children under five years in Kenya. In this study, we sought to determine the prevalence and the distribution of PMQR determinants mediating CIP non-susceptibility in E. coli and Klebsiella spp isolates from children being discharged from the hospital. In addition, we identified correlates of carriage of CIP non-susceptible isolates. This in-depth analysis will help inform the burden of CIP resistance carriage in children under five years being discharged from hospitals in SSA.

Results

Participant population and baseline characteristics

Six hundred and fifty-one isolates (406 E. coli and 245 Klebsiella spp) were isolated from 568 children and subjected to AST (Fig. 1). Among the 568 children, 343 (60.4%) were discharged from Kisii Teaching and Referral Hospital (KTRH) and 225 (39.6%) were discharged from Homabay County Referral Hospital (HCRH), 348 (61.3%) were less than two years of age and 230 (40.5%) were female (Table 1). Prior to discharge, the median duration of hospitalization was 3 days [interquartile range (IQR) 2,6 days] with 203 (35.7%) children being hospitalized for ≥ 4 days. The majority of the children (97.7%) had their human immunodeficiency virus (HIV) status determined with 12 (2.1%) being HIV-infected while 70 (12.3%) were HIV-exposed. There were more HIV-exposed children in HCRH 51 (22.7%) compared to KTRH 19, (5.5%). Common diagnoses at discharge were pneumonia 164 (28.9%), malaria 155 (27.3%), and diarrhoea 108 (19%) (Table 1). The majority of the children 502 (88.4%) had an antibiotic prescribed during their hospitalization with penicillins 359 (63.2%) being the most prescribed antibiotic followed by gentamicin 316 (55.6%) and ceftriaxone 187 (32.9%). Fluoroquinolones were rarely administered during hospitalization with CIP being prescribed to only 2 (0.9%) children in HCRH and none among children from KTRH. Children at KTRH 328 (95.6%) compared to HCRH 174, (77.3%) were more likely to be prescribed an antibiotic. Nearly half of the children, 266 (46.8%) had CIP non-susceptible isolates (46.3% E. coli and 37.6% Klebsiella spp). Of the 83 children who had both E. coli and Klebsiella spp isolated, 14 (2.46%) had CIP non-susceptibility in both isolates.

Fig. 1
figure 1

Flowchart of participants isolates. AMR = Antimicrobial resitanace; AST = Antimicrobial susceptiability testing; CIP = Ciprofloxacin; E. coli = Escherichia coli; n = Number

Full size image
Table 1 Participant baseline characteristics
Full size table

Correlates for carriage of fluoroquinolone non-susceptible E. coli or Klebsiella at hospital discharge

Children who received an antibiotic during hospitalization were 69% more likely to have a CIP non-susceptible E. coli isolate (PR 1.69, [95%CI = 1.09, 2.63], p = 0.01) and over two times more likely to have a CIP non-susceptible Klebsiella spp isolate (PR 2.61, [95%CI = 1.05, 6.53], p = 0.01). The presence of ESBL carriage was also associated with the presence of either a CIP non-susceptible E. coli or Klebsiella spp isolate. Length of hospital stay was associated with CIP non-susceptible E. coli and children with hospitalizations extending more than 4 days were nearly 40% more likely to have CIP non-susceptible E. coli (PR 1.38 [95%CI 1.07, 1.78] p = 0.01). The use of either CIP or ceftriaxone were equally associated with CIP non-susceptible Klebsiella or E. coli. Children hospitalized for diarrhoea were 27% less likely to have a CIP non-susceptible E. coli compared to those who did not present with diarrhoea (PR 0.73 [95%CI 0.53, 1.0] p = 0.03). Similar magnitudes of association for hospital length and diarrhoea diagnosis were observed in Klebsiella isolates but were not statistically significant (Table S1).

Distribution of CIP non-susceptible isolates

Among the 266 children with CIP non-susceptible isolates, we isolated 280 CIP non-susceptible isolates: 188 and 92 E. coli and Klebsiella spp, respectively. Of the 266 children, 14 had both a CIP non-susceptible E. coli and Klebsiella spp isolated. Among the 92 Klebsiella spp, 86 were K. pneumoniae and 6 were Klebsiella oxytoca (K. oxytoca) (Fig. 1).

CIP MICs values for the 280 CIP non-susceptible isolates ranged between 0.25 – 32 µg/mL with CIP MIC50 (µg/mL) of 32 µg/mL. Of the 280 isolates, 214 (76.4%) were resistant while 61 (21.8%) were intermediate. Among the resistant isolates, high-level CIP resistance (MIC ≥ 32 µg/mL) was common in almost half of the isolates (46.4%) most commonly among E. coli 102 (54.3%) and slightly less in Klebsiella spp (28/92, 30.4%). The CIP MIC distribution, MIC50 and MIC90 among the isolates are shown in Table 2.

Table 2 Distribution of MIC (µg/mL) per organism
Full size table

Distribution of PMQR determinants

Six different PMQR determinants: qnr (qnrB, and qnrS), enzyme modifying aac(6′)-Ib-cr, and efflux pumps (qepA, oqxA, and oqxB) were detected (Table 3). At least one of the PMQR genes was detected in nearly all (224/280, 80%) of the screened isolates. Most E. coli and Klebsiella spp (40%) isolates had at least one qnr determinant detected. Of the qnr genes, qnrB was the most commonly detected qnr gene (20/188, 11%) and 47/92(51%) in E. coli and Klebsiella spp, respectively. In E. coli, qnrS was the most detected qnr gene (27/188, 14%). The qnrA gene was not detected in any of the E. coli or Klebsiella spp isolates.

Table 3 Distribution of PMQR determinants per organism
Full size table

The most predominant PMQR gene was aac(6’)-lb (167/280, 59%) identified in more than half of all CIP non-susceptible isolates. Klebsiella spp had more aac(6’)-lb positive isolates with (78/89, 85%) Klebsiella compared to (89/188, 47%) of E. coli isolates. All isolates carrying the aac(6’)-lb gene were positive for the cr variant. The qepA gene was only detected in (16/188, 9%) E. coli CIP non-susceptible isolates. The oqxAB complex was the most dominant efflux pump detected (61/92, 66.3%) in Klebsiella, however, none were detected in E. coli. Only (2/188, 1%) E. coli isolates had the oqxA gene, however, both of these isolates lacked the oqxB gene therefore none of the E. coli isolates carried the oqxAB complex.

QepA sequence analysis

DNA sequencing of the qepA gene from 16 E. coli isolates revealed amino acid substitutions at codons 95 and 134. Double amino acid substitutions F95L and V134I were common in 9/16 (56.3%) E. coli isolates. Six qepA positive E. coli isolates carried had no amino acid substitution while one isolate had only V134I amino acid substitution (Table 4). Assigned accession numbers for the qepA gene from 8 representative isolates submitted to GenBank are as follows: OP918677, and OQ031499-OQ031505.

Table 4 MICs and amino acid changes in qepA E. coli isolates
Full size table

Distribution of co-carriage of PMQR determinants per organism

A total of 225/280 (80%) isolates had at least one PMQR gene including; all 92 Klebsiella spp and most E. coli 133/189 (70.37%). Interestingly, qnrB and qnrS co-occurred in two K. pneumoniae (0.71%) isolates. Co-carriage of qnrB with acc (6’) lb-cr was present in 12/188 (6.4%) E. coli and aac(6′)-Ib-cr with oqxAB detected in 47/92 (51%) Klebsiella spp were the most prevalent combination of PMQR gene combinations. E. coli isolates had three notable combinations of different PMQR determinants with qnr combination being predominant in the co-existence of genes. On the other hand, Klebsiella spp had as many as nine different combinations and similarly, qnr gene combinations were predominant in the different combinations of determinants. The most common co-carriage in both bacterial species was qnrB with acc (6’) lb-cr found in 56/280 (20%). Additionally, acc(6’)-lb-cr co-existed with a majority of PMQR genes in both E. coli and Klebsiella spp isolates (Table 5).

Table 5 Co-carriage of PMQR determinants per organism
Full size table

Discussion

Fluoroquinolone resistance in Enterobacterales in children under five years recently discharged from the hospital is of great public health concern due to the risk of transmission of these bacteria to the community, and treatment failure, which may require re-hospitalization during the post-hospital discharge period. This study sought to determine the prevalence and correlates of CIP non-susceptibility, and distribution of PMQR genes in E. coli and Klebsiella spp, isolated from children under five years recently discharged from hospitals. We observed a high level of CIP resistance among children being discharged from two hospitals in western Kenya and multiple fluoroquinolone resistance genes in Klebsiella spp and E. coli. Our findings show high levels of MICs to CIP in a majority of the CIP non-susceptible isolates, which is disturbing due to the relationship between increasing MICs leading to fluoroquinolone non-susceptibility and fluoroquinolone treatment failure [17]. This is particularly important given that fluoroquinolones are recommended therapies for the treatment of enteric infections such as shigellosis and salmonellosis [18, 19].

CIP non-susceptibility was detected in commensal E. coli (46%) and Klebsiella spp (38%) isolated from children being discharged from hospital and there was concomitant high carriage of PMQR genes (80%) among the isolates. This is despite less than 1% of the hospitalized children receiving fluoroquinolone antibiotics during hospitalization. The high fluoroquinolone non-susceptibility observed could be attributed to co-selection pressure mediated by non-fluoroquinolone antibiotics especially cephalosporins such as ceftriaxone, facilitating the selection and carriage of PMQR genes as previously reported [20]. This situation is further exacerbated by resistance pressure mediated by ESBL production which is evident in this study; the presence of ESBL production is highly associated with CIP non-susceptibility carriage [21]. Almost all children in the study population received an antibiotic during their in-patient stay with penicillin, gentamicin, and ceftriaxone being the most prescribed antibiotics consistent with other findings in Kenya [22]. Antimicrobial usage has been associated with selective pressure for AMR in gut bacteria [23]; our findings show that there is a strong correlation between antibiotic use and the carriage of CIP non-susceptible bacteria. Notably, children presenting with diarrhoea were less likely to carry a CIP non-susceptible E. coli. Children with diarrhoea may be less likely to be treated with an antibiotic [24], treated with fewer antibiotics, and/or treated with shorter courses of antibiotics compared to sepsis, malnutrition and pneumonia. Despite being the guideline-indicated antibiotic, CIP is rarely used for diarrhoea in this setting due to lack of availability and cost. Another reason could be during diarrhoeal episodes, there is gut dysbiosis that alters the gut composition [25] including resistant bacteria colonizing the gut. Thereby children presenting with diarrhoea would have a lower chance of carrying CIP non-susceptible E. coli compared to children without diarrhoea who have had the gut intact with colonized resistant bacteria.

PMQR genes facilitate low-level fluoroquinolone resistance, however, they select for higher-level resistance mediated by mutations on genes encoding gyrase and topoisomerase enzymes [13]. In this study, we detected six (qnrB, qnrS, aac(6’)lb-cr, qepA, oqxA, and oqxB) PMQR genes mediating fluoroquinolone resistance in both E. coli and Klebsiella spp. One of the six PMQR genes (qepA) had not previously been detected in clinical isolates in Kenya. The qepA gene has been reported in very few studies within the SSA region: Chad, Malawi, Egypt, Sierra Leon, and Nigeria [26,27,28,29,30]. Other PMQR determinants that have been identified in Kenya are aac(6’)-lb-cr, qnrB, qnrS in E. coli [31], qnrS, and oqxAB in K.pneumoniae [32, 33]. Intestinal carriage of PMQR genes in these bacteria has been reported in several studies in SSA and globally [34,35,36]. This is particularly worrying due to their potential transfer of these genetic determinants through mobile genetic elements to pathogenic bacterial species, thereby mediating the transmission of resistant bacteria that may result in treatment failure.

The aac(6′)-Ib-cr gene was the most commonly detected PMQR determinant. Most of the Klebsiella spp (84.78%) harboured the gene aminoglycoside modifying enzyme while almost half of the E. coli isolates (90/188, 48%) harbored the gene which is consistent with findings from previous studies [37]. The aminoglycoside acetyltransferase enzymes have not only been associated with reduced susceptibility to fluoroquinolones but also to aminoglycosides, thus limiting effective antibiotic treatment [38]. Qnr genes were the second most widely detected PMQR determinants associated with resistance to fluoroquinolones with prevalence rates of 40%. The prevalence of qnr genes (40%) was higher compared to previous studies from Kenya which reported (2%) and (8.4%) [33, 39]. The qepA gene, one of the most recently identified PMQR determinants, has been associated with decreased susceptibility to fluoroquinolones and increased MIC levels [13]. This determinant was detected in (16/188, 8.5%) E. coli isolates which was slightly lower compared to previous studies in Nigeria (18.5%) and Sierra Leon (23%) [28, 30]. DNA sequencing confirmed the existence of the qepA gene among E. coli isolates in Kenya with F95L and V134I amino acid substitution consistent with amino acid substitution reported in the qepA4 allele [40]. To our knowledge, this is the first report of the detection of qepA in Kenya; this is worrisome for public health and calls for more active fluoroquinolone resistance surveillance.

Co-carriage of PMQR plays a critical role in multidrug resistance as it influences increased MICs leading to decreased susceptibility to fluoroquinolone antibiotics that may lead to treatment failure. We observed high multiple co-carriage in both E. coli and Klebsiella fluoroquinolone non-susceptible isolates with co-existence of aac(6’)lb-cr and qnrB, or qnrS genes, being the most predominant co-carriage in both bacteria. The prevalence of co-carriage between qnrB and aac(6’) lb-cr (20%) was found to be higher compared to findings from previous studies in SSA [41]. We observed co-carriage of qnrB with qnrS in two K. pneumoniae isolates, a phenomenon that has previously been reported in Klebsiella spp, however, its prevalence in this study was much lower compared than 18.75% reported in Togo [21]. This co-carriage could be attributed to multiple plasmids carrying the different qnr genes within the same genetic environment as has been previously demonstrated [42]. Notably, 12 isolates with the rare qepA co-existed with other PMQR genes (qnrS or aac(6’)lb-cr) which was consistent with findings from other previous studies [43]. Interestingly, one K. pneumoniae isolate co-harboured all determinants detected in this study except qnrA and qepA. This is concerning as the co-existence of multiple PMQR genes has been linked to resistance to multiple antibiotic classes due to the carriage of multiple plasmids carrying resistance determinants to other classes of antibiotics.

This is one of the few studies that has characterized AMR determinants in children post-hospital discharge in SSA settings, including screening for a wide range of PMQR genetic determinants, highlighting the greater diversity and distribution of fluoroquinolone resistance genes. In addition, the focus on commensal E. coli and Klebsiella spp, two commonly isolated Enterobacterales associated with the carriage of AMR determinants as indicator organisms for AMR carriage rather than pathogenic bacteria was important due to their ability to transfer AMR genetic elements.

This study had some limitations. CIP non-susceptibility was determined in a subset of isolates from children after discharge, which means there could be more non-susceptible isolates that were not screened. Only two children in this study received CIP or rather a fluoroquinolone during admission, this may not be sufficient to clearly show the role of fluoroquinolone resistance in poor patient outcomes during the post-discharge period. Limiting the analysis to PCR detection only, other mechanisms mediating resistance such as point mutations which could be detected by comprehensive whole-genome sequencing analyses were not captured. Being a cross-sectional study at the point of hospital discharge, we were unable to determine whether AMR was acquired at the community or nosocomially. Having isolates at admission, the point at which treatment decisions are also made would be able to answer this question of the timing of AMR acquisition.

Conclusion

This study detected multiple PMQR genes and the first report of the qepA gene among CIP non-susceptible clinical E. coli and Klebsiella spp. The study observed high levels of CIP non-susceptibility and fluoroquinolone resistance carriage which could form a reservoir for the community spread of resistance, thus posing a great challenge in the effective treatment during hospital stays and subsequently during the post-hospital discharge period. We recommend enhanced surveillance for fluoroquinolone resistance carriage which will be vital to inform interventions to control antimicrobial-resistant bacteria and antimicrobial stewardship in rural and peri-urban populations.

Materials and methods

Study design

This was a cross-sectional nested study from the Toto Bora trial [44] that utilized E. coli and Klebsiella spp isolates recovered from fecal samples of children under five years discharged from two hospitals in western Kenya. Children being discharged from KTRH and HCRH aged between 1 -59 months were recruited in the parent study [44] to assess the effects of Azithromycin on mortality and rehospitalization in children under five years. The nested study used clinical, sociodemographic, and health history information collected during physical examination from children enrolled in the parent trial or interviews with their caregivers at hospital discharge. Faecal or rectal swab samples were collected before Azithromycin administration was done. The swabs were cultured, isolates recovered and biochemically identified as previously described [44].

Parent trial

Bacterial isolation, identification, and AST

After laboratory culture, E. coli and Klebsiella spp isolates were identified and AST was performed by disc diffusion as previously described [45]. Briefly, a rectal swab or whole stool was collected from the enrolled child and was inoculated on MacConkey Agar and incubated at 37 °C for 24 h within 24 h of specimen collection time. Mucoid, lactose, or non-lactose fermenting colonies suspected to be E. coli or Klebsiella spp were isolated and the API 20E (bioMérieux, Inc, Durham, NC, United States) system confirmed the species of bacteria. A total of 568 children were randomly selected in the parent study, from whom 406 E. coli and 245 Klebsiella spp isolates were recovered and selected to have AST performed. The isolates were subjected to AST by disc diffusion to 5 μg of CIP (Oxoid, Hampshire, England) representing the fluoroquinolone antibiotic class. The isolates were also screened for ESBL production by the combined disc diffusion test and interpreted using the criteria from the Clinical and Laboratory Standards Institute (CLSI) [16]. E. coli ATCC 25922 and E. coli NCTC 13351 were used as negative and positive controls respectively for ESBL screening. E. coli or Klebsiella spp isolates with intermediate (22-25 mm) or resistant (≤ 21 mm) phenotype zone size interpretations for CIP were considered CIP non-susceptible [16].

Nested study

CIP MIC determination by E-test

In the nested study, MICs for CIP were determined by the E-test method on isolates that were CIP non-susceptible by disc diffusion per CLSI guidelines [16]. Bacterial colonies were suspended in 0.85% normal saline to a turbidity equivalent to 0.5 McFarland standard (bioMérieux, Inc, Durham, NC, United States). The bacterial suspension was inoculated on Mueller Hinton agar (Oxoid, Hants, United Kingdom) plates, and the CIP E-strips were placed at the center of the agar followed by incubation at 35 °C for 16 -18 h. Concentration ranges for MICs for E-test strips for CIP (0.002–32 µg/mL) (bioMérieux Marcy l’Etoile, France) and non-susceptibility interpreted according to 2021 CLSI guidelines [16]. MIC results were classified as follows: susceptible (≤ 0.25 µg/mL), intermediate (0.5 µg/mL), or resistant (≥ 1 µg/mL). E. coli ATCC 25922 was used as quality control for determining MICs by the E-test method.

DNA extraction and PMQR characterization

Genomic DNA was extracted using the boiling preparation method [46]. Extracted DNA was subjected to a series of single and multiplex PCR reactions to identify PMQR determinants: qnrA, qnrB, qnrS, aac(6')-Ib, qepA, oqxA, and oqxB. PCR reactions were performed using previously described primers and PCR conditions [47,48,49] (Table S2). All isolates positive for the aac (6’)-lb gene were further analyzed to determine carriage of the (-cr) variant associated with CIP resistance [48]. The PCR products and known positive strains were digested with the restriction enzyme BstCI (New England Biolabs, Ipswich, MA) to identify aac (6’)-lb-cr which lacks the BstCI restriction site present in CIP susceptible isolates as previously described [48]. The positive controls used in screening for PMQR genes were in-house isolates with confirmed target genes by whole genome sequencing and sequence analysis [32]. The identity of the amplified qepA gene was confirmed by amplicon sequencing of both the forward and reverse strands. PCR products positive for the qepA gene were purified using DNA Clean & Concentrator™-25 Kit (Zymo Research, Orange, CA, USA) and sequenced by the Sanger sequencing method using an ABI 3730 DNA analyzer. The consensus nucleotide sequences were analyzed and compared to available sequences deposited in the GenBank at National Center for Biotechnology Information (NCBI) using the Basic Local Alignment Search Tool (BLAST) program (http://blast.ncbi.nlm.nih.gov). The sequences of the qepA gene were submitted to GenBank.

Statistical analysis

Fluoroquinolone resistance was defined by combining resistant and intermediate interpretative breakpoints for CIP. Risk factors for fluoroquinolone resistance previously associated with AMR from published data were chosen to test for association [50]. Patient characteristics including age, sex, hospital site, duration of hospitalization, antibiotic use at admission, HIV status, diagnosis at admission, and ESBL carriage were assessed. Poisson regression was used to determine prevalence ratios (PRs) and associated 95% confidence intervals (CIs) while Chi-square test was used to determine p-values. Associations were considered statistically significant at an alpha of 0.05. Analysis was performed in R software version 4.1.3.