Introduction

Between August 1st and November 26th, 2017, a total of 2414 clinically suspected plague cases were reported to the Central Laboratory for Plague (CLP) at the Institut Pasteur de Madagascar, including 1878 (78%) pulmonary plague (PP), 395 (16%) bubonic plague (BP), one (< 1%) septicaemia and 140 (6%) cases with unspecified clinical form1. This predominantly urban plague epidemic was characterised by a large volume of notifications in two major urban areas (Antananarivo and Toamasina) and by an unusually high proportion of pneumonic forms. According to the 2006 WHO standard plague case definitions and using the results of three types of diagnostic tests assessed (rapid F1-antigen diagnostic test, (RDT), molecular amplification method, and culture)2, 386/1,878 (21%) were probable and 32/1,878 (2%) were confirmed cases among the notified PP cases. The magnitude of this PP outbreak is likely to have been smaller than suggested by notified suspected cases1; and its severity indicated by the case fatality rate among confirmed plus probable cases (about 9%) was substantially lower than observed in the last previous 18 years (25%)3. Over-reporting of PP cases due to limited clinical experience in the two most affected areas, and the difficulty to clinically diagnose PP through respiratory signs was speculated. The clinical diagnosis of PP from polymicrobial sputum associated with other potential causes of pneumonia remains a challenge because the isolation of Yersinia pestis (Y. pestis) is more complicated compared to other bacteria.

Klebsiella pneumoniae (K. pneumoniae) is a Gram-negative bacterium naturally resistant to amoxicillin and carbenicillin. K. pneumoniae complex members comprise 7 phylogroups (Kp1 to Kp7) that have been given taxonomic status as K. pneumoniae sensu stricto, K. quasipneumoniae subsp. quasipneumoniae, K. quasipneumoniae subsp. similipneumoniae, K. variicola subsp. variicola, K. variicola subsp. tropica, ‘K. quasivariicola’, and K. africana, respectively4. K. pneumoniae sensu stricto has become an important multidrug resistant pathogen of the last decade with multiple resistance determinants, mostly for aminoglycosides, cephalosporins and carbapenems5. It is commonly isolated from hospital-acquired infections including pneumonia, bloodstream infection, urinary tract infection, and community acquired infections such as pyogenic liver abscess, meningitis and pneumonia. The capsule is an important virulence factor that protects K. pneumoniae from phagocytosis, with over 79 defined capsular serotypes. Isolates with K1 and K2 capsular serotypes are associated with virulent infections. However, not all K2 capsular isolates are virulent6. Virulence factors associated with hypervirulent Klebsiella infections also include siderophores, including aerobactin and salmochelin, which are typically encoded on virulence plasmids; yersiniabactin (typically chromosomally encoded in an integrative and conjugative element), and the hypermucoviscosity factor rmpA gene (also typically on the virulence plasmid)7,8,9,10,11,12.

Differential diagnosis is important and should be included in the diagnostic procedures in order to detect and identify other pathogens among the suspected but not confirmed cases of plague. K. pneumoniae is one of the pathogens that cause severe bacterial lung infections and which should be considered in non-confirmed suspected pulmonary plague cases. However, K. pneumoniae was not considered among suspected bubonic plague cases13.

The purpose of this study was to characterize K. pneumoniae isolates from some of clinically suspected plague patients during the plague outbreak in Madagascar in 2017. We aimed to analyze their population structure and clonal diversity using core genome Multilocus Sequence Typing (cgMLST). Further, our aims were to analyze their resistance and virulence genes, and the association of virulence factors and phenotype with clonal background.

Material and methods

Patients and bacterial isolates

Patients with confirmed K. pneumoniae, isolated during nine days of the plague epidemic (from the 6th of October 2017 till the 14th of October 2017), were included in this sub-study. Epidemiological, clinical and lab data of patients were extracted from the plague national surveillance system database of Institut Pasteur de Madagascar between August 1st and November 26th in 20171. Y. pestis was isolated from biological samples (bubo aspirates for BP, sputum for PP) by direct culture on Yersinia selective Cefsulodin-Irgasan-Novobiocin (CIN) agar medium (Oxoid Ltd., United Kingdom) and Y. pestis detection by PCR was performed on all samples. All methods were carried out according to the 2006 WHO recommendations14. Culture incubation was done at 26–28 °C for 48 h or longer as Y. pestis grows slower than other bacteria. Colonies obtained within 24 h on CIN medium and which did not have Y. pestis morphology were identified on MALDI-TOF MS (Biotyper version 3.3, Bruker Daltonics, Champs-sur-Marne, France). Colonies identified as K. pneumoniae were further purified on Simmons Citrate Agar Inositol (SCAI) medium15. The Central Laboratory for Plague of the Malagasy Ministry of Health is hosted at the Plague Unit-WHO Collaborating Centre of the Institut Pasteur and all methods used in this study were performed in accordance with the relevant guidelines and regulations. The data reported here are parts of the plague national surveillance system and no specific additional ethics approval was necessary. All information on individual patients has been anonymized for presentation.

Phenotype detection

The hypermucoviscosity phenotype of the K. pneumoniae isolates was determined using the string test, in which a standard bacteriological loop is used to stretch a mucoviscous string from each colony cultured on SCAI. The formation of a viscous string > 5 mm in length was regarded as a positive test result12. The string test results were confirmed using colonies grown on blood agar.

Bacterial susceptibility testing

Antibiotic resistance profiles were determined by the standard disc diffusion method according to CASFM-EUCAST V2-0-May2017 guidelines and using breakpoints for Enterobacteriaceae16 (http://www.sfm-microbiologie.org). Isolates were tested against 17 commonly used antimicrobial agents, namely amoxicillin, amoxicillin-clavulanate, piperacillin/tazobactam, cefalotin, cefoxitin, cefotaxime, ceftazidime, cefepime, aztreonam, imipenem, ertapenem, tobramycin, gentamicin, nalidixic acid, ciprofloxacin, trimethoprim-sulfamethoxazole and tetracycline. In addition, extended spectrum β-lactamase (ESBL) production was tested using the standard double disc synergy test.

Genome sequencing and analysis

Genomic DNA of the K. pneumoniae isolates was extracted using DNeasy Blood & Tissue kit (Qiagen, Germany) and was subjected to whole genome sequencing. Genomic libraries were constructed using the Nextera XT DNA library preparation kit with dual indexing (Illumina, San Diego, USA). The libraries were sequenced on an Illumina NextSeq-500. Genome assembly was performed de novo using Spades Genome Assembler (Version 3.10.0). Genome analyses for core genome MLST (cgMLST) on 632 core genes were performed17. Sequence types (ST) were determined with an in silico MLST pipeline that assembles and compares sequences against allele data derived from the public MLST database at https://bigsdb.pasteur.fr/. Virulence genes and capsular serotypes (K-types) were assigned using the K. pneumoniae database hosted through the BIGSdb web application of the Institut Pasteur in Paris (https://bigsdb.pasteur.fr/klebsiella). Antimicrobial resistance genes were identified from genome sequences using the Resfinder (version 3.2)18. Plasmid replicon types were determined by using the Plasmidfinder (version 2.0)18 tool at https://genomicepidemiology.org. Parsnp was used for the core genome alignment19, while the Gubbins20 software tool was used to remove the single nucleotide polymorphisms (SNPs) from recombined regions and to create a refined phylogenetic tree; this tool uses RaxML21 to build the maximum likelihood phylogenetic tree on the recombination-free regions. The tree was subsequently annotated with iTOL (http://itol.embl.de/itol.cgi)22.

Buboes K. pneumoniae screening by real-time PCR

We performed a real-time PCR targeting the zur-khe intergenic region (called the ZKIR qPCR assay)23 on the bubo samples in order to confirm the presence of K. pneumoniae DNA in the bubo and to exclude any technical contamination during culture. Bacterial DNA was extracted from the bubo samples using DNeasy Blood & Tissue kit (Qiagen, Germany). The real-time PCR assay was performed as previously described with the difference that we used 10 µl of SsoAdvanced universal SYBR Green Supermix (Bio-Rad, USA)23. Amplifications were performed using the CFX-96 (Bio-Rad, USA) platform. The positive controls consisted of DNA from K. pneumoniae UAA2239 and UAA2016 which are reference strains from the National Reference Center for Antibiotics from the Institut Pasteur in Paris, the negative control was plain molecular grade water.

Ethics statement

The Ethics Committee/IRB authorized the use of the patient samples in this study, as long as they are anonymised/de-identified (reference number 261 MSANP/SG/AMM/CERBM).

No additional data was collected.

All patients provided oral consent and voluntarily agreed for sampling for diagnostic purposes.

Nucleotide sequence accession numbers

WGS data have been deposited at the National Center for Biotechnology Information (NCBI) under BioProject PRJNA565154.

Results

Case presentation and K. pneumoniae antimicrobial susceptibility

Twelve clinical samples (2 bubo aspirates for BP, 10 sputum for PP) out of 496 collected between 06 and 14 October 2017 in Antananarivo (N = 362) and Toamasina (N = 134) screened for Y. pestis presence by culture on CIN medium gave rise to abundant colonies (> 103 CFU/ml) with a typical K. pneumoniae morphology (moist, dome-shaped) after 24 h incubation. One representative isolate per plate was selected and the twelve isolates were identified as K. pneumoniae by MALDI-TOF MS. Using PCR, Y. pestis DNA was not detected in samples. We did not perform serology for antibody detection in the confirmed and suspected cases. This period was reported as the peak of the plague epidemic curve, with essentially PP cases. Patients had early clinical signs suggesting pulmonary, secondary pulmonary or bubonic plague in an epidemic setting.

The samples containing K. pneumoniae were 10 sputum samples and two bubo aspirates and were negative for Y. pestis colonies. The description of the 12 patients is shown in Table 1. There was no mortality among these patients (Table 1). Two-thirds of the patients (N = 8) were from Antananarivo and one-third (N = 4) was from Toamasina. Seven patients were men. Five patients were younger than 18 years, and six were 19 to 27 years old and one was 46 years old. Fever status was reported for 10 patients, eight of them had body temperatures > 37.5 °C. According to the clinical forms, nine patients were suspected of having PP, one was defined to suffer from secondary PP, one patient was suspected of BP and for another patient, data about the clinical form was lacking.

Table 1 Study population.
Full size table

Four of the patients coughed for at least 5 days and complained of chest pain, although they were in an overall good state of health. Two patients had signs of hemoptysis, and one of them was in weak health. Two patients coughed without further complaints but one was in weak health.

Four patients were under antibiotic treatment at the time of sample collection: two with trimethoprim-sulfamethoxazole; one with doxycycline and one with gentamycin. One patient who received trimethoprim-sulfamethoxazole was treated in addition with amoxicillin.

The presence of K. pneumoniae DNA in the two bubo samples was also detected by real time-PCR. Melting curve values for the detection of K. pneumoniae were 79 °C and 80 °C for the positive controls and the DNAs extracted from buboes, respectively (Fig. 1). Of the 12 isolates, four had a positive string test. Five isolates were ESBL producers. Six isolates were resistant to sulfonamides and trimethoprim. Three and two isolates were resistant to gentamycin and tobramycin, respectively. One isolate was resistant to ciprofloxacin (Table 2).

Figure 1
figure 1

Melting curve results from ZKIR region detection obtained after 40 cycles using the ZKIR quantitative PCR system. DNA from two positive controls from K. pneumoniae strains (blue: UUA2239, violet: UUA2016). DNA extracted from patients’ buboes is in pink for PP2 and in red for PP8. Orange color corresponds to negative control.

Full size image
Table 2 MLST profiles, cps K-types, resistance profile, resistance genes and virulence genes.
Full size table

Genome analysis

Whole genome sequencing of the K. pneumoniae isolates allowed us to characterize cgMLST alleles, virulence genes, capsular loci and resistance genes. All K. pneumoniae isolates were K. pneumoniae sensu stricto (Kp1) (Fig. 2). The 12 isolates had 11 different sequence types (STs): ST23 (N = 1); ST86 (N = 2); ST65 (N = 1); ST280 (N = 1); ST327 (N = 1); ST380 (N = 1); ST716 (N = 1); and ST3012 (N = 1), and three new STs: ST3441 (N = 1), ST3442 (N = 1), ST3443 (N = 1) (Table 2). Comparative genomic analysis of the two Kp ST86 isolates showed that they differed from each other by 123 alleles out of 632 scgMLST gene loci and are therefore unrelated epidemiologically.

Figure 2
figure 2

Phylogenetic tree of K. pneumoniae isolates obtained from 12 patients. The tree includes the sequence types (ST) of each isolates, the frequency of K-types in indigo and the string test positive in blue sky.

Full size image

The virulence genes identified in most isolates corresponded to the colibactin locus (clb), siderophores (iroBCDN), iron uptake systems and regulators (kfu, kvgA, respectively), and yersiniabactin (fyuA, irp1/2 and ybt) (Supplementary data). The PP3, PP9, PP11 and PP12 isolates were positive for the string test. These isolates belonged to ST320 (PP3), ST86 (PP9 and PP12), and ST3443 (PP11). All the ST23, ST65, ST86 and ST380 isolates had the gene rmpA associated with the hyperproduction of the capsule and also carried the iucABCD genes coding for the synthesis of aerobactin (Supplementary Data).

Isolates belonging to ST23, ST65, ST86 and ST380 were susceptible to all antibiotics tested, with the exception of amoxicillin to which K. pneumoniae is intrinsically resistant.

A total of five isolates were ESBL producers (Table 2). The ST280 and ST3441 isolates carried the blaCTX-M-15 gene. The ST3442-KL1 isolate harboring virulence genes (fyu, irp1/2, ybt) was ESBL producer and carried the cassette comprising qnrB66, aac (3) -IIa, blaSHV-27 and tet (A), as well as the IncFIBK replicon marker (Table 2).

Discussion

Unexpectedly, 12 cases of K. pneumoniae infection were detected and identified among individuals clinically suspected to have plague. No Y. pestis was identified from their clinical samples, but three patients (PP8, PP10 and PP11) yielded a positive result on RDT. However, culture is the gold standard for the identification of Y. pestis and RDT could provide false results24. All patients had no epidemiological relationships and no family member or contact had been recorded with plague.

Although the selective medium for Yersinia was not intended for K. pneumoniae isolation, K. pneumoniae does grow on the CIN medium in 24 h. The selectivity of this medium is reported as being partial, as other Gram-negative bacilli can grow on CIN medium, including other species of Enterobacterales able to ferment mannitol25. Therefore, full species identification is recommended.

It is not surprising to isolate K. pneumoniae from pneumonia cases. Community-acquired K. pneumoniae infections are common, including in Africa26,27. In contrast, to our best knowledge, K. pneumoniae isolated from buboes aspirates were never reported previously. Additionally to culture, the presence of K. pneumoniae in buboes aspirates was confirmed by PCR. The advantage of melting curve analysis over Taqman based real time-PCR is its lower costs without losing specificity28. Further studies are needed to evaluate this method, which could be used in screening for K. pneumoniae in buboes or other suspected biological samples.

K. pneumoniae has the capacity to acquire resistance genes and to become increasingly more difficult to treat. One of the K. pneumoniae isolates detected in one of two patients who were treated by combination of amoxicillin and trimethoprim-sulfamethoxazole developed resistance. However, K. pneumoniae is known to be intrinsically resistant to ampicillin due to the presence of the chromosomal β-lactamase genes blaSHV-1, blaSHV-11 or similar29. At the same time, we identified in this strain (ST280), the two genes dfrA14 and sul2, associated to resistance to trimethoprim and sulfamethoxazole, respectively. In addition, among all isolates, it was the only K. pneumoniae isolates to be resistant to piperacillin/tazobactam (TZP), which is concordant with the presence of the blaTEM-1B gene30.

Two of the five ESBL isolates harbored blaCTX-M15. This gene was commonly found in ESBL-producing K. pneumoniae isolated in Madagascar26,27. blaSHV27 and blaSHV101 were found in the other ESBL-producer.

According to MLST analysis, we observed the presence of five STs known to be associated with hypervirulence, including STs ST23 (N = 1), ST65 (N = 1), ST86 (N = 2) and ST380 (N = 1). In addition, a K. pneumoniae isolate having the new ST3443 differs by a single locus from ST86 on the tonB locus (allele 18 instead of 27). Whole genome analysis of K. pneumoniae showed that the isolates with common ST differed from each other by alleles occurring outside the 7 household genes. The strain ST3443 was tested string positive, as was also the case for the ST380 strain and the two ST86 strains. The presence of common virulence genes in the ST23 isolate which were ICEKp10 encoding clb 2 sequence variants, ybt 1 and rmpA/rmpA2 suggests its belonging to CG23 sublineage I (CG23-I)31.

MLST typing of K. pneumoniae isolated in different countries revealed that ST23, ST65, ST86 and ST380 were responsible for pyogenic liver abscess cases and other invasive community-acquired infections32. These isolates have been reported particularly in Asia, but their diffusion outside Asia has been described31. Among virulence factors, rmpA and aerobactin are the most important ones33. The presence of genes responsible for the hypermucoviscosity phenotype, notably rmpA, plays an important role in the virulence of K. pneumoniae isolates. This gene is often associated with serotype K1 and K2. Expression of rmpA allows the bacteria to escape the host's defense system and colonize the mucous membranes. Epidemiological studies have shown that the majority of ST23 are related to K1 capsular serotypes and liver abscesses31,34, while K2 is the second capsular serotype resulting in community-acquired pneumonia33. Yersiniabactin, a virulence gene (Ybt), detected in the three K. pneumoniae isolates serotyped KL1 has been reported as the iron absorption system in highly virulent Y. pestis35, and was later shown to have evolved ancestrally within the Klebsiella genus25. Several studies have shown isolates belonging to these STs (23, 65, 86 and 380), with the same combination of virulence factors, to be virulent in mouse models12,31.

During an epidemic, knowledge of the etiology is essential in order to provide the most adapted treatment to patients. Microbiological diagnosis can improve the effectiveness of treatments, avoid long-term complications for the infected patient, and in addition avoid widespread overuse and misuse of antibiotics. Early diagnosis can help to prevent or stop an outbreak too. One of the reasons for a possible treatment failure could arise during inaccurate diagnoses and inappropriate treatments. Similar symptoms can lead to routine treatments based on syndromic approaches which are often applied in developing countries, hence the importance of including differential diagnosis in laboratory diagnostic procedures in order to identify the etiology. As the physician is rarely able to make an etiological diagnosis on clinical grounds alone, treatment should ideally be based on the result of bacteriological examination. In this case, bacteriological diagnosis could be complicated by the fact that the respiratory tract could be infected by K. pneumoniae36. Although the population we included in our study is young, the clinical signs of a few patients warned us of possible serious infections due to K. pneumoniae such as bloody sputum and a chest pain which were among typical signs of pestis pneumonia. However, Y. pestis was not found in culture. Typically, the plague is better known by its three clinical forms: bubonic, septicemic and pulmonary plague while hypervirulent K. pneumoniae strains are known to cause pneumonia, sepsis, liver abscesses and meningitis37,38,39.

We acknowledge the following limitations of our study. First, we studied a limited number of samples in a short duration of the epidemic, which is far from being representative of all the negative samples for Y. pestis. Second, detailed data about clinical characteristics and outcomes were lacking due to the outbreak emergency context. Finally, we did not confirm the virulence of the K. pneumoniae strains using mouse models.

Conclusions

Although few samples were studied, within a short duration of inclusion (9 days from 06/10/17 to 14/10/17), our results show that some plague-suspected patients in fact acquired a pneumonia caused by K. pneumoniae. Bacterial identification proved useful for determining the etiology. WGS and AST results showed that among the 12 K. pneumoniae isolates, there were ESBL producers and virulent strains. This study reports on the genomic characterization of K. pneumoniae isolates isolated from patients during an epidemic of plague. These results could serve to warn clinicians regarding the most adequate treatment. Likewise, we reported the importance of bacteriological diagnosis for improving patient management.