Geographical distribution, disease association and diversity of Klebsiella pneumoniae K/L and O antigens in India: roadmap for vaccine development

Abstract

Klebsiella pneumoniae poses a significant healthcare challenge due to its multidrug resistance and diverse serotype landscape. This study aimed to explore the serotype diversity of 1072 K. pneumoniae and its association with geographical distribution, disease severity and antimicrobial/virulence patterns in India. Whole-genome sequencing was performed on the Illumina platform, and genomic analysis was carried out using the Kleborate tool. The analysis revealed a total of 78 different KL types, among which KL64 (n=274/1072, 26 %), KL51 (n=249/1072, 24 %), and KL2 (n=88/1072, 8 %) were the most prevalent. In contrast, only 13 distinct O types were identified, with O1/O2v1 (n=471/1072, 44 %), O1/O2v2 (n=353/1072, 33 %), and OL101 (n=66/1072, 6 %) being the predominant serotypes. The study identified 114 different sequence types (STs) with varying serotypes, with ST231 being the most predominant. O serotypes were strongly linked with STs, with O1/O2v1 predominantly associated with ST231. Simpson’s diversity index and Fisher’s exact test revealed higher serotype diversity in the north and east regions, along with intriguing associations between specific serotypes and resistance profiles. No significant association between KL or O types and disease severity was observed. Furthermore, we found the specific association of virulence factors yersiniabactin and aerobactin (P<0.05) with KL types but no association with O antigen types (P>0.05). Conventionally described hypervirulent clones (i.e. KL1 and KL2) in India lacked typical virulent markers (i.e. aerobactin), contrasting with other regional serotypes (KL51). The cumulative distribution of KL and O serotypes suggests that future vaccines may have to include either ~20 KL or four O types to cover >85 % of the carbapenemase-producing Indian K. pneumoniae population. The results highlight the necessity for comprehensive strategies to manage the diverse landscape of K. pneumoniae strains across different regions in India. Understanding regional serotype dynamics is pivotal for targeted surveillance, interventions, and tailored vaccine strategies to tackle the diverse landscape of K. pneumoniae infections across India. This article contains data hosted by Microreact.

Impact Statement.

Klebsiella pneumoniae produces polysaccharide capsules, which serve as both epidemiological markers and significant virulence factors. The increasing accessibility of whole genome sequencing has made it easier than ever to investigate this capsule diversity. This study is the first of its kind in India to comprehensively investigate the serotype diversity of K. pneumoniae strains and their association with disease severity, antimicrobial resistance/virulence patterns, and geographical distribution across various regions of the subcontinent. This multi-dimensional analysis not only provides valuable insights into the molecular epidemiology of K. pneumoniae in India but also offers crucial data for developing targeted interventions, including vaccine formulations tailored to address the prevailing serotypes. These findings serve as a foundation for informed decision-making in the management and prevention of infections, ultimately contributing to improved public health outcomes in the region.

Data Summary

• All the sequenced data has been submitted to the European Nucleotide Archive (ENA) under the Bioproject numbers PRJEB29740 and PRJEB50614. Run Accessions and Biosample numbers are provided in Supplementary Table 1 with corresponding metadata for each sample used in the study.

• The Microreact link for the genomic analysis is provided.

  (https://microreact.org/project/oqKM84GBszEPW9Emt2FKnP-klebsiella-pneumoniae-indian-serotypes).

• The pipelines used in the study are published in gitlab (https://gitlab.com/cgps/ghru/pipelines).

• The tools’ details and the implementation of the pipelines are described in protocols.io (https://www.protocols.io/view/ghru-genomic-surveillance-of-antimicrobial-resista-bp2l6b11kgqe/v4).

• The R scripts used with all the input files used for each script have been published in Figshare (https://doi.org/10.6084/m9.figshare.25414807.v2).

• The isolates also have been deposited in Pathogenwatch and the collection is available here: https://pathogen.watch/collection/845w6lse0ezf-k-pneumoniae-1072-wgs-study-nihr-ghru-india

Introduction

Klebsiella pneumoniae is a Gram-negative bacterium that is a major cause of nosocomial infections [1], including pneumonia, sepsis, and urinary tract infections [2,3]. Its increasing resistance to antimicrobials poses a serious public health threat [4]. Genome-based surveillance is urgently needed to control the emerging threat of K. pneumoniae [5]. Recent advances in understanding the population structure of K. pneumoniae have revealed an immense genomic diversity, providing a framework for pathogen tracking [6,7]. The emergence of multidrug-resistant (MDR) K. pneumoniae strains, resistant to multiple antimicrobials, is a significant concern, especially in countries like India, where treatment becomes challenging [8,11]. Hypervirulent K. pneumoniae (hvKp) strains represent a distinct subset of K. pneumoniae characterised by their heightened virulence potential compared to classical strains. These hypervirulent variants typically harbour a specific combination of virulence factors, such as specific capsular polysaccharides (KL1 and KL2 serotypes), siderophores (e.g. aerobactin), and regulators of mucoid phenotype A (rmpA), enabling them to cause severe infections, often in otherwise healthy individuals [4]. The hvKp strains have been predominantly reported in East Asian countries, particularly Taiwan [12], China [13], and South Korea [14], where it has emerged as a significant public health concern due to its association with severe infections and increased mortality rates. In these regions, hvKp is frequently implicated in community-acquired infections, including liver abscesses and invasive syndromes. Reports from India have highlighted the presence of hvKP, particularly in nosocomial settings, exhibiting drug resistance. However, these studies have primarily relied on phenotypic and biochemical methods rather than genomic analysis [15,16]. There are also reports of MDR hvKp strains which have been emerging, spawning a new generation of hypervirulent ‘superbugs’ [17,18]. Newer therapeutics, such as vaccines or monoclonal antibodies, are needed to combat MDR and hvKp strains. Vaccines could help to prevent infections from occurring in the first place and/or they could help to reduce the severity of infections in affected individuals. Human monoclonal antibodies, on the other hand, can rapidly progress to innovative prophylactic and therapeutic solutions [19].

Capsular polysaccharides (CPS, encoded by the K locus) and lipopolysaccharides (LPS, encoded by the O locus) are major virulence factors of K. pneumoniae, and they are responsible for protecting the bacterium from the host’s immune system. Bacterial capsule-targeted vaccines, such as those designed for Streptococcus pneumoniae, Neisseria meningitidis, and Haemophilus influenzae, have demonstrated significant effectiveness in preventing illnesses caused by these encapsulated pathogens [20]. Currently, efforts are in place to develop vaccines or monoclonal antibodies against K. pneumoniae. The increased immunogenicity and enhanced surface exposure of CPS and LPS in K. pneumoniae render them appealing candidates [20,22]. A bioconjugation approach based on glycoengineered Escherichia coli expressing K. pneumoniae KL1 and KL2 antigens led to the production of IgG against both glycans in mice conferring protection against lethal challenges with KL1 and KL2 strains [23]. Affinivax is working on an LPS-based formulation that combines antigens O1, O2, O3, and O5 with the type III fimbriae adhesion MrkA to generate a multiple-antigen presenting system (MAPS) [24]. However, the wide range of CPS and LPS types makes it challenging to provide comprehensive coverage [25]. The structural variability, defective capsule or O-antigen production, and variations in the geographic distribution of serotypes limit the potential coverage of CPS/LPS-based vaccines or monoclonal antibodies [26].

There are over 100 capsular serotypes (encoded by the K loci) of K. pneumoniae [27,28], and the prevalence of these serotypes varies from region to region. While only the first 80 K-loci have serotype data associated with them, subsequent loci rely solely on genomic sequence information [29]. Understanding the distribution of K. pneumoniae serotypes is important for vaccine and monoclonal antibody development, as a vaccine targeting the most prevalent serotypes in a given region is more likely to be effective. Among the various capsular (KL) types, KL1 and KL2 are often linked to high virulence, and more concerningly, isolates with KL47 and KL64 are often linked to both hypervirulence and carbapenem resistance [30], which present significant challenges for antimicrobial therapy [31,32]. These isolates are referred to as hypervirulent carbapenem-resistant K. pneumoniae (hv-CRKP) emphasising the urgent need to design and develop broad-spectrum therapeutic drugs or vaccines against K. pneumoniae isolates of serotypes KL1, KL2, KL47, and KL64 [33]. In India, there are few studies with a limited number of samples. These have reported KL51 and KL64 as the most common K. pneumoniae serotypes, which are associated with specific sequence types (STs), ST231-KL51 and ST147-KL64 [34,35]. The frequency of other KL types varies in different studies based on the sample sizes. The globally prevalent, hypervirulent serotypes KL1, KL2, and KL20 were rarely found in these Indian studies.

The O-antigen is a significant virulence factor for K. pneumoniae. It aids the bacterium in evading the host’s immune system and attaching to host cells. The various O-antigen serogroups have different antigenic properties, which can affect the bacterium’s ability to cause disease. The O-antigen of K. pneumoniae can be divided into different groups based on their unique structures and antigenic properties. K. pneumoniae has eleven characterised LPS serogroups. Just four serogroups: O1, O2a, O3, and O5 are expressed by over 80 % of all isolates [20,22]. In India, O1 and O2 are the most prevalent types, collectively constituting over 70 % of the isolates, as indicated in prior studies [10,34, 36]. However, it is worth mentioning that the number of studies conducted in the Indian setting is limited and these studies represent smaller-scale investigations with relatively modest sample sizes. For example, Sundaresan et al. (2022) analysed a dataset comprising only 153 isolates from the PATRIC database submitted from India [35]. Similarly, Wyres et al. (2020) included 102 isolates from South Asia (including India and Nepal) in their study [36]. However, such sample sizes may not fully capture the true diversity of K. pneumoniae in India. This underscores the necessity for more extensive and comprehensive research efforts in this context to provide a more representative understanding of the genomic landscape of K. pneumoniae in India.

Understanding the prevalence of K. pneumoniae serotypes in India will help the design of an effective vaccine that covers the main circulating lineages in the country. For that reason, in this study, we aim to identify the K and O loci of a large collection of country-wide K. pneumoniae strains causing infection in India. Furthermore, our study explores the relationship between these serotypes and disease severity, geographic distribution, and virulence and antimicrobial resistance (AMR) characteristics, which adds to our understanding of their intricate interactions in the Indian setting. This multi-pronged approach will help India to develop effective and specialised vaccines against K. pneumoniae.

Methods

Bacterial isolates and phenotypic characterization

The bacterial isolates used for this study comprised 1072 putative K. pneumoniae isolates primarily sourced from hospital infections and obtained from the years 2014 to 2022 across India. This includes 307 isolates from our previous study [34]. All sequenced isolates in our study represent unique patient infection episodes, with only one isolate per patient. The phenotypic characterization was done at the Central Research Laboratory, Kempegowda Institute of Medical Sciences (KIMS) using the VITEK 2 (bioMérieux, Marcy-l'Étoile, France) compact system. The ethical approval for the study was obtained from the KIMS ethical committee with the study number KIMS/IEC/27/2017. The strain details are provided in Table S1, available in the online version of this article. The isolates were classified into two categories: invasive and non-invasive based on the site of collection and specimen source from which they were isolated. Specimens collected from normally sterile body sites, such as blood, cerebrospinal fluid (CSF), and body fluids, etc. were categorised as invasive. Conversely, specimens obtained from non-sterile sites, including sputum, urine, wounds, etc. were considered non-invasive as described by Pacoza et al. [37] based on the impact of site-specific isolation in K. pneumoniae epidemiology and pathogenesis.

Sequencing and genomic analyses

Whole-genome sequencing, assembly, and annotation

Genomic DNA was extracted and isolated from the bacterial isolates using the QIAamp DNA mini kit (Qiagen, Hilden, Germany) and quantified using the Qubit double-stranded DNA kit (ThermoScientific, Massachusetts, United States) as instructed by the manufacturer. Double-stranded DNA libraries with 450 bp insert size were prepared using the ultraFS-II kit (New England Biolabs, London, United Kingdom). The quality control (QC) check for the prepared libraries was done on an Agilent Tapestation (Santa Clara, California, USA) and libraries were sequenced on the Illumina MiSeq platform (Illumina, San Diego, California, United States of America) with paired-end reads of 250 bp length. All the Whole Genome Sequence (WGS) data generated as a part of this study were submitted to the European Nucleotide Archive (ENA) under the Bioproject numbers PRJEB29740 and PRJEB50614 with accession IDs provided in Table S1.

The bioinformatic analysis was conducted using Nextflow pipelines created as part of the Genomic Surveillance of Antimicrobial Resistance-AMR project available at protocols.io [38]. The pipeline performs trimming, and read correction and merges the reads before assembly using SPAdes assembler v3.12 [39]. Quality control of sequence data was evaluated for the following parameters: (i) the basic statistics of raw reads, (ii) the assembly statistics, (iii) contamination due to single nucleotide variants (SNV) and sequences from different species, (iv) species prediction using Bactinspector v0.1.3 and (v) overall QC as pass, warning or fail for each isolate based on these different parameters as described in the pipeline. All the quality metrics were combined using Multiqc v1.7 and Qualifyr v1.4.4 to generate web-based reports [40]. The contigs were considered as the final assemblies and were annotated with Prokka v1.5 [41].

Insilico genomic characterization

Kleborate v2.3.2 (https://github.com/katholt/Kleborate) is a designated genotyping tool developed for Klebsiella spp. It integrates multiple analysis steps, including multi-locus sequence typing (MLST), and identification of virulence and acquired resistance genes, to provide a comprehensive genotypic profile of the isolates [42]. KL and O antigens were identified using Kaptive [29,43]. Kleborate also generates categorical scores for both virulence and resistance, for each sample and the scores are calculated based on the criteria as described in the framework provided by Lam et al. [42] and is also provided on the GitHub page of the Kleborate tool here https://github.com/klebgenomics/Kleborate/wiki/Scores-and-counts. The raw output of the Kleborate tool has been provided in Table S2.

Variant detection and phylogenetic analysis

Genome mapping of the 1072 isolates to the reference genome of K. pneumoniae (strain NTUH-K2044, GCF_009497695.1) was done using the GHRU-SNP phylogeny pipeline v1.2.2 (https://gitlab.com/cgps/ghru/pipelines/snp_phylogeny). The mobile genetic elements (MGEs) were masked in the pseudo genome alignment using MGEmasker [44], and the recombinant regions of the genome were removed using the Gubbins algorithm v2.0.0 [45]. A maximum-likelihood tree was built utilising the non-recombinant SNPs using IQ-tree [46] with 100 bootstrap replicates and parameters -czb to collapse near-zero branches, and a general time-reversible (GTR) model. Phylogeographic analysis and visualisation were performed on Microreact and the project link is https://microreact.org/project/oqKM84GBszEPW9Emt2FKnP-klebsiella-pneumoniae-indian-serotypes#r769-overallview [47].

Statistical analysis and plots

We used the Rstudio server v2023.03.0 build 386 for descriptive and statistical analysis. Plots were generated using the ggplot package [48] and for data interpretation, the dplyr [49] and tidyr [50] packages were used. We performed the Kruskal-Wallis rank sum test which is a non-parametric test used to determine if there are statistically significant differences between three or more independent groups to evaluate whether there are differences in virulence scores across different categories of K locus, O antigen and invasiveness of the isolates. Fisher’s exact test was used to test the association between virulence factors and antibiotic resistance genes. All the code used in the study is published on figshare and the link is provided (https://doi.org/10.6084/m9.figshare.25414807.v2).

Assembly de-replication

In order to evaluate the impact of clonality on the overall results, we de-replicated the genomes in a separate analysis (see Supplementary Material 1). All analyses in this de-replicated dataset were performed as originally described in the manuscript.

Results

Overview of the collection

The collection included 1072 isolates from 38 different hospitals located in 19 different states across India from 60 % (646/1072) male and 40 % (426/1072) female population. Of the 1072 isolates, 65 (1 %) isolates were obtained from children <24 months of age, and overall patient age ranged from 1 to 96 years with a median age of 48 years. The geographical representation and the timeline of the samples are shown in Fig. S1 and also provided in Table S1.

Among the invasive specimen types, blood samples emerged as the predominant source, yielding 217 K. pneumoniae isolates followed by endotracheal aspirates (ETA) with 147 isolates and body fluids with 50 isolates. In the non-invasive specimen types, urine samples were predominant, with 257 isolates. Pus specimens were followed by 154 isolates and sputum samples with 140 isolates. Ninety percent of the isolates came from just six specimen types suggesting that they are less commonly found in other specimen types. The complete distribution of different specimens is represented in Fig. 1.

Fig. 1. Distribution of specimen types, classified as invasive and non-invasive, from which Klebsiella pneumoniae was recovered. Abbreviations: ETA, Endotracheal Aspirate; BAL, Bronchoalveolar Lavage; CSF, Cerebrospinal Fluid.

K and O loci diversity

In our collection of 1072 genomes, we identified 78 distinct KL. The most prevalent K-loci were KL64 (25 %, 274/1072), KL51 (23 %, 249/1072), KL2 (8 %, 89/1072), and KL10 (5 %, 51/1072). The top three K-loci contributed to 57 % of the isolates. The top 11 major KL types are provided in Table 1 and the complete distribution of each K-loci is provided in Table S3. The major KL types in both invasive and non-invasive specimen sources were similar, i.e. KL64 was a major KL in invasive sources (n=122) and also in non-invasive sources (n=152) (Fig. 2) (Table S3). Hence, we assessed the association between KL sample types (invasive and non-invasive) using Fisher’s exact test [51]. The most prevalent K-loci, namely KL64 (P-value: 0.441), KL51 (P-value: 0.469), and KL2 (P-value: 0.266) were not statistically associated either with invasive or non-invasive disease. Only KL10 was significantly associated with non-invasive disease (P-value: 0.041), indicating a potential link.

Table 1. KL distribution among invasive and non-invasive specimen types. The top 11 KL of frequency >15 are shown individually and the lesser predominant KL <15 counts are grouped. The top ten STs are shown individually and others are grouped as ST_other.

K-locus	Invasive	Non- Invasive	STs (n=count)	Total
KL64	122	152	ST231 (n=9), ST147 (n=103), ST395 (n=90), ST14 (n=6), ST_other (n=13), ST2096 (n=53)	274
KL51	121	128	ST231 (n=213), ST147 (n=20), ST_other (n=3), ST16 (n=13)	249
KL2	46	43	ST14 (n=55), ST_other (n=17), ST101 (n=2), ST11 (n=5), ST15 (n=9)	89
KL10	16	35	ST147 (n=47), ST_other (n=4)	51
KL36	22	20	ST_other (n=1), ST437 (n=41)	42
KL81	17	20	ST147 (n=2), ST_other (n=5), ST11 (n=1), ST16 (n=29), ST11 (n=1)	37
KL17	8	14	ST_other (n=2), ST101-1LV (n=11), ST101 (n=9)	22
KL112	8	13	ST15 (n=21)	21
KL52	8	9	ST_other (n=1), ST38 (n=11), ST437 (n=6)	18
KL62	10	7	ST_other (n=6), ST48 (n=11)	17
KL102	9	7	ST_other (n=2), ST307 (n=14)	16
KL Other(count <15)	111	125	ST231 (n=4), ST147 (n=7), ST395 (n=4), ST14 (n=1), ST_other (n=154), ST101 (n=1), ST1710 (n=12), ST23 (n=12), ST469 (n=12), ST307 (n=1), ST11 (n=14), ST15 (n=8), ST16 (n=7)	236

STSequence type

Fig. 2. Distribution of KL and O types among invasive and non-invasive specimens. The bars are stratified by the number of samples in each KL and O type. The bars are stacked to 100 %.

The theoretical coverage provided by multi-valent vaccines targeting increasing numbers of KL (ordered by KL frequency in the population) is shown in Fig. 3a. The diversity of KL types within our sample collection was assessed using Simpson’s Diversity Index (D). This index gives a numerical representation of the diversity of KL in the population, ranging from 0 (no diversity) to 1 (highest diversity). The KL were very diverse in age groups <2 years (d-value: 0.912), indicating that there is no specific KL associated with paediatric infections. The diversity among other age groups was also very high and the diversity calculation for each age group is provided in Table S4.

Fig. 3. The cumulative coverage of the KL (a) and the O (b) types represents the percentage of isolates covered. The KL and O types are sorted with the highest to lowest frequency. The orange dotted line shows cumulative carbapenemase-positive strains.

Our study assessed the prevalence and associations of different O types with disease. Our collection had a diverse distribution of O antigens with 13 different O types. The most prevalent O-loci were O1/O2v1 (44 %, 471/1072) and O1/O2v2 (33 %, 354/1072), OL101 (7 %, 67/1072), which were collectively the top three O-loci constituting 83 % (892/1072) of the collection. Other O-loci, such as O3b (52/1072), O4 (52/1072), and O3/O3a (49/1072), were also identified as given in Table 2. In the assessment of associations with sample sources, a Fisher’s exact test [51] for O1/O2v1 (P-value: 0.388), O1/O2v2 (P-value: 0.696), and O101 (P-value: 0.445) revealed no significant associations, implying a similar distribution between invasive and non-invasive isolates. Serotype distributions of KL and O types showed no significant differences by age group, as confirmed by Fisher’s exact test (P-value: 0.424). The cumulative coverage of O types shows that if five types are incorporated in a vaccine formulation, it will cover 90 % of the isolates from different specimen types and disease conditions (Fig. 3b).

Table 2. O types distribution among the invasive and non-invasive isolates and their associated STs. The top ten STs are shown individually and others are grouped as ST_other.

O antigen	Invasive	Non invasive	STs	Grand total
O1/O2v1	212	259	ST231 (n=9), ST147 (n=105), ST14 (n=62), ST395 (n=90), ST_other (n=69), ST101 (n=9), ST101-1LV (n=11), ST11 (n=9), ST15 (n=37), ST16 (n=6), ST2096 (n=53), ST48 (n=11)	471
O1/O2v2	162	192	ST231 (n=217), ST147 (n=21), ST395 (n=4), ST_other (n=79), ST101 (n=3), ST11 (n=4), ST23 (n=12), ST307 (n=14)	354
OL101	35	32	ST147 (n=4), ST_other (n=10), ST11 (n=5), ST16 (n=30), ST307 (n=1), ST38 (n=11), ST437 (n=6)	67
O3b	28	24	ST147 (n=3), ST_other (n=24), ST16 (n=13), ST1710 (n=3), ST469 (n=12)	52
O4	27	25	ST_other (n=9), ST11 (n=2), ST437 (n=41)	52
O3/O3a	16	33	ST147 (n=45), ST_other (n=4)	49
OL104	8	4	ST1710 (n=12)	12
O5	4	2	ST_other (n=6)	6
O12	4		ST_other (n=4)	4
OL103	1	1	ST_other (n=1), ST15 (n=1)	2
O1/O2v3	1		ST_other (n=1)	1
Unknown (O1/O2v1)		1	ST_other (n=1)	1
Unknown (O3/O3a)	1		ST147 (n=1)	1

STSequence type

Sequence types and their serotypes association

The 1072 genomes of K. pneumoniae encompass 114 different STs, some persisting for several years. Most of them have been reported globally while others are mainly restricted to Southeast Asia, such as ST231 (n=226), ST147 (n=179), ST395 (n=94) and ST14 (n=62) (Table S1). Analysis of the KL types revealed a distinct pattern of variation, where some STs were associated with single K-locus types and others had multiple K-locus types (Fig. 4). For example, within ST231, KL51 is the dominant allele while KL64 is present in a small subset. The second most prevalent ST, ST147, also exhibited remarkable K-locus heterogeneity, hosting KL64 (n=103), KL51 (n=20), and KL10 (n=47) alleles and three additional K-loci, highlighting the intricate diversity within this clonal group. On the other hand, ST395 exclusively harboured KL64, while ST14 exhibited a strong association with KL2 (P<0.05) as per Fisher’s exact test. This suggests lineage-specific selective pressures or functional constraints shaping KL diversification within the bacterial population.

Fig. 4. The circular view of the SNP tree was constructed using 100 bootstrap replicates. The tree is midpoint rooted and the scale bar represents the SNPs per variable site. The inner ring represents the K-loci types and the outer ring represents the O antigen types. The nodes are coloured by ST.

O serotypes were strongly linked with STs. O1/O2v2 was seen in 220 isolates of ST231 and 21 of the ST147 isolates but ST147 also carried clade-specific associations with O1/O2v1 and O3/O3a. ST395, ST14, ST15, ST2096 and ST101 only carried O1/O2V1. ST16 was associated with three different O-types namely O101, O1/O2v1 and O3b. Other minor STs had varied O-types but some were associated with single O-types. This pattern of O-types suggests ancient independent acquisition of O-types with subsequent clonal spread. The distribution of K- and O-loci among different STs is shown in Fig. 4 and also in the microreact views (microreact.org/klocusvsst and microreact.org/olocusvsst).

Geographical and temporal distribution of KL and O antigens across India

We also examined the geographical distribution of KL and O antigens across India, revealing the diversity of serotype prevalence. The distribution of KL and O antigens was heterogeneous, with some loci being more prevalent in particular parts of the nation than others. The ST147 clone with KL64 and O1/O2V1 locus and the ST395 clone with KL64 and O1/O2V1 locus were the two most prevalent types in the Northern part of the subcontinent. Twelve isolates of ST1710 having KL169 and O104 or O3B were seen in only one sentinel site from Gujarat. From the Simpson’s diversity index, the calculated overall D value of 0.866 indicates a relatively high level of diversity among KL types. On the other hand, the regional variation in the K loci diversity is quite remarkable. Table S5 reveals the Simpson’s Diversity Index for six major political zones across India. We observed that specific zones exhibited a higher degree of diversity in KL types compared to the others. Notably, the Southern and Central zones displayed the highest diversity in K loci, with Simpson’s Diversity Index values of 0.914 and 0.892, respectively. The diversity was relatively lower in the Northern and Western zones (0.758 and 0.794, respectively) (Table S5). The major KL types KL64 and KL51 and O types O1/O2v1 and O1/2v2 were distributed across all the zones and the sentinel sites.

The analysis of temporal trends within our dataset revealed that the predominant STs and KL types remained remarkably stable over the study period and have consistently maintained their prevalence (Figs S2 and S3) (microreact.org/ST_overtime, microreact.org/KL_overtime). Interestingly, we identified newer STs and KL types within each sentinel site over time. Notably, these new KL types were not observed in previous years, indicating dynamic changes in the genomic landscape of K. pneumoniae within the hospital environments. Within individual sentinel sites, we observed an increase in diversity from the beginning to the end of the study period. Despite this expansion, the predominant STs and KL types within each site remained stable throughout the study, underscoring emerging variants' co-existing with established strains (Fig. S4) (microreact.org/KL_within_hospital). The O antigens remained consistent over time and were consistent across all hospitals (microreact.org/Otype_vs_timehttps://microreact.org/project/oqKM84GBszEPW9Emt2FKnP-klebsiella-pneumoniae-indian-serotypes#pwvi-ol-vs-time, microreact.org/Otypevshospital).

Virulence factors associated with capsular (KL) and O antigen serotypes

We screened for all six known virulence factors in K. pneumoniae, i.e. yersiniabactin (ybt), aerobactin (iuc), salmochelin (iro), the genotoxin colibactin (clb), and the hyper mucoid locus rmpADC and rmpA2.

The predominant KL type, KL64, was less virulent than KL51 with 96 % (264/274) of them having a virulence score of 1 (score 1=carriage of only yersiniabactin and no other virulence factors) and 4 % (10/274) of them having virulence score of 0 (score 0=negative for yersiniabactin, colibactin, aerobactin). The KL51 was more virulent with 70 % (173/249) with a virulence score of 4 (score 4 = aerobactin with yersiniabactin [without colibactin]), 27 % (69/249) with a virulence score of 1, one of the isolates with virulence score 3 (score 3=aerobactin [without yersiniabactin or colibactin]) and only six of them with a score of 0. The globally known hypervirulent strain KL2 predominantly carried only yersiniabactin 79 % (70/89) with virulence score 1 and only 12 % (11/89) of them carried aerobactin, salmochelin, rmpADC and a truncated rmpA2. Among the 14 KL1 isolates, 12 carried aerobactin with 9/14 having a virulence score of 5 (score 5 = yersiniabactin, colibactin and aerobactin). These isolates also carried the other three virulence factors salmochelin, rmapA2, rmpADC, but rmpA2 was truncated in all of them. There were 13 other KL types which carried aerobactin (41/1072) including the hypervirulent serotypes KL20 (7/41) and KL57 (3/41). Twenty-one percent (228/1072) of the isolates had a virulence score of 0, which included 63 different KL types (Table S6). The distribution of virulence factors and the virulence score is shown in the microreact view here microreact.org/virulence. The distribution of virulence scores per KL types was significantly different (Kruskal-Wallis chi-squared = 585.99, p-value <0.05). We also observed that there was a significant association of yersiniabactin and aerobactin with certain KL types (Table S7) as confirmed by Fisher’s exact test (P value <0.05).

Due to the less diverse O types, we observed varied virulence scores within an O type. The Kruskal-Wallis rank sum test result suggested there are significant differences in virulence scores among the different O types (Kruskal-Wallis chi-squared = 30.028, p-value = 0.002). Particularly, the two most prevalent O types carried more virulence factors than others, 22 % (105/471) of O1/O2v1 carried aerobactin with virulence score 3 (7/105), virulence score 4 (98/105) and 69 % (324/471) virulence score 1. A significant proportion of O1/O2v2, i.e. 58 % (207/354) also carried aerobactin with a virulence score of 3 (12/354), virulence score of 4 (186/354) and virulence score of 5 (9/354). The other 11 lesser predominant O types had a virulence score of 0 (139/247) or a virulence score of 1 (98/247) (Table S8).

Kruskal-Wallis rank sum test was also conducted to assess the relationship between virulence scores and invasiveness. The analysis revealed no significant difference in virulence scores between invasive and non-invasive isolates (Kruskal-Wallis chi-squared = 0.297, p-value = 0.875) nor there was an association of any particular virulence factor as confirmed by Fisher’s exact test (Table S9 and Figure S5). These findings underscore the complex and varied nature of virulence among K. pneumoniae isolates, highlighting the need for further research to elucidate the factors influencing pathogenicity in this bacterium.

Resistance profiles associated with capsular (KL) and O antigen serotypes

In this study, Kleborate identified 64 distinct genes associated with antimicrobial resistance. These genes are linked to resistance across ten antimicrobial classes. All isolates carried at least one genetic resistance determinant and 73 % (786/1072) were MDR having resistance to more than three classes of drugs. Notably, a significant majority, comprising 78 % (841/1072), exhibited specifically carbapenem-resistance markers. The majority of the isolates in our collection (70%, 748/1072) had a resistance score of 2 (score 2 = carbapenemase without colistin resistance [regardless of ESBL genes or OmpK mutations]), followed by 14 % (158/1072) of the isolates having a resistance score of 1 (score 1=ESBL, no carbapenemase [regardless of colistin resistance]), notably 4 % (43/1072) had the highest resistance score of 3 (score 3=carbapenemase with colistin resistance [regardless of ESBL genes or OmpK mutations]) and only 12 % (123/1072) had a resistance score of 0 (score 0=no ESBL, no carbapenemase [regardless of colistin resistance]). The distribution of resistance scores and carbapenamase genes carried among the isolates is shown in the microreact view: microreact.org/resistance_scores.

Forty (12 %, 129/1072) of the KL types did not carry any carbapenemase genes. The other 38 KL types carried at least one carbapenamase gene. The predominant KL types, KL51 and KL64, exhibited varied associations with different carbapenemase genes, with KL64 primarily linked to bla_OXA-48-like variants (particularly bla_OXA-181 [63/274] and bla_OXA-232 [124/274]), while KL51 strains carried predominantly bla_OXA-232 (183/249), followed by co-carriage of bla_NDM and bla_OXA-232 (38/249). The carbapenemase genes were more associated with ST than with a specific KL serotype. The carbapenemase gene distribution among different KL types is provided in Table S10. Using Fisher’s exact test, we found there was a significant association between some O types and carbapenemase genes (P<0.05). The predominant O types, O1/O2v1 and O1/O2v2 carried both bla_OXA-48-like and bla_NDM genes. Four different O types comprising 1.4 % (16/1072) of the isolates did not carry any carbapenemase genes (Table S11). The acquisition of the carbapenemase genes was independent of serotypes and largely driven by the STs. The cumulative distribution of the KL and O types in Fig. 3 showed that the vaccine formulas against K. pneumoniae in India need to incorporate either ~20 KL types to cover >85 % of the carbapenemase-producing population or four O types to cover the same population.

Discussion

Klebsiella pneumoniae is a human commensal and opportunistic pathogen that can cause severe hospital-acquired infections, especially among patients with compromised immune systems. K. pneumoniae infections are tough to cure due to the organism’s thick capsule [52] and the emergence of MDR strains has made the majority of current antimicrobials ineffective [53]. Effective measures for preventing K. pneumoniae infections are desperately needed and vaccines are one of the proposed alternatives [25]. Capsular polysaccharides in K. pneumoniae have emerged as promising targets for vaccine development due to their role in virulence and the bacteria’s resistance mechanisms [26]. This study aimed to comprehensively investigate serotype prevalence and epidemiology in India, a region grappling with antimicrobial resistance and healthcare-associated infections.

The analysis of 1072 K. pneumoniae isolates from 2014 to 2022 in India revealed significant genomic diversity within the KL and O types. The study highlights a varied genetic landscape among K. pneumoniae strains circulating in Indian hospitals. The extensive diversity observed within the KL of K. pneumoniae poses a considerable challenge for vaccine development. With the identification of 78 distinct KL types and the top three KL (KL51, KL64, KL2) collectively representing only 59 % of the isolates, it is evident that creating a vaccine or monoclonal antibodies targeting a wide spectrum of prevalent KL types will be complex. The recently described development of anti-KL64 antibodies would only cover about 26 % of the Indian K. pneumoniae isolates [19]. Several studies have consistently reported a high diversity of KL types in K. pneumoniae strains across different regions [36,54,57]. In Southeast (SE) Asia particularly, studies have highlighted the regional disparity of KL types in K. pneumoniae strains causing infection. For instance, a study by Holt and colleagues [7] found that KL2 and KL47 were predominant in Thailand, while KL107 was more prevalent in Cambodia. Similarly, a study by Wyres and colleagues [36] reported variations in KL types among K. pneumoniae isolates in Malaysia, with KL1, KL2, and KL102 being the most common. Consistent with previous findings in India, our study also identified KL64 and KL51 as the predominant serotypes [35,36, 58]. The development of the Inventprise 25-valent pneumococcal conjugate vaccine candidate, utilising the patented Hz-PEG-Hz linker technology platform, represents a significant advancement in vaccine technology [59]. It is anticipated that this vaccine candidate will offer the broadest coverage against pathogenic pneumococcal serotypes encountered by populations worldwide, irrespective of geographical location. Similarly, if a K. pneumoniae vaccine targeting 25 KL types were to be developed, it could potentially provide coverage for up to 90 % of the Indian K. pneumoniae population along with 85 % of CRKP clones (Fig. 3a), illustrating the profound impact such a vaccine could have on public health and disease prevention efforts.

Research has highlighted the variability in disease associations linked to specific KL types. Some studies have identified certain KL types, such as KL1 and KL2, as being more prevalent and associated with poorer disease outcomes [13,60, 61], particularly in hypervirulent strains causing community-acquired invasive infections. Conversely, others have noted the rarity of these KL types associated with severe infections in certain geographic regions [35,57, 62]. Interestingly, it was also observed that the conventionally described hypervirulent clone KL2 did not carry the hypervirulence marker aerobactin [63], while other regional serotypes (KL51) exhibited the presence of this hypervirulence marker. This observation challenges conventional assumptions about the relationship between capsular types and hypervirulence, suggesting a more nuanced understanding of K. pneumoniae pathogenicity. Moreover, our comparison between invasive and non-invasive isolates highlighted the prevalence of common KL types across both categories, suggesting that the presence of hypervirulent markers or specific KL types may not solely dictate the pathogenic potential of K. pneumoniae strains [64,65]. Rather, other factors such as host immunity, environmental conditions, and strain-specific genetic determinants may play significant roles in determining disease outcomes [66]. This finding further complicates our understanding of the complex relationships between K. pneumoniae virulence factors and disease manifestation, emphasising the need for further research to elucidate the mechanisms underlying KL-associated pathogenicity.

We discovered a considerable diversity of 14 different O types among K. pneumoniae isolates compared to previous studies [36,67]. However, a few strong O-loci, especially O1/O2v1, O1/O2v2, and O101 stood out, accounting for a significant proportion (83 %) of the isolates analysed. The cumulative coverage analysis revealed that incorporating only five types of O antigens (O1, O2v1, O2v2, O101 and O3b) would cover approximately 90 % of the isolates across different specimen types and disease conditions in India. Previous studies have also explored the inclusion of a narrower selection of O types in vaccine formulations. For instance, Wyres and colleagues [36] suggested that incorporating a subset of prevalent O antigens could provide significant coverage against K. pneumoniae infections. Studies have highlighted the importance of targeting specific O types, such as O1, O2, and O3b, in vaccine development strategies due to their high prevalence and association with disease manifestation [20]. A recent study reported a heptavalent O-antigen bioconjugate vaccine [68] which exhibits promising efficacy against some, but not all, K. pneumoniae isolates. While some studies [69] say that O antigen is accessible by antibodies irrespective of the capsule type other studies are highlighting that hyperproduction of CPS may inhibit the vaccine-induced O-antigen antibody binding [69,70]. A more recent study evaluating monoclonal antibodies against K. pneumoniae ST147_NDM-1 concluded that highly bactericidal anti-O-antigen antibodies are not protective against this hypervirulent and pan-drug-resistant strain [19]. This suggests more studies are needed to clarify the effectiveness of anti-O-antigen vaccine or monoclonal antibody formulations against the broader K. pneumoniae population.

Conclusion

The findings provide crucial insights into the genetic diversity, evolution, and potential adaptation mechanisms of K and O loci within the K. pneumoniae Indian population, with implications for understanding its epidemiology. Additionally, the insights regarding its diversity underscore the challenges in formulating vaccines or monoclonal antibodies that adequately cover the diverse array of K. pneumoniae strains including the carbapenemase-producing ones. The lack of a straightforward correlation between specific serotypes and virulence factors challenges conventional assumptions about hypervirulence clones and necessitates further exploration into the multifaceted nature of virulence determinants in K. pneumoniae. Understanding the epidemiology of K. pneumoniae can help tailor effective prophylactic and therapeutic solutions against KL and O antigens in India.

Abbreviations

AMR

Antimicrobial resistance

BAL

Bronchoalveolar lavage

CPS

Capsular polysaccharides

CSF

Cerebrospinal fluid

ESBL

Extended-Spectrum Beta-Lactamase

ETA

Endotracheal aspirates

ETA

Endotracheal aspirates

hv-CRKP

hypervirulent carbapenem-resistant Klebsiella pneumoniae

hvKP

Hypervirulent Klebsiella pneumoniae

KL

K Locus

K. pneumoniae

Klebsiella pneumoniae

LPS

Lipopolysaccharides

MDR

Multidrug-resistant

ST

Sequence type