Complexity of Genomic Epidemiology of Carbapenem-Resistant Klebsiella pneumoniae Isolates in Colombia Urges the Reinforcement of Whole Genome Sequencing-Based Surveillance Programs

Abstract Background Carbapenem-resistant Klebsiella pneumoniae (CRKP) is an emerging public health problem. This study explores the specifics of CRKP epidemiology in Colombia based on whole genome sequencing (WGS) of the National Reference Laboratory at Instituto Nacional de Salud (INS)’s 2013–2017 sample collection. Methods A total of 425 CRKP isolates from 21 departments were analyzed by HiSeq-X10®Illumina high-throughput sequencing. Bioinformatic analysis was performed, primarily using the pipelines developed collaboratively by the National Institute for Health Research Global Health Research Unit (GHRU) on Genomic Surveillance of Antimicrobial Resistance (AMR), and AGROSAVIA. Results Of the 425 CRKP isolates, 91.5% were carbapenemase-producing strains. The data support a recent expansion and the endemicity of CRKP in Colombia with the circulation of 7 high-risk clones, the most frequent being CG258 (48.39% of isolates). We identified genes encoding carbapenemases blaKPC-3, blaKPC-2, blaNDM-1, blaNDM-9, blaVIM-2, blaVIM-4, and blaVIM-24, and various mobile genetic elements (MGE). The virulence of CRKP isolates was low, but colibactin (clb3) was present in 25.2% of isolates, and a hypervirulent CRKP clone (CG380) was reported for the first time in Colombia. ST258, ST512, and ST4851 were characterized by low levels of diversity in the core genome (ANI > 99.9%). Conclusions The study outlines complex CRKP epidemiology in Colombia. CG258 expanded clonally and carries specific carbapenemases in specific MGEs, while the other high-risk clones (CG147, CG307, and CG152) present a more diverse complement of carbapenemases. The specifics of the Colombian situation stress the importance of WGS-based surveillance to monitor evolutionary trends of sequence types (STs), MGE, and resistance and virulence genes.

. Other carbapenemases identified in CRKP are VIM and NDM-1 (reported since 2011 and 2012 respectively) [12,17,18]. The presence of different evolutionary mechanisms in the challenged healthcare system of Colombia, a medium-income country, may have created the conditions for a significant CRKP epidemic [5].
The present study aimed to extend the scope of previous CRKP reports in Colombia through WGS of CRKP isolates from the 21 most populated departments of Colombia, reported to the National Reference Laboratory (NRL), Instituto Nacional de Salud (INS) between 2013 and 2017. The objective was to increase our understanding of CRKP epidemiology in Colombia to support prevention and control strategies.

Bacterial Isolates
The NRL manages Colombia's antimicrobial resistance (AMR) surveillance program and adjusts the criteria for isolate submission according to the epidemiological behavior observed in the country and epidemiological alerts from the Pan American Health Organization (PAHO) (Supplementary Methods).
Between 2013 and 2017, the NRL received 811 clinical isolates of K. pneumoniae nonsusceptible to carbapenems with different sample size from the 21 departments of the country (see Figure 1). After sample selection (Supplementary Materials), 425 isolates remained that met the criteria: (1) originating from healthcare-associated infections (HAI); (2) year of isolation between 2013 and 2017; (3) phenotypic confirmation of K. pneumoniae bacterial identification by Vitek-2 (bioMérieux); (4) resistance to carbapenems determined by AST disk diffusion using CLSI guidelines for the corresponding year; and (5) genotypic confirmation of carbapenemases in K. pneumoniae.

Whole Genome Sequencing
Genomic DNA was extracted from single-colony cultures using PureLink™ Genomic DNA Mini Kit (Invitrogen, USA). The DNA integrity was checked by electrophoresis in a 1% (w/v) agarose gel, the DNA purity was estimated using NanoDrop® spectrophotometer (Nanodrop Technologies Inc., USA), and the concentration was determined with a Qubit Fluorometer (Invitrogen, USA). Subsequently, 1000 ng of DNA was shipped to the Wellcome Sanger Institute (Hinxton, UK) to sequence using Illumina HiSeq-X10 (San Diego, USA). DNA concentrations were confirmed using AccuBlue Broad Range assay (Biotium, Inc., Fremont, USA), followed by normalization and DNA library construction. DNA was sheared into 450-550 bp fragments using Covaris LC 220 ultrasonicator (Brighton, UK), followed by polymerase chain reaction (PCR)-based library preparation using Illumina adaptors and 384-indexed tags (NEBNext Ultra II FS DNA library kit). Afterward, sizeselection, amplification, purification, and multiplexing were carried out, and libraries were pooled. Pool was quantified and normalized down to 4 nM using Biomek NXP workstation for (automated liquid handling; California, USA), Agilent Bioanalyzer 2100 (California, USA), and Roche Lightcycler 480) (Utah, USA), before denaturation and loading on the Illumina platform.

Quality Control and Bioinformatic Analysis
All genome sequence data were processed using versioned Nextflow workflows and associated Docker containers covering de novo assembly, mapping-based SNP phylogeny, MLST assignment, and AMR determinant detection. The workflow structure is summarized in protocols [19]. Major quality control parameters are available in Supplementary Table 1, including file size, number of contigs, genome size, N50. AMR determinants were detected [19], and virulence factor prediction was performed using ARIBA software (github.com/sangerpathogens/ariba) in conjunction with the VFDB database (github.com/haruosuz/vfdb). MSLT prediction was assigned with ARIBA software using PubMLST databases downloaded on 20 December 2019. The isoforms of Tn4401 were determined with TETyper (github.com/aesheppard/TETyper) and annotated with Prokka v1.14.5. Additional AMR determinants, virulence factors, and typing (K-locus, wzi, and O-locus) were detected and species predicted using Kleborate and Kaptive as implemented on the PathogenWatch platform (https://pathogen.watch) [20,21].
Single-nucleotide polymorphism (SNP) variants were called relative to a reference genome (K. pneumoniae strain K2044 [NCBI RefSeq NZ_CP026011.1]) and a maximum likelihood phylogeny inferred from pseudogenome alignment of the variants using IQ_TREE with GTR+G model and 1000 bootstrap replicates [19,22]. High-quality views of trees and data were rendered using the ggtree R package (bioconductor.org/packages/ggtree). For clustering, clonal groups were defined using core genome SNP analysis using Prokka v1.14.5 gene annotation and a 70-90% ANI between isolates.

Geographic Spread of High-Risk Clones in Colombia
The 425 CRKP isolates analyzed in this study originated from 21 departments of Colombia, representing 90% of the total population ( Figure 1A). These isolates belonged to 129 institutions (114 hospitals and 15 microbiology laboratories). The level of medical services was determined for 102 institutions, of which 68, 19, and 15 provided high, medium-high, and medium levels, respectively.
All but 1 of the 7 main CGs showed increased and recent dissemination compared to previous reports in Colombia. Of CG258, ST11 was reported in 10 departments in this study, contrasting with previous reports, in which it was identified in only 2 departments [16,30]. ST512, which was previously only detected in Medellin, is now reported in 4 departments [5,16]. CG147 and CG307, which had been sporadically reported in Antioquia, were reported in 11 and 6 departments, respectively [16,26,29,31]. CG152 is reported for the first time in Colombia in 5 departments (Supplementary Figure 1A). In this study, the prevalence of CG14/15 (5.2%) is lower compared to previous reports, of 30% between 2005 and 2007 and 8.3% between 2012 and 2014 [5]. This study represents the broadest geographical analysis of CRKP in Colombia and confirms their endemic status.
Most of the STs found in Colombia in this study have been reported globally. ST258, the dominant high-risk ST worldwide, was first reported in Colombia in 2006 following global clonal propagation [15,21,24,25]. CG258 (ST258, ST11, ST512) has been identified as responsible for 68% of 112 K. pneumoniae outbreaks worldwide, reported between 2005 and 2016 [32]. The percentage of CRKP that belongs to CG258 in this study is lower than those reported in the United States (70%) by the Centers for Disease Control and Prevention (CDC), or in Israel (90%) [33][34][35]. ST11 is a prevalent clone in Asia, although its spread in Latin America had been only reported in Brazil [36][37][38][39]. ST512 is endemic in Italy, Israel, and Greece, ST147 in India, Italy, Tunisia, and Greece, and CG307 in Italy and the United States [16,29,31,[40][41][42]. The local samples behave as clones for ST258 and ST512 [35,41], whereas ST11 shows local expansions and some diverged lineages of closely related samples against other clones reported globally [39]. Because differences in fitness, epidemicity, and host-related factors between CGs have been reported, molecular identification of specific highrisk clones in a region is of the utmost importance to implement appropriate infection control programs [32].

Detection and Distribution of Antimicrobial Resistance Mechanisms and Mobile Genetic Elements
Carbapenemase genes were identified in 389 CRKP genomes; 36 isolates were non-carbapenemase producers.

bla KPC
Among carbapenemase, bla KPC was identified in 294 isolates, and bla NDM and bla VIM in 63 and 20, respectively. Coproduction of bla KPC+NDM was found in 11 isolates and co-production of bla KPC+VIM in 1 isolate ( Figure 1A). These findings confirm that, unlike in other regions of the world, bla KPC continues to dominate in Colombia, and bla OXA-48-like does not play a role in carbapenem resistance. They also indicate a recent expansion of bla NDM in Colombia.
We found 2 variants of bla KPC-bla KPC-3 (39%, n = 168) and bla KPC-2 (30%, n = 126)-across 64 different STs in multiple departments ( Figure 1A). Additionally, a wide variety of mobile genetic elements with bla KPC were identified. The most common plasmid replicons in bla KPC -positive isolates were IncFIB(K) (86.7%), IncFII(K) (56.8%), ColRNAI (53.1%), IncR (39.1%), and IncFII (32.3%) (Table 1, Figure 2C). From the KPC-carrying isolates, 68% showed the presence of the Tn4401 transposon. In the isolates carrying bla KPC-2 , the gene was located in 3 Tn4401 isoforms and in non-Tn4401 elements (NTEKPC) (n = 52) (Table1, Figure 2D). In contrast, bla KPC-3 was localized in Tn4401a in ST512 and ST1519 isolates and in Tn4401b in 82.5% of ST258. The clonal lineage ST258/512 accounts for 47% of all KPC genes reported in this study, whereas this percentage is >70% in the EuSCAPE sample collection [43]. The variant bla KPC-2 is associated with a greater variety of mobile genetic elements than in previous reports and is found in different CGs, including CG307, CG14/15, and CG147 (Table1), suggesting that its propagation is associated with successful Horizontal Gene Transfer (HGT) [5,17]. In contrast, bla KPC-3 is correlated with the presence of 2 types of Tn4401 elements in the clonal lineage ST258/512, suggesting dissemination by clonal expansion. These observations confirm and extend the suggested compartmentalization of the genes encoding for the 2 KPC variants in Colombia by Rojas et al [5]. This horizontal transfer and promiscuity of KPC-2-carrying mobile genetic elements might have appeared in Colombia by selective pressure due to increased use of carbapenems [7].

bla NDM
Two variants of bla NDM were found: bla NDM-1 (n = 60) and bla NDM-9 (n = 3). This is the first report of bla NDM-9 in the Americas. Previous local bla NDM reports were found in 6 departments (n = 15) [12]. Additionally, the first description of bla NDM-1 genetic context was described in a Bogotá outbreak (n = 6) [18]. However, it is the first time that the dissemination of bla NDM-1 is reported in 12 departments in Colombia ( Figure  1A). Also, this is the first description of bla NDM-9 prevalence and genetic context. The 2 variants were associated mainly with ST152 (33%) and ST11 (27%) (Supplementary Figure  1B). An IncC plasmid was carried by 98.4% of bla NDM isolates. Colocation of both determinants (IncC-bla NDM ) in the same contig could be confirmed in 15.9% of isolates, using Artemis (Supplementary Figure 2). The IncC replicon is common in bla NDM isolates worldwide and has been previously reported in Colombia [28,44]. Other isolates showed incomplete replicons due to short read sequencing limitations. The genetic context was identical in all isolates, including variants bla NDM-1 and bla NDM-9 (Supplementary Figure 2). A TnEc-like transposon from Tn3 family was identified in all genomic contexts with NDM genes (within 3093 bp) [28]. We conclude that the presence of NDM is associated with the expansion of the clones ST152 and ST11 and horizontal transfer of NDM-1 via the vehicle of an IncC plasmid.

bla VIM
Three bla VIM variants-bla VIM-2 (n = 13), bla VIM-4 (n = 6) and bla VIM-24 (n = 1) -were reported in 20 isolates from 7 departments, belonging to 8 STs, mainly ST147 (23.8%, n = 5), ST11 (19%, n = 4), and ST15 (19%, n = 4). The 3 bla VIM variants were found within 2 different class 1 integrons, known to be responsible for the mobilization of VIM genes (Supplementary Figure  2) [42,45]. Our results contrast with those reported globally, where bla VIM-1 is the most common variant in VIM-carrying CRKP [42]. Bla VIM-24 was reported previously in an extremely resistant CRKP 2011 case report (n = 1) [17]. The observation that bla VIM-2 is the most common allele of VIM in P. aeruginosa in Colombia suggests that HGT may have been responsible for the intra-species transfer of the gene [45]. Further work will be carried out to investigate this hypothesis and the potential directionality of the transfer.

Co-productions
The 11 isolates with co-production of bla KPC+NDM came from 3 departments and were associated with 8 STs. In the isolates co-harboring bla KPC-3+NDM-1 , it was observed that bla KPC-3 incorporated with Tn4401a exhibited a genetic environment,  including [trpR (Tn3)-ISPsy42(Tn3)-ISKpn7(IS21)-blaKPC-3-ISKpn6(IS1182)-aacA4], and that the vehicle for bla NDM-1 was similar (TnEc-like). Coproducers of bla KPC-2+NDM-1 contained the same bla NDM-1 platform (TnEc-like), but the bla KPC-2 genetic environment revealed a diversity of genomic contexts. In the isolate co-producers of bla KPC-3+VIM-24 , the bla VIM-24 gene was found associated with class-1 integron in IncC, and bla KPC-3 exhibited Tn4401b complemented with Tn2 transposases (Supplementary Figure 2). CRKPs with co-production of at least 2 genes have been previously reported worldwide, including bla KPC and bla VIM in Italy, bla KPC-2+NDM-1 in China and Brazil, and bla NDM-1+OXA-181 in Singapore [46]. The 3 coproductions reported in this study have been previously sporadically reported in Colombia [11,12,28], but this is the first report of the full replicon. Monitoring the dynamics of CRKPs with coproductions in the surveillance program is essential because of the potential risks they present, and because the way they emerge and whether they are only transient or have transmission capacity remain uncertain [43,46].

CRKP Isolates not Harboring Carbapenemases
In 36 CRKP isolates not harboring a carbapenemase gene in their genomes, the resistance to carbapenems could be attributed to the presence of extended spectrum β-lactamase (ESBL) genes (86%, n = 31), especially bla CTX-M-15 (22/31); in 24 of the 31 isolates, these ESBL genes were in combination with porin truncations (Supplementary Table 2). Porin truncations or mutations were found in 52% of all isolates. These truncations result in a nonfunctional pore that acts in concert to lower carbapenem concentrations, limiting antibiotic influx [47]. In 1 isolate, no resistance mechanisms were identified (Supplementary Table 3).
The enzymatic hydrolysis of carbapenems by carbapenemases was identified as the primary mechanism of carbapenem resistance, and the vast majority of carbapenemase producer isolates (91%, n = 385) were resistant simultaneously to all screening carbapenems (IPM, MEM, ETP) (Supplementary Figure 3). Possible explanations include the preferential use of  carbapenems in hospitals for the treatment of serious infections caused by carbapenem resistance isolates, as well as the circulation of bla KPC-2 and bla KPC-3 -harboring isolates promoting the progression of carbapenem resistance in Colombia, which is a KPC endemic country. Additional AMR determinants to aminoglycosides (n = 390), fluoroquinolones (n = 378), phenicols (n = 317), ESBL and/or AmpC (n = 248), and tetracyclines (n = 219) were observed in the majority of isolates (Supplementary Table 2).
Our results outline the increasing complexity of the CRKP population, resistance mechanisms, and lineages in Colombia, which will require adaptations of the conventional infection control measures that were successful in areas where the objective was to curb the expansion of a single dominant clone.

Detection of Virulence Mechanisms and Serotypes
In general, we observed acquired virulence mechanisms in <30% of CRKP isolates in this study, which is in line with global trends [48]. However, we additionally observed an over-representation of colibactin genes among the isolates and the first observation in Colombia of a hypervirulent CRKP isolate.
It is remarkable to note that isolates of ybt17/ colibactin (clb3) lineages represented 25.2% of all isolates, while the spread of clb3 among CRKP is usually around 10% [52]. However, most isolates ybt17+clb3+ST258 have been reported in the literature as not being colibactin producers, due to a deletion in clbJ/clbK [50,52].
The 2 cases of virulence elements reported in this study illustrate the risk of convergence of resistance and virulence, either through the acquisition of virulence plasmid(s) by CRKP or through the acquisition of resistance mobile genetic element(s) by virulent clone. The latter appears to be more difficult [52]. This emphasizes the need for genomic surveillance programs to include both resistance and virulence locus information.

Genetic Diversity Among CRKP Isolates
The phylogenetic analysis by SNP reflected the MLST distribution into STs as described above. Additionally, it highlights the genomic diversity within each ST. Of interest, we observed ST11, ST147, ST152, and ST307 with low nucleotide divergence (0.05-0.47%), suggesting clonality ( Figure 1B). However, a deeper analysis by ST showed distinct clades within each ST (Supplementary Figure 4).
To our knowledge, this is the first report of bla NDM +VIMcarrying ST11 isolates in Colombia, although they have been reported worldwide [42,44,53]. There were several distinct clades of ST11, with different serotypes, resistance genes, and plasmid replicons (Supplementary Figure 4). Importantly, ST11 holds isolates with bla KPC , bla NDM , or bla VIM . From those, bla NDM was observed alongside IncC1 and Tn552-like, although bla KPC was associated to IncFIB and Tn4401, and VIM isolates with IntI1, all 3 with no clear phylogenetic relationship (see microreact). This opens the possibility of multiple introductions or the emergence of ST11 into Colombia. After comparing with globally reported ST11 samples, isolates relate mostly to Brazil and European samples [39]. This CRKP diversity in ST11 calls for further investigation in their introduction pathways.
The ST147 showed several clades with differences in carbapenemase genes, related to Inc-type elements, discarding the clonality of the group (Supplementary Figure 4). Similarly, ST307 showed different clades some with bla KPC , bla NDM , or bla VIM , alongside IncFIB mobile elements and ESBL genes. The apparent clonality may come from the acquisition of carbapenemase genes in multiple events, related to mobile elements. This is in accordance with ST147 found related to Inc-type plasmids, and ST307 to bla OXA48-like plasmid genes, both related to carbapenemase genes HGT [43].
ST152 was found as an emerging clone with a genomic profile characterized by bla NDM , several Inc-type mobile elements and hypervirulence provided by the presence of yersiniabactin, similar to that reported in Saudi Arabia and other countries as a hypervirulent ST of importance in urinary tract infections [54]. All previous analyses enhances the importance of WGS for AMR surveillance, as differential findings between STs and genes reveal nonclonality that may be translated into differential treatment of CRKP infections.
Among this study's limitations were: having only voluntary surveillance isolates at the national level; and limited data of clinical and epidemiological information for a more exhaustive investigation. Nevertheless, the information obtained is essential for national surveillance and provides insights into pathogen genetics for planning effective surveillance.

CONCLUSION
This study allowed a detailed description of CRKP epidemiology in Colombia that should be taken into account when establishing efficient WGS-based surveillance programs in the country.
There are 7 predominant high-risk STs in Colombia. In addition to the previously described CG258 with KPC, we showed a complex, recent and significant expansion of multiple clones of interest and carbapenemases, urging reinforcement of the CRKP genomic surveillance process.
The dissemination of carbapenemase genes is based on a complex combination of mechanisms. KPC-3 occurs by the expansion of CG258, mainly ST258 and ST512 carrying an IncFII replicon plasmid with the insertion of transposon Tn4401-like. The spread of KPC-2 and VIM genes was associated with the horizontal transfer of different plasmid replicons and mobile elements. NDM-1 dissemination was associated with the expansion of clones CG152 and ST11 (CG258) and possible HGT related to IncC. The high proportion of HTG mechanisms in disseminating carbapenem resistance genes could have resulted from higher selective pressure due to overuse of antibiotics in a challenged healthcare system. This result shows the importance of including mobile genetic elements in the CRKP surveillance program in Colombia.
The hypervirulent strain report and the over-representation of the colibactin virulence factor urge close monitoring of the potential convergence of virulence and resistance of CRKP in Colombia.

Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes
The authors acknowledge all public health laboratories from the National Reference Laboratory (NRL) for providing the isolates CRKP obtained under the previous National Surveillance by the Laboratory of Antimicrobial Resistance Surveillance Network.
Financial support. This work was supported by Official Development Assistance (ODA) funding from the National Institute for Health Research [grant number 16_136_111] and the Wellcome Trust grant number 206194. This research was commissioned by the National Institute for Health Research using Official Development Assistance (ODA) funding. The views expressed in this publication are those of the authors and not necessarily those of the NHS, the National Institute for Health Research or the Department of Health. MK reports support from Centre for Genomic Pathogen Surveillance (CGPS). M. A. reports support from Wellcome Connecting Science.
Potential conflicts of interest. M. K. reports grants or contracts, support for attending meetings/travel, and receipt of equipment, materials, drugs, medical writing, gifts or other services from National Institute for Health Research (NIHR). All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed. Supplement sponsorship. The supplement is sponsored by the UK National Institute for Health Research Global Health Research Unit on Genomic Surveillance of AMR.