Clinical samples were collected during a 22-month period (May 2010–February 2012) from The Children’s Hospital & The Institute of Child Health, Lahore (Lahore, Pakistan), the largest tertiary care hospital in the region (Figure 1, panel A). The hospital had a capacity of 650 beds during the study period but is under pressure to handle up to 2,000 inpatients at any given time. The primary catchment area is Lahore (population ≈10 million); the hospital also receives patients from the greater area of Punjab province (population ≈100 million) (Figure 1, panel A). The Ethical Committee of The Children’s Hospital & Institute of Child Health, Lahore, approved the study.

A total of 44,260 samples were collected in the course of routine sampling from children; 5,475 (12.4%) resulted in laboratory-positive cultures. Of these, 710 (13.0%) samples were positively identified as K. pneumoniae, the third most dominant isolate after Escherichia coli (1,336 [24.4%]) and coagulase-negative staphylococci (724 [13.2%]) (Figure 1, panel B). We screened all K. pneumoniae isolates for resistance to ceftazidime (30 μg disc, zone of inhibition <17 mm) or cefotaxime (30 μg disc, zone of inhibition <22 mm). We further tested K. pneumoniae isolates that were resistant to any of these indicator drugs using the Clinical and Laboratory Standards Institute combined-disc confirmatory test (6); extended-spectrum β-lactamase (ESBL) production was confirmed when the zone of inhibition by either cephalosporin drug increased by >5 mm in the presence of clavulanate. A total of 214 of K. pneumoniae isolates were ESBL-positive (Figure 1, panel C); most were isolated from children with bloodstream infections (Figure 1, panel D). The outcomes were severe, especially among neonatal patients (Figure 1, panel D); 56 died, 31 were taken home against medical advice, and 127 were discharged (Figure 1, panel D). Almost all patients infected with ESBL Klebsiella had received an intravenous line (209 [97.7%]) (Figure 1, panel E,F), and a high number received a urinary catheter (46 [21.5%]).

We performed whole-genome sequencing on 178 isolates (Technical Appendix Table 1). We prepared Illumina sequencing libraries (Illumina, San Diego, CA, USA) with a 450-bp insert size according to the manufacturer’s protocols and sequenced them on an Illumina HiSeq2000 with 100-bp–long paired-end reads before assembly using an open-source high-throughput assembly and improvement pipeline as described (7) (https://github.com/sanger-pathogens/) and annotated using prokka (8). Initial clustering using mash (9) enabled aligning of these isolates to published reference sequences (Technical Appendix Figure 1, panel A). The clustering indicated a strong structure for the isolates that fell within the species K. pneumoniae (Technical Appendix Figure 1, panel B). However, the analysis also revealed a large group of sequences most similar to K. quasipneumoniae; closer inspection focusing on this species showed strongest similarity to subspecies similipneumoniae (Technical Appendix Figure 1, panel C) (10). We combined several independent datasets: a large global collection (1); 2 hospital outbreaks obtained in a comparable time frame, 1 of which was based in Nepal in 2012 (3); and a hospital study from Spain that also focused on diversity within ESBL-producing strains (11) (Technical Appendix Table 2). We applied the pan-genome pipeline Roary version 3.7.0 (12) with a blastp (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE=Proteins) percentage identity of 90% and a core definition of 99%, resulting in a core gene alignment comprising 1,793 genes for all studies (Figure 2) and 3,486 genes for the strains of this study (Technical Appendix Figure 2). We first extracted single-nucleotide polymorphisms using snp-sites version 2.3.2 (13), then calculated a maximum-likelihood tree using RAxML version 8.2.8 (14) with the general time-reversible model and 100 bootstrap repeats. The core gene phylogeny (Figure 2) shows a wide distribution of the isolates from Pakistan across different lineages rather than 1 clonal lineage. The diversity of our strain collection is further emphasized through the diversity of multilocus sequence types (STs). No single ST dominates (Figure 2 outer ring; Technical Appendix Figure 2); however, a large group of isolates belongs to ST15, which is known to be problematic. The presence of K. quasipneumoniae isolates agrees with an overall lower percentage of reads mapped against K. pneumoniae (Technical Appendix Table 1) and with recent descriptions of virulent K. quasipneumoniae strains (1,9,15). Assessing the metadata in phylogenetic context highlights the association of the K. quasipneumoniae lineage with patients in the neonatal ward, suggestive of its nosocomial residency (Technical Appendix Figure 2). However, other main lineages (e.g., ST15, ST48) show a dynamic spread across wards and age groups, indicating against >1 resident lineages but instead a frequent movement of K. pneumoniae through the hospital, general population, or both.

The high number of K. quasipneumoniae isolates, even if potentially restricted to most sequences derived from a lineage potentially resident in a specific ward, highlights the importance of a diverse set of sampling sites to be studied. It also highlights the need for continued monitoring of new emerging strains and that our knowledge of the diversity of potentially problematic lineages is far from exhaustive.

The Klebsiella isolates in this study represented the Klebsiella isolates routinely present in infections over a protracted period. Our findings highlight a consistent problem with ESBL-encoding strains belonging to a multitude of lineages. We observed sporadic single-isolate lineages, as well as smaller, related clusters of 5–10 strains per lineage, in addition to 2 larger clusters of strains. More studies are needed to better delineate the distinguishing features for successful spread and persistence of lineages such as the ST15 cluster. Also, the large spread of K. quasipneumoniae is unusual. Further intense monitoring of LMIC hospital environments is urgently needed to prevent the persistence of resident lineages with very high base-level drug resistance, which, through the inevitable acquisition of a few more genes, would lead to untreatable infections.

Details of sequenced Klebsiella pneumoniae strains from children, Pakistan; GenBank accession numbers of published K. pneumoniae strains included in this analysis; whole-genome clustering; patient metadata in phylogenetic context.