Klebsiella pneumoniae, an opportunistic pathogen known for its ability to cause a wide range of nosocomial and community-acquired infections, has emerged as a significant public health threat due to its strain-specific, extensive arsenal of resistance and virulence factors (Wyres et al., 2020; Antimicrobial Resistance Collaborators, 2022). Infections caused by multi-, extensively-, and pandrug-resistant strains result in high mortality due to limited response to antibiotic therapy, which poses an increasing threat (Ventola, 2015; Navon-Venezia et al., 2017; Avgoulea et al., 2018). Apart from classic strains, a hypervirulent K. pneumoniae (hvKp) pathotype occurs and is characterized by invasive, often life-threatening and multiple site infection, characteristically in healthy patients from the general population (Russo and Marr, 2019). In addition, convergent types that successfully combine resistance and hypervirulence represent a “perfect storm” and have been increasingly reported in recent years (Heiden et al., 2020; Lan et al., 2021; Eger et al., 2022).

Beyond typical resistance mechanisms against various antimicrobials, functional resistance mechanisms have been elucidated that lead to antimicrobial treatment failure and foster the development of relapses and persistent infections (Ster et al., 2017). The formation of a biofilm matrix represents one of these mechanisms that facilitates antibiotic tolerance and the generation of bacterial persister cells (Ster et al., 2017). Interestingly, it has been demonstrated that a decrease in capsule biosynthesis, which is crucial for hypervirulent phenotypes, leads to increase in vitro biofilm formation and intracellular persistence (Ernst et al., 2020). Another non-classical mechanism leading to antibiotic tolerance is the formation of the small colony variant (SCV) phenotype. SCVs are subpopulations of bacteria that exhibit slow growth, reduced colony size, and altered phenotypic properties compared to their normal-growing counterparts, making them difficult to detect and treat effectively (Proctor et al., 2006; Becker, 2023). Their ability to evade the host’s immune surveillance and to undermine the effectiveness of antimicrobial interventions by host cell internalization results in intracellular persistence, which contributes significantly to the recurrence and chronicity of the infection (Tuchscherr et al., 2011; Kahl et al., 2016). Intracellular persistence have been shown for different human and animal cell types including endothelial and epithelial cells such as keratinocytes and osteoblasts (von Eiff et al., 2001; Strobel et al., 2016). Another pivotal attribute facilitating this phenomenon is their capability to modulate metabolic processes and virulence characteristics (Kriegeskorte et al., 2014; Proctor et al., 2014). Hypermutator SCVs characterized by higher mutation frequencies than wild-type strains and isolated especially from cystic fibrosis (CF) patients (Oliver et al., 2000; Prunier et al., 2003) have also been associated with antibiotic resistance (Schaaff et al., 2003; Besier et al., 2008) and biofilm formation (Morelli et al., 2015).

To date, research has focused on staphylococcal SCVs, while SCVs of Gram-negative bacteria have been investigated in only a few studies and case reports (Proctor et al., 2006). Although the formation of small colonies in K. pneumoniae has been noticed during resistance studies against cephalosporins in the mid-1960s (Benner et al., 1965), this issue has not received sufficient attention and detailed research has not been conducted on this subject. The first clearly defined SCV of K. pneumoniae (SCV-Kp) in literature was obtained by in vitro exposure to gentamicin (Musher et al., 1979). SCV-Kp were also isolated from a patient treated with aminoglycoside antibiotics (Murray and Moellering, 1982). Smaller and non-mucoid colonies were obtained as a result of conjugation-induced mutation in the outer membrane protein of a hypervirulent K. pneumoniae isolate (Srinivasan et al., 2012). Another study showed that biofilm-forming K. pneumoniae developed heteroresistance to colistin by presenting slow-growing SCV-Kp (Silva et al., 2016).

Here, we report on K. pneumoniae isolates displaying 14 different resisto- and morpho-types obtained from an immunocompetent male patient, who had sustained a traumatic injury caused by shrapnel shell fragments. The isolates comprise an initial, mostly susceptible K. pneumoniae isolate with typical morphological characteristics isolated from the patient’s urinary specimen. From the urine and tissue samples, 13 additional phenotypes with different combinations of resistance and morphological characteristics including K. pneumoniae SCV phenotypes were isolated.

Overall, 14 distinct phenotypes were determined ( Table 1 ). From the urine, two phenotypes (1-A and 1-B) exhibiting a normal colony size and glistening surface but differing in the color of their colonies displaying whitish or grey colonies, were isolated. All other phenotypes (n = 12) were isolated from tissue specimens. Strains numbered 1-A, 2-A, 3-A, 4-B, 5-B, numbered 1-B, 2-B, 3-B, 4-C, 5-C, and numbered 4-D, 5-D, displayed identical morphological attributes each, distinguished by whitish, glistening, and smooth ( Figure 1B ), grey, glistening, and smooth ( Figure 1C ), and grey, dry, and rough colonies ( Figure 1D ), respectively ( Supplementary Figure S1 ). These strains revealed a normal colony size of 2.4 mm on average (range, 1 – 5.5 mm). The isolates displaying the SCV phenotype, numbered 4-A and 5-A, exhibited similar morphological characteristics, and colony sizes were smaller than 0.5 mm ( Figure 1 , Supplementary Figure S1 ). No discernible variation in terms of colony clustering was observed among the various agar plates. There were no obvious differences in size or shape of cells between different phenotypes except bacteria from grey, dry, rough phenotype 5-D, in which cells were clearly elongated ( Supplementary Figure S2 ).

Initially, largely antibiotic-susceptible K. pneumoniae phenotypes exhibiting whitish and grey colony morphologies on Columbia agar plates were isolated from the urine sample. Following antibiotic treatment, MDR K. pneumoniae strains displaying the normal colony size were isolated from tissue samples, again characterized by subsequent whitish or grey colony formations. Subsequently, SCVs of K. pneumoniae were isolated from tissue samples. Subcultivation of different SCV colonies consistently yielded a division into four distinct colony morphotypes including one SCV phenotype that resembled the initial SCV, along with three normal-sized phenotypes distinguished by variations in colony color and visual attributes. While normal-sized phenotypes exhibited stability following each round of re-cultivation, SCV isolates displayed instability and recurrently diverged into the four phenotypes described above. We have designated the emergence of these multiple phenotypes as “hyper-splitting”. Despite minor variations in MIC values, these “hyper-splitting” phenotypes exhibited multidrug resistance ( Table 1 ).

Except for isolates 1-A and 1-B, all isolates were resistant to the tested carbapenems. Initially, during routine diagnosis, isolate 2-B was found to be carbapenem-resistant by VITEK^® 2 AST, and to harbor bla _OXA-48 gene by LAMP. After subcultivation of this isolate for MIC determination, this resistance disappeared and the isolate became susceptible to all tested beta-lactam antibiotics except piperacillin. We assume that a mobile genetic element harboring bla _OXA-48 gene was lost upon subcultivation. Only isolates 1-A and 1-B were susceptible to piperacillin, and only isolate 4-B was not resistant to the cephalosporins tested. Interestingly, only isolates 4-A and 5-A, which demonstrated the SCV phenotype, were resistant to amikacin and trimethoprim-sulfamethoxazole. Another remarkable finding was the observed increase in the MIC values of cefiderocol and trimethoprim-sulfamethoxazole over time ( Table 1 ).

Whole-genome sequence (WGS) analysis revealed that all isolates belonged to sequence type (ST) 147. Lipopolysaccharide antigen (O) loci were O1/O2v1 and capsule biosynthesis (KL) loci were KL64 for all isolates except isolate 4-D, which could not be assigned, as it missed most genes of this locus. Isolates 1-A, 1-B and 2-B showed lower Kleborate resistance score than the other isolates (resistance: 0 vs. 2). The resistance score of 0 indicates that the isolate(s) did not carry any genes for extended-spectrum beta-lactamases (ESBL) or carbapenemases and a score of 2 correlated with the presence of carbapenemase genes without colistin resistance genes (Lam et al., 2021). In accordance with the resistance scores, we detected several beta-lactamase genes, such as bla _SHV-11, bla _TEM-1 and bla _OXA-9, ESBL genes, such as bla _CTX-M-15 and bla _OXA-1, and the carbapenemase genes bla _NDM-1 and bla _OXA-48. bla _SHV-11 was found in all isolates whereas bla _TEM-1 and bla _OXA-9 were present in all isolates except 1-A and 1-B. However, bla _CTX-M-15 was not found in isolate 4-A. In isolate 4-B, bla _CTX-M-15 and bla _NDM-1 genes were initially detected by WGS, however, after sub-cultivation, a discrepancy between AST and WGS results was observed. Re-testing by LAMP at this later time point revealed the loss of both genes ( Table 1 ). Genes associated with sulphonamide (sul1) and chloramphenicol (catB3) resistance were also detected in all isolates except 1-A, 1-B and 2-B. Note that we did not detect any common cefiderocol resistance genes.

The isolates exhibited clonality as emphasized by the low number of SNPs among them ( Supplementary Tables S1 , S2 ). Especially isolates from the same time point showed no difference in the core genome alignment (5,360,988 bp) with the exception of 2-A and 2-B (six SNPs) and 5-D (one additional SNP compared to 5-A–C). The largest distance with 17 SNPs was between 2-A and 5-D ( Supplementary Table S1 ).

When evaluating the results, we can roughly identify three distinct outcomes. The first significant observation concerns the emergence of resistance development chronologically within a K. pneumoniae strain, originating from a patient subjected to continuous, uninterrupted antibiotic intervention. This scenario promptly elicits contemplation of the subject concerning within-host adaptive evolution of bacteria. In fact, in-host resistance evolution, either due to plasmid mediation or chromosome mutations, has been observed even shortly after the initiation of antimicrobial treatment (Jin et al., 2021).

The second notable observation in our study is the occurrence of SCVs from patient specimens following the detection of normal-sized morphotypes. SCVs demonstrate remarkable abilities to invade and persist within host cells, thus evading the surveillance mechanisms of the immune system (Tuchscherr et al., 2020). The existence of SCVs, mostly observed in Staphylococcus spp., has been documented since the onset of the 20th century and has gained increasing attention due to its potential implications for both clinical and basic research (Jacobsen, 1910; Proctor et al., 2006). Regarding the SCVs of Gram-negative bacteria, studies have particularly focused on Burkholderia and Pseudomonas spp. isolated from CF patients (Oliver et al., 2000; Haussler et al., 2003a, Haussler et al., 2003b). However, there are only sparse data on the occurrence of SCV in Klebsiella spp (Benner et al., 1965; Musher et al., 1979; Murray and Moellering, 1982; Srinivasan et al., 2012; Silva et al., 2016).

Basically, SCVs have been determined as a subpopulation characterized by their distinct phenotypic properties, such as atypical colony morphologies including the reduced colony size (Proctor et al., 1995). Their decreased growth rate is thought to contribute to their inherent resistance, given that the decelerated growth dynamics potentially hinder the effectiveness of antibiotics geared towards rapidly proliferating cell populations (Proctor and von Humboldt, 1998). Furthermore, this phenomenon concurrently signifies decreased metabolic activity, which may engender modifications in cell wall permeability, drug uptake, or the modulation of efflux pump expression (Mitsuyama et al., 1997).

For electron transport chain-defective staphylococcal SCVs, lower efficacy of aminoglycosides known to be taken up through electrical potential across the cytoplasmic membrane (ΔΨ) was demonstrated, which is attributable to low ΔΨ (Baumert et al., 2002). These alterations could collectively contribute to enhancing resistance patterns. In this study, we observed an increase in the MIC values of amikacin, cefiderocol, and trimethoprim-sulfamethoxazole in the isolates recovered over time. This MIC increase was especially pronounced for amikacin in SCV phenotypes. Moreover, most antibiotics penetrate into host cells poorly, so the concentrations required to kill intracellularly persistent SCVs cannot be achieved (Proctor et al., 2006).

SCVs, known for their inducible formation through in vitro processes involving various agents, including antibiotics (Benner et al., 1965), have exhibited a propensity for increased persistence and adaptability when confronted with challenging environments (Li et al., 2016). An enhanced ability to form biofilms on biotic and abiotic surfaces has been shown for SCVs of different bacterial species (Haussler et al., 2003b; Webb et al., 2004; Al Laham et al., 2007; Allegrucci and Sauer, 2007; Millette et al., 2023). The substantial implication of SCVs extends to their involvement in biofilm development, as biofilms effectively shield bacteria from harsh host environments, thereby complicating the elucidation of drug resistance mechanisms within biofilm structures (Craft et al., 2019). Biofilms not only confer protection against host immune defenses but also serve as reservoirs for persistent infections and recurrent episodes (Mirani et al., 2015). The impact of SCV phenotype on biofilm formation in in Klebsiella remains to be elucidated in further studies.

Furthermore, the emergence of SCVs could plausibly be due to selection pressure from antibiotic regimens or other host-associated factors, e.g., host cationic peptides. Consistent with the case that was the subject of our study, the higher frequency of SCVs in isolates from chronic and recurrent infections compared to acute infections suggests a potential role for these variants in evading host immune responses and antimicrobial treatments (Proctor et al., 2006). In the context of our study, the emergence of SCVs after the initiation of cefiderocol treatment while already undergoing antibiotic therapy could be construed as a form of in vivo or in host induction.

The third noteworthy finding from our study underscores the inherent instability of SCVs. This dynamic interplay between stable and unstable SCVs is still poorly understood and its elucidation may contribute to a deeper understanding of their role in infection in general and persistence phenomena in particular (Becker et al., 2006). Despite comprehensive explorations largely focusing on staphylococci, a lack of investigations concerning Klebsiella spp. persists, and requires attention.

The observed instability among SCVs, combined with distinct antibiotic susceptibility profiles across phenotypes, increases the significance of investigating SCV plasticity (Proctor et al., 1995). Stable SCVs represent a long-term adaptation strategy, whereas their unstable counterparts may arise as stress-induced variants that result from rapid adaptation to fluctuating environments (Tuchscherr et al., 2010, Tuchscherr et al., 2011, Tuchscherr et al., 2015). This inherent instability potentially serves as a mechanism for evading host immune responses and circumventing antibiotic interventions (Tuchscherr et al., 2015). Furthermore, the involvement of epigenetic modifications, including alterations in DNA methylation patterns, could significantly influence SCV stability (Guerillot et al., 2019). In addition, regulatory systems, such as two-component systems and quorum sensing, play a crucial role in SCV formation by modulating bacterial behavior and adaptation. Disruption or dysregulation of these systems could lead to the emergence of SCVs with altered phenotypic properties (Pader et al., 2014). Due to instability, slow-growing SCVs may generate mutants that exhibit a faster growth rate than usual (Brandis et al., 2017). In instances of reversion to the wild type, rapidly growing mutant revertants may demonstrate either the loss or preservation of antibiotic resistance (Brandis et al., 2017).

A high mutation rate might favor the emergence of SCVs (Schaaff et al., 2003) and also explain the emergence of antibiotic resistance as a result of antibiotic selective pressure and the adaptation of hypermutable strains in patients, especially CF patients (Prunier et al., 2003). CF-like chronic infections have been shown to specifically contribute to the development of bacterial mutations (Smith et al., 2006). Hypermutation could result in a subpopulation of bacteria that temporarily does not grow, thus leading to persistence (Witzany et al., 2022). Additionally, an increase in the prevalence of mutator bacterial strains with deficient DNA mismatch repair (MMR) system has been detected in CF patients, who are used as a reservoir for mutation (Mena et al., 2008). To our best knowledge, we were unable to identify any instance in the available literature wherein a solitary SCV colony has given rise to four distinct colonies exhibiting disparate morphologies. Accordingly, we suggest the designation “phenotypic hyper-splitting” for this distinctive phenomenon.

We described in this study unprecedented phenotypic attributes and primarily focused on in vitro experiments. Therefore, the clinical relevance of our findings necessitates validation through animal models and clinical sample analyses. In this context, macrophage and neutrophil assays would be valuable for assessing both the extent of immune response and the presence of persistent cells. Moreover, the determination of the auxotrophism (Kriegeskorte et al., 2014; Becker, 2023) of K. pneumoniae SCVs and of the molecular mechanisms that drive SCV formation and the resulting antibiotic resistance in this species require further investigation. Integrating a comprehensive range of approaches encompassing genomics, transcriptomics, and proteomics, the utilization of experimental evolutionary models can yield valuable insights into the genetic determinants and regulatory networks orchestrating SCV phenotypes.

The genomic analysis conducted in this study has revealed clonality among all 14 isolates. Further exploration is warranted to uncover the intricate molecular mechanisms underlying phenotypic hyper-splitting and to elucidate the potential pathogenic implications of this phenomenon. To better understand the formation of the SCV phenotype especially in Gram-negative pathogens, efforts need to be intensified (i) to improve the detection and characterization of SCVs recovered from clinical samples and (ii) to elucidate their clinical impact.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ebi.ac.uk/ena, PRJEB71325.

Ethical approval was not required for the study involving humans in accordance with the local legislation and institutional requirements. Written informed consent to participate in this study was not required from the participants or the participants’ legal guardians/next of kin in accordance with the national legislation and the institutional requirements.

ED: Writing – original draft, Writing – review & editing, Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Software, Visualization. KSy: Data curation, Formal analysis, Investigation, Methodology, Software, Writing – review & editing. SH: Data curation, Formal analysis, Software, Visualization, Writing – review & editing. EE: Writing – review & editing, Investigation, Data curation. GW: Validation, Writing – review & editing. RP: Validation, Writing – review & editing. JB: Validation, Writing – review & editing. EI: Writing – review & editing, Validation, Project administration, Methodology, Conceptualization. KSc: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing – review & editing. KB: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing – review & editing.