This study was conducted on 204 ESBL-producing E. coli isolates collected in 2018 from four health facilities in Ethiopia (Figure 1). These isolates were collected in line with the ongoing antimicrobial resistance surveillance which was launched in 2017 by the Ethiopian Public Health Institute under the supervision of the Federal Ministry of Health (Ibrahim et al., 2018). This national AMR surveillance is following the World Health Organization (WHO) Global Antimicrobial Resistance Surveillance System (GLASS) guideline and reports the phenotypic antimicrobial susceptibility profiles for priority AMR pathogens to the local, national and international stakeholders. All isolates were identified according to WHO guidelines operable in a basic clinical bacteriology laboratory for the identification of fermenter Gram-negative bacterial pathogens (Vandepitte et al., 2003). All isolates from the AMR sites are collected at the National Reference Laboratory (Addis Ababa) were they are stored at −80°C.

Critically, the national AMR surveillance is not tasked with characterizing the molecular mechanisms of antimicrobial resistance. Therefore, we used this substantial isolate collection as a resource to initiate a characterization of the molecular mechanisms behind ESBL production. For this study purpose, 204 non-duplicated E. coli strains resistant to at least one third generation cephalosporin were randomly selected and confirmed for ESBL production. The identity of these isolates was confirmed using a VITEK 2 compact automated ID/AST instrument according to the manufacturer’s instructions (bioMérieux, Marcy l’Etoile, France). All isolates were stored at the NRL at –80°C in trypticase soy broth containing 20% of glycerol.

The antimicrobial susceptibility tests and phenotypic detection of ESBL producing E. coli were based upon the standardized Kirby Bauer disk diffusion method on Mueller Hinton agar (Oxoid LTD, Basingstoke, and Hampshire, England) and the double disk synergy test, respectively. Each test was performed as recommended by the Clinical and Laboratory Standard Institute (CLSI) (Philadelphia, PA, United States) (Clinical and Laboratory Standards Institute, 2018). Bacterial isolates that were resistant to at least one agent in three different antimicrobial categories were recognised as MDR (Magiorakos et al., 2012). We used E. coli ATCC 25922 (ESBL negative) and K. pneumoniae subsp. pneumoniae ATCC 700603 (ESBL positive) as reference strains (Microbiologics Inc., Saint Cloud, MN, United States).

Sigma-Aldrich Co (Dorset, England) synthesized all oligonucleotides.

The entire strain collection was screened for the presence of bla_TEM, bla_SHV, bla_OXA, and bla_CTX–M by multiplex PCR using primer combinations as previously described (Dallenne et al., 2010). Carbapenem resistant isolates were screened using multiplex PCR for the presence of bla_{OXA48–like,} bla_IMP, bla_VIM, and bla_KPC (Dallenne et al., 2010). All bla_SHV and bla_CTX–M and selected bla_TEM and bla_OXA positive PCR products were sequenced to determine the genetic variation within ESBL genes. The β-lactamase gene variants were identified by sequence alignment with known GenBank sequence using the Basic local alignment search tool (BLAST)¹ (Altschul et al., 1990; Ye et al., 2006).

Phylogenetic grouping of ESBL producing E. coli was performed using Clermont’s revised PCR protocol (Clermont et al., 2013). All B2 phylogenetic group isolates were screened for international pandemic sequence type clones and their sub clones using established PCR protocols (Clermont et al., 2009; Johnson et al., 2009; Banerjee et al., 2013; Colpan et al., 2013).

A subset of 40 randomly selected strains, including 20 belonging to the ST131 PCR positive B2 group, were randomly selected and subjected to Multi-Locus Sequence Typing (MLST) according to the Achtman seven housekeeping genes (adk, fumC, gyrB, icd, mdh, purA, and recA) scheme as previously described (Wirth et al., 2006) and updated at https://enterobase.readthedocs.io/en/latest/mlst/mlst-legacy-info-ecoli.html. The 20 strains belonging to B2 phylogeny group with PCR positive indicator for ST131 were chosen to confirm the accuracy of the PCR procedure for specific detection of this high risk clone. The remaining 20 isolates belonging to different phylogeny groups were chosen as a pilot sample to gain a glimpse of the clonal distribution of E. coli in Ethiopia. When resources become available, this sample size will be increased to include all isolates.

Phylogenetic tree construction used Phyloviz software and the neighbour joining algorithm (Francisco et al., 2012). Tree branch-length minimization utilized the neighbour joining method of Saitou and Nei (Saitou and Nei, 1987). For the phylogenetic dendrogram, a WPGMA (Weighted Pair Group Method with Arithmetic mean) hierarchical clustering was performed by selecting the criterion of minimal dissimilarity. The default parameters of a 1% tolerance and an 85% similarity index were considered for clustering purposes.

The data was prepared using Excel spreadsheets (Microsoft Office) and imported to SPSS version 20.0. The frequencies of antimicrobial susceptibility and ESBL producers among different variables were calculated. Cross-tabulation was used to present the different relation between data. The compressions of ESBL distribution between studies sites and among clones were evaluated using the Chi-square test. The tests were two sided and P values < 0.05 were considered statistically significant.

Ethical clearance for this study was obtained from the Ethiopian Public Health Institute scientific and ethical review board (EPHI-IRB-054-2017) and Addis Ababa University, the College of Natural and Computational Science (IRB/039/2019).

We characterized 204 E. coli ESBL producers from four health facilities in Ethiopia. In Table 1, we present the patient demographics associated with these isolates. A small majority of the isolates were recovered from female patients (n = 118, 57.8%), and from all age groups. The majority of isolates were recovered from urine (n = 166 at 81.4%) and pus (n = 30, 14.7%).

Antimicrobial susceptibility testing revealed that all 204 ESBL producers in the strain collection were resistant to ampicillin and all generations (first to fourth) of cephalosporins, but remained highly susceptible to meropenem (n = 202, 99%) and amikacin (n = 201, 98.5%) (Figure 2A). Moreover, 93.1% (n = 190) of E. coli strains were resistant to sulfamethoxazole-trimethoprim and 89.7% (n = 183) to ciprofloxacin (Figure 2A). In addition, the prevalence of resistance to amoxicillin-clavulanate, gentamicin and cefoxitin were 72.5% (n = 148), 50% (n = 102), and 17.6% (n = 36), respectively (Figure 2A). Significantly, all 204 isolates were resistant to more than one agent in three separate antimicrobial categories (Figure 2B), which is indicative of MDR according to Magiorakos and colleagues (Magiorakos et al., 2012). In fact, most isolates were resistant to antibiotics belonging to six or seven antimicrobial categories (Figure 2B).

All 204 E. coli strains with ESBL phenotype were characterised genetically to identify the β-lactamase gene type. We identified at least one β-lactamase gene in all 204 isolates. CTX-M group-1 was detected in 94.7% (n = 179) of all CTX-M positive isolates (n = 188) (Table 2). Of these, 87.7% (n = 157) harboured the CTX-M-15 subtype (Table 2). The non-ESBL TEM-1 and OXA-1 β-lactamase genes were detected in these CTX-M-15 positive isolates at a frequency of 72.2% (n = 115) and 82.2% (n = 129), respectively (Table 2). Of the remaining isolates with a detectable group-1 CTX-M (n = 22, 12.3%), we could specify CTX-M-55, CTX-M-101, CTX-M-103, CTX-M-142, CTX-M-180, CTX-M-182, CTX-M-225, and CTX-M-227 types (Table 2). One isolate was even positive for both CTX-M-182 and CTX-M-227. In addition, CTX-M group-9 was detected in 6.4% (n = 12) of the 188 CTX-M positive strains, of which 91.7% (n = 11) harboured CTX-M-27, while the remaining isolate was positive for CTX-M-14 (Table 2). Two CTX-M-27 isolates were also positive for CTX-M-15. Our analysis also revealed eight SHV positive isolates, of which 62.5% (n = 5) contained SHV-11 and the remaining three isolates contained one of SHV-7, SHV-34 or SHV-61, respectively (Table 2). Association with CTX-M-15 was found in these isolates excepting the SHV-61 isolate. Finally, the bla_OXA–1 and bla_TEM–1 genes were identified at a respective frequency of 78.9% (n = 161) and 73.5% (n = 150) (Table 2). Conversely, no carbapenemase genes were detected despite meropenem resistance in two isolates. However, one of these isolates harboured genes for both CTX-M-15 and CTX-M-27, and the other genes for both CTX-M-182 and CTX-M-27. According to others (Tsai et al., 2013; Wang et al., 2016), being positive for one CTX-M from group-1 and CTX-M-27 from group-9 might be a reason these two isolates are resistant to carbapenem. However, resistance might also be due to porin loss in combination with ESBLs or AmpC β-lactamase production (Doumith et al., 2009).

Crucially, β-lactamases were detected in isolates obtained from all four study sites. Focusing on the frequently identified CTX-M-15 type, we observed no significant difference (P > 0.05) in the distribution frequency of this type among the four study sites (Table 2). However, we note that 81.8% (n = 9) of CTX-M-27 positive isolates originated from the geographically linked NRL and TASH, while 65.5% (n = 5) of SHV positive isolates originated from JUH (Table 2). Whether these associations are significant is unclear given the low numbers of isolates in these categories.

Next we applied the modified Clermont quadruplex PCR protocol (Clermont et al., 2013) to assign 97.0% (n = 198) of the 204 isolates into known phylogenetic groups. The largest phylogenetic group was B2 occurring at a frequency of 44.9% (n = 89) (Table 3). Other phylo-groups were detected at the frequencies of 24.7% for group A (n = 49), 16.7% for group C (n = 33), 4.5% for group D (n = 9), 3.5% for group F (n = 7) and 3.0% for group B1 (n = 6) (Table 3). The production of group 1 CTX-Ms, including CTX-M-15, were detected in all phylogenetic groups, with the highest detection rate in phylo-group B2 (43.9%, n = 69) (Table 3). On the other hand, group-9 CTX-M distribution was restricted to phylo-groups A, B2 and C, although phylo-group B2 dominated (66.7%, n = 8) (Table 3). Phylo-group B2 was also most commonly associated with isolates containing bla_OXA–1 (47.2%, n = 76) and bla_TEM–1 (42.7%, n = 64) (Table 3).

We further characterized the B2 phylo-group to provide novel information on the occurrence and distribution of the international pandemic ST131 clonal group and its sub-clones. Out of 89 isolates within phylo-group B2, 83.1% (n = 74) were positive for the mdh and gyrB ST131 markers. Of these 74 isolates, serotype O25b accounted for 83.8% (n = 62), and all of these 62 were positive for fimH30. Of these, 67.7% (n = 42) were identified as the H30Rx clade.

To examine for the prevalence of genes encoding the different CTX-M β-lactamases as well as the extent of ciprofloxacin (fluoroquinolone) resistance among the ST131 isolates (n = 74), we first classified them into ST131-H30 (n = 62) and non-ST131-H30 (n = 12) (Figure 3A). Interestingly, the PCR amplification and sequencing of the isolates revealed a high prevalence of the CTX-M-15 encoding gene, and to a lesser extent the CTX-M-9 encoding gene. There was no significant difference (P > 0.05) in prevalence of group-1 CTX-Ms, including CTX-M-15, between these two groups (Figure 3A). On the other hand, group-9 CTX-Ms genes, including the gene encoding CTX-M-9, were all associated with ST131-H30 isolates (Figure 3A). Although both bacterial groups were resistant to ciprofloxacin, the ST131-H30 group were significantly more resistant (P < 0.05) than the non-ST131-H30 group (Figure 3A). We also classified the most prevalent ST131-H30 (n = 62) subclone into H30-Rx (n = 42) and other H30 (n = 20), and observed that genes encoding group-1 CTX-M’s, including CTX-M-15, were detected in both groups with similar frequency (Figure 3B). However, group-9 CTX-M’s were more commonly associated with other H30 strains compared to H30-Rx (P < 0.05) (Figure 3B). Moreover, the patterns of ciprofloxacin (fluoroquinolone) resistance were more pronounced in H30-Rx strains (P < 0.05) (Figure 3B).

Forty isolates belonging to different phylogenetic groups were selected for MLST analysis. We detected 15 different sequence types (ST) using the allelic profile of seven housekeeping genes (adk, fumC, gyrB, icd mdh, purA, and recA). The most dominant ST identified was ST131 (50.0%, n = 20) followed by ST410 (12.5%, n = 5) (Figure 4). ST167 and ST450 were detected in two isolates, while a further eleven STs were identified for the remaining isolates (Figure 4). A BLAST analysis using the Enterobase database (Zhou et al., 2020) classified the 15 ST’s into eight sequence complexes (SCs) or clonal complexes (CCs) in which 21 (52.5%) belonged to CC131 (SC131), 5 (12.5%) to CC23, and 3 (7.5%) to CC10 (Figure 5). Five isolates were identified as CC38, CC86, CC405, CC440, and CC648 (Figure 5). On the other hand, the allelic profile of ST8529, ST3059, ST2259, ST2974, and ST450 were branded as singletons, as these STs could not be assigned to any SC or CC within the Enterobase database (Figure 5). Based on these CC assignments, ST networks were constructed using the PHYLOVIZ software² (Francisco et al., 2012). The epidemiological links and phylogenetic relationships among the STs are illustrated in Figure 6.

CC131 was most prevalent and contained all ST131 strains. These were positive for O25b and H30 subclone specific PCR, and 85% (n = 17) were positive for H30-Rx subclone specific PCR (Figure 5). All belonged to phylo-group B2, and 90% (n = 18) were positive for CTX-M-15, 5% (n = 1) positive for CTX-M-27, and 5% (n = 1) positive for both CTX-M-182 and CTX-M-27 (Figure 5). One isolate was also positive for SHV-11 (Figure 5). Most ST131 isolates were recovered from urine samples, but exhibited variable resistance levels to the non-cephalosporin antibiotics, as well as non-ESBL enzyme OXA-1 and TEM production (Figure 5). CC23 was also prevalent, and to which all ST410 strains belonged. All produced CTX-M-15; despite belonging to different phylogenetic groups, but not all produced non-ESBL enzymes, and they displayed variable resistance levels to non-cephalosporine antibiotics (Figure 5). All except one isolate was recovered from urine. Finally, CC10 was also prevalent, and comprised one ST617 and two ST167 isolates. All were in phylo-group A, and all were positive for CTX-M-15 and the non-ESBL enzymes OXA-1 and TEM (Figure 5).