This study utilized a total of 583 non-duplicate Enterobacteriaceae acquired between 2016 and 2019. Isolates comprised 487 E. coli, 87 Klebsiella species, and nine Citrobacter species and were recovered from human clinical sample: stool (n = 60) and urine (n = 35), human hospital environment (n = 15), rectal swabs of camels (n = 40), cattle (n = 36), dogs (n = 42), pigs (n = 65), and cloacal swabs of poultry (n = 250) and clinical samples (liver) of poultry (n = 40) were used in this study (Ngbede et al., 2020). Procedures for sample collection, isolation, and identification have been previously reported (Ngbede et al., 2020). Ethical approval for the collection of samples from humans was provided by the Health Research and Ethics Committee of the facilities (FMH/FMC/MED.108/VOL.I/X and BSUTH/MKD/HREC/2013B/2018/0027).

Minimum inhibitory concentrations (MICs) of carbapenem (meropenem, ertapenem, and imipenem) were determined on the 583 isolates using the broth dilution method as recommended by the Clinical and Laboratory Standards Institute (CLSI) (Clinical and Laboratory Standards Institute, 2020). Briefly, colonies of each isolate from an overnight growth on tryptone soya agar were suspended in 5 mL normal saline to make an inoculum the equivalent of a 0.5 MacFarland standard. The turbidity of the inoculum was measured using a densitometer (BioScan). A 20 μL volume of the inoculum was dispensed into 2 mL of Mueller–Hinton broth containing different concentrations of the respective carbapenems and incubated at 35°C for 24 h. E. coli ATCC 25922 and Pseudomonas aeruginosa ATCC 27853 served as controls.

The Carbapenem inactivation method (CIM) was used to screen the isolates for carbapenemase production as described (van der Zwaluw et al., 2015). Briefly, a loopful (≈10 μL) of the isolate was suspended in 400 μL of double distilled water and followed by a 10 μg meropenem disk (Oxoid, United Kingdom) which was immersed in the suspension and incubated for 2 h at 35°C. The disk was removed from the suspension with an inoculation loop and placed on a Mueller–Hinton agar plate inoculated with a 0.5 McFarland standard E. coli strain ATCC 29522 (a susceptible indicator strain) using a sterile cotton swab; it was subsequently incubated at 35°C for 24 h. A positive results was provided when isolates with carbapenemase production inactivates meropenem in the disk, allowing uninhibited growth of the susceptible indicator strain. A negative result occurred when meropenem disks were incubated in suspensions of isolates without carbapenemases, yielding a clear inhibition zone of the susceptible indicator strain (van der Zwaluw et al., 2015).

Antimicrobial susceptibility profiles of the isolates were determined by the disk diffusion method (Clinical and Laboratory Standards Institute, 2020) using 10 antimicrobial agents sourced from Oxoid, United Kingdom: Amoxycillin (10 μg), amoxicillin/clavulanic acid (30 μg), cefoxitin (30 μg), ceftriaxone (30 μg), ciprofloxacin (5 μg), chloramphenicol (10 μg), gentamicin (10 μg), doxycycline (30 μg), enrofloxacin (5 μg), and sulfamethoxazole-trimethoprim (25 μg). The results were interpreted based on the guidelines of the CLSI (Clinical and Laboratory Standards Institute, 2020).

Fifty out of the 53 isolates co-resistant to colistin and carbapenem, were randomly selected and subjected to WGS. Only 50 isolates were randomly selected for the WGS due to its associated cost and the selection of isolates represented the different species and susceptibility patterns encountered among the 53 co-resistant (carbapenem-colistin) isolates recovered from humans, animals and the environment. Genomic DNA (gDNA) was extracted from overnight cultures of the isolates using the Wizard Genomic DNA Purification Kit (Promega, United States) following the manufacturer’s recommendations. Isolates were sequenced using next-generation sequencing on an Illumina MiSeq platform (OE Biotech, Shanghai, China) using the V2 paired-end chemistry (2 × 250 bp).

Quality of the sequencing was assessed using QUAST v5.1¹ (Gurevich et al., 2013) before de novo assembly using the SeqMan Pro v.11.2.1 (DNASTAR, United States) followed by annotation using PROKKA v1.14.5. Species identity of the isolates was further confirmed with the WGS data using the KmerFinder v3.2² (Hasman et al., 2014; Larsen et al., 2014; Clausen et al., 2018). Genetic relatedness of isolates was analyzed by phylogroup (E. coli only), multilocus sequence types (MLST) and core genome MLST (cgMLST). MLST for each isolate was predicted using the pubMLST³ while the assignment of core genome MLST and phylogroups for the E. coli isolates were carried out using the cgMLSTFinder v1.1⁴ (Zhou et al., 2020) and the online Clermont tool v20.03⁵ (Beghain et al., 2018; Clermont et al., 2019).

In silico serotyping for the E. coli and Klebsiella isolates was performed using the SeroFinder 2.0⁶ (Joensen et al., 2015) and Kaptive⁷ (Wick et al., 2018).

The presence of acquired antimicrobial resistance genes were investigated using the ResFinder v4.0⁸ (Bortolaia et al., 2020) and CARD RGI 5.2.0⁹ (Alcock et al., 2020). Plasmids were identified using PlasmidFinder 2.1¹⁰ (Carattoli et al., 2014) and MGEFinder v1.0.3¹¹ (Johansson et al., 2021). Virulence genes were identified using the VirulenceFinder 2.0 (Lui et al., 2019).

Mutations in the genes previously identified as responsible for resistance to colistin (mgrB, prmAB, phoPQ, arnT, ccrB) and carbapenem (ompC, ompF, ompK/35/36/37, marR, acrR, ramR) were investigated by alignment with wild type reference genomes of E. coli strain K-12 substrain MG1655 (NC_000913.3) and K. pneumonia subspecies pneumoniae MGH 78578 (NC_009648.1). PROVEAN v1.1.3¹² was employed to predict the possible role/effect of observed amino acid substitutions (mutation/alteration) on protein functions i.e., colistin or carbapenem resistance (Choi and Chan, 2015). The standard PROVEAN cutoff score of ≤-2.5 and >-2.5 was used to categorize the mutation with deleterious and/or neutral effects on protein function, respectively (Choi et al., 2012; Choi and Chan, 2015).

The conjugative transferability of colistin and carbapenem resistance determinants was assessed using the solid mating conjugation assay with sodium azide-resistant E. coli J53 as the recipient (Ojo et al., 2016). Overnight cultures of the donor and recipient were mixed in a 1:4 ratio in tryptone soya broth and centrifuged at 14,000 × g for 1 min. The pelleted cells were resuspended in 15 μL of 0.85% NaCl, spotted onto MHA, incubated at 37°C for 20–24 h. Bacteria growing on the MHA were resuspended in 1 mL of 0.85% NaCl and a 100 μL serially diluted aliquot. Each dilution was placed on Brain Heart Infusion (BHI) agar supplemented with sodium azide (150 μL/mL) + colistin (2 μg/mL), meropenem (2 μg/mL) + sodium azide (150 μL/mL), colistin (2 μg/mL) + meropenem (2 μg/mL) + sodium azide (150 μL/mL) for the selection of transconjugant. The conjugation, or transfer, the frequency was measured based on the ratio of the observed transconjugant CFU (T) divided by the recipient CFU (T/R).

Of the 583 isolates tested, 110 (18.9%) comprising 53 E. coli and 57 Klebsiella species were resistant to at least one of three carbapenems (meropenem, 18.7%; 109/583; ertapenem, 16.8%; 98/583; imipenem, 18.2%; 106/583) and originated from pigs (n = 14), poultry (n = 39), humans (n = 33), environment (n = 3), cattle (n = 10), camel (n = 1), and dogs (n = 10). Based on colistin resistance findings determined in a former study (Ngbede et al., 2020) and this study, 53 isolates (9.1%; 53/583) were considered resistant to both colistin and at least one of the carbapenems. Of the 53 isolates expressing concurrent carbapenem-colistin resistance, 50 were randomly selected for further investigation: 23 human isolates (16 K. pneumoniae, 2 K. quasipneumoniae, and 5 E. coli), one hospital environmental isolate (K. quasipneumoniae), and 26 animal isolates (1 K. pneumoniae, 24 E. coli, and 1 C. werkmanii). The MIC of carbapenems ranged between 2 and 32 μg/mL (Supplementary Table 1).

The resistance profile is shown in Table 1 for the Citrobacter and E. coli isolates and Table 2 for the Klebsiella isolates. The 50 carbapenem/colistin resistant strains were negative for carbapenemase production, with 98% of the isolates resistant to the β-lactam antibiotics. This particularly included amoxicillin and amoxicillin/clavulanic acid as all the isolates except one E. coli, were resistant to amoxycillin. Similarly, 48, 66, and 78% of the isolates were resistant to aminoglycosides, sulphamethaxazole/trimethoprim, and fluoroquinolones, respectively (Tables 1, 2).

Quality assessment of the genome reads revealed the Citrobacter isolate had a total genome length of 5.0 MB distributed over 60 contigs with an N₅₀ and average GC content of 48,746 bp and 52, respectively. The total genome length of the E. coli isolates ranged from 3.7 to 4.2 MB distributed over 50–1465 contigs with an N₅₀ length of 2,642–296,122 bp and an average GC content (mol%) of 50 while the Klebsiella isolates had an average GC content (mol%) of 50, total genome and N₅₀ length of 5.0–5.6 MB and 126,035–462,504 bp, respectively distributed over 24–195 contigs. The WGS data further confirmed the identity of the 50 isolates as C. werkmanii (n = 1), E. coli (n = 29), K. pneumoniae (n = 17), K. quasipneumoniae (n = 3).

The WGS data confirmed that none of the isolates harbored any known carbapenemase genes. However, we found novel deletions, including the deletion of 24 amino acids from Met1 to Val24 (M1_V24del) in the ompC gene of all the E. coli isolates (29/29) (Table 3). Previously known substitutions in the ompC gene mediating carbapenem resistance were also detected among E. coli isolates (D192G/K in 17/29 isolates; N47D in 5/29 isolates) but, no mutations associated with carbapenem resistance were detected in the ompF gene. Among the Klebsiella isolates, we detected a HYTH insertion between amino acid Met233 and Thr234 (M233_T234insHYTH) (10/20) insertion in the ompK37, substitution A217S (15/20) and N218H (6/20) in ompK36, and substitution I70M (20/20), I128M (20/20), N230G (10/20) and T261A (1/20) in ompK37 (Table 4).

Multiple substitutions were detected in the global regulator proteins marR and ramR, but not in the regulators AcrR, MarA, RamA, and SoxR. Only the M1V substitution in E. coli marR and K. pneumoniae ramR was associated with carbapenem resistance (Tables 3, 4).

Sixteen (one C. werkmanii, eight E. coli, and seven K. pneumoniae isolates) of the 50 carbapenem-colistin resistant isolates harbored the plasmid mediated colistin resistant genes: mcr-1 (n = 9), mcr-1 and +mcr-5 (n = 1), mcr-8 (n = 4), mcr-1, and +mcr-8 (n = 2) (Tables 3, 4). No point mutations associated with colistin resistance were detected in the pmrABC and phoPQ genes. Alterations in the mgrB were detected in all the isolates viz E. coli (29/29), Klebsiella (20/20), and Citrobacter (1/1) (Tables 3, 4). Novel substitutions were found in some the E. coli isolates, including V336M and T411I in the etpA and A278T, T157A, L485F, and W100M in the arnT (Table 3). The majority of the Klebsiella isolates harbored the N195S amino acid substitution (18/20), and some harbored the N141I/H (5/20) and W140L/S (5/20) substitutions in the ccrB, which were previously reported to mediate colistin resistance (Cheng et al., 2016). Finally, novel mutations as predicted to mediate colistin resistance, G164S in arnT, and the M1_L56del deletion were detected in one Klebsiella isolate (Tables 3, 4).

The 50 sequenced isolates harbored a plethora of resistant genes including β-lactamases (bla_CTX–M, bla_SHV, bla_OXA–1, bla_TEM–1, bla_CMY, bla_DHA), aminoglycosides [aac(3)-IIa, aac(6′)-Ib3, aac(3)-VIa, aph(3′)-Ia, aph(3″)-Ib, aph(6)-Ic, aph(6)-Id, armA, aadA1, aadA2b, aadA5, aac(6′)-Ib-cr], fluoroquinolones [qepA1, qepA4, oqxAB, qnrS1, qnrS13, qnrB4, qnrB17, qnrB19, qnrB34, aac(6′)-Ib-cr], sulfonamides (sul1, sul2, sul3), trimethoprim (dfrA1, dfrA10, dfrA12, dfrA14, dfrA15, dfrA17 dfrB4), phenicols (catA2, catB3, cmlA1, floR), tetracyclines [tet(A), tet(B), tet(D), tet(J), tet(M)] as well as mutations in acrR, gyrA, and parC for fluoroquinolones and ramR for tetracyclines particularly tigecycline (Tables 1, 2).

Genes bla_TEM (15/30) and bla_CTX–M (14/30) were found to be the predominant beta-lactamases among the E. coli isolates (Table 1) while bla_SHV was frequently observed among the Klebsiella isolates (15/20) (Table 2). Some isolates (19/50) harbored more than one bla in different combinations. The single Citrobacter isolate in the study harbored the only AmpC type bla detected i.e., CMY-98 genotype (Table 1). Fluoroquinolone resistance was mostly mediated by plasmid mediated quinolone resistance qnr (16/29) and chromosomal mutations (S83L, D87N) of the gyrA (15/29) in the E. coli. Amongst the Klebsiella isolates, quinolone resistance was mostly due to a mutation of the acrR gene (18/20). Seven isolates harbored the aac(6)-Ib-cr gene which confers resistance to both aminoglycosides and fluoroquinolones.

Gene tet(A) (20/29) was the most common tetracycline resistance gene carried by the E. coli isolates while tet(A) (5/20) and tet(D) (n = 5/20) were the most common among the Klebsiella species. The tet(M) commonly restricted to Gram-positive bacteria was detected in one E. coli and a K. pneumoniae isolated from a cloacal swab from poultry. Additionally, one of the Klebsiella isolates (L18) harbored the ramR mutation (A19V) responsible for tigecycline resistance.

Multilocus sequence and phylogroup typing showed that the isolates were polyclonal and genetically diverse (Supplementary Tables 2, 3). The E. coli isolates clustered within six of the eight known phylogroups: A, B1, B2, C, D, E with phylogroup B1 been the predominantly occurring (12/29) based on the Clermont typing profile. MLST assigned the isolates into 23 sequence types (STs) (and one1 unknown ST) with four of the STs occurring more than once: ST191 (n = 3), ST224 (n = 2), ST410 (n = 3), and ST2485 (n = 3) (Supplementary Table 2). cgMLST assigned them into 24 different cgSTs due to genetic differences between the two ST224 strains which were identified as cgST72904 (isolate L17) and cgST38401 (isolate L22) (Supplementary Table 3).

In silico serotyping using the WGS data assigned the isolates into 23 serotypes, O128ab/ac:H20 (n = 2). Although the “O”-group antigen of eight isolates were unknown, the isolates belonged to the H9 (n = 3), H21 (n = 1), H23 (n = 1), H30 (n = 1), H37 (n = 1), and H45 (n = 1).

Similarly, the Klebsiella isolates were distributed into 14 different multilocus sequence types (STs) with three STs occurring more than once: ST45 (n = 2), ST307 (n = 3), and ST340 (n = 3). We were however unable to infer the ST for one of the K. quasipneumoniae isolate (Supplementary Table 3). In silico serotyping of the Klebsiella isolates assigned them into 15 different serotypes with the most commonly occurring serotype being KL102:O2v2 (n = 4).

Thirty different plasmid replicon types were detected via analyses of the WGS data using PlasmidFinder (Supplementary Tables 2, 3). The majority of the isolates harbored at least two plasmid replicon types with the most abundant being the IncFIB (AP001918) in E. coli (n = 17) and IncFIB(K) in Klebsiella (n = 16). Other dominant replicon types were IncFII and IncFIA in E. coli (n = 9 each), IncFII(K) and IncR in Klebsiella (n = 11 each). No plasmids were detected in three E. coli and two Klebsiella isolates.

Although, the mcr-1 was located in a unique contig that had the same sequence as the IncX4 plasmid backbone, no other resistance genes were found on the IncX4 plasmid in this study. Other resistance genes harbored on plasmids include sul2 and aph(3″)-Ib on IncQ1, blaTEM-1B on IncX1, blaTEM-1C on IncFII(K).

A total of 42 different virulence genes were detected across the 50 isolates (Supplementary Tables 2, 3). Carriage of extra-intestinal pathogenic E. coli (ExPEC) virulence associated genes (VAGs): cva, cvi, hylF, iroN, iss, iutA, ompT, sitA, traT was widespread among the E. coli isolates (Supplementary Table 2). Pathotype specific VAGs including astA (EAEC), afaD, hlyA, ihA, fyuA (UPEC), vat (APEC), and papC (APEC/UPEC) were also detected (Supplementary Table 2).

The conjugation assay confirmed transfer of the colistin resistant determinant from two isolate (H2 and L40) to the recipient E. coli J53 with conjugation frequencies of 3.4 × 10^–1 cfu/recipient cfu and 4.6 × 10^–1 cfu/recipient cfu, respectively. None of the isolates transferred the carbapenem resistant determinant. The colistin MIC for the two transconjugants was 4 μg/mL compared with 0.5 μg/mL for the E. coli J53. Furthermore, the transconjugants expressed two resistant phenotypes (L40: AMC-AMX-CIP-CEF-DOX-GEN; H2: AMC-AMX-CEF-GEN).