A post-translational modification of fimbriae drives pathogenicity in Klebsiella pneumoniae

Abstract

Antimicrobial resistance is a severe public health burden. Especially concerning are multi-drug resistant (MDR) infections, which restrict treatment options and significantly increase mortality risk. A major cause of MDR infections worldwide is carbapenem-resistant Kp (CRKp). The predominant CRKp sequence type worldwide is ST258. However, the factors underlying ST258’s epidemic success are not well defined. Genomic analyses of clinical isolates of CRKp have found that the two-component regulatory system CrrAB is a genomic feature of ST258, suggesting that it may contribute to its global dominance. Despite this, the molecular details underpinning CrrAB’s contribution to ST258 Kp biology and pathogenicity are poorly understood. We used RNA-sequencing to identify the regulon of CrrA and found that CrrAB induces the expression of a novel gene, encoding Crr-regulated fimbriae modifying protein (CfmP), that is essential for pathogenesis driven by this two-component system. We performed mass spectrometry analyses of fimbriae purified from Kp expressing or lacking cfmP and found that CfmP induces a novel oxidation to a histidine residue in the major pilin subunit of fimbriae, FimA. We demonstrate that this oxidation significantly increases host cell adhesion and high bacterial loads within the host. CrrAB also drives high antibiotic resistance in CRKp. Thus, our results place CrrAB at the intersection of pathogenicity and antibiotic resistance supporting its function as an important regulatory system driving the global dominance of ST258.

Introduction

Multi-drug resistant (MDR) Gram-negative bacterial infections are among the most urgent health threats¹. Klebsiella pneumoniae (Kp) is a major cause of MDR infections, accounting for the third most global deaths attributable to or associated with bacterial MDR¹. Especially concerning is carbapenem resistant Kp (CRKp), which is a leading cause of invasive, drug-resistant infections in many human body sites, with high rates of morbidity and mortality^2–8. Resistance has emerged in many Kp clonal groups, but only a few lineages have come to dominate MDR Kp infections worldwide. In particular, the global spread of CRKp has been driven in main part by sequence type 258 (ST258)^2,8,9. ST258 represents a particular challenge for healthcare providers, as it exhibits high levels of multi-class drug resistance, severely limiting treatment options⁸. Notably, the reasons underlying ST258’s world-wide dissemination and the molecular mechanisms underpinning its pathogenic success are unclear^2,8,9.

Genomic analyses performed on ST258 isolates have identified several loci that are associated with pathogenicity in ST258^4,10–15. However, molecular studies to determine the relevance of these potential virulence determinants are lacking. Notably, the understudied two component system (TCS) CrrAB was recently identified as a genomic feature of ST258^16,17 and has been suggested to confer pathogenic and physiological benefits to ST258^17,18. Further, missense mutations in CrrAB were identified in clinical isolates of ST258 CRKp, which were found to confer high antibiotic resistance and high virulence^19–22. These CrrAB mutations induce resistance to last-line polymyxin antibiotics, cationic non-ribosomal peptides that disrupt the bacterial cell envelope. CrrAB was shown to induce polymyxin resistance (PR), in part, through the activation of eptA and the arn operon. These genes encode enzymes that catalyze the addition of modifications to the lipopolysaccharide (LPS), the primary surface molecule of Gram-negative bacteria, protecting against polymyxin attack^19–22. Intriguingly, polymyxin antibiotics function similarly to host-produced antimicrobial peptides (AMPs). This suggests that CrrAB may function in ST258’s response to the host environment. However, EptA- and Arn-driven LPS modifications do not fully correlate with increased virulence in Kp, suggesting that CrrAB functions to increase ST258 pathogenicity through additional, unknown mechanisms²².

Here, we report that a previously uncharacterized gene in the CrrA regulon, cfmP, drives pathogenesis in ST258 Kp. We show that CfmP induces the addition of a novel post-translational modification to the major pilin FimA of type I pili, or fimbriae, on the cell surface. We found that this modification, an oxidation to a conserved histidine in FimA, leads to increased host cell attachment and bacterial proliferation within the host, to cause high pathogenicity. Only a few examples of translational modifications to type I pili have been demonstrated^23,24 and their implications for fimbriae function and host-pathogen interactions have not been characterized. Our findings represent the first example of an oxidative posttranslational modification to fimbriae and broaden our understanding of how modifications to the pilus shaft impact its function in pathogenicity. Overall, these findings expand our mechanistic understanding of how CrrAB, a genomic feature of ST258, contributes to the pathogenicity of this high-risk clonal group.

Results

crrAB is conserved in ST258 CRKp

ST258 has been the dominant driver of CRKp infections globally. The CrrAB TCS (Fig. 1a) was previously reported to be a genomic feature of ST258^16,17, suggesting that it may contribute to its pathogenic success. We analyzed the genomes of a collection of 918 CRKp isolates from Columbia University Irving Medical Center^25–27. We found that 58% (535) belonged to 258 clonal group (CG258), containing ST258 and its single-locus variants ST11, ST340, and ST512, while the remaining 383 isolates (42%) were distributed amongst 125 different sequence types (Fig. 1b). Among all CG258 isolates, we performed genomic searches for crrA and crrB. Out of 271 isolates (249 ST258, 15 ST11, 6 ST340, 1 ST512), 240 contained both genes (Fig. 1b). For the 31 isolates lacking crrA and crrB, we manually examined their assembled genomes to assess whether we could determine how or why crrAB was lost. We found that 19 of these isolates were lacking crrAB but contained genes immediately up- and downstream in the chromosome (Figure 1A, bottom panel) and had insertion sequences (IS) nearby, suggesting IS-driven transpositions may have removed crrAB. Overall, 96% of the CG258 isolates we searched contained crrAB or showed remnants of the genomic locus. Further, we searched publicly available genomes for crrA and crrB and found it is conserved in ST258 genomes worldwide (SI Appendix Table S1), supporting that CrrAB is a conserved feature of CG258 globally.

Figure 1: CrrAB transcriptionally activates five uncharacterized genes in its surrounding locus.

(A) Genomic searches performed on CRKp isolates collected from CUIMC (n=918). 535 isolates belonged to CG258 (ST258, ST11, ST340, ST512), while the rest (n=383) were distributed among 126 different sequence types. Among CG258 isolates collected from unique patients (n=271), a majority (n=259) contained crrAB or genes in the crr locus. The nineteen isolates with insertion sequences (IS) within the crr locus are shown in the schematic; the red dashed lines represent the genomic regions lost via IS interruption and the number represents the number of isolates with that particular IS interruption. (B) Schematic of the CrrB histidine kinase containing the location of the L94M mutation and the CrrA response regulator. (C) Volcano plot of RNA-seq data generated from comparing the transcriptional profile of the clinical mutant crrB*L94M vs. wildtype (pink) and crrB*L94M ΔcrrA vs. crrB*L94M (grey). Genes in the crr locus are highlighted. (D) Quantitative reverse transcription PCR (RT-qPCR) was used to confirm the fold change in transcription of the genes in the crr locus in the indicated strains. (E) A schematic of the crr loci of Kp and other Enterobacteriaceae species. Searches were performed using publicly-available genomes (GenBank accession numbers are noted in parentheses). CrrB was used to generate the sequence similarity network used for genomic neighborhood analyses of the crr locus.

CrrAB upregulates its surrounding crr locus.

The CrrAB TCS was previously shown to induce high pathogenicity in Kp²². CrrB is a membrane-anchored sensor histidine kinase, while CrrA is a DNA-binding response regulator. To identify the full landscape of transcriptional changes induced by CrrAB and the specific genes involved in the increased pathogenicity driven by CrrAB, we performed RNA sequencing (RNAseq). We compared the transcriptomes of wildtype Kp and an isogenic strain into which we introduced a clinically-identified, hyperactive mutant of CrrB, crrB*L94M^22,25 (Fig. 1a,c). It has been shown that CrrAB and two other TCSs, PhoPQ and PmrAB, share transcriptional targets in Kp^19,22. Therefore, to determine CrrAB-specific regulation, we compared the transcriptional profiles of crrB*L94M in the absence of the response regulators of each individual signaling system (crrB*L94M ΔcrrA, crrB*L94M ΔpmrA, and crrB*L94M ΔphoP). In agreement with previous studies, our results showed that crrB*L94M, in coordination with PmrAB and PhoPQ, increases transcription of the arn operon and eptA to drive PR²⁸ (SI Appendix Fig. 1). Intriguingly, we found that CrrA, but not PhoP or PmrA, leads to the dramatic upregulation of five genes of unknown function in the genomic locus surrounding crrAB, which we have termed the crr locus (Fig. 1b,c,d). Importantly, we found that the same CrrA regulon identified in our isogenic strains is also upregulated in an original clinical Kp isolate identified harboring crrB*L94M^22,25 (SI Appendix Fig. 2a). These genes were previously identified to be associated with crrAB in Klebsiella species and in some Citrobacter and Enterobacter species^16,17, although their roles in the pathogenicity of Kp and other Enterobacteriaceae are unknown. We found that this locus was conserved in all strains that contained crrAB, suggesting that the genes within the crr locus are evolutionarily conserved with this TCS. Gene neighborhood analysis was performed^29,30 and further identified the crr locus in diverse Enterobacteriaceae species, including several human pathogens (Fig. 1e), suggesting that it may contribute to CrrAB’s function in virulence.

CrrAB-regulated cfmP confers high virulence in crrB*L94M

We next investigated whether genes in the crr locus are responsible for the drug resistance and virulence phenotypes observed in hyperactive CrrAB clinical mutants, such as CrrB*L94M²². As previously shown, CrrC is necessary for PR in crrB*L94M (Fig. 2a). crrC is directly upstream of crrA and suggested to encode a connector protein, mediating signaling cross-talk between CrrAB and PmrAB^19,22. Using the Galleria mellonella virulence model, we found that a deletion of only one of the genes in the crr locus, cfmP, completely reduced the high virulence of crrB*L94M back to wildtype levels (Fig. 2b, SI Appendix Fig. 3b). To further investigate the mechanisms underlying CrrAB and cfmP-driven virulence, we determined the bacterial loads over the course of G. mellonella infections. We showed that, in addition to causing high mortality of G. mellonella (Fig. 2b,c), CrrAB’s regulation of cfmP causes high bacterial proliferation in the host (Fig. 2c). Further, we found that overexpression of cfmP increases virulence and bacteria loads in wildtype Kp (Fig. 2d). Importantly, virulence was restored when cfmP was complemented in trans (Fig. 2e, SI Appendix Fig. 3a and 3b), and the decrease in virulence of crrB*L94M ΔcfmP was not due to off-target effects on the transcription of the crr locus (SI Appendix Fig. 3b). Further, we found that deletion of cfmP in the original Kp clinical isolate harboring the crrB*L94M mutation also completely decreases the virulence of this isolate (SI Appendix Fig. 2b).

Figure 2: A novel gene in the CrrA regulon and crr locus, cfmP, drives the high virulence of CrrB*L94M.

(A) A polymyxin B broth microdilution (BMD) plate of the indicated mutants. (B) G. mellonella infections using the indicated strains. (C) G. mellonella survival curve and G. mellonella bacterial load quantification of infections with wildtype, crrB*L94M, crrB*L94M ΔcrrA, and crrB*L94M ΔcfmP. (D) G. mellonella infections of wildtype, crrB*L94M, and wildtype harboring empty vector (ev) or overexpression plasmids (pUC19-P_lac::optRBS-cfmP). (E) G. mellonella infections of wildtype, crrB*L94M, crrB*L94M ΔcfmP and crrB*L94M ΔcfmP harboring empty vector (ev) or complementation plasmids (pACYC184-P_lac::natRBS-cfmP). cfmP overexpression, complementation, and empty vector (ev) strains were grown in the presence of 500 μM IPTG. 20 G. mellonella larvae per strain were injected on day 0. Three larvae were sacrificed daily to calculate bacterial proliferation; average CFU count per larva is reported. One representative experiment is shown, from at least three biological replicates.

We found that crrB*L94M and CfmP induce high virulence in G. mellonella, causing rapid melanization and death. This severe phenotype made us wonder whether CfmP may increase host immune signaling and induce a detrimental cytokine storm. To test this, we measured how cytokine secretion and human Toll-like receptor 4 and 2 (TLR4/TLR2) signaling is impacted CrrB*L94M and CfmP. We found that wildtype Kp increases cytokine signaling in human macrophages (SI Appendix Fig. 4a) and increases TLR4 and TLR2 activity (SI Appendix Fig. 4b). However, we saw no altered immune response between wildtype and crrB*L94M or crrB*L94M ΔcfmP (SI Appendix Fig. 4).

CfmP is a member of the TupA-like ATP-grasp family, with an amidoligase-like fold

cfmP is a gene of unknown function. Therefore, to investigate the function of CfmP, we performed protein sequence homology searches using the HHpred server^31,32. We found that CfmP has a predicted ATP-grasp domain at its C-terminus (Fig. 3a, SI Appendix Fig. 5a). ATP-grasp domains are a unique ATP-binding fold found in a large family of enzymes involved in diverse metabolic and cellular functions³³. We performed a multi-sequence alignment of the C-terminal ATP-grasp domain of CfmP with other confirmed ATP-grasp domains and identified several conserved residues necessary for ATP binding and enzyme function (SI Appendix Fig. 5a)^34,35. To confirm that the ATP-grasp domain is necessary for CfmP’s function, we mutated the conserved E269 residue of CfmP (E269Q) to disrupt ATP-binding and found that this mutant completely reduced bacterial loads in G. mellonella to wildtype levels (Fig. 3b,c). 6x-Histidine tags were introduced at the C-terminus of CfmP*E269Q and wildtype CfmP, to confirm that CfmP*E269Q is stably produced (SI Appendix Fig. 5b,c).

Figure 3: CfmP has predicted structural homology to amidoligases involved in modifying cell envelope polysaccharides.

(A) Alphafold2-generated model of CfmP, containing a predicted ATP-grasp domain (cyan) and domain of unknown function (magenta) (left). Foldseek-generated superimpositions of the predicted structure of CfmP with DdlB from P. aeruginosa (UniProt Q9LCT6) (center) and with WemR from P. mirabilis (UniProt D9YZ35); in yellow are the structures of DdlB/WemR, in blue is the structure of CfmP. (B) Multiple sequence alignments of CfmP with WemR and CfmP homologs found in other Enterobacteriaceae species. The alignment is colored by sequence similarity, with the conserved glutamic acid (E269) starred. The N-terminal domain of unknown function of CfmP is boxed in a solid red line and the sequence of Kp CfmP is boxed in a dashed red line. (C and D) Survival curves (left) and bacterial proliferation plots (right) of G. mellonella infections with the indicated strains. The crrB*L94M ΔwbaP strain is deleted for the first step in capsule synthesis, and crrB*L94M ΔwecG strain is deleted for Enterobacterial common antigen (ECA). 20 G. mellonella larvae per strain were injected on day 0. Three larvae were sacrificed daily to calculate bacterial proliferation; average CFU count per larva is reported. One representative experiment is shown, from at least three biological replicates.

Notably, our HHpred search did not identify any structural homology to the N-terminus of CfmP. To gain insight into the enzymatic activity of CfmP, we used AlphaFold2^36,37 to generate a predicted structure of CfmP (Fig. 3a) and used this as an input for the Foldseek server³⁸. Searches using the full sequence of CfmP clearly identified homology to ATP-grasp domains, such as that found in the D-Alanine-D-Alanine ligase DdlB (Fig. 3a), but failed to find matches to the N-terminal domain of CfmP. However, when we input only the first 61 amino acids of CfmP, we identified homology to members of the TupA-like ATP-grasp protein family (Pfam: PF14305), such as WemR from Proteus mirabilis (Fig. 3a)³⁹. TupA-like ATP-grasp enzymes are often found within genomic loci encoding genes involved in the biosynthesis of cell surface polysaccharides and are predicted to catalyze the addition of amino acid decorations to these cell surface polymers^39,40.

CfmP-induced pathogenicity does not depend on cell surface polysaccharides

Considering the predicted structural homology of CfmP to the TupA family, we tested whether the pathogenicity induced by CfmP is dependent on the presence of cell surface polysaccharides. The cell envelope of Kp is composed of an inner membrane phospholipid bilayer, a structural layer of peptidoglycan, and an asymmetric outer membrane (OM) composed of an inner leaflet of phospholipids and an outer leaflet of LPS (SI Appendix Fig. 6a). The OM is further decorated by the capsule and the Enterobacterial common antigen (ECA). The LPS, capsule, and ECA are all known to be immunogenic and mediate host-pathogen interactions^41,42. Thus, we wondered whether CfmP adds a modification to one of these structures to increase pathogenicity. To test this, we individually deleted the first genes responsible for the biosynthesis of the capsule (ΔwbaP) and the ECA (ΔwecG) (Fig. 3d, SI Appendix Fig. 6b) in crrB*L94M. We observed no decrease in virulence (Fig. 3d), showing that CfmP must not modify these polysaccharides to induce virulence in Kp. Additionally, our wildtype isolate of Kp contains an IS disrupting the O-antigen flippase and therefore produces no O-antigen (SI Appendix Fig. 6b). Thus, CfmP must not affect virulence through O-antigen modifications. Finally, we performed thin layer chromatography analysis of purified lipid A from wildtype, crrB*L94M, and crrB*L94M ΔcfmP and observed no changes dependent on the presence of cfmP (SI Appendix Fig. 6c). We performed a protein sequence alignment of WemR with Kp CfmP and homologs found in other Enterobacteriaceae and notably found that the N-termini of CfmP differed markedly from WemR, exhibiting only 38% percent identity (Fig. 3b). This supports that, although WemR and CfmP homologs contain similar predicted structure, their precise enzymatic activities may differ.

Fimbriae are necessary for the pathogenicity induced by CfmP.

After ruling out the role of cell envelope polysaccharides in the pathogenicity caused by CfmP, we wondered whether CfmP may be modifying a different cell surface-associated factor. We returned to our RNAseq data of the highly virulent crrB*L94M mutant and searched for upregulation of other virulence factors or cell surface proteins or polymers. We found that the fim locus encoding type I pili, or fimbriae, is upregulated in crrB*L94M (Fig. 4a,c). Fimbriae are surface exposed proteinaceous filaments important for host cell adhesion, colonization, and virulence in Gram-negative bacteria^43–45. Fimbriae are composed of a long, helical filament of the major pilin FimA, followed by the minor pilins FimF and FimG, and the adhesin FimH at the tip of the pilus (Fig. 4a). The pilus is anchored in the OM by the usher FimD, which also aids in the export and polymerization of the FimAFGH structure⁴³. Expression of the fim operon is known to be driven by an invertible fimS promoter, which under normal laboratory growth conditions is in the “off” orientation, keeping fimbriae production minimal⁴⁶. However, we found that an insertion sequence (IS) is disrupting the coding region of FimE, a negative regulator of fim expression, in our isolate background and in the original clinical isolates harboring crrB*L94M (SI Appendix Fig. 7a). This results in consistent production of fimbriae, even during logarithmic growth (SI Appendix Fig. 7b,c).

Figure 4: Fimbriae are essential for CfmP-driven virulence in crrB*L94M.

(A) Volcano plot of RNA-seq data generated from comparing the transcriptional profile of crrB*L94M vs. wildtype. Genes in the fim locus are highlighted. (left). Schematic of a type I pilus, or fimbriae (right). (B) G. mellonella survival curve (top) and bacterial proliferation plot (bottom) of the indicated strains. (C) RT-qPCR was used to confirm the increase in transcription of the fim operon in crrB*L94M and to confirm upregulation of fim expression in the strain harboring an introduced optimized ribosome binding site and constitutive −10/−35 sequence upstream of fimA (left). Significant differences in upregulation were tested by one-way ANOVA, followed by Dunnett’s T3 multiple comparison test (****p<0.0001). No significant differences in upregulation were observed between genes in the fim operon in crrB*L94M. G. mellonella bacterial proliferation curve of indicated strains (right). (D) G. mellonella survival (left) and bacterial proliferation (right) curves of cfmP overexpression strains in wildtype (top) versus ΔfimD (bottom) strains. One representative experiment is shown, split into two sets of plots for clarity. All data points are derived from the same experiment, with crrB*L94M and PBS samples shown twice, on both plots, for comparison. 20 G. mellonella larvae per strain were injected on day 0. Three larvae were sacrificed daily to calculate bacterial proliferation; average CFU count per larva is reported. One representative experiment is shown, from at least three biological replicates. Significant differences in bacterial proliferation on the final day of all replicates of G. mellonella infections of indicated strains were calculated via Mann-Whitney U test (****p<0.0001).

To investigate the role of fimbriae in crrB*L94M pathogenicity, we deleted fimD, to completely prevent the production of pili on the surface of the cell. The crrB*L94M ΔfimD mutant had a virulence defect compared to crrB*L94M in G. mellonella (Fig. 4b), indicating that fimbriae are necessary for virulence in crrB*L94M. Notably, deleting the usher FimD may cause a build-up of un-polymerized FimA within the cell, which could lead to fitness defects and decrease pathogenicity. To rule this out, we tested bacterial growth kinetics and found no difference in growth between crrB*L94M, crrB*L94M ΔfimD, and crrB*L94M ΔfimA (SI Appendix Fig. 7d). We wondered whether the increase in fim expression in crrB*L94M is responsible for the higher virulence of this mutant, or whether a post-transcriptional mechanism is responsible. Thus, we introduced an optimized ribosome binding site and a constitutive −10/−35 RNA polymerase binding site from P_tac upstream of the fim operon in the chromosome of wildtype Kp to artificially increase fim expression (SI Appendix Fig. 7a). We found that this significantly increased fim expression but had no effects on virulence (Fig. 4c). This suggested that a post-transcriptional aspect of fimbriae production may be responsible for the pathogenicity of crrB*L94M. Given the necessity of both fimbriae and CfmP for the pathogenicity of crrB*L94M, we hypothesized that CfmP’s function may be to post-translationally modify fimbriae to increase pathogenicity. To test this, we overexpressed cfmP in wildtype and in the absence of fimbriae and monitored bacterial loads and killing in the G. mellonella model. We found that the increase in virulence we saw in response to cfmP overexpression in wildtype did not occur in cells lacking fimbriae (Fig. 4d), indicating that fimbriae are necessary for CfmP-driven virulence.

CfmP induces a post-translational modification to FimA to drive virulence.

Together, our data indicates that CfmP is necessary for type I pili-mediated pathogenesis in Kp. To test whether fimbriae are post-translationally modified in a CfmP-dependent manner, we purified fimbriae from the surface of wildtype, crrB*L94M, crrB*L94M ΔcfmP, and the cfmP-complemented and empty vector strains and performed untargeted liquid chromatography tandem mass spectrometry (LC-MS/MS). Analyses found several peptides in FimA, FimF, and FimH containing post-translational modifications (PTMs) in crrB*L94M and crrB*L94M ΔcfmP with cfmP complementation (SI Appendix Fig. 8a,b). However, only one, an oxidized histidine in FimA (FimA-H11) was replicated and identified in a second biological replicate. Regardless, to test all modifications, we mutated the affected residues to alanine to prevent their modification and assessed their effects on crrB*L94M virulence in G. mellonella. Consistent with our MS results, only mutating the oxidized histidine in FimA (FimA-H11) completely abrogated the virulence of crrB*L94M (SI Appendix Fig. 8c, Fig. 5a). Importantly, we confirmed that all crrB*L94M fim* mutants still produce fimbriae to the same levels as the parental crrB*L94M strain (SI Appendix Fig. 9a). In an Alphafold⁴⁷-generated polymer of Kp FimA, this histidine occurs in FimA’s N-terminal extension and points inwards towards the lumen of the polymer (Fig. 5c). Notably, the modified histidine in FimA is conserved in other Enterobacteriaceae species that also contain the crr locus (SI Appendix Fig. 9b).

Figure 5: CfmP induces a post-translational modification to FimA to increase host cell adhesion and virulence.

(A) G. mellonella survival curve (left) and bacterial proliferation plot (middle) of the indicated strains. 20 G. mellonella larvae per strain were injected on day 0. Three larvae were sacrificed daily to calculate bacterial proliferation; average CFU count per larva is reported. One representative experiment is shown, from at least three biological replicates. Significant differences in bacterial proliferation on the final day of all replicates of G. mellonella infections of indicated strains were calculated via Mann-Whitney U test (****p<0.0001, ***p<0.001). (B) Kp FimA 10 subunit polymer, generated by the AlphaFold server. Kp FimA polymer is colored in teal, except for one subunit in purple that is highlighted. All FimA-H11 residues are shown as red sticks. (C) C3H/HeN mice infected with 10⁸ CFUs of the indicated strains. Urine, bladder, and kidney titers enumerated 24 hpi (n = 10 mice total from two independent replicates). Mann–Whitney U test p-values are indicated, solid lines indicate geometric mean. (D) Bar charts of the indicated bacterial strains’ percent adhesion to Caco-2 monolayers after 30 minutes of incubation. Percent adhesion was calculated as the number of CFUs/mL of post-incubation bacterial suspension divided by the input CFUs/mL for each strain. Significant differences in adhesion were calculated via Mann–Whitney U test (****p<0.0001, ***p<0.001, *p<0.02). (E) Representative confocal microscopy z-stacks of wildtype and crrB*L94M adhered to Caco-2 monolayers. Bacterial strains were incubated with Caco-2 monolayers as described in the Methods. Caco-2 cells were stained for 1 h with 10 μg/mL Hoechst 33342 and for 10 m with 1X CellMask^™ Deep Red Plasma Membrane stain. Bacterial cells harbor a plasmid containing a Venus cytoplasmic marker. Scale bar, 10 μm. Shown are the maximum intensity projections (MIP) of all channels of each stack, and a montage of slices from each stack to show relative positions of bacteria (indicated by white arrows), Caco-2 plasma membrane (surface), and Caco-2 nuclei (interior). Each slice corresponds to 0.5 μm.

CfmP-dependent oxidation of FimA-H11 increases adhesion to the intestinal epithelium.

Fimbriae are important for mediating contact between pathogenic bacteria and host cells^43,44. In particular, fimbriae have been extensively studied in the context of urinary tract infections (UTI), in which they are critical for initial binding to the bladder epithelium to establish infection⁴⁸. Thus, we hypothesized that the CfmP-driven oxidation of FimA-H11 may cause increased bacterial loads in a murine model of acute UTI. We transurethrally infected C3H/HeN mice with wildtype, crrB*L94M, crrB*L94M fimA*H11A, and crrB*L94M ΔfimD strains of Kp and enumerated urine, bladder, and kidney titers 24 hours post infection (hpi) (Fig. 5c). We found that crrB*L94M caused a modest increase in Kp CFUs compared to wildtype in the urine and bladder, and that this was dependent on the presence of fimbriae (Fig. 5c). However, mutating the oxidized H11 residue in FimA did not decrease the bacterial load back to wildtype levels. This suggests that high CrrAB signaling may increase fitness in a UTI model, perhaps through the upregulation of the fim operon in crrB*L94M. However, the CfmP-driven oxidation of FimA does not contribute to UTI pathogenicity.

In E. coli and Kp studies have also shown that fimbriae help to mediate intestinal colonization^49–54. This intestinal carriage is thought to act an important bacterial reservoir that increases infection risk, especially in the context of MDR infections⁵⁵. To test whether the CfmP-dependent oxidation on FimA-H11 increases host-pathogen interactions in the gastrointestinal tract, we performed adhesion assays using Caco-2 human intestinal epithelial cells⁵⁶. We incubated monolayers of Caco-2 cells with wildtype, crrB*L94M, crrB*L94M ΔcfmP, crrB*L94M fimA*H11A, and ΔfimD Kp, washed away unadhered bacteria, plated the remaining mix of Caco-2/bacteria, and compared the number of colony forming units (CFUs) from the input bacterial cultures with the post-Caco-2 incubation output CFUs. We found that in the absence of fimbriae (ΔfimD) adhesion is significantly decreased compared to wildtype (Fig. 5d). Further, crrB*L94M exhibits much higher adhesion than wildtype Kp, which is restored to wildtype levels in crrB*L94M ΔcfmP and crrB*L94M fimA*H11A (Fig. 5d). This adhesion assay could not rule out whether output counts of crrB*L94M were higher due to increased internalization of the Kp cells. To test this, we performed confocal microscopy of the Caco-2 cells post-incubation with Kp. We found that Kp cells were located on the surface of the Caco-2 monolayer (Fig. 5e), supporting that the CfmP-driven modification to fimbriae increases cell adhesion, and does not impact the ability of Kp to invade host cells.

In E. coli, it is established that FimH binds mannose on host cells to mediate adhesion. To test whether the FimA-H11 oxidation directly impacts lectin binding, we tested its ability to agglutinate guinea pig erythrocytes. In agreement with previous results^57,58, we found that our Kp strains do not exhibit detectable mannose-sensitive agglutination. To circumvent this, we introduced a point mutation (FimH*S62A) that shifts FimH into a more favorable mannose-binding conformation⁵⁷ into our Kp strains. We found that, while fimH*S62A increases mannose-sensitive agglutination, neither crrB*L94M nor the presence of CfmP affects agglutination titers (SI Appendix Fig. 9b). This suggests that CfmP does not directly affect FimH binding to mannose. Notably, it was shown that the surface glycan profile of Caco-2 intestinal cells changes dramatically during differentiation, with a significant decrease in mannose over time⁵⁹. This supports that Kp fimbriae and the post-translational modification driven by CfmP do not function in the context of a UTI, but instead may play an important role in intestinal adhesion and colonization.

Discussion

Together, our data supports a model in which CrrAB induces high pathogenicity in Kp through its post-transcriptional regulation of cfmP. We found that CfmP induces the oxidation of a highlyconserved histidine residue in the major pilin of fimbriae, FimA. CfmP-dependent modification of FimA causes increased adhesion to host cells and bacterial proliferation within the host, thus increasing pathogenicity (Fig. 6).

Figure 6: CrrAB regulation of cfmP induces the post-translational modification of fimbriae to increase the pathogenicity of Kp.

The CrrAB TCS induces the expression of cfmP in the crr locus. CfmP (dark green) induces the oxidation (blue dot) of a conserved histidine in FimA (light green), the major pilin of type I pili, also known as fimbriae. The oxidized, surface-exposed fimbriae in the outer membrane cause increased adhesion to host cells. This increased adhesion then results in high bacterial proliferation in the host and a subsequent increase in pathogenicity.

Altogether, our results are consistent with the oxidation of FimA-H11 being CrrAB- and CfmP-dependent (Fig. 5, SI Appendix Fig. 8). Predicted structural homology indicates that CfmP is a novel member of the TupA-like ATP-grasp enzyme family (Fig. 3a). TupA-like ATP-grasp proteins have diverse enzymatic activities, with a subset being predicted amidoligases^33,39. For several members of this family there is biochemical evidence supporting their amidoligase activity, such as DdlA/B and RimK^60,61. However, the N-terminal (non-ATP-grasp) domain of CfmP differs significantly from these enzymes. Instead, CfmP’s predicted structure is more similar to WemR and other enzymes hypothesized to add amino acid modifications to OM polysaccharides (Fig. 3a)^39,40. Despite this, our data shows that CfmP activity does not require the presence of OM polysaccharides, indicating that CfmP does not post-translationally modify any surface polysaccharide in Kp to induce high pathogenicity. Consistent with this, the N-terminal domains of WemR and CfmP are not identical; many of the most highly conserved residues among CfmP homologs are not conserved in WemR (Fig. 3a,b), supporting that CfmP may have a unique function. Additionally, WemR and its homologs are often found in genomic neighborhoods with their predicted substrate, such as K- and O-antigen clusters^39,40. cfmP, however, is conserved within the crr locus, and not found within polysaccharide biogenesis loci, further indicating that it may have a novel function. In contrast, we identified a CfmP-dependent oxidation on fimbriae, a proteinaceous extracellular polymer (Fig. 4). Our data does not rule out the possibility that the CfmP-dependent modification of fimbriae requires an additional factor to fulfill the enzymatic requirements for oxidation of FimA-H11. Biochemical analyses will be required to determine whether CfmP directly oxidizes FimA, or whether there are intermediate or interacting partners necessary for this post-translational modification. This is one of the first instances discovered of a PTM to fimbriae and one of the only examples of a regulated oxidative PTM on histidine in bacteria. Further studies will be integral to fully uncover the mechanistic details of how this oxidation is catalyzed in vivo and the specific activity of CfmP in this process.

Our data shows that CfmP increases adhesion to host cells in an intestinal epithelium model, suggesting that this leads to high bacterial loads during colonization and pathogenicity (Fig. 5,6). Notably, we found no post-translational modifications on the adhesin FimH that impact Kp virulence. Rather, it is a CfmP-dependent oxidation of FimA-H11 in the pilus rod that is necessary for pathogenicity. The oxidized histidine is located within FimA’s N-terminal extension, which is critical for FimA polymerization⁴⁹. Interestingly, the histidine is predicted to face inward towards the polymer lumen and does not obviously participate in subunit-subunit contacts⁴⁹ (Fig. 5c). The FimA pilus is a helical polymer; the flexibility and force required to extend the helix has important implications for adhesion and significantly impacts gut colonization and urinary tract infections in E. coli^{49,50,62–65}. Further, it has been shown that the identity of the FimA pilus rod, rather than the FimH adhesin, can significantly alter host cell adhesion^66,67. Thus, we hypothesize that oxidation of FimA-H11 alters the conformation and/or dynamics of the FimA rod, allowing for better adhesion to host surfaces and subsequent colonization. Notably, Kp intestinal colonization is often a reservoir for infection^68,69. This suggests that fimbriae also play a role in establishing or maintaining the intestinal reservoir of Kp thereby increasing the risk of infection and transmission. Future work will focus on optimizing a reliable murine model for intestinal colonization of ST258 and other CRKp to further test the impact of CrrAB and CfmP in this context.

Genomic analyses we and others^16,17 performed show that crrAB is a key regulatory system of ST258 CRKp, and we hypothesized that it may be a major factor in explaining the global dominance of this important ST. Here, we utilized a hyperactive mutant of CrrAB isolated from a human infection to further our understanding of the function of this important signaling system. We identified and characterized a novel protein under the transcriptional control of CrrAB, CfmP. We found that CfmP induces a novel post-translational modification to fimbriae that functions to increase host cell adhesion. Together this data increases our understanding of the role CrrAB and CfmP play in Kp biology and pathogenicity. The environmental conditions that are necessary for the upregulation of cfmP by the CrrAB TCS are unknown and the activating signal of the CrrB histidine kinase has not been discovered. Given the impacts CrrAB signaling has on host-pathogen interactions, it is tempting to speculate that stresses encountered during infection may impact CrrAB signaling. In addition, we and others have shown that CrrAB increases resistance to polymyxins^16,21,22,25, antibiotics with structural and functional similarities to host-produced AMPs. Significantly, we found that crrB*L94M also increases resistance to the human AMP LL-37 (SI Appendix Fig. 10). Thus, our data indicates that CrrAB activity both increases resistance to host innate immunity effectors and increases the capacity of Kp to colonize the host environment. Together, this study supports that this TCS plays a pivotal role in the pathogenicity of Kp ST258.

Materials and Methods

Strains, plasmids, and routine growth conditions

Unless indicated, all Klebsiella pneumoniae (Kp) strains were derived from the parental NR5452 ST258 clinical isolate, from the Columbia University Medical Center Clinical Microbiology Laboratory²². Cells were grown in Cation Adjusted Mueller Hinton broth (CAMHB) at 37°C, unless otherwise specified. Antibiotic concentrations were used at: 1 mg/mL zeocin and 50 μg/mL nourseothricin for Kp. Antibiotic concentrations used for cloning in DH10β were 40 μg/mL zeocin, 50 μg/mL nourseothricin, and 100 μg/mL ampicillin.

Strains, plasmids, oligonucleotide primers, and details on strain and plasmid construction can be found in the SI Appendix.

Sequence typing and crrAB genomic searches

Previously-sequenced carbapenem resistant Kp isolates collected from Columbia University Irving Medical Center from 2009–2023 were used for genomic analyses^25–27. Multi-locus sequence typing (MLST) and genomic searches for crrAB were performed using SRST2⁷⁰. For crrAB searches, the first ST258 Kp isolate collected per patient was included in the analysis. Hybrid genome assemblies were constructed from short- (Illumina) and long-read whole genome sequencing data using Unicycler⁷¹, followed by annotation using Prokka v1.12⁷².

Gene ontology analysis

Genomic neighborhood analysis was performed using the Enzyme Function Initiative Genome Neighborhood Tool5 (EFI-GNT v1.0) available at: https://efi.igb.illinois.edu//efi-gnt/index.php. The sequence of CrrB was used as the query for a BLAST search of the UniProt database, using the “Retrieve neighborhood diagrams” option within EFI-GNT. Gene neighborhood diagrams (GND) were generated to visualize the 20 nearest genes surrounding crrB.

RNAseq

Bacterial strains were grown to mid-log phase in Cation-adjusted Mueller Hinton Broth (CAMHB) at 37°C and cells were harvested and normalized to 500 μL of OD₆₀₀ = 0.5. Cell pellets were resuspended in RNAprotect (Qiagen) and RNA was extracted using the RNeasy kit with an on-column DNase treatment (Qiagen). RNAseq libraries were prepared using the QIAseq UPXome RNA Library Kit for Bacteria, with FastSelect rRNA depletion (Qiagen). Sequencing was performed on an Illumina MiSeq using a v2 300 cycle reagent kit. Quality-filtered reads (trimmomatic⁷³) were mapped to the NR5452 wildtype single contig reference genome using bowtie2⁷⁴, and samtools⁷⁵ and htseq⁷⁶ were used to generate read count tables. DESeq2⁷⁷ in R was used to identify genes with significantly altered transcription. Heatmaps were generated in R using ggplot2⁷⁸.

Quantitative reverse transcription PCR (RT-qPCR)

Bacterial strains were grown to mid-log phase in Cation-adjusted Mueller Hinton Broth (CAMHB) at 37°C and cells were harvested and normalized to 500 μL of OD₆₀₀ = 0.5. Cell pellets were resuspended in RNAprotect (Qiagen) and RNA was extracted using the RNeasy kit with an on-column DNase treatment (Qiagen). cDNA was generated from 500 ng input RNA using LunaScript RT SuperMix Kit (NEB) and was used as a template for RT-qPCR reactions using SsoAdvanced Universal Inhibitor-Tolerant SYBR Green Supermix (BioRad). Log fold changes were calculated using the 2^−ΔΔC_T method⁷⁹, with rpsL used as the housekeeping gene.

Drug susceptibility testing

Minimum inhibitory concentrations (MICs) were determined using broth microdilution (BMD) according to Clinical and Laboratory Standards Institute guidelines (CLSI, 2017). Suspensions of each bacterial strain were made to a MacFarland standard of 0.5, diluted 1:100 in Cation-Adjusted Mueller-Hinton broth (CAMHB), and inoculated into a 96 well plate containing increasing concentrations of antibiotic/compound in CAMHB and incubated for 24 hours at 37°C.

Galleria mellonella infections

Bacterial strains were grown in Cation-adjusted Mueller Hinton Broth (CAMHB) at 37°C to mid-log phase, and cells were harvested and normalized to OD₆₀₀ of 0.05 in phosphate buffered saline (PBS). 10 μL of each bacterial suspension (corresponding to 1 × 10⁷ colony forming units (CFUs)/mL, or 10⁵ CFUs total per larva) was injected into the last left proleg of 20 G. mellonella larvae per strain⁸⁰. 20 larvae were injected with 10 μL of PBS as a negative control. For experiments in which larvae were additionally plated out for bacterial loads, 20 additional larvae were injected per strain for a total of 40 larvae per strain. Serial dilutions of each strain suspension were plated to calculate CFUs and ensure equivalent inoculums. Larvae mortality was monitored over the course of 3 to 7 days by tactile stimulus⁸⁰. To calculate bacterial proliferation, 3 larvae per strain, per day were each homogenized in 500 μL PBS, and serial dilutions were plated on KPC Chromagar^™ agar plates to select for the injected carbapenem resistant Kp. Larvae injected with only PBS showed no growth on KPC plates. 2 μL of 10⁻¹ to 10⁻⁶ serial dilutions of homogenized G. mellonella were plated and dilution at which growth stopped was used calculate total bacterial load per larva.

Cytokine production panel

THP-1 monocytes were obtained from ATCC (TIB-202). Cells were maintained in ATCC-formulated RPMI-1640 medium containing 10% fetal bovine serum (FBS), 100 U/mL penicillin, and 100 μg/mL streptomycin at 37°C with 5% CO₂. THP-1 cells were seeded into 12-well culture-treated plates at 5×10⁵ cells per well and were incubated with 200 nM phorbol 12-myristate 13-acetate (PMA) for 72 hours to differentiate into macrophages. 24 hours prior to infection/incubation with antigen, differentiated and adhered THP-1 cells were washed twice with phosphate buffered saline (PBS) and wells were replenished with fresh media, without PMA.

Bacterial strains were grown to mid-log phase and 1.5 mL of OD₆₀₀ = 0.05 was harvested and resuspended to concentrate in 60 μL of PBS. 20 μL of bacteria were added to each well of differentiated THP-1 macrophages, for an approximate MOI = 10. As a positive control, purified LPS from Escherichia coli 0111:B4 (InvivoGen) was added to a final concentration of 100 ng/mL. Plates were then incubated at 37°C with 5% CO₂ for 4 hours. Input bacterial suspensions were serially diluted and plated for CFUs to confirm equal bacterial loads between strains. After 4 hours, supernatant containing secreted cytokines was collected from wells, filter sterilized (0.22 μm), and stored at −80°C. Samples were analyzed using the Human Cytokine Proinflammatory Focused 15-Plex Discovery Assay Array (HDF15) from Eve Technologies.

TLR4/TLR2 signaling assays

HEK-Blue^™ hTLR4 and hTLR2 cells were obtained from InvivoGen. Cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM) containing 10% fetal bovine serum (FBS), 100 μg/mL Normocin^™, 100 U/mL penicillin, 100 μg/mL streptomycin, and HEK-Blue Selection^™ at 37°C with 5% CO₂.

For signaling assay, HEK-Blue cells were suspended in HEK-Blue^™ Detection medium at 140,000 cells/mL and 180 μL was added per well (corresponding to 25,000 cells) into 96-well culture-treated plates. Bacterial strains were grown to mid-log phase and normalized to OD₆₀₀ = 0.03125 in PBS (corresponding to a multiplicity of infection of 5 in 20 μL). Bacteria were heat killed at 80°C for 20 minutes and serially diluted to the appropriate MOI in equivalent volumes (ranging from 0.01 to 5). Plates were incubated at 37°C with 5% CO₂ for 18 hours. Plates were briefly shaken, and OD₆₄₀ measurements were taken.

Sequence Alignment

Protein sequences of ATP-grasp domain containing proteins and FimA were obtained from the UniProt KnowledgeBase⁸¹ (UniProtKB; https://www.uniprot.org/) or in-house isolated and sequenced clinical strains. Multi-sequence alignments were performed using either MUSCLE or Clustal Omega⁸². Pairwise sequence alignments were performed and percent identity calculated using EMBOSS Needle⁸². Alignments were visualized in Jalview⁸³.

Immunobloting

The OD₆₀₀ was recorded for each culture, and cells were harvested to a final, normalized OD₆₀₀ of 10 for equivalent loading. Pellets were collected and re-suspended in 1X sodium dodecyl sulfate (SDS) sample buffer (10% glycerol, 50mM TrisHCl pH 6.8, 2% SDS, 0.005% bromophenol blue) containing 10% β-mercaptoethanol. Cells were heated at 95°C for 10 min prior to loading. Proteins were separated by SDS-PAGE on 4–20% TGX polyacrylamide gels (Biorad), transferred onto nitrocellulose membranes (Biorad), and blocked in 3% bovine serum albumin (BSA) in phosphate-buffered saline (PBS) with 0.5% Tween-20. The blocked membranes were probed with anti-His (1:5,000) (GenScript) or anti-FimA (1:5,000) (Invitrogen) diluted into 3% BSA in 1x PBS with 0.05% Tween-20. Primary antibodies were detected using horseradish peroxidase-conjugated goat anti-mouse or anti-rabbit IgG (BioRad) and the Clarity^™ Western ECL Substrate (Biorad). Signal was detected using a ChemiDoc MP Imaging System (Biorad).

Cell envelope polysaccharide detection

Bacterial strains were grown to mid-log phase and cells were harvested and normalized to 500 μL of OD₆₀₀ = 1.0. Cell pellets were resuspended in 50 μL 1x SDS sample buffer (10% glycerol, 50mM TrisHCl pH 6.8, 2% SDS, 0.005% bromophenol blue) containing 10% β-mercaptoethanol. Samples were boiled 10 minutes, then incubated overnight at 55°C with 300 μg/mL Proteinase K (Qiagen). Samples were run at 150V for 1 hour on a 4–12% Bis-Tris polyacrylamide gel (Biorad) and capsule / LPS was labeled using the Pro-Q^™ Emerald 300 Glycoprotein Gel and Blot Stain Kit (Invitrogen) following the manufacturer’s instructions.

Lipid A radiolabeling and thin layer chromatography

To evaluate any changes in the lipid A domain of LPS, E. coli and K. pneumoniae were grown in the presence of 2.5 μCi/ml of ³²P_i (Perkin-Elmer). The lipid portion of LPS was released using mild acid hydrolysis (pH 4.5, 100°C, 30 min), which cleaves the bond between lipid A and the first residue of the core oligosaccharide. Lipid A species were then isolated following established protocols^84,85 and resolved by thin-layer chromatography (TLC) using a solvent mixture of chloroform, pyridine, 88% formic acid, and water (50:50:16:5, vol/vol). Radiolabeled species were detected by phosphorimager analysis. E. coli K-12 strains W3110 and WD101 were included as controls. W3110 produces unmodified lipid A and is polymyxin sensitive, whereas the lipid A of WD101 is highly modified and is polymyxin resistant.

Fimbriae purification

Bacterial strains were grown to late-log phase in CAMHB at 37°C, and 125 mL of OD₆₀₀ of 1.0 was harvested as normalized input. Cells were pelleted at 5000 xg for 30 minutes and washed twice in 2.5 mL 10 mM Tris pH 8. Suspensions were incubated at 65°C for 1 hour to release fimbriae from the cell surface. Cells were pelleted and supernatant containing fimbriae was retained. NaCl was added to 500 mM final and MgCl₂ was added to 250 mM final, and supernatants were incubated with rocking at 4°C overnight. Precipitated protein was spun at 14,000 xg, high salt buffer was removed, and fimbriae were resuspended in 10 mM Tris pH 8 to re-solubilize proteins. Protein concentration was measured using a BCA Protein Assay Kit (Pierce). Equal volumes of purified fimbriae from each strain were run on 4–12% Bis-Tris polyacrylamide gel (Biorad) and stained with GelCode Blue Stain Reagent (ThermoFisher).

Transmission electron microscopy (TEM)

Bacterial strains were grown to log phase in CAMHB at 37°C. Cultures were concentrated to 100 μL of OD₆₀₀=10 and resuspended in 500 μL of 2% glutaraldehyde in 0.1X PBS for fixation. Cells were fixed for 1 hour at 4°C and then washed 5 times in 500 μL 0.1X PBS. Cell suspensions were spotted onto 200 copper mesh grids coated in pure carbon (Ted Pella) and imaged on a FEI Talos F200X Transmission / Scanning Electron Microscope.

Mass spectrometry and data analysis

Mass spectrometry analyses were performed at the Mass Spectrometry Facility, UMass Chan Medical School. For details see SI Appendix.

Mouse infections

Urinary tract infections were performed as previously described⁸⁶. Briefly, 50 μl of 1×10⁹ CFUs of bacteria (suspended in phosphate buffered saline) were introduced into the bladders of anesthetized 8- to 16-week-old female C3H/HeN mice (Charles River Laboratories, strain number 025) through a soft polyethylene catheter (0.61 mm outer diameter; BD Biosciences). After 24h, CFUs in urine (collected directly from mice), bladder, and kidney homogenates were quantified by serial dilution on LB agar plates. Mouse work was performed according to approved IACUC protocol at Columbia University Medical Center.

Hemagglutination titers

Bacterial hemagglutination was performed with guinea pig erythrocytes (Colorado Serum Company). Erythrocytes were washed and resuspended in PBS to an OD₆₄₀=2.0. Bacterial cultures were grown in CAMHB at 37°C to mid-log (Kp strains) or under static, 2×24h conditions (UTI89) and normalized to OD₆₀₀=10 in PBS. 25 μL of bacterial suspensions were serially diluted in PBS with and without the addition of 4% methyl α-D-mannopyranoside. An equal volume of erythrocytes was added and gently mixed. Plates were incubated at 4°C overnight.

Epithelial cell adhesion assay

Caco-2 colon epithelial cells were obtained from ATCC (HTB-37). Cells were maintained in ATCC-formulated Eagle's Minimum Essential Medium (EMEM) containing 20% fetal bovine serum (FBS), 100 U/mL penicillin, and 100 μg/mL streptomycin at 37°C with 5% CO₂. To prepare cells for the bacterial adhesion assay, Caco-2 cells were seeded at 2.1×10⁵ cells/mL per well in a culture-treated 12-well plate. Caco-2 layers were differentiated for 14 days, with media renewed every three days. Prior to adding bacteria, old media was removed from the Caco2 cells, and fresh media without antibiotics was added to each well.

Bacterial strains were grown in CAMHB at 37°C to mid-log phase and normalized to OD₆₀₀ = 0.25 in 500 μL of PBS. 20 μL of bacteria were added to each well of Caco-2 cells (for an approximate MOI = 5) and were incubated at 37°C for 30 minutes. Input bacterial suspensions were serially diluted and plated for CFUs. After incubation, wells were washed 5 times with 500 μL of PBS, and contents were resuspended in 1 mL ice-cold 0.01% Triton X-100. Suspensions were serially diluted and plated for output CFUs. Adhesion/internalization was calculated as the % remaining of total input bacterial counts, post-washes for each strain.

Confocal microscopy

Caco-2 cells were cultured as described for the “Epithelial cell adhesion assay” but were grown in a 1.5 uncoated coverslip 10 mm diameter 12-well plate (MatTek). 1 hour prior to infections, 10 μg/mL Hoechst 33342 (BD Biosciences) was added to the cultures. Bacterial strains were grown in CAMHB at 37°C to mid-log and normalized to OD₆₀₀ = 0.25 in 500 μL of PBS. 20 μL of bacteria were added to each well of Caco-2 cells (for an approximate MOI = 5) and were incubated at 37°C for 30 minutes. Input bacterial suspensions were serially diluted and plated for CFUs. After incubation, wells were washed 2 times with 500 μL of PBS, then stained for 10 minutes with 5 μg/mL CellMask^™ Deep Red Plasma Membrane Stains (Invitrogen) in 1X PBS. Wells were washed another 3 times with 500 μL of PBS prior to imaging. For imaging details and settings, see SI Appendix.

Significance Statement.

Multi-drug resistant Gram-negative bacteria, including carbapenem-resistant Klebsiella pneumoniae (CRKp), are a public health emergency. The predominant CRKp sequence type worldwide is ST258, but the molecular underpinnings contributing to its pathogenicity are poorly defined. Here, we show that the two-component system crrAB is a genomic feature of ST258 and drives virulence in this sequence type through its regulation of an uncharacterized protein, CfmP. We provide evidence that CfmP induces a novel post-translational modification to fimbriae that increases host cell adhesion and colonization, thereby increasing the pathogenic potential of ST258 CRKp. Moreover, we found that CrrAB and CfmP are conserved in other multi-drug resistant Enterobacterales, broadly implicating this system in driving virulence in this important class of pathogens.

Data, Materials, and Software Availability

All genomic data (including results from whole-genome sequencing and RNA sequencing), and electronic data will be deposited on NCBI’s GenBank, Sequence Read Archive, or Gene Expression Omnibus databases. All other data are included in the article and/or SI Appendix.