Background

Extended-spectrum β-lactamases (ESBLs) are bacterial enzymes capable of degrading most β-lactam antibiotics, rendering them therapeutically useless. Many environmental [1] and animal [2] reservoirs of ESBLs have been reported, however, the heavy use of antimicrobials in hospitals remains the main driver of ESBL-mediated antimicrobial resistance (AMR) [3, 4]. The problem of ESBLs is emerging globally [5, 6] and represents one of the most dire current public health threats in both community and hospital settings [7,8,9]. ESBLs have been identified and described in both industrialized and resource-constrained countries. For example, a high ESBL burden has been observed in France and China, where local hospitals reported prevalence figures of 17.7% [10] and 68.2% [11], respectively. Likewise, studies among hospital patients in Africa show ESBL occurrences as high as 62.3% in Mali [12], 70% in Burkina Faso [13], and 84% in Cote d’Ivoire [14]. In reports from Zambia, ESBL-associated multidrug resistance (MDR) was observed in all 45 Klebsiella pneumoniae isolates obtained from neonates [15] and in all 15 diarrhoeagenic Escherichia coli isolates from under-five children [16] at the University Teaching Hospital (UTH).

Over 200 ESBLs have been documented [17], of which the majority are derivatives of TEM, SHV, or CTX-M. While TEM and SHV are common ESBL variants, the CTX-M family has recently emerged to become the preponderant type worldwide [18]. This epidemiologic transition has been paralleled by the success and rapid dissemination of the pandemic clone E. coli O25b:H4-ST131, which is frequently associated with blaCTX-M-15 [19]. Additionally, the efficient spread of blaCTX-M genes has also been facilitated by plasmids, integrons, and insertion sequences (IS) (e.g., ISEcp1, IS1, IS5, and IS26) [20]. blaCTX-M genes are usually associated with MDR, since they primarily reside on plasmids carrying other AMR genes. Considering that plasmids often incur a fitness cost in the absence of antibiotic selection pressure [21], it is possible to combat blaCTX-M-associated MDR through the prudent use of antimicrobials. However, a series of studies have reported the occurrence of blaCTX-M genes on the chromosomes of various Enterobacteriaceae [22,23,24,25,26,27], although their placement with other AMR genes was not examined. With a growing number of studies reporting the chromosomal location of blaCTX-M, it is of considerable interest to explore possible links between chromosomally-located blaCTX-M and MDR.

In this study, using phenotypic and genotypic characterization of clinical Enterobacteriaceae isolates from Zambia, we identified the probable clonal spread of blaCTX-M among selected strains. Furthermore, by reconstructing and analyzing nearly complete genome sequences, we identified chromosomally-integrated plasmid segments co-harboring ISEcp1-blaCTX-M-15 and diverse AMR genes in Enterobacter cloacae and E. coli. Most of the strains harboring these insertions exhibited MDR phenotypes corresponding with the AMR genes in the insertions. Based on these findings, we speculate that ISEcp1-mediated transposition may be responsible for the spread of MDR determinants among Enterobacteriaceae. Furthermore, stable maintenance of these chromosomal insertions may enhance the dissemination of multiple MDR Enterobacteriaceae species, with potentially detrimental effects on patient outcomes.

Methods

Strain selection

Between June and October 2018, 46 non-repetitive cefotaxime resistant clinical Enterobacteriaceae strains were obtained from the UTH, Zambia. These strains were previously isolated from blood (n = 1), cerebrospinal fluid (CSF) (n = 2), high vaginal swab (HVS) (n = 1), pus (n = 3), sputum (n = 4), stool (n = 30), and urine (n = 5) during routine clinical investigations. Cefotaxime resistance was confirmed by plating each strain on LB agar supplemented with 1 μg/ml cefotaxime (Sigma-Aldrich, USA). In addition, the minimum inhibitory concentration (MIC) of cefotaxime against each strain was determined by broth microdilution.

MIC determination

Antimicrobial susceptibility profiles of strains toward 10 different antimicrobials were determined by broth microdilution using breakpoints specified in Additional file 1: Table S1. Briefly, overnight cultures were diluted 104-fold, added to twofold serial dilutions of antibiotics in a 96-well plate in triplicate, then incubated at 37 °C for 18 h with shaking at 1600 rpm. Next, optical densities at 595 nm (OD595) were measured using the Multiskan FC Microplate Photometer (Thermo Scientific, USA), with OD595 values of at least 0.1 indicating positive bacterial growth. The MIC was defined as the minimum antibiotic concentration producing an OD595 value of less than 0.1. Two laboratory strains, E. coli MG1655 and E. coli 10-β (NEB, USA), were used for quality control.

Growth rate determination

Bacterial cultures prepared in plain LB were monitored for growth by measuring OD600 in real-time over a period of 16 h. To this end, overnight cultures were diluted 103-fold and added to a 96-well plate in duplicate, then subjected to OD monitoring using the Varioskan LUX Multimode Microplate Reader (Thermo Scientific, USA) at 37 °C, while shaking at 600 rpm. Growth rates were estimated from the slopes obtained by fitting parametric models to the resulting data using the R package grofit version 1.1.1 [28].

Whole genome sequencing

Genomic DNA was extracted from overnight cultures prepared in LB supplemented with 1 μg/ml cefotaxime using a QIAamp PowerFecal DNA Kit (QIAGEN). Libraries were prepared using NexteraXT (Illumina) and sequenced using MiSeq (Illumina) to obtain 300 bp paired-reads. The same genomic DNA samples were also subjected to long-read sequencing by MinION (Oxford Nanopore Technologies) using a Rapid Barcoding Kit (SQK-RBK004) and an R9.5 flowcell (FLO-MIN107). Short-read trimming and adapter sequence removal were performed using Trim Galore version 0.4.2 with options of "–paired –nextera" (https://github.com/FelixKrueger/TrimGalore). Basecalling of long reads was performed using Guppy Basecalling Software version 3.4.5 and the reads were assembled using Canu version 1.8 [29] with the "corOutCoverage = 1000 genomeSize = 6 m" option applied. Redundant repeats at terminal ends of contigs were identified using Gepard version 1.40 [30] and manually trimmed down to one copy. Trimmed contigs were then subjected to base-error correction with Illumina reads using Pilon version 1.23 [31]. Contigs longer than 2 Mb were classified as chromosomal while circular contigs of less than 500 kb and containing plasmid replicons were classified as plasmids. To categorize the rest of the contigs as either chromosomal or plasmid-based in origin, a local database was generated for each strain using known chromosomal and plasmid segments. Unclassified contigs were then subjected to BLASTn searches against the local database and sequences matching known chromosome or plasmid regions with at least 70% identity were regarded as redundant and removed from the data pool. Sequences exhibiting hits with less than 70% identity were retained and subjected to NCBI BLASTn searches against the nt database. A sequence was regarded as chromosomal if at least 70% of the top 10 hits were chromosomes, and as plasmid-based if at least 70% of the top 10 hits were plasmids.

Phylogenetic analysis

To elucidate the evolutionary relationships among strains, whole genome-based phylogenetic trees were constructed using Parsnp version 1.2 [32] and imported into MEGA version 7.0 [33] for visualization and editing. Multilocus sequence typing (MLST) was performed in silico by uploading raw Illumina reads to a public MLST server (www.cbs.dtu.dk/services/MLST) [34].

Detection of plasmid replicons, strain serotypes, and AMR genes

Plasmid replicons and O:H serotypes were determined by interrogating contigs with PlasmidFinder [35] and EcOH [36] databases, respectively, using ABRicate software version 0.8.10 (https://github.com/tseemann/abricate) with options –mincov 90 and –minid 90 specified. To identify AMR genes, the AMRFinderPlus tool [37] was used with option -i 0.7 engaged.

Determination of clustering patterns among strains

To unravel the mechanisms of blaCTX-M spread at the UTH, strains were compared in terms of AMR phenotype, AMR genes, and plasmid replicons by exploiting clustering patterns and correlations using the R package ComplexHeatmap [38].

Sequence alignment and determination of chromosomal insertion locations of bla CTX-M

Strains carrying chromosomal blaCTX-M were annotated using DFAST version 1.2.4 [39] and aligned to reference sequences using Mauve [40]. Zam_UTH_18, Zam_UTH_26, and Zam_UTH_41 were aligned against the reference strain E. coli MG1655 (GenBank accession no. NC_000913.2), while Zam_UTH_44 was aligned against E. cloacae ATCC 13047 (GenBank accession no. NC_014121.1). Zam_UTH_42 and Zam_UTH_47 were aligned against E. coli ST648 (GenBank accession no. CP008697.1), while Zam_UTH_43 was aligned against another strain (Zam_UTH_08) in our data. Comparison, visualization, and exploration of the data were performed using genoPlotR [41].

PCR and sanger sequencing

To validate and further characterize chromosomal insertions, junctions between chromosomes and plasmids were amplified by PCR using primers shown in Supplementary Additional file 1: Table S2. Where possible, primers external to the insertion were used and amplicon size was compared to that obtained from a control strain. Also, strains for which the exact blaCTX-M allele was unidentifiable due to low Illumina read number were subjected to PCR using primers shown in Additional file 1: Table S3, followed by Sanger sequencing. Briefly, PCR was performed using KOD One Master Mix (TOYOBO, Japan) and the products were purified using a MinElute PCR Purification Kit (QIAGEN). Sequencing PCR was performed using a BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, USA), followed by Sanger sequencing using a 3130 Genetic Analyzer (Applied Biosystems, USA). Raw sequence data was assembled using SnapGene software (Insightful Science, available at snapgene.com) and the obtained consensus sequences analyzed by the AMRFinderPlus tool [37] with option -i 0.7.

Results

MLST revealed high genetic diversity among strains

This study examined 46 cefotaxime-resistant Enterobacteriaceae strains isolated from various clinical samples among patients at the UTH, Zambia (Table 1). To elucidate the genetic variability and relationships among these strains, WGS was used to reconstruct nearly complete genome sequences. In silico MLST from the WGS data identified the strains as E. cloacae (1/46, 2.2%), K. pneumoniae (9/46, 19.6%), and E. coli (36/46, 78.3%) (Table 1).

Table 1 Description of 46 strains used in this study
Full size table

Although E. coli were represented by 12 sequence types (STs), 25/36 (69.4%) strains belonged to 4 STs (i.e., ST69, ST131, ST617, and ST405) (Fig. 1a), suggesting the expansion of a few eminent clones. ST131 is known to rapidly spread due to its diverse virulence and AMR mechanisms [42], but its prevalence in this study, at 6/36 (16.7%), was second to that of ST69. ST69 predominated, with a prevalence of 9/36 (25.0%), highlighting its important contribution to the burden of ESBLs in the local hospital. Of the 12 E. coli STs identified in this study, one was novel and has since been designated ST11176. Among the K. pneumoniae isolates, which were classified into 3 STs, the most prevalent was ST307, representing 6/9 (66.7%) strains (Fig. 1b).

Fig. 1
figure 1

Phylogenetic analysis. Whole genome-based phylogenetic trees for 36 E. coli and 9 K. pneumoniae strains. a E. coli: a total of 12 STs were identified, including one novel type (marked with *). Four STs constituted 25/36 (69.4%) of E. coli strains, the most common being ST69 (9 strains), followed by ST131 (6 strains). b K. pneumoniae: a total of 3 STs were identified, with ST307 alone representing 6/9 (66.7%) strains

Full size image

Chromosomal bla CTX-M genes were identified by in silico analysis

To identify AMR gene patterns among the strains, in silico detection was performed using AMR gene sequence databases. A total of 58 AMR genes belonging to 12 classes were identified (Additional file 1: Table S4). Among the β-lactamase genes, the blaCTX-M family predominated, with 45/46 (97.8%) strains carrying at least one of the following alleles: blaCTX-M-14, blaCTX-M-15, blaCTX-M-27, and blaCTX-M-55 (Fig. 2). These blaCTX-M genes were located on chromosomes in 7 strains (7/45, 15.6%; one E. cloacae and six E. coli); one of these strains harbored an extra blaCTX-M gene on a plasmid, and the remaining 38 strains (38/45, 84.4%) harbored blaCTX-M genes exclusively on plasmids. In agreement with previous studies [43], blaCTX-M-15 was the most prevalent ESBL gene (28/45, 62.2%). Despite documentation of a close association between blaCTX-M-15 gene and E. coli ST131 [44], only one E. coli ST131 (1/6, 16.7%) possessed blaCTX-M-15 (Fig. 2), with the remaining five E. coli ST131 strains possessing plasmid-borne blaCTX-M-27. In silico serotyping of all six E. coli ST131 strains revealed that none of them belonged to the pandemic clone O25b:H4-ST131, with the 5 blaCTX-M-27-carrying strains belonging to Onovel31:H4 and the blaCTX-M-15-carrying strain belonging to O107:H5.

Fig. 2
figure 2

AMR phenotypes, AMR genes and plasmid replicons. All but one strain displayed resistance to at least three antimicrobial classes. There was no phenotypic or genotypic resistance to imipenem, however, one strain (Zam_UTH_40) exhibited phenotypic resistance to colistin. A total of 12 AMR gene classes were identified. Within the β-lactamase gene class, the blaCTX-M family showed the most diversity, with blaCTX-M-15 being the most common variant. Most blaCTX-M genes were located on plasmids, however, 7/45 (15.6%) strains harbored the genes on the chromosome. A total of 24 plasmid replicons were detected, with the most prevalent being IncFIB(AP001918)_1, which was present in 30/46 (65.2%) strains. Hierarchical clustering showed aggregation of strains of the same ST. Cefotaxime (CTX) is not shown here since all strains were selected using CTX. AMP; ampicillin. CHL; chloramphenicol. CIP; ciprofloxacin. CST; colistin. DOX; doxycycline. GEN; gentamicin. IPM; imipenem. NAL; nalidixic acid. NIT; nitrofurantoin

Full size image

To determine antimicrobial susceptibility profiles, MICs of 10 antimicrobials were measured against each strain. Forty-five out of 46 (97.8%) strains displayed MDR phenotypes, defined as a lack of susceptibility to at least one antimicrobial from at least three antimicrobial classes [45]. The highest non-susceptibility rate was to ampicillin (46/46, 100%), followed by gentamicin (43/46, 93.5%), ciprofloxacin (41/46, 89.1%), and nalidixic acid (41/46, 89.1%). While no carbapenem resistance was detected, one K. pneumoniae strain (1/46, 2.2%) displayed low-level resistance (MIC = 4 μg/ml) to the drug of last resort, colistin (Fig. 2). In almost all cases the probability of phenotypic resistance was high (positive predictive value ≥ 80%) given a positive genotypic result (Additional file 1: Table S5). While previous studies have shown a significant correlation between resistance range and fitness cost [46], we did not find any correlation between the number of AMR genes and bacterial growth rate (Additional file 1: Fig. S1A).

E. coli ST69 strains possess an IncH plasmid harboring bla CTX-M-14

To characterize the plasmid content of the strains, WGS data were compared against the PlasmidFinder database [35]. In total, 24 plasmid replicons were detected, the most common being IncFIB(AP001918)_1 (30/46, 65.2%), IncFIA_1 (27/46, 58.7%), Col(MG828)_1 (16/46, 34.8%), and IncB/O/K/Z_2 (13/46, 28.3%) (Fig. 2). Most E. coli ST69 strains (8/9, 88.0%) in this study carried an IncF plasmid possessing dfrA12 and sul2 genes, encoding resistance to trimethoprim and sulfamethoxazole, respectively. This was expected since previous studies have associated E. coli ST69 with IncF plasmids encoding trimethoprim-sulfamethoxazole (SXT) resistance [47]. However, all the E. coli ST69 strains in this study also carried two additional MDR plasmids, including a 225 kb IncH plasmid which is not commonly found in this ST [47, 48]. This IncH plasmid contained blaCTX-M-14 and shared over 80% homology with an IncH plasmid in a K. pneumoniae ST37 strain (Zam_UTH_04) from our data pool (Additional file 1: Fig. S2A), indicating that these plasmids could share a common origin. The two plasmids exhibited similar genetic contexts around blaCTX-M-14 gene, however, both plasmids contained several insertions that carried various AMR genes as well as genes related to fitness traits (Additional file 1: Fig. S2B). For instance, the IncH plasmid from E. coli ST69 possessed the mucAB operon which is associated with resistance to DNA-damaging agents such as ultraviolet radiation [49]. Furthermore, this plasmid also harbored the mer operon that is involved in resistance to organomercury compounds [50].

Analysis of the other three dominant E. coli STs (i.e., ST131, ST617, and ST405) showed a close association between IncF plasmids and blaCTX-M genes, as previously reported [51]. However, comparison of IncF plasmids from strains of different STs revealed low nucleotide sequence homology (Additional file 1: Fig. S2C), ruling out horizontal gene transfer as a mode of blaCTX-M gene propagation.

The spread of bla CTX-M genes is driven by clonal expansion

To establish the mode of blaCTX-M propagation among strains, hierarchical clustering was conducted using AMR phenotypes, AMR genes, and plasmid replicons as variables. The results indicated that strains can be grouped into 4 main clusters, which we designated as EC1, EC2, EC3, and KP (Fig. 2). The largest cluster, EC1, was predominated by E. coli ST69 and was characterized by the presence of blaCTX-M-14 and IncHI. In addition, EC1 contained more AMR genes than what was observed in clusters EC2, EC3, and KP (P < 0.01) (Additional file 1: Fig. S3). Cluster EC2 was closely related to EC3, with both clusters containing qacEdelta1, mph(A), and aad5 genes that were absent in clusters EC1 and KP. Lastly, cluster KP, formed by K. pneumoniae and E. cloacae, was defined by the presence of four plasmid-mediated quinolone resistance (PMQR) genes, oqxA, oqxB, oqxB19, and qnrB1. Overall, the analysis showed the aggregation of strains belonging to the same ST. Taken together with our MLST findings, these results suggest the possible spread of blaCTX-M by clonal expansion rather than horizontal gene transfer.

Large chromosomal insertions co-harbor bla CTX-M-15 and other AMR genes

To determine the genetic contexts of chromosomally-located blaCTX-M in 7 strains (one E. cloacae and six E. coli), chromosomal insertions were identified by alignment against reference sequences using Mauve [40]. In all cases, the insertions were bounded by ISEcp1 at one end (Figs. 3, 4, 5), implying that this element is responsible for mobilization. However, some insertions also harbored various IS elements (e.g., IS1, IS6) and transposons, suggesting that other sophisticated mechanisms may play a role. The inserted segments, which were confirmed by PCR (Additional file 1: Fig. S4), were genetically distinct, with sizes ranging from ~ 3 kb to ~ 41 kb. Furthermore, these insertions showed high nucleotide sequence similarity to plasmids available in the NCBI GenBank, implying the transposition from plasmid to chromosome. In one E. cloacae and three E. coli strains, blaCTX-M-carrying insertions were larger than 10 kb and harbored other types of AMR genes. As expected, strains possessing these MDR insertions showed resistance to multiple antibiotics of clinical importance (Figs. 4, 5).

Fig. 3
figure 3

blaCTX-M genes present on short chromosomal insertions in E. coli. a Zam_UTH_41. This E. coli ST8767 strain carried blaCTX-M-14 on a 3,095 bp chromosomal insertion with ISEcp1 located 249 bp upstream of blaCTX-M-14. Zam_UTH_26 (not shown) also had a similar genetic context for its insertion. b Zam_UTH_43. This E. coli ST131 O107:H5 strain harbored blaCTX-M-15 on a 6,036 bp chromosomal insertion, with ISEcp1 located 255 bp upstream of blaCTX-M-15. About 2.5 kb downstream of this insertion was another insertion that harbored genes conferring resistance to chloramphenicol (cat), β-lactams (blaOXA-1), and aminoglycosides/quinolones (aac(6′)-Ib-cr5). F1, F2, F3, R1, R2, R3; primers used for confirmation of insertions

Full size image
Fig. 4
figure 4

blaCTX-M genes present on large chromosomal insertions in E. coli. a Zam_UTH_18. This E. coli ST3580 strain possessed blaCTX-M-15 on an 11,383 bp chromosomal insertion, which was very similar to plasmid pF609 (GenBank accession no. MK965545.1). blaCTX-M-15 was closely associated with ISEcp1, which was located 255 bp upstream. The insertion also harbored the quinolone resistance gene qnrS1, located 4639 bp downstream of blaCTX-M-15. The phenotypic AMR profile of this strain showed resistance to ampicillin and susceptibility to quinolones. b Zam_UTH_42. This E. coli ST648 strain harbored blaCTX-M-15 on a 14,328 bp chromosomal insertion that was very similar to plasmid p13ARS_MMH0112-2 (GenBank accession no. LR697123.1). This insertion carried genes associated with resistance to aminoglycosides (aac(3)-IIa), aminoglycosides/quinolones (aac(6′)-Ib-cr5), β-lactams (blaOXA-1, blaTEM-1), and chloramphenicol (catB3). ISEcp1 was located 255 bp upstream of blaCTX-M-15, however, unlike in other strains, ISEcp1 in this strain was truncated by IS1 and transposase. The phenotypic AMR profile of this strain was consistent with the AMR genotype for the insertion. Zam_UTH_47 (not shown) also had a similar genetic context for its insertion. F4, F5, F6, F7, R4, R5, R6, R7; primers used for confirmation of insertions. White; susceptible. Black; resistance phenotype in the absence of corresponding AMR gene. Red; β-lactam resistance. Brown; chloramphenicol resistance. Green; aminoglycoside and/or quinolone resistance

Full size image
Fig. 5
figure 5

blaCTX-M gene on a large chromosomal insertion in E. cloacae. Zam_UTH_44. This E. cloacae ST316 strain carried blaCTX-M-15 on a ~ 41 kb chromosomal insertion that exhibited nucleotide sequence homology with plasmid pCRENT-193_1 (GenBank accession no. CP024813.1). ISEcp1 was located 255 bp upstream of blaCTX-M-15. The insertion also included diverse AMR genes encoding resistance to six antimicrobial classes, namely; aminoglycosides (aac(3)-IIa), quinolones (qnrB1), aminoglycosides/quinolones (aac(6′)-Ib-cr5), β-lactams (blaOXA-1), trimethoprim (dfrA14), chloramphenicol (catB3), and tetracyclines (tet(A)). The phenotypic AMR profile of this strain was consistent with the AMR genotype of the insertion. F8, F9, R8, R9; primers used for confirmation of insertions. White; susceptible. Black; resistance phenotype in the absence of corresponding AMR gene. Red; β-lactam resistance. Brown; chloramphenicol resistance. Green; aminoglycoside and/or quinolone resistance. Pink; tetracycline resistance

Full size image

Chromosomal insertions were bracketed by 5-bp direct repeats in all six E. coli strains, which is typical of ISEcp1-mediated transposition. However, ISEcp1 was truncated by IS1 and transposase in two strains (Zam_UTH_42 and Zam_UTH_47) (Fig. 4b), suggesting that the interrupted ISEcp1 may no longer be functional, and that mobilization of blaCTX-M-15 may have been induced by IS1 and/or transposase. In all strains, the relative position of the downstream blaCTX-M gene was constant for each allele (i.e., blaCTX-M-14 or blaCTX-M-15), indicating possible allele-specific ancestral origins. For instance, in Zam_UTH_26 and Zam_UTH_41 (both E. coli ST8767), blaCTX-M-14 was located 249 bp downstream of ISEcp1 on a 3,095 bp insertion. Similarly, in 5 strains, blaCTX-M-15 was observed 255 bp downstream of ISEcp1 on insertions of varying lengths. More specifically, Zam_UTH_43 (E. coli ST131) (Fig. 3b) and Zam_UTH_18 (E. coli ST3580) (Fig. 4a) carried blaCTX-M-15 on 6,036 bp and 11,383 bp insertions, respectively, while Zam_UTH_42 and Zam_UTH_47 (both E. coli ST648) (Fig. 4b) carried blaCTX-M-15 on a 14,328 bp insertion. Furthermore, Zam_UTH_44 (E. cloacae ST316) (Fig. 5) carried blaCTX-M-15 on a chromosomal insertion longer than 41 kb.

Of the 7 strains carrying chromosomal blaCTX-M, 4 strains (i.e., Zam_UTH_18, Zam_UTH_42, Zam_UTH_44, and Zam_UTH_47) harbored the gene on insertions longer than 10 kb that exhibited high nucleotide sequence similarity to plasmids available in the NCBI GenBank (Figs. 3, 4, 5). Notably, the insertion in Zam_UTH_18 also appeared in the chromosome sequences of Salmonella enterica (GenBank accession no. CP045038) and two E. coli ST38 strains (GenBank accession no. CP010116 and CP018976). Furthermore, on the 4 insertions larger than 10 kb, we found additional AMR genes downstream of blaCTX-M-15. Zam_UTH_18 was found to possess the qnrS1 gene, which is associated with quinolone resistance, while Zam_UTH_42 and Zam_UTH_47 had genes associated with resistance to aminoglycosides (aac(3)-IIa), aminoglycosides/quinolones (aac(6′)-Ib-cr5), β-lactams (blaOXA-1, blaTEM-1), and chloramphenicol (catB3). Similarly, Zam_UTH_44 possessed 7 genes known to confer resistance to aminoglycosides (aac(3)-IIa), quinolones (qnrB1), aminoglycosides/quinolones (aac(6′)-Ib-cr5), β-lactams (blaOXA-1), trimethoprim (dfrA14), chloramphenicol (catB3), and tetracyclines (tet(A)).

Phenotypic AMR profiles of the strains carrying chromosomal blaCTX-M were in agreement with the presence of AMR genes on insertions except for Zam_UTH_18 which showed quinolone susceptibility despite possessing qnrS1. Additionally, these strains had cefotaxime MICs that were at least 64-fold higher than the clinical breakpoint of 2 μg/ml published by the Clinical and Laboratory Standards Institute [52]. Though Zam_UTH_26 was closely related to Zam_UTH_41 (Fig. 1a), the latter had a higher cefotaxime MIC (≥ twofold) possibly due to the presence of plasmid-borne blaCTX-M-15, in addition to blaCTX-M-14 on the chromosome. This difference was also reflected by the lower growth rate (μ = 0.136) of Zam_UTH_41 compared to Zam_UTH_26 (μ = 0.189), suggesting a fitness cost imposed by the extra blaCTX-M-15-carrying plasmid. Likewise, Zam_UTH_18, which lacked plasmids, grew at a rate higher than the 75th percentile of the estimated E. coli growth rates (Additional file 1: Fig. S1B; Additional file 2: Fig. S5).

Discussion

In this study, we sought to investigate the phenotypic and genotypic characteristics of clinical Enterobacteriaceae strains from the UTH, Zambia. Our results suggest that the dissemination of blaCTX-M has been mediated mainly by clonal expansion. While the majority of strains possessed blaCTX-M genes on plasmids, 7 strains harbored the genes on chromosomes. Data obtained in previous studies indicate that the chromosomal incorporation of blaCTX-M is usually mediated by ISEcp1; however, the co-occurrence of ISEcp1-blaCTX-M and additional AMR genes has never been reported in E. cloacae or E. coli. Despite approaches enabling examination of the immediate flanking regions of chromosomal blaCTX-M, limitations of Southern hybridization and WGS using short reads make it difficult to identify and resolve large chromosomal insertions. By reconstructing nearly complete genomes, our analysis of the genetic context of chromosomally-located blaCTX-M was significantly improved compared to previous analyses. In one E. cloacae and three E. coli strains, we found blaCTX-M-15 in large chromosomal insertions whose nucleotide sequences significantly resembled plasmids available in the NCBI GenBank. We also found that these insertions were bounded by ISEcp1 at one end and possessed additional AMR genes. Phenotypic AMR profiles of these strains showed near-perfect agreement with the observed chromosomal MDR genotypes. These findings suggest that ISEcp1 contributes to the dissemination of blaCTX-M-15-associated MDR determinants among various Enterobacteriaceae species.

Phylogenetic analysis highlighted the predominance of four E. coli STs (ST69, ST131, ST617, and ST405) and one K. pneumoniae ST (ST307), suggesting the clonal dissemination of a few particular strains. Fitting with this hypothesis, strain comparisons based on AMR phenotypes, AMR genes, and plasmid replicons generally showed clustering of strains according to ST. Most E. coli ST131 strains belong to the O25b:H4 pandemic clone which frequently harbors blaCTX-M-15 [42, 53]. However, the six E. coli ST131 strains examined in this study belonged to either Onovel31:H4 or O107:H5. Newer variants of E. coli ST131 have recently emerged [54], including Onovel31:H4 [55] detected in this study. To our knowledge, this is the first report of E. coli ST131 with serotype O107:H5.

The most abundant E. coli ST in this study, ST69, belonged to a clade that possessed more AMR genes compared to other clades (Additional file 1: Fig. S3). Generally, ST69 strains lack ESBL genes [56], are associated with community acquired infections [57], and carry IncF plasmids encoding SXT resistance [47]. Interestingly, the ST69 strains in this study carried blaCTX-M-14 on an IncH plasmid and showed the highest prevalence among the examined hospital isolates. This rare combination of IncH plasmids and E. coli ST69 was also reported in an Egyptian strain isolated from raw milk cheese [58], although the IncH plasmid (GenBank accession no. CP023143) in this strain was substantially different in AMR gene content. The observed predominance of ST69 in our study may be a consequence of selection pressure arising from the use of SXT for prophylaxis against Pneumocystis jirovecii (previously carinii) pneumonia (PCP). In Zambia, HIV infected or exposed patients at risk of developing PCP receive SXT prophylaxis for several months, or even years, until they are no longer at risk [59]. While SXT prophylaxis is highly beneficial, this practice has been associated with the development of SXT resistance in various bacterial species, including E. coli [60,61,62]. We speculate that the high proportion of ST69 observed here was a result of SXT-mediated selection, followed by acquisition of the blaCTX-M-14-containing IncH plasmid. The resulting ESBL phenotype, coupled with survival mechanisms such as error-prone mutagenesis mediated by mucAB operon, probably fostered the emergence of ST69 as a successful hospital clone.

The phenotypic and genotypic resistance observed in this study may be attributed to poor local antimicrobial stewardship. In Zambia, policies promoting the judicious use of antimicrobials are hardly adhered to and most drugs can be accessed without a prescription [63]. Moreover, a study conducted at the UTH reported a 53.7% antibiotic prescription rate [64], clearly surpassing the 30% rate recommended by the WHO [65]. Most serious bacterial infections among UTH inpatients are treated with third-generation cephalosporins (e.g., cefotaxime) [66], while quinolones (e.g., ciprofloxacin) are commonly used for outpatient infections [64]. Our results show that nearly 90% of cefotaxime-resistant strains were also resistant to quinolones. Meanwhile, carbapenem use at UTH is strictly monitored and prescription usually depends on robust laboratory evidence. Encouragingly, we did not detect carbapenem resistance either phenotypically or genotypically. However, it is noteworthy that carbapenem resistance has continued to emerge globally, even in Africa [67], resulting in the gradual re-introduction of colistin as a drug of last resort [68]. Despite this transition in treatment by many countries, there is no current report of the clinical use of colistin in Zambia. Nonetheless, we found borderline phenotypic colistin resistance in a K. pneumoniae strain (Zam_UTH_40) isolated from CSF. While this finding is of clinical significance, further studies are needed to verify this observation.

Systematic characterization of the diverse genetic environments of chromosomally-located blaCTX-M revealed a close association with ISEcp1, as previously reported [69]. Intriguingly, one E. cloacae and three E. coli strains carried both ISEcp1-blaCTX-M-15 and other AMR genes on large chromosomal insertions that significantly resembled publicly available plasmid sequences. While this work was in progress, Goswami et al. described an E. coli strain possessing blaCTX-M-15 and four other AMR genes on a chromosomal insertion similar to a plasmid region, however, in contrast to our work, the mobilizing unit was IS26 [47]. More recently, in South Korea, Yoon et al. observed the co-occurrence of ISEcp1-blaCTX-M-15 and other AMR genes on chromosomal insertions in K. pneumoniae [70]. In contrast, none of the K. pneumoniae strains in our study harbored chromosomal blaCTX-M genes, probably due to geographic variations between Zambia and South Korea. Here we report, for the first time, chromosomal integration of ISEcp1-blaCTX-M-15-associated MDR elements in E. cloacae and E. coli. Since the identified MDR insertions were all bounded by ISEcp1 at one end, it is likely that this IS element was responsible for mobilization. However, the presence of other mobile elements, as well as the interruption of ISEcp1 in 2 strains, also potentially suggests the occurrence of other complex genetic events. Phenotypic susceptibilities confirmed that chromosomally-located MDR determinants were capable of conferring AMR to multiple drugs. For instance, Zam_UTH_44, which carried exclusively chromosomal AMR genes, displayed resistance to β-lactams, quinolones, gentamicin and chloramphenicol. However, Zam_UTH_18 was susceptible to quinolones despite harboring chromosomal qnrS1. We did not investigate qnrS1 expression, however, this observed contradiction could be due to the fact that PMQR determinants (such as qnrS1) are associated with only small reductions in quinolone susceptibility [71]. Nevertheless, such small changes are usually sufficient to cause treatment failure in patients, and thus some researchers advocate for the revision of quinolone breakpoints [72].

While the benefits of chromosomal integration of ISEcp1-blaCTX-M-15-associated MDR elements are not fully understood, it may help accelerate the spread of MDR determinants as an intermediary reservoir. As previously observed, plasmids belonging to the same incompatibility group are unable to stably coexist in the same bacterial host cell line [73]. Interplasmid gene transfer would thus be optimized by transient chromosomal incorporation. Furthermore, chromosomal integration of ISEcp1-blaCTX-M-15-associated MDR elements warrants stable dissemination of resistant strains regardless of the presence or absence of antibiotic selection pressure [70]. Since bacteria are likely to lose plasmids when the cost of maintaining them outweighs their benefits [74, 75], chromosomal integration and the stable maintenance of crucial AMR genes seem to be a viable safeguard for survival and continued AMR spread. This hypothesis is supported by our observation that Zam_UTH_18 did not carry any plasmids but possessed a chromosomally-incorporated 11 kb segment bearing blaCTX-M-15 and qnrS1. The high growth rate exhibited by this strain may suggest a fitness advantage conferred by its plasmid-free status. We speculate that Zam_UTH_18 may have possessed a plasmid that was lost after chromosomal integration of the blaCTX-M-15-carrying segment. To verify this observation, more studies need to be conducted on growth and competitive performance using a reference strain harboring the blaCTX-M-15-containing segment on a plasmid.

Table 2 Accession numbers for reads and assemblies
Full size table

Conclusion

We characterized AMR phenotypes and genotypes of cefotaxime-resistant clinical Enterobacteriaceae and noted that the main mode of blaCTX-M transmission was through the spread of clones of a few resilient STs. In one E. cloacae and three E. coli strains, chromosomally-integrated ISEcp1-blaCTX-M-15 transposition unit existed on large insertions that concurrently harbored diverse AMR genes and originated from plasmids. Our findings suggest that ISEcp1 facilitates chromosomal integration of blaCTX-M-15-associated MDR determinants in diverse Enterobacteriaceae species. Stable maintenance of these determinants on chromosomes may promote the propagation of MDR clones among multiple Enterobacteriaceae species, jeopardizing the treatment efficacy of available antimicrobials.