The genus Klebsiella, a member of the Enterobacteriaceae family, includes Gram-negative, non-motile (except K. aerogenes) and non-spore-forming capsulated bacteria. Bacteria belonging to the genus Klebsiella are found in water, soil and plants, and as commensals in the gut of animals including humans (Schmitz et al., 2002; Brisse et al., 2006; Caltagirone et al., 2017). In humans, Klebsiella species are frequently associated with hospital-acquired infections and are increasingly multidrug-resistant (Paczosa and Mecsas, 2018). Klebsiella oxytoca is the second most common Klebsiella species causing disease in humans, after K. pneumoniae (Broberg et al., 2014). K. oxytoca carries a chromosomally encoded β-lactamase gene (bla_OXY) that confers resistance to amino- and carboxypenicillins (Fournier and Roy, 1997). This gene was shown to have diversified in parallel to housekeeping genes, and variants were classified into seven groups (bla_OXY–1 to bla_OXY–7) (Granier et al., 2003a, b; Fevre et al., 2005; Izdebski et al., 2015). K. oxytoca phylogenetic lineages were named Ko1, Ko2, Ko3, Ko4, Ko6, and Ko7 reflecting which bla_OXY variant they carry; note that Ko5 was not defined, as isolates carrying bla_OXY–5 represent a sublineage of Ko1 (Fevre et al., 2005). Taxonomic work has shown that K. oxytoca (sensu lato, i.e., as commonly identified in clinical microbiology laboratories) is in fact a complex of species, with K. oxytoca (sensu stricto) corresponding to phylogroup Ko2, K. michiganensis to Ko1 (Saha et al., 2013) and K. grimontii to Ko6 (Passet and Brisse, 2018). The closely related K. huaxiensis (Hu et al., 2019) represents yet another phylogroup, which we here denominate as Ko8 and which carries bla_OXY–8. Phylogroups Ko3, Ko4, Ko7, and K. huaxiensis were so far described only based on a single strain (Fevre et al., 2005; Hu et al., 2019), which has limited our ability to define their genotypic and phenotypic characteristics. While analyzing a large number of Klebsiella strains from multiple human, animal and environmental sources in and around the Northern Italian town of Pavia, we identified 3 Ko3, 13 Ko4, and 3 K. huaxiensis strains. The aim of this work was to define the taxonomic status of K. oxytoca phylogroups Ko3 and Ko4 and provide identification biomarkers for all members of the K. oxytoca species complex.

The phylogenomic analysis based on the concatenation of 3,814 core genes (Figure 1) showed six distinct and highly supported branches. The thirteen Ko4 strains were clustered with Ko4 reference strain SG266 (SB3355) and this group was related to, but clearly distinct from, K. grimontii (Ko6). The three Ko3 strains (SPARK_350_C1, SPARK_775_C1 and SPARK_1442_C2) formed a well-defined cluster with Ko3 reference strain SG271 (SB3356, Figure 1), whereas the remaining three strains (SPARK_1445_C2, SPARK_1448_C1, SPARK_1495_C1) clustered with K. huaxiensis, which formed a distinct phylogroup that we here name Ko8. We therefore identified novel strains of these three phylogroups, which were each previously recognized based on a single strain. Furthermore, genome-based phylogeny revealed that Ko4 shares a common ancestor with K. grimontii, K. michiganensis and K. oxytoca, whereas Ko3 and K. huaxiensis share a common ancestor distinct from the Ko1/Ko4/Ko6 one (Figure 1).

To determine how previously used phylogenetic markers (Brisse and Verhoef, 2001; Granier et al., 2003a, b; Fevre et al., 2005) would group these novel strains, the sequences of internal portions of the housekeeping genes gyrA (383 nt) and rpoB (501 nt), as well as the rrs (1,454 nt) sequence coding for 16S rRNA, were extracted from genomic sequences and compared to previously characterized sequences of reference and type strains from the K. oxytoca complex (Table 1). The clustering of Ko4 strains and Ko3 strains was supported by phylogenetic analysis of combined gyrA and rpoB gene sequences (Figure 2), as well as by single gene phylogenies (Supplementary Figures S1, S2), showing that either gene used alone would allow reliable identification. The phylogeny of the chromosomal OXY beta-lactamase gene (Supplementary Figure S3) was also in concordance with previous phylogenetic analyses. However, phylogroup Ko1 and Ko3 each harbored two different types of bla_OXY, coding for OXY-1/OXY-5 and OXY-3/OXY-9, respectively (Supplementary Figure S3). As previously reported (Boye and Hansen, 2003; Naum et al., 2008; Passet and Brisse, 2018), the phylogeny based on the rrs gene was not reliable for species or phylogroup identification (type strain sequences were >97.8% similar), with only a few informative variable sites (Supplementary Figure S4).

Average nucleotide identity was estimated between Ko3 and Ko4, and the type strains of species of the K. oxytoca complex (Table 2). The three Ko3 strains, including SPARK_775_C1^T, shared high identity (above 98%) with the Ko3 strain SG271 (SB3356) (data not shown). The ANI values of SPARK_775_C1^T (Ko3) strain with K. huaxiensis (WCHKl090001^T), K. michiganensis (W14^T), K grimontii (06D021^T) and K. oxytoca (ATCC 13182^T) were 90.7, 88.4, 88.3, and 87.9%, respectively (Table 2). The novel Ko4 strains showed approximately 98% ANI with Ko4 strain SG266 (SB3355). The ANI values of SPARK_836_C1^T (Ko4) with K. grimontii, K. michiganensis, K. oxytoca and K. huaxiensis were 95.5, 93.3, 90.6, and 87.1%, respectively (Table 2). Finally, the three Ko8 strains presented ANI values >99% with the type strain of K. huaxiensis (WCHKl090001^T), showing that they belong to this recently described species. The isDDH relatedness range between the Ko3 and Ko4 type strains and other species was 36.3–44.1% and 34.3–67.8%, respectively. In conclusion, both ANI and isDDH values were below the thresholds proposed (Rossello-Mora and Amann, 2015) for species distinction (∼95–96% in the case of ANI, ∼70% in the case of isDDH), indicating that Ko3 and Ko4 represent two new species.

The phenotypic characteristics of Ko3 and Ko4 strains were analyzed and compared with those of other Klebsiella isolates. We confirmed that all strains were non-motile by microscopy and that all isolates were positive for indole, lactose, mannitol, malonate, lysine decarboxylase, and the ONPG test, and reduced nitrate to nitrite, whereas they were all negative for ornithine decarboxylase. Ko8 and Ko3 isolates were negative for Voges–Proskauer test and Ko3 isolates were urease positive (similar to Ko2). To define further the biochemical features of the K. oxytoca phylogroups, their carbon source utilization profiles were analyzed. Among 190 substrates, several appeared useful for differentiating the phylogroups among themselves and to differentiate Ko3 and Ko4 strains from other groups (Table 3 and Supplementary Figure S5). The inability to metabolize L-proline and tricarballylic acid differentiated Ko3 strains from other phylogroups except Ko8, which can be differentiated based on its unique ability to utilize 3-O-methyl-glucose. Ko4 had a weak but unique capacity to utilize glyoxylic acid, and differed from Ko6 (K. grimontii) by its inability to metabolize D-melezitose; Ko4 was otherwise similar to Ko6 for many features, consistent with their phylogenetic association.

We also analyzed the MALDI-ToF MS peak patterns of the different members of the K. oxytoca complex. Based on the MALDI Biotyper Compass database version 4.1.80 (Bruker Daltonics, Bremen, Germany), the thirty strains were identified either as K. oxytoca (23 strains, all belonging to Ko1, Ko2, Ko4, and Ko6) or as Raoultella ornithinolytica (7 strains, all strains of Ko3 and Ko8). These misidentifications can be explained by the lack of reference spectra of most phylogroups in the reference database. Supplementary Figure S6 summarizes the peak positions found in each strain. A total of 31 biomarkers (2383–10152 m/z) associated with specific members of the K. oxytoca complex were identified (Supplementary Table S1 and Supplementary Figure S6). Consistent with genetic and biochemical findings, we also observed that Ko4 shared most of its spectral peaks with Ko1 and Ko6, presenting only one specific peak (which was variably present) at 3681 m/z, whereas Ko3 shared six peaks with only Ko8 and presented two unique peaks at 5178 and 6795 m/z. For the remaining phylogroups, specific peaks were observed for Ko2 and Ko8, whereas Ko1 and Ko6 could be identified by specific peak combinations. Based on the current dataset, the specificity and sensitivity of their distribution among phylogroups ranged between 60–100% and 80–100%, respectively (Supplementary Table S1). This finding paves the way to identify isolates of the K. oxytoca complex at the species (or phylogroup) level based on MALDI-ToF MS analysis, pending incorporation of reference spectra of the various taxa into reference spectra databases.

Based on the above genomic, phenotypic and proteomic characteristics, we propose Ko3 and Ko4 to be considered as two novel species, which we propose to name K. spallanzanii and K. pasteurii, respectively.

The nucleotide sequences generated in this study were deposited in European Nucleotide Archive (ENA) and are available through the INSDC databases under accession numbers MN091365 (SB6411T = SPARK775C1T), MN091366 (SB6412T = SPARK836C1T), MN104661 to MN104677 (16S rRNA), MN076606 to MN076643 (gyrA and rpoB), and MN030558 to MN030567 (blaOXY). Complete genomic sequences were submitted to European Nucleotide Archive under the BioProject number PRJEB15325.

The approval of Ethical Committee of the San Matteo Hospital in Pavia was granted under number 20170001787 in date 25/05/2017, proceeding number is 2017000759. The internal code of the project is 0890170117. The ethical procedure includes written informed consent from all the patients participating in the study.

CM, MC, PM, CB, and DS isolated Klebsiella from diverse sources. CM, CR, VP, and MC performed the microbiological characterization of isolates. CM, HT, TK, and DS performed the genomic sequencing. CM, CR, HT, and TK analyzed the sequence data. CR and VP performed the MALDI-TOF analyses. VP and SB performed the phenotypic microarray analyses. CM, CR, and SB wrote the initial version of the manuscript. All authors revised the manuscript. EF, SB, JC, CB, and DS acquired funding for this study.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.