Case Report

In February 2022, a 62-year-old patient who was positive for SARS-CoV-2 was transferred from a long-term care facility to Policlinico Umberto I (PUI) in Rome. The patient was hospitalized for 95 days initially in the COVID-19 intensive care unit (ICU) and then in the general ICU until death, which was caused by gastrointestinal bleeding. The patient’s medical history was notable for bipolar disorder, obesity, inflammatory bowel disease, type 2 diabetes, respiratory failure, chronic kidney failure, and heart failure. Six months before transfer to PUI, the patient had a percutaneous endoscopic gastrostomy performed because of severe gastrointestinal bleeding caused by underlying inflammatory bowel disease. During the patient’s hospitalization at PUI, we isolated a total of 5 different K. pneumoniae strains from patient rectal swab samples and respiratory tract specimens.

Isolation of K. pneumoniae Strains and Susceptibility Testing

We identified the 5 K. pneumoniae strains by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (Bruker Daltonik GmbH, https://www.bruker.com). We identified carbapenemase genes by PCR using the GeneXpert system (Cepheid, https://www.cepheid.com) and used the lateral flow immunochromatography systems (NG-Test CARBA 5; Hardy Diagnostics, https://www.hardydiagnostics.com). We tested antimicrobial drug susceptibility by using the MicroScan WalkAway system (Beckman Coulter, Inc., https://beckman.com). We used gradient strips (Liofilchem, https://www.liofilchem.com) to test MVB and CZA MICs and the Compact Antimicrobial Susceptibility Panel (ComASP; Liofilchem) to test FDC MICs.

Whole-Genome Sequencing and Assembly

We performed whole-genome sequencing for each isolate. We purified genomic DNA by using the Isolate II Genomic DNA Kit (Bioline, https://www.bioline.com). We sent DNA to an external service to perform short-read Illumina sequencing (Illumina Inc., https://www.illumina.com). We obtained long reads by using the MinION Mk1C sequencing platform (Oxford Nanopore Technologies, https://nanoporetech.com). We extracted DNA for long reads by using the Wizard HMW DNA Extraction Kit (Promega, https://www.promega.com) and prepared libraries by using the Rapid Barcoding Kit 96; we sequenced libraries on R9.4.1 flow cells (Oxford Nanopore Technologies). We performed long-read assemblies by using Flye (16) with standard parameters. We integrated Illumina reads and Oxford Nanopore Technologies assemblies by using the Unicycler tool (17) in normal bridging mode and refined results by using the Bandage tool (18).

Genomic and Phylogenetic Analyses

We analyzed single-nucleotide polymorphisms (SNPs) among the 5 sequenced genomes in this study by using the Snippy tool (https://github.com/tseemann/snippy). We annotated the 5 genomes by using Rapid Annotation using Subsystem Technology and compared by using SEED software (19).

We used Prokka software (20) to annotate 133 genomes belonging to ST512: 5 sequences from this study, 12 from a previous study performed at PUI (8), and 116 downloaded from the Pasteur Institute BIGSdb database (https://bigsdb.pasteur.fr/klebsiella). We analyzed the resulting general feature formats by using Roary software (21) to build a core genome alignment and generated a consensus phylogenetic tree by using 1,000 ultrafast bootstraps (22) in IQ-TREE 2 (23) and the transversion model plus base frequency plus proportion of invariable sites nucleotide substitution model (24). We visualized tree and metadata by using Microreact (25) and adjusted those data by using open source InkScape software (https://www.inkscape.org). We assessed all genomes for replicons by using PlasmidFinder (26), for capsular polysaccharide and lipopolysaccharide loci by using Kaptive (27), and for virulence and resistance gene content by using Kleborate (28) software.

Cloning bla_KPC-154 in pCR-Blunt II TOPO Vector

We cloned the novel β-lactamase bla_KPC-154 allele from isolate 6099 into pCR-Blunt II TOPO-NeoR Vector and transformed TOP10 E. coli cells (both Thermo Fisher Scientific, https://www.thermofisher.com); we confirmed correctness of the cloned insert by Sanger sequencing. We tested the KPC-154 TOP10 E. coli clone for antimicrobial drug susceptibility by using the MicroScan system and measured CZA MICs by gradient tests (Liofilchem) as previously described.

OmpK36 Variant Modeling

We predicted structures of the outer membrane porin (Omp) K36 from isolate 0296 in silico by using Alphafold2 on the European Galaxy server (https://usegalaxy.eu) and analyzed those structures by using UCSF ChimeraX (29) for both the cartoon (ribbon) and surface images. We compared the structures with chain B in the crystal structure of OmpK36 from a K. pneumoniae ST258 clinical isolate that has a GD amino acid insertion (30). We used the Orientations of Proteins in Membranes database (31) to obtain spatial arrangements of the protein structures in lipid bilayers.

Data Availability

We submitted the sequences of the strains analyzed in this study to GenBank. Whole-genome sequences are under Bioproject no. PRJNA992043 and complete plasmid sequences under accession nos. OQ096263 (pKpQIL-6099), OQ282880 (pIncA-6379), and OQ282881 (pIncA-0296).

Ethics Approval

According to the hospital’s routine practice, the patient or his relatives gave informed consent to share data for research purposes during hospital admission. The study protocol was approved by the Ethics Committee of Azienda Ospedaliero-Universitaria Policlinico Umberto I (approval no. 109/2020).

K. pneumoniae Strain Descriptions

Figure 1

Figure 1. Timeline of colonization and infection of 1 patient by Klebsiella pneumoniae clones in study of genotypic evolution of K. pneumoniaesequence type 512 during ceftazidime/avibactam,...

On the first day of the patient's hospitalization at PUI, surveillance rectal swab samples tested positive for K. pneumoniae (strain 6379) that produced both KPC and Verona integron-encoded metallo-β-lactamase (VIM) (Figure 1). The patient had been treated with CZA at the long-term care facility for a previous carbapenem-resistant K. pneumoniae bloodstream infection; treatment was discontinued upon admission to PUI.

After 17 days of hospitalization, a respiratory tract infection caused by Providencia stuartii (>100,000 CFU/mL in bronchoalveolar lavage fluid) developed in the patient, accompanied by pleural effusion, which we treated with CZA. After 1 week of treatment, we isolated 2 CZA-resistant, meropenem-susceptible K. pneumoniae strains from patient rectal swab samples but not respiratory tract specimens; both strains produced KPC but not VIM and had dimorphic colony phenotypes: white (strain 1186W) and transparent (strain 1186T). Because of potential pleural infection, we maintained CZA therapy for ≈40 days, resulting in complete eradication of P. stuartii from the respiratory tract.

On day 56, the patient was SARS-CoV-2 negative and was transferred to the general ICU at PUI. Respiratory tract samples were negative for P. stuartii but positive (>1,000 CFU/mL lavage fluid) for CZA-resistant K. pneumoniae (strain 6099) that produced KPC. CZA therapy was stopped, and MVB treatment was begun and continued until day 75. On day 75, respiratory tract samples tested negative for K. pneumoniae but positive for carbapenem-resistant A. baumannii (1,000 CFU/mL lavage fluid). Because of subsequent worsening respiratory conditions and chest imaging suggestive of new onset pneumonia, we replaced MVB therapy with FDC therapy (day 79) to treat A. baumannii pulmonary infection. On day 88, during FDC treatment, respiratory tract samples tested positive for both A. baumannii and K. pneumoniae (strain 0296) (1,000 CFU/mL lavage fluid each); strain 0296 was positive for KPC and VIM carbapenemases.

Phylogeny, Antimicrobial Resistance, and General Features of ST512

Figure 2

Figure 2. Phylogenetic analysis and genetic features of Klebsiella pneumoniaeST512 in study of the strain’s genotypic evolution during ceftazidime/avibactam, meropenem/vaborbactam, and cefiderocol treatment, Italy. Phylogenetic tree was constructed from 4,654...

We assigned the 5 K. pneumoniae isolates from the patient (6379, 1186W, and 1186T from rectal swab samples, 6099 and 0296 from respiratory tract samples) to ST512 by using in silico multilocus sequence typing of whole-genome sequences. We aligned the complete genomes of those strains against 116 ST512 genomes retrieved from the Pasteur Institute K. pneumoniae BIGSdb database (December 7, 2022) and 12 ST512 genomes previously identified at PUI (8,32). Phylogeny of 4,654 core genes showed how isolates from the patient clustered together in the same clade with ST512 strains isolated at PUI during 2018–2020 (Figure 2) and were separated by 0–11 SNPs (Appendix 1 Table 1).

All ST512 isolates are defined by conserved features: the wzi allele 154 (assigned to the KL107 capsule), typical of the main clade II of clonal group 258 (33); O1/O2v2 O locus, serotype O2afg; 35Q mutation in the chromosome-encoded SHV-11 β-lactamase (28); premature stop at codon 89 in the OmpK35 protein, in most isolates coupled with a GD amino acid insertion in the OmpK36 eyelet (34); and chromosome mutations in the gyrA (S83I) and parC (S80I) genes potentially conferring quinolone resistance (35). The 5 isolates in this study had additional common features diverging from the average ST512 strain: a SNP in the mgrB gene, which stops translation at aa 29 of the MgrB regulator protein, conferring resistance to colistin (36); a SNP in the uhpB gene, which stops translation at aa 206 of the UhpB protein (37), conferring resistance to fosfomycin; and the presence of the locus encoding yersiniabactin siderophore, designated as YbST157 (38). The yersiniabactin siderophore is not common in ST512 isolates because this locus could only be found in 31/133 (23%) ST512 genomes from the BIGSdb collection.

The 5 K. pneumoniae strains carried multiple acquired resistance genes (Appendix 2 Table 1). Plasmid content of the 5 ST512 strains from the patient was not homogeneous: strains 6379 (first strain isolated) and 0296 (last strain isolated) both carried the pKpQIL plasmid containing the bla_KPC-3 gene and IncA plasmid carrying the bla_VIM-1 gene (Figures 1, 2). Strains 1186W, 1186T, and 6099 lacked the IncA-bla_VIM-1 plasmid.

KPC Variants Conferring Resistance to Ceftazidime/Avibactam

On day 24, K. pneumoniae strains 1186W and 1186T colonized the patient (isolated from rectal swab samples). Those strains were negative for VIM-1 (loss of plasmid pIncA-bla_VIM-1) but reached high levels of CZA resistance through the production of KPC-31. The bla_KPC-31 gene replaced bla_KPC-3 on an otherwise indistinguishable pKpQIL plasmid; the plasmid was integrated in the chromosome only in strain 1186W. The KPC-31 variant is characterized by a D179Y substitution in the Ω loop, previously shown to confer CZA resistance but restores carbapenem susceptibility (39).

On day 51, CZA-resistant K. pneumoniae strain 6099 caused a respiratory tract infection in the patient (Table). This strain hosted a pKpQIL plasmid encoding a novel KPC variant that was assigned by GenBank as KPC-154 (accession no. OQ096263) and was negative for the VIM-encoding IncA plasmid. Compared with KPC-3, KPC-154 has an RAPNKDDKYS amino acid duplication in position 263–273, corresponding to the 270 loop (40), but no differences are present in the Ω loop. When compared with an isogenic KPC-31–carrying strain, the strain carrying KPC-154 had higher MICs for several β-lactams (Appendix 2 Table 2).

Cefiderocol Resistance in ST512 Isolates Co-Producing KPC and VIM

Figure 3

Figure 3. In silico 3-dimensional structure predictions for mutated OmpK36 porin in Klebsiella pneumoniae sequence type 512 strain 0296 in study of K. pneumoniaegenotypic...

K. pneumoniae strain 0296 (pulmonary infectious agent, isolated on day 88) was genotypically related to strain 6379 (colonizer, isolated on day 1). Both strains had similar plasmid content, including pKpQIL carrying bla_KPC-3 and pIncA carrying bla_VIM-1 (Figures 1, 2). However, isolate 0296 carried a pKpQIL-pKPN fused plasmid carrying catA1, dfrA12, aadA2, aph(3′)-Ia genes, 2 copies of mph(A), sul1, and ΔqacE genes (Appendix 2 Table 1), and the previously described putative gene involved in iron acquisition (fec) (41), all of which were not present in plasmid pKPN from isolate 6379. Within the core genome, strain 0296 had 7 SNPs and 3 deletions when compared with strain 6379 (Appendix 1 Table 2); 1 SNP created a premature stop codon at position 133 in the iron transporter protein CirA (Table; Figure 1) (42). Strain 0296 showed 4-fold higher FDC MIC than strain 6379 (6379 FDC MIC = 8 mg/L; 0296 FDC MIC = 32 mg/L). Moreover, in strain 0296, a conservative in-frame deletion resulted in a novel mutation within OmpK36; the deletion was 26 aa (residues Thr263 to Tyr289) according to the reference OmpK36 crystal structure in the Protein Data Bank (no. 6RCP; https://www.rcsb.org) (30). The in silico predicted 3-dimensional structure of the mutated OmpK36 porin showed a deep lateral cave, probably favoring MVB permeability from the extracellular site into the periplasmic space (Figure 3). Strain 0296 carrying the mutated OmpK36 protein showed a 5-fold reduction in MVB MICs compared with strain 6379 (6379 MVB MIC = 1.5 mg/L; 0296 MVB MIC = 0.047 mg/L).