A Klebsiella pneumoniae ST307 outbreak clone from Germany demonstrates features of extensive drug resistance, hypermucoviscosity, and enhanced iron acquisition

Antibiotic-resistant Klebsiella pneumoniae are a major cause of hospital- and community-acquired infections, including sepsis, liver abscess, and pneumonia, driven mainly by the emergence of successful high-risk clonal lineages. The K. pneumoniae sequence type (ST) 307 lineage has appeared in several different parts of the world after first being described in Europe in 2008. From June to October 2019, we recorded an outbreak of an extensively drug-resistant ST307 lineage in four medical facilities in north-eastern Germany. Here, we investigated these isolates and those from subsequent cases in the same facilities. We performed whole-genome sequencing to study phylogenetics, microevolution, and plasmid transmission, as well as phenotypic experiments including growth curves, hypermucoviscosity, siderophore secretion, biofilm formation, desiccation resilience, serum survival, and heavy metal resistance for an in-depth characterization of this outbreak clone. Phylogenetics suggest a homogenous phylogram with several sub-clades containing either isolates from only one patient or isolates originating from different patients, suggesting inter-patient transmission. We identified three large resistance plasmids, carrying either NDM-1, CTX-M-15, or OXA-48, which K. pneumoniae ST307 likely donated to other K. pneumoniae isolates of different STs and even other bacterial species (e.g., Enterobacter cloacae) within the clinical settings. Several chromosomally and plasmid-encoded, hypervirulence-associated virulence factors (e.g., yersiniabactin, metabolite transporter, aerobactin, and heavy metal resistance genes) were identified in addition. While growth, biofilm formation, desiccation resilience, serum survival, and heavy metal resistance were comparable to several control strains, results from siderophore secretion and hypermucoviscosity experiments revealed superiority of the ST307 clone, similar to an archetypical, hypervirulent K. pneumoniae strain (hvKP1). The combination of extensive drug resistance and virulence, partly conferred through a “mosaic” plasmid carrying both antibiotic resistance and hypervirulence-associated features, demonstrates serious public health implications.

In this study, we analyzed carbapenemase-producing ST307 isolates, which have been recovered from screening and clinical samples within the course of an outbreak [19] that took place in four medical facilities in Western Pomerania, Germany, from June 2019 to October 2019. Additional cases were detected after the actual outbreak in the beginning of 2020. These isolates were characterized as carrying NDM-1 and OXA-48 carbapenemaseencoding genes, mostly simultaneously, and tested colistin-resistant. We performed whole-genome sequencing and phenotypic experiments to enable functional genomics for the in-depth understanding of this XDR outbreak clonal lineage.

Sequenced isolates and metadata
Between June 2019 and February 2020, we investigated 56 enterobacterial isolates from 25 different patients involved in the outbreak. In addition to K. pneumoniae, we included all other Enterobacteriaceae that matched the carbapenem-resistant phenotype. Most isolates were obtained from rectal (n = 23) or throat/tracheal secretion (n = 17) swabs collected as part of an extensive surveillance program in the affected clinical institutions (Additional file 2: Table S1). Initial antibiotic susceptibility testing (AST) was performed using the VITEK 2 (bioMérieux, Nürtingen, Germany) system and 96-well plate broth microdilution (Merlin, Bornheim-Hersel, Germany). Bacterial species were initially identified by MALDI-TOF MS (VITEK MS, bioMérieux, Nürtingen, Germany). For rapid detection of carbapenemase genes, a loop-mediated isothermal amplification (LAMP) assay (eazyplex SuperBug CRE, AmplexDiagnostics, Gars, Germany) was subsequently included in the University Medicine laboratory diagnostic program [19]. Almost 59% (33/56) of the isolates were assigned to "infection" samples. Clinical data and outcomes of 17 initial cases have been published elsewhere [19]. Most of these had severe underlying diseases. Patients were treated with ceftazidime-avibactam/aztreonam with synergistic activity. By October, six patients had died. Causal associations with the outbreak clone as well as clinical outcomes of other cases are still under investigation, however. Epidemiologic links among the different institutions were identified (Additional file 2: Table S1). Samples were incubated overnight on CHROMID CARBA and CHROMID ESBL agar plates (bioMérieux, Nürtingen, Germany), and single colonies were picked for identification with VITEK MS (bioMérieux, Nürtingen, Germany). Antimicrobial susceptibility testing was carried out using VITEK 2 (bioMérieux, Nürtingen, Germany), and in addition, 96-well plate broth microdilution was performed for determination of colistin MICs (Merlin, Bornheim-Hersel, Germany).
3.13.1/3.14.0 [21]. As part of the pipeline, trimmed reads were subsampled to assemble at a maximum coverage of 100×. Besides the polishing step as part of the shovill pipeline, assemblies underwent an additional polishing step. For this, all trimmed reads were mapped back to the contigs using bwa v. 0.7.17 [22]. The obtained SAM/BAM files were sorted with Samtools v. 1.9 [23] and optical duplicates marked with GATK v. 4.1.2.0 [24]. Finally, variants were called with Pilon v. 1.23 [25]. The genomes of strains for which additional long-read sequencing data were obtained were hybridassembled with Unicycler v. 0.4.8 [26]. To verify the "hybrid" nature of plasmid 1, we mapped the long-reads of PBIO1953 back to the assembly using minimap2 v. 2.17 [27] and visualized the alignment with Tablet v. 1.19.09.03 [28] (Additional file 1: Fig. S1). The assembly graphs of putative plasmid recipient isolates and the closed reference isolate PBIO1953 were inspected with Bandage v. 0.8.1 [29] and its integrated BLAST hit (Megablast, ≥ 99% identity, E value 1e −10) visualization (Additional file 1: Fig. S2). Genome quality and completeness were assessed with CheckM v. 1.0.13 [30]. We used Prokka v. 1.14.1 [31] to annotate draft and finished genomes automatically.

Pathway analysis of variants
Single-nucleotide polymorphisms detected among the K. pneumoniae ST307 genomes were filtered for missense, frameshift, and stop-gained variants and respective genes subjected to metabolic pathway analysis using the EcoCyc (https://ecocyc.org/) [52]

Growth curves
Growth curves in LB medium were performed using standardized protocols. Experiments were performed using three technical replicates and three biological replicates [56]. An E. coli K-12 strain (W3110) was used as control. Growth rates were calculated as follows: μ = (ln(CFU/mL t 1 ) − ln(CFU/mL t 0 ))/t 1 − t 0 .

Hypermucoviscosity
Hypermucoviscosity experiments were performed using the string test. Strings of 5 mm or longer that formed after stretching on the tip of a sterile inoculation loop were defined positive [6]. Experiments were performed with three technical replicates and three biological replicates.

Siderophore secretion
We analyzed the study's set of isolates for their ability to secrete siderophores using an adapted method described by Schwyn and Neilands [57]. Fifty microliters of overnight cultures of the isolates was grown in 5 mL fresh LB medium to an OD 600 of 0.6. Five microliters of this culture was put on agar plates containing chrome azurol S-iron(III)-hexadecyltrimethylammonium bromide and incubated overnight at 37°C. Iron uptake was determined visually by color shift from blue to yellow the following day. One ST131 E. coli strain (IMT21183) and one K-12 E. coli strain (W3110) were included as controls [58]. Experiments were performed with three technical replicates and three biological replicates.

Biofilm formation
We used the same set of isolates as described above and performed a biofilm formation experiment as described in previous reports [59]. Ten microliters of overnight cultures of all isolates was transferred to 1 mL LB medium and cultured at 37°C until OD 600 values of 0.6-0.8 were reached. One hundred microliters of culture per well was transferred to a 96-well microtiter plate. After 3 days of static culture at 37°C, planktonic cells were removed from the liquid medium. The wells were washed three times with 150 μL of double-distilled water, and the majority of the biofilms was stained with 150 μL of 0.1% crystal violet (CV) for 30 min. Then, the unbound dye was removed, and the plates were again washed. Finally, the CV binding to the biofilm was dissolved in 150 μL of 95% ethanol for 30 min, and biofilm formation was quantified by measuring the absorbance at OD 590 with a microplate reader (Fluostar Omega, BMG Labtech, Ortenberg, Germany). Experiments were performed with three technical replicates on individual 96-well microtiter plates and three biological replicates. One biofilm-negative (PBIO729) and one positive control (W3110) [56] were included.

Serum resistance
We performed serum resistance experiments in human serum (Pan-Biotech GmbH, Germany) as described previously [60]. We inoculated 5 μL of overnight culture in 495 μL fresh LB medium and incubated at 37°C for 1.5 h. Inoculum was centrifuged for 3 min and resuspended in 1 mL of sterile 1× PBS. Thirty microliters was added in triplicates to 96-well microtiter plates containing 270 μL of 50% human serum. Thirty microliters of the sample was collected from each well, serially diluted, plated on LB plates, grown at 37°C overnight, and counted the next day (0-h count). The 96-well microtiter plates were incubated for 4 h at 37°C. Following incubation, 30 μL of the culture was plated, incubated, and again counted (4-h count). Growth in serum was obtained by determining differences in the CFU after 4 h of incubation compared to the 0-h count. Experiments were performed with three technical replicates on individual 96-well plates and three biological replicates. An E. coli serum-resistant control (PBIO1289, ST1159 [61]) was included.

Desiccation tolerance
Desiccation tolerance experiments for the study's set of isolates were performed as described previously with some modifications [59]. A single colony was cultured in 10 mL liquid LB broth until bacterial cells reached an OD 600 value of 0.6-0.8. One hundred microliters was serially diluted, plated on LB plates, grown at 37°C overnight, and counted the next day (C0d, 0-day count). Another 100 μL of the same culture was transferred into a 96-well microtiter plate. Subsequently, the plate was transferred into a sterile dryer with dehydrated silica gel. The dryer was placed in a sterile incubator (Mytron, Heilbad Heiligenstadt, Germany), which was kept at a constant temperature of 37°C. After 6 days of drying, the 96-well microtiter plate was removed, 100 μL/well of fresh medium was added, and the plate was cultured with 200 rpm shaking at 37°C for 3 h. One hundred microliters was collected from each well, serially diluted, plated on LB plates, grown at 37°C overnight, and counted the next day (C6d, 6-day count). Experiments were performed with three technical replicates on individual 96-well microtiter plates and three biological replicates. An E. coli K-12 strain (W3110) was included as control.

Heavy metal tolerance
Overnight cultures of the study's set of isolates were adjusted to McFarland standard 0.5, and 50 μL of a 1:200 dilution of adjusted suspensions in Mueller-Hinton broth (Roth, Karlsruhe, Germany) was used as inoculum for incubations for 16 to 20 h at 37°C in heavy metalcontaining microtiter plates (Merlin Biocide plates, Bornheim-Hersel, Germany). The plates contained a wide range of concentrations of three heavy metals: zinc chloride [4-8192 μg/mL], copper sulfate [32-8192 μg/ mL], and silver nitrate [0.5-64 μg/mL]. We used a sealing tape to prevent dehydration of the plates. After incubation, the minimum inhibitory concentration was determined visually and reported as the tolerance breakpoint. Experiments were performed with three technical and three biological replicates. E. coli ATCC25922 was used as control isolate.

Statistics
Statistics were performed using GraphPad Prism 8.0 (https://www.graphpad.com/). After investigating Gaussian distributions, the non-parametric Kruskal-Wallis test [62] was applied for multiple comparisons of bacterial groups using median values. Bonferroni adjustment was applied, which resulted in corrected p values of p < 0.016 to assess significant changes [63]. Pairwise comparison of growth rates (μ) between all K. pneumoniae ST307 isolates and PBIO1961 was performed using the Mann-Whitney U test (p < 0.05).

Genomic analysis and phylogeny
The XDR K. pneumoniae outbreak clone was first detected at the University Medicine Greifswald on June 25, 2019 (Additional file 2: Table S1), following bacterial screening of a tracheal secretion sample [19]. Of 52 K. pneumoniae, 44 belonged to sequence type (ST) 307, three to ST395, three to ST11, and one each to ST405 and ST147. The two E. coli were ST405 and ST362, whereas C. freundii was a ST153 and E. cloacae a ST45 isolate (Additional file 2: Table S1). We focused mainly on the phylogenetics and phenotypes of the ST307 outbreak but included accessory and non-ST307 genomes to investigate transmission of resistance plasmids within the bacterial species and to others.
When comparing all 44 ST307 against the closed reference genome of PBIO1953, our analysis revealed 22 singlenucleotide polymorphisms (SNPs) at most (minimum, 6 [PBIO1958, PBIO1960, PBIO1932]; maximum, 22 [PBIO2004]; median, 11). As expected, the phylogram shows a homogenous picture ( Fig. 1; Additional file 1: Figure S4) with several sub-clades. There are sub-clades that include only isolates originating from the same patient (clades B [8-11 SNPs compared to PBIO1953] and C [6-8 SNPs compared to PBIO1953]) and sub-clades that comprise isolates from different patients (clades A [14-17 SNPs compared to PBIO1953] and D [11-15 SNPs compared to PBIO1953]) suggesting recent transmission between patients. Epidemiologic data support these results, for example for sub-clade D: PT17 with isolates PBIO1965, PBIO1969, PBIO1974, PBIO1991, PBIO1995, and PT21, from whom we obtained isolates PBIO1936 and PBIO1970, stayed on the same ward during the same time (September 2019). On the contrary, PT06 with isolate number PBIO1964 was present on a different ward during September 2019. Note, however, that PT06 was transferred later on. Interestingly, this patient underwent endoscopy examination with the same endoscope used for PT17 and PT21. It thus seems possible that K. pneumoniae ST307 was transmitted among patients either by crosscontamination through healthcare workers and surfaces or by an endoscope as has been described previously [64].
When we placed the outbreak phylogeny in a global context (Additional file 1: Fig. S4), we noticed that a cluster of KPC-producing ST307 genomes originally obtained from the United Kingdom (UK) (Additional file 2: Table S2) was the phylogenetically closest to our ST307 outbreak isolates.
We then explored the distribution and character of SNPs among the ST307 genomes (with PBIO1953 as reference) further (Additional file 1: Fig. S5). All polymorphisms that were not in coding sequences (CDS) or not assigned as missense, frameshift, or stopgained variants were excluded from our subsequent analysis. We further excluded all insertion and/or deletion mutations (Indels). Variants accumulated almost uniquely in chromosomally encoded genes (59/ 66). When analyzing the data of 44 annotated genes after exclusion of hypothetical proteins in EcoCyc, we noticed that pathways related to membrane transport (14/44), regulation and signal transduction (9/44), amino acid and sugar metabolism (12/44), DNAreplication/conjugation (5/44), and lipopolysaccharide biosynthesis (4/44) were often affected by mutations. Thirty-seven genomes harbored a missense SNP in the conjugation gene traI (plasmid 3) and, simultaneously, the topoisomerase gene gyrA, which are both involved in plasmid conjugation and transfer [65,66]. Interestingly, all potential K. pneumoniae ST307 plasmid donors belonged to this set of genomes.
We often found variants in genes encoding for membrane efflux. While in sub-clade C, sotB was affected, in other clusters, we observed missense mutations in phoE, gltC, and ompC. This might be an example for phenocopy in isolates of different sub-clades. The nitrate/nitrite sensor gene narX, differentiating sub-clades B (PT19) and D (various patients), and sensor protein pmrB, differentiating sub-clade C (PT23) from other sub-clades (various patients), were repeatedly and independently mutated among different patients. In addition, two other genes, narI and narJ, displayed SNPs; both are involved in the regulation of anaerobic respiratory gene expression in response to nitrate and nitrite.
Several genomes obtained from the same patient demonstrated identical variants not present in other genomes, which could be explained by either a disruption of the infection chain between patients or the non-advantageous character of the mutation for dissemination. One example is the variant in btsT, a gene involved in pyruvate uptake and present in two isolates from the same patient over a period of 7 days. Interestingly, the number of missense/nonsense SNPs did not significantly increase over time during the course of the outbreak compared to the earliest isolate PBIO1953 (for example PBIO1955, 8 SNPs; PBIO1956, 11 SNPs; PBIO1957, 11 SNPs; PBIO2011, 11 SNPs; PBIO2012, 13 SNPs; and PBIO2018, 12 SNPs; Additional file 1: Fig. S5, top).
All ST307, with the exception of PBIO2003, carried bla NDM-1 , bla OXA-48 , and bla CTX-M-15 resistance genes simultaneously, which was consistent with their phenotypes. Due to the fact that mcr-genes were not present, phenotypic resistance against colistin could not be explained by the expression of such. We identified several missense mutations in the two-component systems PhoP/PhoQ (phoQ: 89T>A  His188Pro]). Interestingly, one missense mutation in pmrB (604C>A [Gln202Lys]) was exclusively present in isolates from patient PT23. While the amino acid substitutions in PmrA/PmrB (Ala41Thr/Leu213Met; Gly256Arg), together with an insertional inactivation of mgrB, were previously reported in a colistin-resistant K. pneumoniae ST307 isolate in 2015 [14], both this study's colistin-susceptible and colistinresistant isolates showed an uninterrupted mgrB, whose gene product is a small negative regulator of PhoQ. The colistin-resistance phenotype is possibly explained by the combination of several mutations in chromosomally encoded genes.
When comparing these plasmid sequences to the other genomes (Fig. 3, Additional file 1: Fig. S2, and Additional file 1: Fig. S3B), we noticed that two non-K. pneumoniae ST307 isolates were also positive for plasmid 1 (NDM-1): (i) PBIO1963 is an E. coli ST362 isolate obtained on September 10, 2019, from patient PT07, who also carried K. pneumoniae ST307 (PBIO1955-obtained on July 24, 2019). We suggest that K. pneumoniae ST307 donated plasmid 1 to PBIO1963 within the patient. (ii) PBIO1961 is a K. pneumoniae ST395 isolate. We did not obtain any additional carbapenem-resistant isolates from this patient (PT12), which indicates interpatient transfer of resistance plasmids (Fig. 3); however, our epidemiologic data do not support this suggestion as no other ST307-positive patient was simultaneously present on the same ward. Plasmids 2 and 4 were also present in some non-ST307 genomes. K. pneumoniae ST405 (PBIO1979) isolated on September 25, 2019, from patient PT21 carried plasmid 2 (CTX-M-15). From the same patient, we isolated K. pneumoniae ST307 (PBIO1936) on September 18, 2019, again suggesting plasmid intra-bacterial species transfer within this patient. Plasmid 4 (OXA-48) was present in K. pneumoniae ST147 (PBIO1999-isolated from patient PT16 on October 2, 2019) and a K. pneumoniae ST307 isolate (PBIO2000) obtained from the same patient on September 4, 2019. Plasmid 4 was further carried by PBIO1966, a C. freundii ST153 isolate obtained on September 20, 2019, from patient PT02, who additionally carried two K. pneumoniae ST307 isolates (PBIO1958: August 5, 2019; PBIO1932: September 14, 2019). It was also present in E. cloacae ST45 (PBIO2014: December 24, 2019), which was the only sequenced isolate from patient PT20. Again, these are examples for putative inter-bacterial plasmid transfers within a patient and among patients, respectively (Fig. 3). Inter-patient transfer for PT20 is further supported by the epidemiologic data: PBIO2012 (ST307) was isolated on December Fig. 2 Synteny plot of hybrid resistance/virulence plasmid 1 (pPBIO1953_NDM-1) and resistance plasmid 2 (pPBIO1953_CTX-M-15). The plot depicts pairwise BlastN comparisons (E value 1e−10) between pPBIO1953 plasmid 1 and 2 and pK2044 and pKCTC2242. Alignment lengths were required to be at least 1% of the smaller replicon in the comparison to be included. Direct comparisons are colored with red hues whereas reverse comparisons are colored with blue hues. Boxes on top depict CDS on the forward strand, and those at the bottom depict CDS on the reverse strand. Boxes with a dotted outline are annotated as pseudo CDS. Plasmid 1 of PBIO1953 combines virulence features, such as aerobactin, with acquired antimicrobial resistance genes (e.g., bla NDM-1 ), which are missing from typical virulence plasmids of hypervirulent strains NTUH-K2044 and KCTC 2242. Note that PBIO1953 plasmid 2 carries several metal resistance genes (pco and sil genes) present on hvKP virulence plasmids but absent from plasmid 1 16, 2019, from a putative donor patient (PT10), who stayed on the same ward as PT20.

Phenotypic experiments
We then investigated the hypothesis that K. pneumoniae ST307 outbreak isolates combined high-level antibiotic resistance with fitness, resilience, and virulence features. In several phenotypic experiments, which were selected based on their relevance in the clinical setting and on the presence of resistance and virulence genes in ST307, we investigated five K. pneumoniae ST307 isolates from different time points during the outbreak and different patients and hospitals (outbreak isolates), two K. pneumoniae ST395, and one K. pneumoniae ST405 (subsequently termed as the internal group). These were compared to four external K. pneumoniae isolates with Fig. 3 Putative plasmid transfer among K. pneumoniae isolates and between different bacterial genera. Solid arrows indicate putative intra-species transfer, whereas dashed arrows indicate inter-bacterial transfer. The putative inter-patient transfer is illustrated through a question mark. Note that from PT12 (PBIO1961), no additional carbapenem-resistant isolates were obtained and that no other ST307-positive patient was simultaneously present on the same ward whereas inter-patient transfer for PT20 (PBIO2014) is further supported by the epidemiologic data. Line colors match the colors of transferred virulence/resistance plasmids and are also shown next to the isolate names as circles. Closed circles-complete plasmid backbone (Megablast hits with identity ≥ 99%: coverage ≥ 99%). Open circles-incomplete plasmid backbone (Megablast hits with identity ≥ 99%: coverage > 77% and < 99% (Additional file 2: Table S4)). a The phylogenetic tree is based on distance-corrected MinHash dissimilarities between sequenced genomes and was inferred with FastME as part of JolyTree. Shown at the branches is the rate of elementary quartets (REQ) for values ≥ 0.75 (circle diameter). b Timeline figure of plasmid transfer between July and December 2019 ST498, ST15, ST307, and ST86 including an archetypal, hypervirulent K. pneumoniae strain (hvKP1), which are unrelated to the outbreak. In addition, we included a partially plasmid-cured variant (PCV1935) and controls for each phenotypic experiment.
Hypermucoviscosity experiments revealed that all K. pneumoniae ST307 outbreak isolates showed strong hypermucoviscosity (≥ 5 mm) (Fig. 5a), whereas internal K. pneumoniae ST395 and ST405 isolates did not (p < 0.0001). Also, the external isolates showed a negative hypermucoid phenotype, with the exception of K. pneumoniae ST15 and, unsurprisingly, hvKP1. Hypermucoviscosity of PCV1935 was not significantly different from wildtype PBIO1935, which is interesting given that both plasmid-encoded rmpA and rmpA2 were lost during the curing process. On the contrary, PBIO1961, which carried plasmid 1 (Figs. 3, 4, and 5, and Additional file 1: Figure S3B and S7B), did not show hypermucoviscosity. This genome demonstrated a different rmpA2-truncation and K locus in comparison to the ST307 clone (Additional file 2: Table S1). Also consider that the different phenotypes might be due to synergy-dependent processes such that plasmids have reduced impacts on other genetic backgrounds than ST307. The hypermucoid phenotype appears to be a fine-tuned process. A recent study [68] showed that loss of rmpC, which is a newly identified gene that contributes to capsule regulation in hvKp, resulted in decreased capsule gene expression while simultaneously retaining hypermucoviscosity. Additional investigations will have to address further which regulatory mechanisms contribute to the hypermucoid phenotype in our outbreak clone. Hypermucoviscosity is associated with invasive and other aggressive types of infection, but recent literature suggests that this characteristic alone is not per se responsible for a hypervirulent phenotype in Klebsiella spp. and that both terms should not be used synonymously [5,69].
Similar to the hypermucoviscosity experiments, we observed significant differences between the K. pneumoniae ST307 outbreak isolates and the internal and external groups regarding their siderophore secretion capacities (p < 0.0001) (Fig. 5b, c). They all exhibited a significant higher secretion with average bleaching zone diameters of 20 mm, compared to 8.4 mm of internal and 10.8 mm of external isolates (positive control, 12 mm). This is likely due to the presence of the NDM-1plasmid-encoded aerobactin, underlined by the results shown for PCV1935, which demonstrated significantly reduced siderophore secretion (p = 0.0025) (Fig. 5b, c). Note that hvKP1 also showed increased siderophore secretion compared to the internal group (p = 0.0126) but in tendency less than the ST307 outbreak isolates, although this difference was not significant (p = 0.35). PBIO1961 registered within the result range of internal and external isolates. This is likely due to a missense mutation of plasmid-encoded iutA and/or the different character of yersiniabactin ("unknown" ybt-Additional file 2: Table S1). Given that hypermucoviscosity, aerobactin secretion, and the metabolite transporter PEG344 [70] are suggested key virulence features of hypervirulent Klebsiella strains during infection, the aboveaverage performance, in addition to extensive antibiotic resistance expression, provides one step closer to explaining why this outbreak evolved. PBIO2009 (external ST307 isolate), which does not possess typical hvKpassociated features like peg-344, iucA, and rmpA [11], showed negative hypermucoid and iron uptake phenotypes (Fig. 5a-c), possibly strengthening our assumption.
We were then interested in whether the K. pneumoniae ST307 outbreak isolates also showed sufficient capacity for survival and resilience in the clinical setting and host. Serum survival experiments revealed high survival rates for all isolates, and together with similar strong biofilm formation capacities (Fig. 4b, c), this suggests that the outbreak K. pneumoniae ST307 isolates had good abilities to resist clinical challenges. Comparable results were also obtained for desiccation resilience (Fig. 4d). Despite that internal and external groups showed a tendency to survive 6 days of desiccation at higher rates than K. pneumoniae ST307, the difference was not significant after applying Bonferroni correction (outbreak vs. internal group: p = 0.5488; outbreak vs. external group: p = 0.0730; internal vs. external group: p > 0.9999).
The genetic characterization revealed several heavy metal efflux genes on plasmid 2. Heavy metal compounds, such as zinc oxide or copper sulfate, are regularly used as feed supplements in livestock, e.g., for prevention of gastro-intestinal disorders and growth promotion in piglets [71], and co-selection of heavy metal and antimicrobial resistance has been increasingly reported [72,73]. We thus investigated the bacteria's tolerance by determining minimum inhibitory concentrations (MICs) to copper, zinc, and silver (Additional file 2: Table S1). A MIC value of 1024 μg/ mL for copper sulfate was obtained for almost all K. pneumoniae isolates, both outbreak, internal, and external group isolates (control ATCC25922, 256 μg/ mL), which is an average copper tolerance for Enterobacteriaceae [74]. Interestingly, hvKP1 showed a reduced MIC for copper sulfate (256 μg/mL). We observed similar results for zinc and silver for all isolates (MIC zinc chloride 512 μg/mL, control 256 μg/ mL; silver nitrate 4 μg/mL, control 4 μg/mL), again with the exception of hvKP1 demonstrating a reduced MIC for zinc (256 μg/mL).

Discussion
To the best of our knowledge, this is the first outbreak in Germany of a K. pneumoniae ST307 clone that produced both NDM-1 and OXA-48 carbapenemases and showed resistance against colistin. It was first detected at ). Yellow areas around colonies indicate siderophore secretion. Abbreviations and symbols: n.s., not significant; ****p value < 0.0001; ***p value < 0.001; **p value < 0.01 the University Medicine Greifswald in June 2019 from a tracheal secretion sample [19]. As we did not detect any mcr-genes, we suggest that colistin resistance is due to chromosomal point mutations including the twocomponent systems PhoP/PhoQ and PmrA/PmrB, which have been previously described in this context [67].
When placing our outbreak clone in a global frame, a cluster of KPC-producing ST307 genomes from the UK was the phylogenetically closest. Interestingly, this cluster was part of a study from 2017 revealing that these genomes harbor genetic features important for clinical and host adaptation, in particular glycogen synthesis [3]. Our outbreak isolates have seemingly developed different resistance phenotypes and virulence strategies, and the UK cluster is thus probably not the true, most likely recent common ancestor of the ST307 outbreak clone.
Although we did not unequivocally verify the hypervirulent character of this clone, it demonstrated hypermucoviscosity, iron uptake, and metabolite transporter capacities-which are relevant for invasive infection and assertiveness in different host environments [41,69,70]-comparable to an archetypal hvKp strain [54]. The fact that these key hypervirulence features in addition to disinfectant resistance are found on mutual virulence/resistance plasmids in extensively drug-resistant isolates is concerning and has tremendous public health implications as these mobile genetic elements may be transferred across different bacteria [75]. Our previous work, and those of others, suggests that the combination of high-level drug resistance and virulence is a good combination for the successful spread of bacterial pathogens [3,53,[76][77][78][79]. On the other hand, the co-carriage of plasmid-encoded heavy metal efflux genes did not significantly impact the phenotypic tolerance in the study's K. pneumoniae outbreak isolates, pointing towards that this capacity is less likely a major contributor to the clone's success in this outbreak situation.
We detected identical plasmids among ST307 and other K. pneumoniae isolates as well as other bacterial genera, exacerbating the threat this clone poses across clinical settings. Note, however, that it is possible that ST307 has a greater tolerance towards possible fitness costs of the carried plasmids, implying that donated plasmids might have reduced impacts in other genetic backgrounds [56]. We suggest that the K. pneumoniae ST307 isolates are the general plasmid donors; they were all isolated at an earlier date than the putative acceptor isolates and more independent variants accumulated in some putative acceptor plasmids compared to the K. pneumoniae ST307 donor plasmids (Additional file 2: Table S3). The fact that ST395 occurred three times among all isolates but only ST395 PBIO1961 was positive for plasmid 1 additionally strengthens this assumption.

Conclusions
While the emergence of XDR, virulent, and fit pathogens is worrisome, our study helps to implement control measures and calls for prospective surveillance strategies that take the emergence of "converging" cKp and hvKp pathotypes into account.
Additional file 2: Table S1. Overview of strains investigated in this study. Table S2. Overview of publicly available Klebsiella pneumoniae ST307 genomes obtained from online sources. Table S3. Overview of accumulated variants in transferred plasmids. Table S4. Coverage of isolate replicons.

Funding
We acknowledge support for the Article Processing Charge from the DFG (German Research Foundation, 393148499) and the Open Access Publication Fund of the University of Greifswald. The funding sources had no influence on the design of the study and collection, analysis, and interpretation of data, and writing of the manuscript. Open Access funding enabled and organized by Projekt DEAL.

Availability of data and materials
The experimental and computational data that support the findings of this research are available in this article and its supplementary information files. The genomic data have been deposited in the European Nucleotide Archive (ENA) at EMBL-EBI under accession number PRJEB37933 (https://www.ebi.ac. uk/ena/browser/view/PRJEB37933) [84].
Ethics approval and consent to participate Ethical approval was given by the ethics committee of the University of Greifswald, Germany (BB 133/20). Informed patient consent was waived as