1.1 The hipB/A operon is conserved across Klebsiella and other Gram-negative bacteria

The hipB/A operon is established as one of the main drivers of antibiotic tolerance in E. coli; therefore, we first compared the amino acid sequence of HipB/A with its putative homolog in K. pneumoniae. Pairwise sequence alignment using BLASTp showed that HipA_kp and HipB_kp shared 69% and 56% of sequence identity with HipA_ec and HipB_ec, respectively (Figs 1A, S1A and S1B). The alignment of the HipA_kp structure as predicted by AlphaFold [22,23] with the experimentally determined HipA_ec structure [24] showed a high degree of conservation (RMSD 0.444 Å), including within the ATP and Mg²⁺ ion binding pockets important for kinase activity (Fig 1B). Furthermore, several residues known to be essential for kinase activity were conserved between HipA_kp and HipA_ec, such as the autophosphorylation site S150, the catalytic residue D309, the residue L181 involved in ATP-binding, and the residues N314 and D332 involved in Mg²⁺ binding [13] (S1A Fig). Alignment of the AlphaFold-predicted structure of HipB_kp with the experimentally determined structure of HipB_kp [24] showed a similar degree of conservation (RMSD 0.405 Å) (S1C Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Bioinformatic analysis of hipA_kp and hipB_kp.

A) Schematic representation of the hipB/A operon in K. pneumoniae and E. coli. The numbers indicate the percent identity of HipA and HipB between the species. B) Alignment of AlphaFold predicted structure of HipA_kp, UniProt ID: A6T928 (cyan), with the experimentally determined structure of HipA_ec, PDB ID: 3DNT (orange), together with ATP (sticks), and Mg²⁺ molecules (yellow spheres), in the conserved pocket. C) Distribution of percentage of sequence identity of HipA_kp homologs within the Klebsiella genus and the number of hits obtained upon analyzing the top 1,000 results from protein BLAST of HipA_kp protein with all the default settings except limiting the organism search to “Klebsiella (taxid:570)” as “Organism” in Standard settings.

https://doi.org/10.1371/journal.ppat.1012759.g001

Protein BLAST analysis showed that the HipA_kp is conserved and present in different species of Klebsiella, with the mean percent identity varying from 95% to 99% in different isolates of the same genus, species and subspecies (Fig 1C). For further visualization of the presence of HipA_kp across all organisms, we plotted the percentage of sequence identity of HipA_kp homologs from the top 5,000 protein BLAST hits and determined that the sequence identity reaches up to 70% in many Gram-negative bacteria such as E. coli and bacteria belonging to the genus Shigella, Serratia and Salmonella. Some of the genera were under-represented due to the large number of hits originating from Klebsiella and E. coli (S1D Fig). Combined, these results showed high structural similarity of HipB/A between K. pneumoniae and E. coli, as well as conservation of HipA among many Gram-negative bacteria.

1.2 Overproduction of HipA_kp is toxic to the cell and can be counteracted by HipB_kp

To investigate the role of hipA_kp, we first analyzed its effect on the growth and viability of E. coli cells in LB medium. As previously reported for hipA_ec, overexpression of hipA_kp was expected to be toxic to the cells. Therefore, we ectopically expressed hipA_kp in E. coli under the control of an arabinose-inducible promoter with an optimized Shine-Dalgarno sequence [25]. We observed that overexpression of hipA_kp was highly toxic to E. coli cells, resulting in reduction of their growth after 1 h post-induction by three-fold and survival by 2.5-fold, as measured by optical density (OD_600nm) and colony forming units (CFU), respectively (Fig 2A). To determine whether the overproduction of the putative antitoxin HipB_kp counteracted the activity of hipA_kp in E. coli, we simultaneously overexpressed hipA_kp and hipB_kp from different plasmids. Compared to the growth of E. coli overexpressing only hipA_kp, simultaneous overproduction of hipB_kp restored the growth of hipA_kp-overexpressing E. coli and therefore reversed the hipA_kp-related toxic phenotype (Fig 2B). We therefore concluded that overproduced HipA_kp and HipB_kp act as a canonical toxin/antitoxin pair in E. coli.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Effect of overexpression of hipA_kp on WT E. coli and WT & ΔhipA K. pneumoniae cells.

A) Growth and viability of E. coli containing the empty pBAD33 vector and pBAD33::hipA_kp. The expression of hipA_kp is under the control of the arabinose-inducible promoter. After the cells reached an OD_600nm of 0.3, expression of hipA_kp was induced with 0.2% arabinose for 1 h. Growth was monitored by absorbance measurements at OD_600nm and viability was determined by CFU quantification. Bacteria were harvested at 1 h post-induction for proteome and phosphoproteome analysis. B) Growth of E. coli carrying pBAD33::hipA_kp alone or together with plasmid pGOOD::hipB_kp, in which hipB_kp is under the control of an IPTG-inducible promoter. Overnight cultures of the bacteria were used to inoculate the cultures for the assay at 0.08 OD_600nm in medium containing 0.2% arabinose and 1 mM IPTG. As a control, one set of samples was left uninduced and one strain carrying and expressing only hipA_kp. The growth was followed via OD_600nm measurements for 5 h in a plate reader (Tecan). C) Growth and viability of WT K. pneumoniae harboring the empty pBAD33 plasmid and pBAD33::hipA_kp and ΔhipA containing pBAD33::hipA_kp with expression under the control of arabinose-inducible promoter. The strains were grown in LB Lennox and induced with 0.2% arabinose at OD_600nm 0.3 for 1 h and growth was followed by optical density at OD_600nm and CFU for viability on plate. Samples were harvested at 1 h post-induction for proteome and phosphoproteome analysis. D) Growth of ΔhipA K. pneumoniae carrying pBAD33::hipA_kp and pGOOD::hipB_kp in which expression of hipA_kp and hipB_kp is under the control of arabinose-inducible and IPTG-inducible promoters, respectively. Cultures were started at 0.05 OD_600nm and grown till OD_600nm of 0.3 and induced with 0.2% arabinose and 1 mM IPTG. Uninduced conditions served as controls. The growth was followed via optical density for approximately 5 h in a plate reader. All the plots are representative of mean values ± SD of three independent experiments.

https://doi.org/10.1371/journal.ppat.1012759.g002

We next overexpressed hipA_kp in the native K. pneumoniae background. To this end, we used the K. pneumoniae isolate ATCC13883 (wild-type, WT), which harbors the hipB/A operon on the chromosome. First, we generated an unmarked hipA_kp gene (ΔhipA) deletion mutant and assessed the impact of the deletion on the growth and viability of K. pneumoniae by measuring the optical density (OD_600nm) and CFU levels. We observed that the growth of the ΔhipA mutant was almost identical to that of the WT in LB over a 24 h incubation period. Likewise, the deletion of hipA did not affect the survival of K. pneumoniae in the exponential phase and resulted in a slight reduction of CFU/ml in the stationary phase (S2A Fig). We next overexpressed hipA_kp from the pBAD33 vector in both WT and ΔhipA K. pneumoniae. The overproduction of HipA_kp decreased growth and viability by three-fold at 1 h post-induction (Fig 2C). We tested if the overexpression of hipB_kp in K. pneumoniae WT + hipA_kp and ΔhipA + hipA_kp could counter the toxic effect of hipA_kp expression. We observed that bacteria expressing both hipA_kp and hipB_kp grew better than ΔhipA + hipA_kp indicating that HipB_kp partially rescued the toxic effect of HipA_kp (Fig 2D). These results point to a similar role of the antitoxin HipB in K. pneumoniae and E. coli cells.

1.3 HipA_kp has Ser/Thr kinase activity and phosphorylates multiple substrates

Based on the sequence similarity of HipA_kp to HipA_ec and the presence of conserved kinase and ATP-binding domains (S1A Fig), we postulated that HipA_kp has kinase activity. In order to identify its putative substrates, we overexpressed hipA_kp in E. coli and performed quantitative phosphoproteome analysis using liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS). At the proteome level, we performed a one-sample t-test with FDR<0.05 to prepare a volcano plot, based on the log₂ ratio of WT E. coli with HipA_kp induced and empty vector (S2B Fig), which showed that overproduced HipA_kp did not affect the regulation of other proteins. We measured a more than eight-fold increase in HipA_kp in the cells upon hipA_kp overexpression for 1 h in comparison with the empty pBAD33 vector control in both replicates (S2C Fig and Tab A in S1 Dataset). These results confirmed the efficiency of the hipA_kp overexpression strategy.

At the phosphoproteome level, we identified 317 phosphorylation sites on 189 proteins in E. coli + hipA_kp 1 h post-induction, with an excellent correlation between the replicates (Pearson’s correlation coefficient value of 0.91) (S2D Fig and Tab B in S1 Dataset). Upon HipA_kp overproduction, we reproducibly detected increased phosphorylation of multiple substrates; among them was GltX, which showed a more than 16-fold increase in phosphorylation at position S239 in all replicates. The direct comparison of phosphoproteome ratios normalized to the proteome and unnormalized confirmed that most of the phosphorylation sites that were more than four-fold up-regulated, remain unaffected by normalization except for HipA_kp autophosphorylation sites (S2E Fig and Tab A in S1 Dataset). The data indicate that HipA_kp has kinase activity and GltX is its major substrate in E. coli.

To investigate HipA_kp activity and identify its potential targets in the native organism K. pneumoniae, we induced expression of hipA_kp from pBAD33 in WT and ΔhipA K. pneumoniae in LB medium for 1 h and performed LC-MS/MS analysis. The reproducibility between the proteome and phosphoproteome data of three replicates in both WT and ΔhipA backgrounds with HipA_kp overproduction was high (overall Pearson’s correlation coefficient >0.7) (S3A and S3B Fig). One sample t-test was performed based on the log₂ proteome ratios from three replicates. The results revealed a more than 16-fold increase in HipA_kp levels in the WT background (Fig 3A) and ΔhipA (Figs 3B and S3C) as compared to bacteria harboring the empty vector (Tab C in S1 Dataset). At the phosphoproteome level, we identified a total of 747 phosphorylation sites, belonging to 417 phosphoproteins (Tab D in S1 Dataset). These phosphorylation sites were filtered for localization probability of >0.75. Upon HipA_kp overproduction, we reproducibly detected increased autophosphorylation of HipA on S150 and T158 by more than 16-fold, as well as phosphorylation of GltX on S239, also by 16-fold, in both WT (Fig 3C) and ΔhipA K. pneumoniae (Fig 3D). The comparison of ΔhipA and WT K. pneumoniae overexpressing hipA_kp phosphoproteome showed a high correlation of 0.83, with several phosphorylation sites more than four-fold upregulated compared to empty vector control (S3D Fig). Up-regulation of these additional sites indicates that HipA_kp acts on multiple substrates that may play a role in the toxic phenotype. All the phosphorylation sites were normalized to the proteome which showed that most of the up-regulated phosphorylation sites remain unaffected by normalization except HipA autophosphorylation sites in WT and ΔhipA K. pneumoniae (Figs 3E, 3F, S3E and S3F).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Effect of overexpression of hipA_kp on the proteome and phosphoproteome of WT and ΔhipA K. pneumoniae.

A and B) Volcano plots showing differential expression to compare the proteome of K. pneumoniae cells overproducing HipA_kp with the empty vector control. Student’s one sample t-test was performed for n = 3 biological replicates. The x-axis represents log₂ fold changes based on the indicated ratio and the y-axis represents negative log₁₀ of the Benjamini-Hochberg corrected (FDR<0.05) t-test p-value. The curve lines indicate confidence levels with p-value = 0.05. Proteins with a p-value of < 0.05 are indicated by black dots and represent the significantly up and down-regulated proteins. C and D) Scatter plot showing the distribution of quantified phosphorylation sites upon hipA_kp overexpression in WT and ΔhipA background based on log₂ ratio vs sum of intensity, with the average median values from three replicates. The phosphorylation sites with log₂ ratio greater than 2 were more than four-fold up-regulated upon HipA_kp overproduction compared to K. pneumoniae harboring the empty vector. E and F) The ratios for phosphoproteome were normalized to proteome and plotted here against the sum of intensity. C)–F) The phosphorylation sites that were more than four-fold up-regulated (black dots) are numbered and listed here. The size of each phosphorylation site in the scatter plot corresponds to its respective peptide score and scale is provided in the top left corner of the plot.

https://doi.org/10.1371/journal.ppat.1012759.g003

1.4 HipA_kp overproduction in K. pneumoniae leads to increased tolerance to ciprofloxacin, but not to gentamicin

We next investigated the effect of HipA_kp overproduction on the growth and survival of K. pneumoniae in the presence of gentamicin or ciprofloxacin, antibiotics commonly used in treatments of K. pneumoniae infections. We first determined the susceptibility of the K. pneumoniae isolate ATCC13883 to gentamicin, an aminoglycoside that inhibits protein synthesis, and found that a concentration above 3 μg/mL led to complete growth inhibition after 24 h of exposure. We then exposed the cells with and without hipA_kp overexpression to 4 μg/mL gentamicin and did not observe any influence of hipA_kp on the survival of K. pneumoniae cells (Figs 4A and S4A). These results were in agreement with a previously published antibiotic tolerance test where it was shown that HipA_ec conferred protection against several different classes of antibiotics but not against the aminoglycoside tobramycin [13]. We next focused on ciprofloxacin, a fluoroquinolone antibiotic that inhibits DNA replication in growing cells. Ciprofloxacin was shown to be effective in the treatment of K. pneumoniae infections [26] and was previously used for testing antibiotic tolerance in E. coli [27]. We determined the susceptibility of the K. pneumoniae isolate ATCC13883 against ciprofloxacin and found that a concentration above 0.5 μg/mL led to complete growth inhibition after 24 h of exposure (S4B Fig). K. pneumoniae WT and ΔhipA expressing hipA_kp from pBAD33 were exposed to 1 μg/mL ciprofloxacin for 2 h. Bacteria under uninduced conditions and bacteria harboring the empty pBAD33 vector were used as negative controls. The proteome measurement confirmed the overproduction of HipA_kp in the induced strains after 2 h of ciprofloxacin treatment. Importantly, only cells expressing hipA_kp showed survival after 2 h of ciprofloxacin treatment with a mean log₁₀ CFU/ml value of 8. In contrast, the viability of the negative controls was six orders of magnitude lower (log₁₀ CFU/ml value of 2) (Fig 4B). To further support the finding that HipA_kp has kinase activity, K. pneumoniae was transformed with pBAD33 harboring a hipA_kp gene that carries a mutation that leads to a D309Q amino acid substitution. The aspartate at position 309 is a catalytic residue in E. coli and the D309Q modification is expected to decrease kinase activity and ciprofloxacin tolerance in K. pneumoniae. Indeed, overproduction of the mutated HipA_kp protein reduced the number of viable K. pneumoniae following ciprofloxacin treatment as compared with bacteria expressing the wild-type hipA_kp allele (S4C and S4D Fig). We conclude from the data that HipA_kp kinase activity contributes to antibiotic tolerance against ciprofloxacin in K. pneumoniae under the tested conditions, as also previously observed for hipA_ec in E. coli [13].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Effect of overexpression of hipA_kp on antibiotic tolerance in K. pneumoniae.

A) Growth of WT and ΔhipA K. pneumoniae strains transformed with pBAD33 and pBAD33::hipA_kp, respectively in which hipA_kp expression was driven by the arabinose-inducible promoter. Strains were grown in LB Lennox medium and expression was induced at OD_600nm of 0.3 for 1 h followed by treatment with 4 μg/mL gentamicin for 2 h. Growth was followed via optical density and viability by measuring colony-forming units. B) Growth of K. pneumoniae WT and ΔhipA strains transformed with pBAD33 and pBAD33::hipA_kp with hipA_kp expression driven by the arabinose-inducible promoter. Strains were grown in LB Lennox medium and expression was induced at OD_600nm of 0.3 for 1 h followed by treatment with 1 μg/mL ciprofloxacin for 2 h. Growth was followed via optical density and viability by measuring colony-forming units. A) and B) Data shown are mean values ± SD of three biological replicates. C) Volcano plots showing differential protein expression to compare the proteome of ciprofloxacin-treated ΔhipA K. pneumoniae cells overproducing HipA_kp with bacteria containing an empty vector. Student’s one sample t-test was performed for n = 2 biological replicates. The x-axis represents log₂ fold changes based on the ratio and the y-axis represents negative log₁₀ of the Benjamini-Hochberg corrected (FDR<0.05) t-test p-value. The curves indicate confidence levels with p-value = 0.05. Proteins above the curve line on the left side are significantly up-regulated. D and E) Scatter plot showing the distribution of quantified phosphorylation sites after 2 h of ciprofloxacin treatment upon hipA_kp overexpression, based on log₂ ratio of ΔhipA + pBAD33::hipA_kp to WT + empty vector for phospho unnormalized (D) and normalized to proteome (E) on the x-axis and log₁₀ of sum of intensity on y-axis, with the average median values from two replicates. The phosphorylation sites with log₂ ratio greater than 2 were more than four-fold up-regulated upon HipA_kp overproduction compared to the empty vector sample. These phosphorylation sites (black dots) were numbered and listed on the right side of the plot in red. The size of each phosphorylation site in the scatter plot corresponds to its respective peptide score and scale is provided in the top left corner of the plot.

https://doi.org/10.1371/journal.ppat.1012759.g004

1.5 Phosphoproteome analysis reveals potential HipA_kp substrates in ciprofloxacin-treated K. pneumoniae

In order to detect the HipA_kp targets potentially involved in the antibiotic survival, we overexpressed hipA_kp in ΔhipA K. pneumoniae, treated the culture with ciprofloxacin for 2 h and compared the phosphoproteome results with uninduced and empty vector controls. At the proteome level, we identified a total of 1,889 proteins and confirmed HipA_kp overproduction upon induction based on the volcano plot showing differential protein expression and correlation plot (Figs 4C and S4E and Tab E in S1 Dataset).

At the phosphoproteome level, we identified 547 phosphorylation sites that showed a high correlation between biological replicates (Pearson’s correlation coefficient 0.88) (S4F Fig and Tab F in S1 Dataset). These phosphorylation sites were filtered for localization probability of >0.75. Following this, many phosphorylation sites were found to be up-regulated following ciprofloxacin treatment in hipA_kp overexpressing cells, with approximately 20 phosphorylation sites consistently observed as up-regulated in both unnormalized and normalized phosphoproteome datasets (Figs 4D, 4E, and S4G and Tab E in S1 Dataset). This included the autophosphorylation sites and GltX phosphorylation at S239, as compared to untreated hipA_kp overexpressing cells. The similarity in up-regulated phosphorylation sites with and without ciprofloxacin treatment indicates that overproduction of HipA_kp induced tolerance to ciprofloxacin by phosphorylating target proteins. All the putative phosphorylation sites of HipA_kp have been listed inS4 Table with the frequency of their occurrence to be more than four-fold increased or only identified upon hipA_kp overexpression (Tabs D, F, and H in S1 Dataset).

1.6 K. pneumoniae phosphoproteome reveals numerous pathways that are potentially regulated at the post-translational level

Our study provided the most comprehensive phosphoproteome dataset for K. pneumoniae so far [28–31], containing a total of 1439 phosphorylation sites from 741 proteins and a total of 2128 proteins from all experiments (S3 and S4 Tables and Tabs I and J in S1 Dataset). 37% phosphoproteins and 27% phosphopeptides occurred in multiple sets of independent experiments (Fig 5A and 5B). Further analysis revealed the distribution of phosphorylated serine, threonine, and tyrosine was 46.2%, 39.7%, and 14.0%, respectively (Fig 5C) Functional enrichment analysis performed on all identified phosphoproteins revealed the cellular processes potentially regulated by phosphorylation (Fig 5D and Tab K in S1 Dataset). Phosphoproteins were distributed across numerous cellular functions, with a significant proportion implicated in translation and RNA-binding. Additionally, many were associated with glycolysis, purine and pyrimidine biosynthesis, and DNA repair. The comparison of all previously published K. pneumoniae phosphoproteomics datasets [28–30] with our combined dataset (Tab I in S1 Dataset) revealed a large number of novel phosphoproteins and phosphopeptides that were previously unreported (Fig 5E and 5F). This dataset will serve as a valuable resource for researchers interested in studying protein phosphorylation in K. pneumoniae.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Overview of the K. pneumoniae phosphoproteome datasets.

A) and B) Venn diagram comparing the three phosphoproteome datasets of K. pneumoniae obtained in this study showing the overlap between quantified phosphoproteins (A) and phosphopeptides (B). C) Distribution of all phosphorylated serine (pS), threonine (pT) and tyrosine (pY) identified in the combined dataset from three experiments, showing the number of phosphorylation sites of each amino acid in bracket. D) Functional enrichment analysis of phosphorylated proteins showing the fold enrichment and number of proteins phosphorylated in the indicated pathway. E) Comparison of all previously published phosphoproteome datasets for K. pneumoniae with our dataset for the number of phosphoproteins and phosphopeptides identified. F) Analysis of the overlap between the number of phosphopeptides identified in K. pneumoniae in our study and those in published K. pneumoniae phosphoproteome studies [28–31].

https://doi.org/10.1371/journal.ppat.1012759.g005

1. Bacterial strains and plasmids

All strains, primers and plasmids used in this work are listed in the table below (S1 and S2 Tables). Due to the high homology of hipA_kp with hipA_ec, the hipA_kp gene was cloned with the same Shine-Dalgarno sequence containing the alternative starting codon GTG as used in previous work [19]. hipA_kp and hipB_kp genes were amplified from the Klebsiella pneumoniae subsp. pneumoniae reference strain ATCC13883 with primers HipA.kpn.SD8.GTG.pBAD33-for and HipA.kpn.pBAD33-rev, and hipBkp.pGOOD-for and hipBkp.pGOOD-rev, respectively. The pBAD33 vector and pGOOD vector were digested with XbaI and BglII restriction enzymes, respectively and finally pBAD33 was ligated with hipA_kp and pGOOD with hipB_kp amplified PCR product using Gibson assembly mix [55]. The resulting plasmids were transformed into Top10 E. coli competent cells and confirmed by PCR for the presence of the gene of interest and sequenced before transforming them into final working strains in both E. coli and K. pneumoniae. The Q5 Site-Directed Mutagenesis Kit (NEB) was used to introduce a GAC to CAG codon exchange into hipA_kp on pBAD33 leading to HipA_kpD309Q according to the manufacturer’s instructions. The plasmid was sequenced and cloned into K. pneumoniae.

Generation of K. pneumoniae ΔhipA mutant

A markerless in-frame deletion of hipA was generated by amplifying approximately 600 bp up- and down-stream of hipA. PCR products were fused with SOE-PCR and cloned with In-Fusion cloning (TakaraBio) into the suicide vector pKNOCK-Km, a gift from Mikhail Alexeyev (Addgene #46262). The sacB gene including its promoter was amplified from the plasmid pEXG2 and cloned into pKNOCK-Km to introduce a counter selection marker. Plasmid inserts were verified by Sanger sequencing. The resulting plasmids were transferred into E. coli S17λpir for the transformation of K. pneumoniae ATCC13883 by conjugation. K. pneumoniae merodiploids were selected on LB agar plates supplemented with 100 μg/ml kanamycin, streaked for single colony isolation and then incubated in LB without selection overnight. Counter selection was performed with 15% sucrose, and kanamycin sensitive colonies were tested for the deletion of hipA by PCR.

2. Bioinformatic analysis

The sequence of HipA_kp from Klebsiella pneumoniae subsp. pneumoniae ATCC13883 was analyzed for its identity within Klebsiella genus and other organisms using protein BLAST at the NCBI server (https://blast.ncbi.nlm.nih.gov/Blast.cgi). For the initial pBLAST search, the search was limited to Klebsiella (taxid:570) as the organism and 1,000 maximum target sequences in the algorithm parameters. Percent identity and number of hits per species and subspecies were extracted from the pBLAST result. For the analysis of HipA_kp identity across all organism without limiting the search to Klebsiella, we performed the pBLAST search for top 5,000 hits and plotted the results. Both the graphs were generated using the online tool, Shiny BoxPlotR (http://shiny.chemgrid.org/boxplotr/). The pairwise sequence alignment between HipA_kp and HiA_ec, or HipB_kp and HipB_ec proteins was performed with EMBOSS Needle [56] using the default settings.

3. Growth experiments

4. Antibiotic tolerance test of K. pneumoniae overexpressing hipA_kp

5. Cell lysis and protein precipitation

Harvested cells were centrifuged at 4,000 g and the pellet was stored at -80°C. The pellet was then resuspended in an SDS lysis buffer containing 40 mg/ml SDS (sodium dodecyl sulfate), 100 mM Tris-HCl pH 8.6, 10 mM EDTA, 5 mM glycerol-2-phosphate, 5 mM sodium fluoride, 1 mM sodium orthovanadate and 1 tablet of complete protease inhibitors (Roche). The cell lysate was sonicated at 40% amplitude for 30 secs cycle at least five times or until a transparent, non-viscous lysate was obtained. The cell debris was pelleted by centrifugation at 13,000 g for 30 mins and the supernatant was collected for protein precipitation using methanol and chloroform method. The obtained protein pellet was air-dried and dissolved in denaturation buffer (6 M urea, 2 M thiourea and 10 mM Tris pH 8.0). The protein concentration was determined by using standard Bradford assay (Bio-Rad).

6. Protein in-solution digestion

For each phosphoproteomics experiment, 3–6 mg protein per sample (strain/condition) was used. Briefly, precipitated proteins were reduced by using 1 mM dithiothreitol (DTT) for 1 h and then alkylated using 5.5 mM iodoacetamide (IAA) for an additional 1 h in the dark with constant shaking at 700 rpm. Half of the protein from each sample was diluted with four times volume of 62.5 mM Tris pH 8.0 and 12.5 mM CaCl₂ and digested with the enzyme chymotrypsin (1:100 w/w) overnight at room temperature (RT) in a shaker. The other half of the protein was pre-digested with the endoproteinase LysC (1:100 w/w) for 3 h and then diluted with four times volume of milli-Q water, adjusted to a pH higher than 8.0 and supplemented with the enzyme LysC (1:100 w/w) (for E. coli samples) and trypsin (1:100 w/w) (for K. pneumoniae samples) for overnight digestion at RT and shaking. The reaction was then stopped by acidification with trifluoroacetic acid (TFA) to pH 2.0 and centrifuged to get rid of precipitates.

7. Solid phase extraction and dimethyl labeling

Acidified peptides were then purified by solid phase extraction on Sep-Pak C18 cartridges (Waters, Milford, MA) and labeled using triplex stable isotope dimethyl labeling as previously described [57]. Briefly, C18 columns were activated with methanol and equilibrated with Solvent A* [2% (v/v) acetonitrile (AcN) and 1% (v/v) formic acid (FA)]. The digested and acidified peptide samples were loaded and later, the column was washed with HPLC Solvent A [0.1% (v/v) FA]. These samples were then labeled with 2.5 ml of the respective labeling solutions: CH₂O (Sigma-Aldrich) and NaBH₃CN (Fluka) for Light label, and CD₂O (Sigma-Aldrich) and NaBH₃CN for Medium label, and C₁₃D₂O (Sigma-Aldrich) and NaBD₃CN (96% D, Isotec) for Heavy label. The labeling solutions were flushed with approximately 10–15 mins contact time through with the column. Labeled peptides were washed again with HPLC Solvent A on the column and eluted with 600 μl HPLC Solvent B [80% (v/v) AcN in 0.5% (v/v) FA].

8. Labeling efficiency and mixing check

For the validation of labeling efficiency and accurate mixing of the labeled peptides, two sets of 5 μg of each eluted labeled sample were used for LC-MS/MS measurements, separately (for labeling efficiency) and mixed in 1:1:1 ratio for a pilot mixing check measurement. Adjustments were made based on the mixing ratios to achieve a target ratio close to 1 to ensure optimal quantification. The labeling efficiency, for all labels, was > = 95%.

9. Phosphopeptide enrichment

Phosphopeptides were enriched by titanium dioxide (TiO₂) beads with a ratio of 1:10 (beads: protein ratio) for five consecutive rounds of enrichment for 10 mins each. After mixing the labeled samples together and taking out an aliquot of 10 μg for proteome analysis, the peptides were acidified. TiO₂ beads were washed with 80% (v/v) AcN and 6% (v/v) TFA and incubated with the samples with constant mixing. The beads with bound phosphopeptides were washed again to remove any unbound or acidified peptides using 80% AcN and 6% TFA. These beads were then loaded onto C8 (Empore) StageTips and further washed with a wash buffer [80% AcN and 1% TFA] and Solvent B [80% AcN and 0.1% FA]. Phosphopeptides were eluted with 30 μl of elution solution I [1.25% (v/v) ammonium hydroxide of pH>10.5 into a tube containing 20 μl of 20% (v/v) TFA. This was followed by 70 μl elution solution II [5% (v/v) ammonium hydroxide in 60% (v/v) AcN (pH>10.5)] and finally with 20 μl of elution solution III [60% (v/v) AcN, 1% (v/v) TFA]. Each elution solution took at least 15 mins, at 1000–1500 rpm to elute. Acetonitrile was evaporated from eluates by vacuum centrifugation and samples were acidified to pH 2.0, and purified by StageTips.

10. Peptide purification by StageTips

Before LC-MS/MS measurements, samples for proteome analysis and eluted phosphopeptides were desalted and purified on C18 StageTips [58]. Briefly, reverse-phase chromatography was applied using C18 discs (Empore). The discs were activated with methanol and equilibrated with Solvent A*. The acidified peptides were loaded onto the discs and washed with Solvent A. Peptides were eluted with 50 μl of Solvent B and vacuum centrifuged to evaporate AcN. The final sample volume was adjusted with Solvent A and final 10% (v/v) of Solvent A*.

11. LC-MS/MS measurement

Purified peptides were separated by an online coupled EASY-nLC 1200 system (Thermo Fischer Scientific) to an Orbitrap Exploris 480 spectrometer (Thermo Fischer Scientific) through a nano-electrospray ion source (Thermo Fischer Scientific). Chromatographic separation was performed on a 20 cm long and 75 μm inner diameter analytical column packed in-house with reversed-phase ReproSil-Pur C18-AQ 1.9 μm particles (Dr. Maisch GmbH). Peptides were loaded onto the column at 40°C, with 1 μl/min flow rate under a maximum backpressure of 850 bar. The gradient was applied using HPLC Solvent A and 10 to 50% Solvent B at a 200 nl/min constant flow rate. Labeling efficiency samples were eluted using 36 mins, mixing check using 60 mins, phosphopeptides using 60 mins and proteome samples using 130 mins or 230 mins gradients. Mass spectrometer was operated in positive ion and data-dependent acquisition mode. The acquisition of all full MS was in the scan range 300–1750 m/z at a resolution of 60k. For proteome measurements, the 20 most intense peptides were picked for HCD fragmentation at 15k resolution and for phosphoproteome at 30k resolution. The normalized collision energy was set to 28% and dynamically excluded the mass of sequenced precursors for 30 secs from repeated fragmentation. The ions with single, unassigned or charge higher than six were also excluded from selection for fragmentation.

12. MS data processing with MaxQuant

The acquired raw files were processed using the MaxQuant software (version 2.2.0.0) [59]. Raw files from each set of experiments were processed separately in a similar manner (S3 Table). In total, we used 130 files of which 110 were files from phosphopeptide enrichment fractions. The obtained peak list was searched using Andromeda search engine integrated in MaxQuant [60] against E. coli K-12 MG1655 proteome (Taxonomy ID 83333) (released 30.01.2024, 4416 entries), and Klebsiella pneumoniae subsp. pneumoniae MGH78578/ ATCC700721 (Taxonomy ID 272620) (released 10.11.2022, 5127 entries), and common potential contaminants list. All search parameters were kept to default except the ones mentioned here. Labeling was set to three multiplicity, with Light: DimethylLys0 and DimethylNter0, Medium: DimethylLys4 and DimethylNter4, and Heavy: DimethylLys8 and DimethylNter8. Phospho (STY) was added as a variable modification for phosphopeptide-enriched files. Proteome and phosphoproteome files were grouped separately to only look for Phospho (STY) modification in phospho-files. pHis and pAsp were not included as variable modifications, due to low pH conditions used during phosphoenrichment. For Lys-C digestion, Lys-C enzyme was selected with a maximum of two missed cleavages allowed, similarly for trypsin with two missed cleavages allowed, and for chymotrypsin five missed cleavages allowed. To increase the number of quantified features, “match between runs” was enabled. The “Re-quantify” option was also enabled to allow the quantification of dimethyl-labeled pairs. Different experiments were processed separately for individual analysis and also together for the final set of phosphoproteome data.

13. Data analysis with Perseus

For the statistical analysis of MaxQuant output data, we used Perseus software (version 1.6.5.0.) [61] and the figures were edited using Adobe Illustrator. All contaminants, reverse hits, and diagnostic peaks were filtered out from the Phospho (STY) table. Average median values of phosphorylation site and proteome ratios were log₂ transformed and plotted against log₁₀ transformed sum of the intensities of phosphopeptides or proteins. Phosphorylation data was also normalized to proteome by dividing phospho ratios to proteome and plotted against log₁₀ of the sum of intensity. Significantly regulated sites were determined by applying a threshold of 2 on a log₂ scale (four-fold). The correlation between the two replicates was plotted with the density estimation feature in Perseus and plotted log₂ ratio of the two replicates and calculated the value of Pearson’s correlation coefficient. Likewise, for the Protein groups files, contaminants, reverse hits and only identified by sites proteins were filtered. After the log₂ transformation of ratios, density estimation was performed. Scatter plots were prepared for the reproducibility between the replicates and Pearson’s correlation was calculated. Further statistical analysis was performed for all proteome data in Perseus. Student’s one-sample t-test was performed to prepare volcano plots showing differential expression of protein and phosphorylation sites based on the T-test difference of log₂ ratio on the x-axis and the negative log₁₀ of the Benjamini-Hochberg corrected (FDR < 0.05) p-value on the y-axis. Functional enrichment analysis of the final phosphoproteome dataset was performed using the online tool ShinyGO 0.77 (http://bioinformatics.sdstate.edu/go/) [62]. All Protein IDs from the combined Phospho (STY) table (Tab I in S1 Dataset) were added to the list and the species was selected as “Klebsiella pneumoniae”. Using default values with an FDR cut-off of 0.05, we performed the enrichment analysis and obtained a table containing results of all enriched pathways, 149 pathways in total (Tab K in S1 Dataset). The data was further filtered using MS Excel for the number of genes per pathway ≥4, fold enrichment ≥4 and enrichment FDR ≤0.01. This resulted in 23 pathways that were used for plotting the highly enriched phosphorylated proteins in our dataset (Fig 5D).