Antimicrobial Resistance and Virulence Characteristics of Klebsiella pneumoniae Isolates in Kenya by Whole-Genome Sequencing

Klebsiella pneumoniae is a globally significant opportunistic pathogen causing healthcare-associated and community-acquired infections. This study examined the epidemiology and the distribution of resistance and virulence genes in clinical K. pneumoniae strains in Kenya. A total of 89 K. pneumoniae isolates were collected over six years from five counties in Kenya and were analyzed using whole-genome sequencing and bioinformatics. These isolates were obtained from community-acquired (62/89) and healthcare-associated infections (21/89), and from the hospital environment (6/89). Genetic analysis revealed the presence of blaNDM-1 and blaOXA-181 carbapenemase genes and the armA and rmtF genes known to confer pan-aminoglycoside resistance. The most abundant extended-spectrum beta-lactamase genes identified were blaCTX-M-15 (36/89), blaTEM (35/89), and blaOXA (18/89). In addition, one isolate had a mobile colistin resistance gene (mcr-8). Fluoroquinolone resistance-conferring mutations in gyrA and parC genes were also observed. The most notable virulence factors were those associated with hyper-virulence (rmpA/A2 and magA), yersiniabactin (ybt), salmochelin (iro), and aerobactin (iuc and iutA). A total of 38 distinct sequence types were identified, including known global lineages ST14, ST15, ST147, and ST307, and a regional clone ST17 implicated in regional outbreaks. In addition, this study genetically characterized two potential hypervirulent isolates and two community-acquired ST147 high-risk clones that contained carbapenemase genes, yersiniabactin, and other multidrug resistance genes. These results demonstrate that the resistome and virulome of Kenyan clinical and hospital environmental K. pneumoniae isolates are diverse. The reservoir of high-risk clones capable of spreading resistance, and virulence factors have the potential to cause unmanageable infection outbreaks with high morbidity and mortality.


Introduction
Klebsiella pneumoniae is a Gram-negative, rod-shaped ubiquitous bacterium that inhabits soil, water, and sewage ecosystems. It is also found on various human body sites and organ systems, including skin, nose, throat, and intestinal tract, as part of the natural microflora [1]. K. pneumoniae is a prominent member of the Klebsiella pneumoniae species complex that consists of seven species that include Klebsiella pneumoniae, Klebsiella quasipneumoniae subsp. quasipneumoniae, Klebsiella quasipneumoniae subsp. similipneumoniae, Klebsiella variicola subsp.

Study Samples
The study analyzed hospital environmental and clinical isolates of K. pneumoniae that were collected between May 2015 and March 2020 from eight hospitals in five counties in Kenya as part of an ongoing antimicrobial resistance surveillance study (KEMRI2767/WRAIR 2089) and an environmental study (KEMRI 3482/WRAIR 2416). The clinical samples were urine, wound swabs, and pus collected from consenting patients with suspected bacterial infections. The environmental samples were collected via swabs of high-touch areas in the participating hospitals. Klebsiella pneumoniae identification and antimicrobial susceptibility testing (AST) were performed on the Vitek2 ® system (bioMérieux, Lyon, France) using the GN-ID and XN05-AST cards. The AST panel consisted of penicillins (piperacillin and ticarcillin/clavulanic acid), monobactam (aztreonam), cephalosporins (cefuroxime, cefuroxime axetil, cefixime, ceftriaxone, and cefepime), carbapenems (meropenem), fluoroquinolones (levofloxacin and moxifloxacin), tetracyclines (tetracycline and minocycline), glycylcycline (tigecycline), phenicol (chloramphenicol), and trimethoprim. The AST results were interpreted according to CLSI guidelines (2018) [27], and isolates were classified as either multidrug-resistant (resistant to three or more drug classes) or non-multidrug resistant and ESBL positive or negative.
The whole genome sequence reads of 40 additional K. pneumoniae isolates, sequenced on an Illumina MiSeq platform at MRSN-WRAIR as previously described [19], were included.

De novo Assembly of Raw Reads and Database Querying
The fastQ files of the long reads were filtered to retain only those with a Q-score ≥ 7. The adapters in the short-pair-ended reads were trimmed using Trimmomatic v0.39 [28]. The trimmed Illumina reads were assessed for quality using FastQC v0.11.9 [29] before de-novo assembly using the default Shovill v1.1.0 [30] pipeline settings. Next, the draft assemblies were polished using pilon v1.24 [31]. Finally, the ONT long-reads were de novo assembled using flye assembler v2.8.1 [32] with the plasmid option, followed by one polishing round using medaka v1.3.2 [33]. All polished draft assemblies from Illumina and ONT sequencing were analyzed in the same way. First, the quality was assessed using QUAST [34]. Then, the draft assemblies were queried using the ABRicate v1.0.1 [35] pipeline against CARD [36] to identify AMR genes, VFDB [37] to identify virulence factors, and PlasmidFinder [38] to identify plasmids. The assemblies were queried against the Klebsiella MLST database using the command line mlst v2.19 [39] pipeline to determine the sequence types, while the capsule (K) and O types were determined using the Kleborate pipeline [40] against the Kaptive database [41]. Next, a new ybt-typing scheme updated in the Kleborate [40] pipeline was explored to assign allelic profiles to yersiniabactin genes. Finally, a maximum-likelihood phylogenetic tree was generated using Parsnp v1.2 [42] and NC_009648.1 as the reference genome. The tables were created using flextable [43] (R package), while the circular tree and heatmaps were generated using Interactive Tree of Life (iToL) v6.3.2 [44] tree annotator.
Of the four members of the K. pneumoniae Complex identified [2], K. pneumoniae subsp. pneumoniae represented the largest proportion of all isolates at 79% (70/89), with 44 recovered from SSTIs, 20 recovered from UTIs, and 6 from the hospital environment. This phylogroup had the largest number of MDR isolates (39%). The second most represented phylogroup was K. variicola subsp. variicola (18%, 17/89) with 15 isolates recovered from SSTIs and 2 isolated from UTIs. There was only 1 MDR isolate (kkp059) in this category. The least represented phylogroups were K. quasipneumoniae subsp. quasipneumoniae (2%, 2/89) and K. quasipneumoniae subsp. similipneumoniae (1%, 1/89) ( Figure 1). Among the isolates from these minor subspecies, kkp022 and kkp034 were isolated from a UTI, while kkp036, the only MDR in this category, was recovered from an SSTI. All K. quasipneumoniae subsp. quasipneumoniae, K. quasipneumoniae subsp. similipneumoniae, and K. variicola subsp. variicola, except for one isolate (kkp078), were from community-acquired infections from different geographical locations ( Figure 1). There was no evident geographical clustering of all the phylogroups by county or infection types. . The triangle symbols in the innermost ring represent the geographic source of the isolates: Nairobi (purple), Kisii (teal), Kilifi (blue), Kericho (orange), and Kisumu (brown). The star symbols represent the isolates recovered from community-acquired (blue) or healthcare-associated infections (yellow). The square symbols represent the multidrug resistance status of the isolates as either multidrug-resistant (black) or non-multidrug-resistant (white). The circle symbols in the outermost ring represent the clinical presentation of the isolates, i.e., skin and soft tissue infections (dark-blue), urinary tract infection (red), and the hospital environment (light-blue).

Genomic Characteristics of K. pneumoniae Isolates
Draft genomes were generated from the 89 isolates: 40 via Illumina short-read sequencing and 49 via MinION-based long-read sequencing. The sizes of the draft genomes Pathogens 2022, 11, 545 6 of 18 ranged from 5.2 to 5.9 Mb with an average G + C content of 57.27%, typical of K. pneumoniae genomes [45] (Table S1). The average N50 for the short and long reads was 226,402 and 5,147,285 base pairs, respectively. The isolates had 0-8 plasmid replicons, averaging 3 per isolate. The highest number of plasmid replicons were in genomes kkp001 (8), kkp018 (8), and kkp0e21 (8) (Table S2), whereas five genomes had no predicted replicons: kkp012, kkp030, kkp070, kkp102, and kkp112. The most abundant plasmid replicon types identified belonged to the Col and Inc family (particularly the IncF type) ( Figure 2). The other Inc-like plasmid replicons identified were IncR, IncH, IncX, IncC, IncN, IncM, and IncY. Seven types of Col plasmid replicons, the second most represented type, were identified, dominated by Col(pHAD28) (Figure 2). Other plasmid types identified included pKP1433 and rep_KLEB_VIR in kkp056 in kkp043 genomes, respectively.

Genomic Characteristics of K. pneumoniae Isolates
Draft genomes were generated from the 89 isolates: 40 via Illumina short-read sequencing and 49 via MinION-based long-read sequencing. The sizes of the draft genomes ranged from 5.2 to 5.9 Mb with an average G + C content of 57.27%, typical of K. pneumoniae genomes [45] (Table S1). The average N50 for the short and long reads was 226,402 and 5,147,285 base pairs, respectively. The isolates had 0-8 plasmid replicons, averaging 3 per isolate. The highest number of plasmid replicons were in genomes kkp001 (8), kkp018 (8), and kkp0e21 (8) (Table S2), whereas five genomes had no predicted replicons: kkp012, kkp030, kkp070, kkp102, and kkp112. The most abundant plasmid replicon types identified belonged to the Col and Inc family (particularly the IncF type) ( Figure 2). The other Inc-like plasmid replicons identified were IncR, IncH, IncX, IncC, IncN, IncM, and IncY. Seven types of Col plasmid replicons, the second most represented type, were identified, dominated by Col(pHAD28) (Figure 2). Other plasmid types identified included pKP1433 and rep_KLEB_VIR in kkp056 in kkp043 genomes, respectively.
No. of isolates with plasmid replicon type Plasmid replicon type

Genetic Determinants of Resistance
Comparison of the isolates' phenotypic and genotypic antimicrobial susceptibility results demonstrated high concordance in the trimethoprim, beta-lactam, and tetracycline antibiotic classes, as demonstrated in Figure 4. In particular, there was strong concordance among the K. pneumoniae isolates between the presence of a bla CTX-M gene and non-susceptibility to beta-lactams and cephalosporins (Figure 4), dfrA gene with trimethoprim non-susceptibility, and tet genes with tetracycline non-susceptibility. Most of the genomes generated from long reads enabled the detection of circularized plasmids bearing AMR genes (Table 2). MDR isolates had more plasmid replicons than the non-MDR isolates which had only 0-3 plasmid replicons (Figure 4).
Three mechanisms for quinolone resistance were detected. First, plasmid-mediated quinolone resistance (PMQR) genes (qnrB or qnrS1) were identified in 26 isolates. Second, chromosomal-encoded efflux pumps (emrR, oqxA, and oqxB) genes were constitutive in all isolates. Third, gene mutations were detected in gyrase A (gyrA)-Ser83Phe, Asp87Aspn, and Ser83Ile-and Deoxyribonucleic acid topoisomerase IV subunit A (parC)-Ser80Ile (Table S6), which are involved in DNA synthesis. A total of 29 isolates (32%, 29/89) had the tetracycline-resistance genes, tetA or tetD, while 15 isolates (17%, 15/89) carried the chloramphenicol resistance genes cat1, catII, and catB3 (encoding chloramphenicol acetyltransferases) as well as floR and cmlA (encoding chloramphenicol efflux pumps). Resistance to sulphonamide and trimethoprim, administered as co-trimoxazole, was mediated by the dfrA trimethoprim resistance gene and sul sulphonamide resistant gene, found in 54 and 55 isolates, respectively. Resistance to other drug classes was conferred by: fosfomycins-fosA; macrolides-mphA, mphE, msrE, and ereA2; rifamycins-arr2 and arr3; and even antiseptics-qacE∆1 and qacL.  Known virulence factors involved in adherence, biofilm formation, capsule synthesis regulation, mucoid phenotype regulation, immune evasion, secretion system, serum resistance, siderophores expression (enterobactin, yersiniabactin, aerobactin, and salmochelin), efflux pump expression, allantoin utilization, and enterotoxin generation were detected among the isolates ( Figure 5). The most ubiquitous were the chromosomal genes fim, mrk, and ecp for adherence and biofilm formation, which were present in all isolates except kkp043 which lacked the fim genes. In addition, some isolates carried multiple copies of the mrk genes in plasmids (Table 2). Other genes identified in all the isolates were those for serum resistance factors that determine the 'O-antigen' lipopolysaccharide serotype, the immune evasion factors which determine the polysaccharide capsule (K antigen) type, capsule synthesis regulation (rcs), efflux pump expression (acrAB), and enterobactin (ent, fep) (Table S5). In addition, all isolates possessed type VI secretion system loci genes except kkp034. Figure 5. A heatmap of the multidrug resistance status, ST, serotype, and virulence gene content of K. pneumoniae isolates (n = 89). The cladogram on the left with colored labels on the edge indicates the clustering of the isolates by phylogroups: K. pneumoniae (pink), K. quasipneumoniae subsp. similipneumoniae (blue), K. quasipneumoniae subsp. quasipneumoniae (gold), and K. variicola subsp. variicola (green). The circular symbols represent the multidrug resistance status of the isolates as either multidrug-resistant (purple) or non-multidrug-resistant (white). The serotypes assigned to the isolates are indicated as follows: multilocus sequence type (MLST) (black), capsule type (purple), and O type (teal). The virulence profile is represented as a gene present (color) or absent (white) for factors: adherence (purple), serum resistance (blue), immune evasion (pink), enterobactin (green), yersiniabactin, aerobactin, salmochelin, regulation of mucoid phenotype, enterotoxin and allantoin utilization; blank spaces indicates unassigned serotypes.
The ybt loci, identified in 24/89 isolates, encode for yersiniabactin siderophores found within conjugative transposons in the chromosome. Based on a new typing scheme [46] updated in the kleborate pipeline [40], this study identified four distinct ybt types: ybt14 found within ICEKp5, ybt15 in an ICEKp11, ybt16 in an ICEKp12, and ybt9 within an ICEKp3. In addition, the analysis revealed three isolates ( Figure 5) with chromosomally encoded genes for allantoin utilization and two isolates (kkp012 and kkp045) ( Figure 5) with magA and K2 capsule types linked to hypervirulence [47].
Several isolates were unique in having plasmid-encoded virulence genes. For example, a hypervirulent isolate (hvKP), kkp043, was identified bearing the repB_KLEB_VIR plasmid containing rmpA and rmpA2 genes, which regulate the expression of the mucoid phenotype, salmochelin (iroBCDN), and aerobactin (iucABCD, iutA). In addition, kkp083 also demonstrated an iroBDEN cluster and a heat-stable enterotoxin (astA) gene carried in an IncF(K)_1 plasmid, in contrast to the chromosomally-bound astA gene in kkp032. This study did not identify MDR hypervirulent isolates with AMR and hypervirulent genes [48]. Furthermore, the MDR isolates did not carry factors associated with hypervirulence, while the hypervirulent isolates were mostly antibiotic susceptible, i.e., they were non-MDR and ESBL-positive ( Figure 5).

Discussion
This study characterized 89 isolates based on their clinical, geographic, genotypic, and phenotypic characteristics. It was noted that all four recognized phylogroups were represented, with Klebsiella pneumoniae subsp. pneumoniae isolates predominating, consistent with findings from other studies [49][50][51]. Some differences were observed in the characteristics of the phylogroups. For example, Klebsiella variicola subsp. variicola isolates are typically linked to bloodstream infections and UTIs [52,53]; however, those identified in this study were mainly associated with SSTIs ( Figure 1) and were largely antibiotic susceptible (Figure 4). In previous studies, Klebsiella quasipneumoniae subsp. similipneumoniae isolates were linked to nosocomial infections such as UTIs [54,55]. Yet in this study, one isolate, kkp034, caused a community-acquired SSTI, potentially indicating a broader distribution of this phylogroup in Kenya than found in other countries, such as sewage in Brazil [56] and a turtle in China [57].
The lineages identified in this study were highly diverse, and most of them were local strains that have not been described in other countries. Globally disseminated high-risk lineages such as ST14, ST15, ST307, and ST607 were identified (Figure 3). These highrisk lineages are bacterial pathogens that easily acquire and disseminate antimicrobial resistance [58]. For example, ST14, ST15, and ST147 have been linked to the spread of carbapenemase resistance genes in many countries [59,60]. In this study, the two extensively drug resistant (XDR) ST147 strains carried a bla OXA-181 (kkp001) and a bla NDM gene (kkp003), respectively, while the MDR ST14/15 strains carried several ESBL genes. Notably, kkp018 (ST15) harbored a mobile colistin-resistant gene, mcr-8, and an AmpC beta-lactamase gene, bla DHA (Figure 4). ST307 and ST607 are emerging strains linked to ESBL infections (Long, et al., 2017), and they were observed in six MDR strains isolated from SSTIs and UTIs (Figure 1 and 5). ST17 is a regional strain that has been implicated in outbreaks in Kilifi [6], Mwanza [61], and Kilimanjaro [62]. This study identified three MDR and ESBL positive ST17 strains, isolated from a community-acquired SSTI, a nosocomial UTI, and a hospital environmental swab (Figures 1 and 5). High-risk clones have also been linked to specific serotypes, e.g., ST607-K25, responsible for a nosocomial outbreak at a neonatal intensive care unit in a hospital in France [63]. Significantly, this study identified three MDR and ESBL-positive ST607-K25 clones associated with SSTIs ( Figure 5).
The presence of these high-risk strains in Kenya indicates the significant clinical and public health threat they pose. However, we noticed that this threat was highest in Kisumu and Nairobi, the two largest cities in the country, where most of the global lineages (79%, 11/14) were identified. These cities are travel hubs with large referral hospitals serving patients from a wide geographical area locally and as well as global travelers.
The diverse patient population could explain the concentration of local, regional and global lineages, the great strain diversity, and novel alleles in isolates from the two counties compared to Kisii, Kilifi, and Kericho Counties.
Plasmids are the main vehicle for AMR gene transmission, and in this study, we found that a majority of the AMR genes were carried in plasmids, particularly IncFIB(K) and IncFII(pKP91) ( Table 2), which are common among Enterobacterales. The predominant plasmid replicon types belonged to the diverse Incompatibility (Inc) family [64], whose host range is mainly limited to Enterobacterales [65] which are known to have large multireplicon plasmids [66]. The multi-replicon IncF plasmids contain the FII, FIA, and/or FIB replicons (Table 2 and Table S2) which account for their high abundance ( Figure 2). Unsurprisingly, antibiotic susceptible KP isolates (kkp012, kkp030, kkp102, and kkp0112) contained few or no plasmids and were ESBL negative and non-MDR. The MDR kkp070 was exceptional because although it possessed no plasmid replicons, it had several resistance genes integrated into a genetic island in the chromosome. These integration events were not uncommon, as plasmid-associated AMR genes were detected in the chromosome of several isolates: kkp005 and kkp006 carried a bla CTX-M-15 gene, kkp109 had a dfrA14 trimethoprim resistant gene, and kkp001 possessed several AMR genes (aac(6 )-Ib9, bla CTX-M-15 , arr-2, and rmtF). The integration of AMR genes into the chromosome is alarming because the resistance is transferred clonally, becomes part of the core genome, and increases the spread and prevalence of non-susceptible K. pneumoniae lineages.
Although the col plasmid family was the second most dominant group, the majority of col-like plasmids did not contain any resistance or virulence genes, except for the abundant colE1 type, Col (pHAD28), which typically carries qnrS1 and other AMR genes [65] as observed in kkp0e21 (Table 2). In addition, col plasmids benefit K. pneumoniae because they produce bacteriocins lethal to rival bacteria. One of the other plasmid families identified was the pK1433 (kkp056) and is associated with the bla KPC2 gene [67] that provides a mechanism for carrying and spreading KPC-type genes. The isolates with plasmids carrying a bla CTX-M gene also carried bla OXA , bla SHV , or bla TEM genes, implying that these resistance genes are transferred together (e.g., ESBL-positive kkp081 and kkp059 isolates possessed bla CTX-M-15 , bla SHV-134 , and bla TEM181 genes). The most intriguing isolates were the ESBL-positive isolates (kkp015, kkp070, and kkp107) which only contained the bla SHV penicillinase gene. We hypothesize that the ESBL phenotypes in these isolates may have been due to alternative resistance mechanisms such as up-regulation of MDR efflux pumps. The expression of genes encoding MDR efflux pumps is noteworthy because their activity induces resistance of different antibiotic classes non-specifically and can cause phenotypic and genotypic discordance.
Examination of the virulence factors among the K. pneumoniae phylogroups highlighted one main difference. K. quasipneumoniae subsp. similipneumoniae did not demonstrate the Type VI protein secretion system found in the other phylogroups (Table S5); instead, it possessed the Type II secretion as previously noted [68]. Siderophores were the most significant virulence genes detected. The ubiquitous enterobactin scavenges iron from host cells; however, their activity is neutralized by human lipocalin-2 protein [69]. In response, K. pneumoniae uses the more virulent yersiniabactin to bind iron and other heavy metals such as copper to avoid metal toxicity and phagocytosis using reactive oxygen species [70]. According to Holt et al. (2015), acquiring yersiniabactin is usually the first step in accumulating more potent siderophores to make them more invasive. Yersiniabactin genes were found among the K. pneumoniae isolates causing various clinical infections and from diverse geographical locations. Most of them were MDR and ESBL positive, making their infections persistent and antibiotic-resistant and likely contributing to their dominance over the other phylogroups.
Classical hypervirulent K. pneumoniae (hvKP) isolates are generally antibiotic susceptible [71], and they are characterized by having rmpA/A2 and magA virulence factors as well as the K2 capsule type [60]. This study identified one potentially hypervirulent isolate (kkp043), which carried rmpA and rmpA2 genes. In addition, it also carried genes for aerobactin and salmochelin that were carried on a repB_KLEB_VIR plasmid ( Table 2). A trade-off between virulence and antibiotic resistance was demonstrated in this potentially hypervirulent isolate (kkp043) as it was non-MDR and ESBL-negative and had only one other plasmid bearing the IncFIA(HI1) replicon, dfrA5, and sul1 genes ( Table 2). Additional examples were the two isolates (kkp012 and kkp045) with magA genes which were ESBL-negative and non-MDR. Furthermore, non-MDR K2 strains possessed the ybt operon, which the MDR K2 strains lacked ( Figure 5). Despite this, there is growing concern about the rise of MDR hvKP globally [46,[72][73][74][75], which cause severe infections with few treatment options [76] emphasizing the significance of monitoring virulence characteristics of K. pneumoniae in Kenya.
The study had several limitations. First, the two sequencing technologies utilized had different strengths and weaknesses, which could introduce bias in the genomic analysis. The draft genomes from the long reads had fewer contigs (<= 12) (Table S1) and were more contiguous, but they had less depth/coverage, while the short reads produced less contiguous draft genomes with greater depth (>40) ( Table S1). More contiguity enabled better plasmids reconstruction, while sufficient depth enabled better assignment of multilocus sequence types. Secondly, susceptibility tests were carried out on the Vitek2 ® platform using a limited number of antibiotics and were not verified using another phenotypic method. Nevertheless, there was good concordance between the phenotypic and genotypic results. Finally, in vivo tests were not conducted to confirm the virulence gene activity, so the virulence results are only predicted.

Conclusions
The findings of this study contribute in several ways to our understanding of the genotypic and phenotypic characteristics of Kenyan K. pneumoniae isolates. The multicenter approach provided more nationally relevant data, unlike prior studies with fewer isolates from single sites. These findings describe a K. pneumoniae population with diverse sequence types, highly abundant and diverse resistance, and virulence profiles. In addition, the presence of high-risk clones in the major cities, Nairobi and Kisumu, enhances their transmissibility within and outside the country. Further research should be conducted to correlate the genotypic findings of virulence with phenotypic data, and additional analysis should investigate the genotypic environment of the acquired antimicrobial resistance genes to determine their risk of spread.