Hyper-capsulation of ST147NDM-1 Kp may explain the reason why, despite applying an antigen-agnostic approach, we identified anti-capsule and anti-OAg bactericidal mAbs as the most immunogenic antigens. Although this work resulted in the identification of a candidate therapeutic mAb against ST147NDM-1 Kp, future research efforts could be directed towards the identification of functional antibodies targeting non-polysaccharidic Kp targets. To this end, implementation of specific screening strategies will be instrumental.

Metadata were downloaded from Pathogenwatch (https://pathogen.watch) in May 2024 for all available Kp strains. Genomes isolated between January 2010 and December 2022 were selected. The ten most frequent sequence types and K loci were identified and grouped by year. K loci with confidence levels ≥'good' were selected following a previously described conservative approach. Frequency percentage values were computed based on the total number of isolates per year. World maps were generated using metadata extracted from Pathogenwatch using GeoPandas 1.0.1 and GeoDatasets 2024.8.0. Data were analysed by Pandas 2.2.2, Matplotlib 3.8.4, Seaborn 0.13.2 in Python 3.9. Data were analysed by Pandas 2.2.2, Matplotlib 3.8.4, Seaborn 0.13.2, and GeoPandas 0.14.4 in Python 3.9.

Clinical protocol Klebsiella01 was approved by the ethics committee of the University Hospital of Pisa (Parere 16526). All enroled patients (Supplementary Table 1) signed a written informed consent form before taking part in the study, which was conducted in accordance with Good Clinical Practice (GCP) guidelines and ethical principles of the Declaration of Helsinki.

Bacterial strains used in this study are listed in Table 1. Kp isolates were grown at 37 °C in LB broth or on LB-agar plates in the presence of the appropriate antibiotic (5 μg ml meropenem for ST147 c.i. 1, ST147 c.i. 2, ST147, ST307, ST13, ST512; 150 μg ml hygromycin for sfmCherry-ST147), 2.5 μg ml gentamicin (ST231) or without antibiotics (ST258, ST493 and ST2096). Where specified, Kp isolates were also grown in M9 medium (M9 salts supplemented with 2 mM MgSO and 0.4% glucose).

The sfmCherry-ST147 recombinant strain was designed on SnapGene 7.0.0 and generated by cloning codon-optimized tandem copies of the sfmCherry gene under the pTAC promoter (GeneArt, Thermo Fisher Scientific) into the HindIII and SpeI restriction sites of the pSEVA23a1 vector, with a hygromycin resistance cassette inserted at BamHI-XbaI (all enzymes were from New England Biolabs). The plasmid was electroporated into ST147 using an Eporator (Eppendorf). Transformants colonies were selected and validated by colony PCR. sfmCherry expression was confirmed by flow cytometry and confocal microscopy.

Peripheral blood mononuclear cells (PBMCs) were isolated from heparin-treated whole blood by density gradient centrifugation (Ficoll-Paque PREMIUM, GE Healthcare). After separation, PBMCs were stained with Live/Dead FSV780 (1:1,000 BD Horizon). Unspecific bindings were saturated with 20% normal rabbit serum (Life Technologies). Following incubation, cells were stained with CD19 APC (1:400, 561742, BD), IgM PeCF594 (1:100, 562539, BD), CD27 APCR700 (1:100, 565116, BD), IgD PE (1:100, 562024, BD), CD3 PE-Cy7 (1:200, 560910, BD), CD14 PE-Cy7 (1:400, 560919, BD) and CD56 PE-Cy7 (1:100, 560916, BD) in staining buffer (PBS/1% fetal bovine serum (FBS)). Cells were resuspended in sorting buffer (PBS/EDTA 2.5 mM). Stained MBC (CD3CD14CD56CD19CD27IgDIgM) were single-cell sorted with a BD FACSAria Fusion (BD Biosciences) into 384-well microplates containing 3T3-CD40L feeder cells and incubated with IL-2 and IL-21 for 14 days as previously described.

Supernatants from sorted MBCs were used as the IgG1 and IgA source for ELISA. ST147 strains (Table 1) were plated into 384-well microplates, fixed with PBS/0.5% paraformaldehyde (PFA), blocked in 2% calf serum (Euroclone, ECS0040L), and incubated with MBCs supernatants. Anti-human IgG and alkaline phosphatase-conjugated anti-human IgA secondary antibodies (1:2000, Southern Biotech) were added, followed by pNPP (p-nitrophenyl phosphate; Sigma-Aldrich) substrate. A was measured on a Varioskan Lux Reader (Thermo Fisher Scientific). After each step, plates were washed three times in PBS/0.05% Tween-20 (PBS-T). Results were analysed using Microsoft Excel 16.97 and GraphPad Prism 10.1.0. Samples with A values at least twofold of the blank were considered positive. Selected samples were lysed in RNAse Out 0.2 U µl, ultrapure BSA 1 mg ml and DEPC HO (Thermo Fisher) and stored at -80 °C until use.

cDNA was synthesized from MBCs lysates. Reverse transcription was performed by adding a mix of 50 ng ml random hexamer primers, dNTPs (10 mM), 0.1 M DTT, 40 U μl Rnase OUT, MgCl (25 mM), 5× buffer, Superscript IV reverse transcriptase (Invitrogen), and DEPC HO. PCR with reverse transcription (RT-PCR) conditions were 42 °C/10 min, 25 °C/10 min, 50 °C/60 min and 94 °C/5 min. After cDNA synthesis, two additional rounds of PCR were performed to obtain the Vh-Vl chains. For PCR I, cDNA, 10 μM of Vh or 10 μM Vl/Vk primer mix (Supplementary Table 3), dNTPs (10 mM), MgCl (25 mM), 5× Kapa Long Range Buffer, and Kapa Long Range Polymerase (Sigma) was amplified as follows: 95 °C/3 min-5 × 95 °C/30 s, 57 °C/30 s, 72 °C/ 30 s-30 × 95 °C/30 s, 60 °C/30 s, 72 °C/30 s-72 °C/2 min. PCR I products were used as a template for PCR II using the same cycling conditions indicated above and primers reported in Supplementary Table 3. PCR II products were purified by Millipore MultiScreen PCR 96-well plate following the manufacturer's instructions. Samples were eluted in DEPC HO and quantified by NanoDrop One (Thermo Fisher).

PCRII products were ligated into human IgG1 expression vectors to generate transcriptionally active PCR (TAP) products as previously described.

To quantify mAb concentrations in supernatants, ELISA plates were coated with 2 µg ml of anti-human IgG at 4 °C overnight, then washed three times in PBS-T and blocked in PBS/BSA 1% (PBS-B). TAP-produced mAbs diluted in PBS-B/0.05% Tween-20 were added to the ELISA plates. Following incubation with alkaline phosphatase-conjugated IgG secondary antibody, A was measured upon pNPP addition. Concentrations were evaluated by linear regression using a standard curve from titrated human IgG (Southern Biotech) and results were analysed in Microsoft Excel 16.97.

Expi293F cells (Thermo Fisher) were transiently transfected with plasmids carrying the heavy and the light chains of each antibody with a 1:2 mass/mass ratio and grown for 6 days at 37 °C with 8% CO, 125 rpm. Enhancers 1 and 2 (Thermo Fisher) were added on day 1 after transfection. mAbs were purified by affinity chromatography on the AKTAgo purification system (Cytiva) using a HiTrap Protein G HP column (Cytiva) and dialysed in PBS pH 7.4 using Slide-A-Lyzer G2 Dialysis Cassette 3.5 K (Thermo Scientific). mAb concentration was determined by Nanodrop.

Bacteria from glycerol stocks were grown overnight on LB-agar plates. Appropriate antibiotics were added at all stages of bacterial growth. Single colonies were cultured overnight at 37 °C in LB, then subcultured (starting OD 0.05, 37 °C, 150 rpm) in LB or M9 medium until cultures reached the exponential phase (OD 0.4-0.6). For L-SBA, 0.5 × 10 bacteria per ml were seeded into 96-well round bottom plates with 12.5% baby rabbit complement (BRC, Cedarlane). Four serial dilutions of TAP-expressed mAbs (1:10, 1:50: 1:250 and 1:1,250) were tested. After incubation at 37 °C, bacteria were pelleted by centrifugation, resuspended in PBS and transferred to a white Optiplate (Perkin Elmer). BacTiter-Glo 1× (Promega) was added and luminescence was measured using Varioskan (exposure: 500 ms). For F-SBA, 0.5 × 10 bacteria per ml in PBS were seeded in 384-well black clear-bottom plates (ViewPlate-384F TC, PerkinElmer) with 12.5% BRC. Purified mAbs were added in threefold step serial dilutions. After incubation, LB/0.025% resazurin (Sigma-Aldrich) was added. Fluorescence (excitation 560 nm, emission 590 nm, exposure: 250 ms) was measured using Varioskan, and IC values were calculated using GraphPad Prism 10.1.0.

Binding of mAbs to different bacterial strains (Table 1) was assessed in the exponential phase of growth. Bacteria were pelleted, resuspended in PBS-B, and plated in round bottom 96-well plates. Upon incubation with 5 µg ml purified mAbs, plates were centrifuged, and pellets were washed in PBS-B. Alexa 488-conjugated anti-human IgG secondary antibody (1:2,000) was added. Samples were fixed in 0.5% PFA, resuspended in PBS-B to OD 0.05. Data were acquired on the BD FACS Canto II and analysed by FlowJo v.10.6.1 (BD Biosciences).

Immunoglobulin genes were identified with NCBI IgBlast v.1.21 and named according to the International Immunogenetics Information System (IMGT) nomenclature. IGHV gene somatic hypermutations were counted from the start of FWR1 until the end of FWR3. Insertions or deletions were counted as one single mutation.

ST147 strains were sequenced using both short- and long-reads technologies. High-throughput-sequencing was performed on the MiSeq platform (Illumina) with 150 bp paired-end reads. Poor-quality reads were filtered using fastp v.0.20. Long-read library was sequenced following manufacturer's instructions using flow cell R9.4.1 (Nanopore). Long-reads quality was reviewed with Filtlong v.0.2.0 (https://github.com/rrwick/Filtlong) applying minimum thresholds of Q8 (quality) 2,000 bp (length). Clean long reads were assembled using the hybrid assembler Unicycler v.0.4.8 and run under the conservative mode. Kleborate v.2.0.0 and Kaptive v.2.0.6 were employed for sequence type and capsule type prediction. Roary v.3.13.0 was used for pangenome analysis, and the core gene alignment was used to generate a core gene single nucleotide polymorphism (SNP) tree. SNPs were concatenated with snp-sites v.2.4.1 and analysed in IQTree v.1.6.8 to create a maximum likelihood tree. The optimal evolutionary model was selected through ModelFinder plus. Raw sequencing data are available under BioProject PRJNA1067287.

Bacteria were prepared as described for serum bactericidal assay. Polysaccharide extracts were obtained by using the LPS isolation kit (Sigma). SDS-PAGE samples were prepared with 1:2 tris-glycine loading buffer and incubated for 10 min at 70 °C. Gels (Novex 12% Tris-Glycine Mini Protein Gels, 1.0 mm, WedgeWell) format were transferred onto PVDF membranes using the iBlot Gel Transfer Device and Stacks (Thermo Fisher). Membranes were blocked in TBS/0.1% Tween-20 (TBS-T)/5% milk, then incubated with 1 µg ml of mAbs overnight at 4 °C. After washing with TBS-T, membranes were incubated with anti-human Fab secondary antibody (1:75,000) in TBS-T/5% milk, washed again, and developed using chemiluminescence.

Bacterial cultures were prepared as described for serum bactericidal assay. Overnight cultures were diluted to OD 0.025-0.05 and seeded into a 96-well Phenoplate (Perkin Elmer, 6055300). After incubation at 37 °C, samples were briefly spun, and adherent bacteria were fixed in PBS/4% PFA (Thermo Fisher) or Cytofix (BD). For single mAb experiments, 0.5 μg ml antibodies were diluted in PBS/1% BSA. Then, samples were further incubated with anti-Human Alexa488 (Thermo Fisher, 1:2,000) and DAPI (Thermo Fisher, 1:2,000). For multiple mAbs experiments, selected mAbs were conjugated with Alexa 488, Alexa 555 and Alexa 647 using Zip Alexa Fluor Kits (Thermo Fisher). After blocking in PBS/1% BSA, samples were incubated with 1 μg ml of each fluorophore-conjugated mAb and DAPI (1:2,000). In all experiments samples were embedded in 1% low-melting point agarose (Sigma). Imaging data were acquired within 24 h.

The wbbO gene of ST147, retrieved by whole-genome sequencing, was synthesized by GeneArt (Thermo Fisher Scientific) and cloned in the pSEVA23a1 vector as described above for the sfmCherry gene. Correct cloning was confirmed by Sanger sequencing (Eurofins). ST147 was transformed by electroporation with the resulting construct according to standard procedures using an Eporator instrument (Eppendorf).

THP-1 cells (ATCC) were maintained in RPMI 1640-GlutaMax, 10 mM HEPES, 1 mM sodium pyruvate, 10% FBS (all Thermo Fisher). For the fluorescence-based OPA, 50,000 cells per well were seeded into a 96-well black-shielded Optiplate (Perkin Elmer) with 20 ng ml phorbol-12-myristate-13-acetate (PMA). The following day, PMA was removed, and macrophages were maintained in fresh medium for 48 h until infection with sfmCherry-ST147. Overnight-grown bacteria were brought to the exponential phase, centrifuged, and resuspended in the medium without FBS. Bacteria were incubated with different mAbs dilutions for 30 min at 37 °C, shaking at 600 rpm, then added to macrophages and centrifuged at 1,000 rpm for 3 min at room temperature. Following 1 h incubation at 37 °C, 5% CO, samples were treated with 150 μg ml streptomycin, then incubated in PBS/0.1% Triton X-100. Bacterial fluorescence was detected using Varioskan (ThermoFisher). For the flow cytometry-based OPA, THP-1 cells were cultured as indicated above, while sfmCherry-expressing bacteria were grown as described for serum bactericidal assay and then washed and resuspended in medium without FBS. THP-1 cells were incubated with bacteria at a multiplicity of infection of 1 in the presence of a panel of mAbs in threefold serial dilutions for 30 min at 37 °C, 5% CO in shaking conditions (600 rpm). After incubation, samples were fixed in 4% PFA on ice, stained with LIVE/DEAD Fixable Violet Dead Cell Stain (Invitrogen) and resuspended in PBS. Flow cytometry data were acquired at the Agilent Novocyte. Gating was performed sequentially to identify singlets of live THP-1 cells and gate on sfmCherry-positive population. The percentage of internalized bacteria in the presence of mAbs was calculated relatively to the control as indicated in figure legends.

sfmCherry-ST147, ST231 and ST2096 were grown as described above. Starting from 100 μg ml, tenfold serial dilutions of mAbs were tested in PBS/12.5% BRC or RPMI/10% FBS medium. Samples were prepared in 96-well Phenoplate and briefly centrifuged at 1,000 rpm. Acquisition started immediately after centrifugation at 37 °C by acquiring frames every 2 min for 2 h.

Imaging was performed using Opera Phenix (Revvity) with a 63× N.A. 1.15 water objective. For mAb binding experiments in fixed conditions, 20 fields of view (FOVs) per well were selected, acquiring 5 z-stacks per FOV (0.5 μm z-step) in confocal mode. Excitation: 425 nm, 488 nm and 561 nm lasers. Emission collected with bandpass filters 435-480 nm, 500-550 nm, and 570-630 nm. For bacterial live imaging, 3 FOVs per well were selected, acquiring 3 z-stacks (0.5 μm z-step) in widefield mode (brightfield or 561 nm laser for excitation); emission collected with 570-630 nm bandpass filter. Images were analysed using Harmony (v.4.9, Revvity) with a custom-made image analysis pipeline (Supplementary Table 6). Individual bacteria were detected using the DAPI channel. For mAbs intensity levels, mean A488 signal intensity was measured within a bacterium-specific ROI (Extended Data Fig. 8). For mAb occupancy area, A488 spots were detected, and their morphology features were measured. A488 spots <0.5 μm were disregarded.

ST147 liquid culture was plated on Worfel-Ferguson-agar plates. Capsule was purified following a published protocol. Sugar content was evaluated by phenol-sulfuric assay, protein content was estimated by microBCA (Thermo Fisher) and DNA presence was measured using the Qubit dsDNA BR Assay Kit (Invitrogen). Protein and DNA contaminants were <1% compared to the sugar concentration. Quality of the estimated molecular size distribution was assessed by size-exclusion chromatography-HPLC (SEC-HPLC) on a Thermo Fisher Vanquish core instrument equipped with a Tosoh TSK gel PWXL guard and G3000 PWXL columns connected in series (4.0 cm × 6.0 mm (808033) and 30 cm × 7.8 mm (808021)) using 0.1 M NaCl, 0.1 M NaHPO, pH 7.2 as running buffer.

O-antigen was extracted by following a published protocol. Total sugar quantification was measured by phenol-sulfuric assay and quality of the estimated molecular size distribution was assessed by SEC-HPLC as described above.

Ten milligrams of purified capsule and O-antigen material was buffer-exchanged to DO by drying with a SpeedVac and resuspending in 99.9 % DO three times in total. NMR spectra (1D H, 2D H-C HSQC) were recorded at 298 and 323 K on a Bruker AVANCE NEO spectrometer, operating at 900 MHz, equipped with triple resonance cryo-probe, and on Bruker AVANCE MHD NMR spectrometers, operating at 400 and 950 MHz (H Larmor frequency), equipped with room temperature proton selective probe and triple resonance cryo-probe, respectively. Spectra were acquired using TopSpin v.4.5.0. Molecular schemes (Extended Data Fig. 5) were designed using CorelDRAW Graphics Suite 2019. The assignment of the sugar protons has been reported on the signals according to literature. The obtained spectra were compared with the data reported in the literature to confirm capsule and O-antigen identity (Extended Data Fig. 6).

Purified O2a-LPS and O2afg-LPS were kindly provided by the group of C. Whitfield. In addition, in-house purified KL64 capsule and O2a and O1 O-antigens were used to coat high-binding 384-well plates (Greiner 781061). Plates were blocked in PBS-B at 37 °C and then incubated with 10 µg ml of each mAb in SB. Anti-human IgG secondary antibody conjugated with alkaline phosphatase was added and absorbance was read by addition of pNPP. After each incubation step, plates were washed three times in PBS-T. Sample buffer was used as blank, and A values threefold higher than the blank were selected.

The protocol performed at Hartford Hospital was approved by the Institutional Animal Care and Use Committee. Specific pathogen-free, female ICR mice (8-9 weeks of age) were obtained from Charles River Laboratories, acclimatized for 48 h, and provided food and water ad libitum. Mice received uranyl nitrate 5 mg kg three days prior to inoculation to produce a controlled degree of renal impairment. ST147, ST231 and ST2096 was frozen at -80 °C in skim milk (BD BioSciences). Before inoculation, two transfers were performed onto trypticase soy agar plates with 5% sheep blood (TSA II; Becton, Dickinson & Co.) and incubated at 37 °C for 24 h. After 18-24 h, a bacterial suspension of ∼10 CFU ml in 5% hog gastric mucin was prepared. Mucin acts as a virulence-enhancing factor for bacteria to establish peritonitis and subsequently bacteraemia. Final inoculum was confirmed via serial dilution and plating. Septicaemia was induced by intraperitoneal injection of 0.5 ml inoculum. 08O09, 05N02 and 05D08 were reconstituted with PBS (pH 7.4) to deliver 1, 5, 10 and 20 mg kg doses. Dosing solutions were kept on ice, syringes were refrigerated until use. Control mice received PBS. For PPX studies, mAbs were administered IP 24 h before inoculation. For treatment studies, 08O09 was administered through intravenously 1 h after inoculation. For PPX-plus-treatment studies, 08O09 was administered intraperitoneally 24 h pre-inoculation, and intravenously 1 h after inoculation. Each group had 10-30 mice, as indicated in figure legends. Sham controls (n = 10) received PBS. One hour after inoculation, the 0 h control group (n = 6) was euthanized for spleen collection to confirm septicaemia. Mortality was monitored for 96 h. Spleens from deceased or euthanized mice were processed for CFU enumeration. Surviving mice were euthanized at 96 h for spleen CFU analysis. Efficacy was measured as spleen bacterial load reduction (log CFU) versus sham control. Mortality rates were recorded and assessed over a 96-240 h study period. Bacterial density reductions were analysed using two-tailed Mann-Whitney test. Survival rates were compared using Kaplan-Meier survival analysis and the log-rank (Mantel-Cox) test.

The protocol performed at Fondazione Toscana Life Sciences was approved by the Italian Ministry of Health (authorization 9AECF.52). Female CD-1 mice (8-9 weeks of age) were obtained from Charles River Laboratories, acclimatized for 48 h, and provided food and water ad libitum. ST147 overnight cultures were subcultured (starting OD 0.05, 37 °C, 150 rpm) until reaching the exponential phase (OD 0.4-0.6). Then, a bacterial suspension of ∼2 × 10 CFU ml was prepared in PBS and 0.5 ml were intraperitoneally injected. 08O09 was reconstituted with PBS (pH 7.4) to deliver 5 mg kg dose. Control mice received PBS. For PPX studies, 08O09 was administered intraperitoneally 24 h before inoculation, whereas for treatment studies it was administered intravenously 1 h after inoculation. Two experiments were performed with 4 and 8 mice each group, respectively. Mortality was monitored for 96 h. Spleens from mice which reached humane endpoint were processed for CFU enumeration. Surviving mice were euthanized at 96 h for spleen CFU analysis. Efficacy was measured as spleen bacterial load and analysed using two-tailed Mann-Whitney test. Survival rates were compared using Kaplan-Meier survival analysis and the log-rank (Mantel-Cox) test.

Sample size choice was selected to achieve robust statistical differences (n = 8-10 mice per group). Animal randomization and experimenter blinding were performed.

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.