Three cotrimoxazole-resistant (CTX), three amikacin-resistant (AMI), three both-resistant (ACB), and three drug-sensitive (SEN) KPN isolates were collected by the microbiology lab at Shanghai Fifth People’s Hospital for this study. The KPN type strains (33259, 13883, and 11296) were purchased from American Type Culture Collection (ATCC, United States). Minimum inhibitory concentrations (MICs) were interpreted according to the Performance Standards for Antimicrobial Susceptibility Testing M100 edition 28 (2018) of the Clinical and Laboratory Standards Institute (CLSI) (Gehring et al., 2021).

For each KPN strain, three KPN monoclonals were inoculated in Luria-Bertani (LB) broth to the exponential phase (OD₆₀₀ = 0.8), and KPN cells were collected and washed in PBS and subjected to global protein extraction using 8 M Urea (PH = 8.0) containing protease inhibitor, followed by 3 min of sonication (3 s on, 3 s off, amplitude 25%). Then the protein concentration was quantified through Bradford method and 100 μg protein was digested overnight following filter-acid sample preparation (FASP) method (Wisniewski et al., 2009) with 3.5 μg trypsin in 50 mM ammonium acid carbonate (PH 8.0) overnight at 37°C. Finally, the purified peptides were acquired after extraction with

50% acetonitrile (ACN) and 0.1% formic acid (FA) following desalination in two layers of Empore 3 M C18 disk with 2 mg packing (3 μm, 150 Å, Agela) in a pipet tip and dried in a vacuum concentrator (Thermo Fisher Scientific).

Proteome analysis was processed on Q Exactive HFX mass spectrometer (Thermo Fisher Scientific, Rockford, IL, United States) coupled with Easy-nLC 1,000 nanoflow liquid chromatography system (Thermo Fisher Scientific). The MS raw files were searched against Klebsiella pneumoniae subsp. Pneumoniae (strain ATCC 700721) database (version 20171126) in the Uniprot Knowledgebase (UniProtKB) using Maxquant (Version 1.5.3.30) (Cox and Mann, 2008). Peptides (minimum length of seven amino acid residues) with 1% FDR and a Mascot ion score greater than 20 were selected for protein identification. Proteins with 1% FDR (with at least one unique peptide) were selected for further analysis. For the proteome quantification, the intensity-based absolute quantification (iBAQ) value was extracted from MaxQuant results and subjected to FOT calculation. FOT was defined as protein’s iBAQ divided by the total iBAQ of all identified proteins in one experiment. Finally, FOT was multiplied by 10⁶ for easy presentation. The geometric mean value of the copy numbers in three proteomic analyses of each macrophage population were calculated and used for protein quantification. The significantly different expressed proteins (DEPs) were filtered with the fold change > 2 and P-value < 0.05 (bilateral Student’s t-test).

Unsupervised hierarchical clustering was carried out using R package “pheatmap” (version 1.0.12). The distances between the rows or columns of a data matrix were computed based on the Euclidean distance. The “complete” method was used in agglomeration process. PCA was performed using R package “gmodels” (version 2.18.1) in the statistical analysis environment R version 4.0.0 based on the relative protein quantification values (FOT) of each sample.

Gene Ontology (GO) enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses of the DEPs between sensitive and drug (AMI or CTX)-resistant isolates was performed with the R Bioconductor package “clusterProfiler” (version 4.0.2). Enrichment significance was determined using Fisher’s exact test.

To search the global protein patterns of different KPN clinical isolates, a total of 15 KPN strains, including three sensitive (SEN), three amikacin-resistant (AMI), three cotrimoxazole-resistant (CTX), and three amikacin/cotrimoxazole both-drug-resistant (ACB) KPN isolates derived from 12 different patients (Table 1) in our hospital, as well as three ATCC type strains, were collected and subjected to LC-MS/MS in single runs by a quadrupole Orbitarp instrument after trypsin digestion for label-free proteomics analysis (Figure 1A). As a result, a deep coverage of 2,266 proteins were successfully identified and quantified in total (Figure 1B and Supplementary Table 1), expanding the experimental coverage of the 5,126 predicted bacterial gene products (in KPN database from UniprotKB, strain ATCC 700721) from 23% (1,156) (Sharma et al., 2019), 32% (1,654) (Keasey et al., 2019) to 44%, providing an extended characterization of KPN proteome. The dynamic ranges of protein quantification values in this study spanned over six orders of magnitude (Supplementary Figure 1A), with the most abundant proteins being rpsU, HupA, tufa, etc. (Supplementary Figure 1B). rpsU, also known as ribosomal protein S21 (rps21), was reported to contribute to fitness, stress-tolerance and host interaction (Koomen et al., 2018). HupA (heat-unstable α, HUα) is a subunit of a heterotypic dimer, Huαβ, resulting in high growth and a lowly pleiotropic phenotype (Abebe et al., 2017). tufA (translational elongation factor EF-Tu) was involved in a change in state or activity of a cell as a result of an antibiotic stimulus (Silk and Wu, 1993). The proteome quantification results of 15 experiments were comparable, as their medians were on the same level (Supplementary Figure 1C).

To obtain the proteome identity of different strains systematically, PCA was carried out based on the protein abundance of KPNs. The result showed a very clear separation between ATCC and hospital isolates on the first component level as shown in x-axis, and captured the significant differences between sensitive and drug-resistant strains on the second component level as shown in y-axis (Figure 1C). An unsupervised hierarchical clustering analysis of the proteome patterns further supported the results (Figure 1D). Therefore, considering the fundamental differences between standard KPN strains from ATCC and clinical isolates from our hospital, we mainly focused on comparative analysis between proteome of three drug-resistant isolates (CTX, AMI, and ACB) and sensitive isolates for further study.

Cotrimoxazole is a combination of trimethoprim/sulfamethoxazole, and widely used for treatment of bacterial infections. The drug is extensively used due to the relatively cheap, available over-the-counter, and well tolerated (Morgan et al., 2011). Because of the injudicious use, the spread of bacteria with this antibiotic resistance has been a major factor (Okeke et al., 1999). Recent studies from the African continent have reported high rates of cotrimoxazole resistance in gram-negative pathogenic bacteria in the range of 50–96% (Manyahi et al., 2017).

In our comparative proteome analysis between CTX-resistant isolates and SEN isolates, we identified 183 up-regulated proteins and 501 down-regulated proteins, respectively (Figure 2A and Supplementary Figure 2A). Further gene enrichment analysis showed an activation of alcohol metabolic process (eutT, dhaD, dhal) as well as a restraint of lipid catabolic process (fadl, fadA, fadB) and tricarboxylic acid (TCA) cycle pathway (ynel, mdh, acnA) in CTX-resistant isolates, which may contribute to ROS reduction and cell survival under drug pressure. It’s also worth noting that although the folate biosynthesis of cells (folA, folX, morA) was attenuated by CTX administration, the usage rate of one carbon unit was not affected accordingly illustrated by activation of purine ribonucleotide biosynthetic process and “one carbon pool by folate pathway (apaH, purD, cpdA). Besides, proteins in polysaccharide biosynthetic and cell wall organization processes (rcsF, wbaP, wecF), which may protect cell from drug perturbation through the physical safeguards” strategies, were up-regulated in CTX-resistant isolates (Figure 2B). There results showed the relative metabolism alteration of CTX-resistant isolates comparing with SEN isolates.

Amikacin shows a particularly broad antimicrobial activity which is used for severe bacterial infections. Nowadays, the emergency of amikacin-resistant KPN occurs dramatically with 20% in Turkey (Gokmen et al., 2016). Based on the proteome dataset, we identified 305 up-regulated proteins and 319 down-regulated proteins, respectively in AMI-resistant isolates comparing with SEN strains (Figure 3A and Supplementary Figure 2B). Gene enrichment analysis of these differentially expressed proteins indicated activation of glutathione metabolism pathway (yfcF, ybgK, pxpA) and restraint of TCA cycle pathway (frdB, sdhC, fumC), contributing to ROS clearance and cell survival under drug perturbation, in AMI-resistant isolates. Additionally, protein metabolic process as well as amino acid biosynthesis process was elevated in AMI-resistant isolates, indicating less affection of AMI administration in these isolates (Figure 3B).

Of note, combined with the results of CTX-resistant isolates, we found TCA cycle pathway was also inhibited in AMI-resistant isolates, and the downregulated proteins in TCA cycle pathway were high-degree overlapped, such as sdhC, fumC, mqo, pcka, and KPN_00489. These results reveal that restraint of TCA cycle pathway may be a widely used strategy of KPN to gain antibiotic resistance.

Combining with the proteome profiling of ACB-resistant isolates, we screened out a set of candidate proteins that were specifically associated with the mechanism of single antibiotic resistance of KPN. As a result, we identified 24 proteins whose expression were simultaneously up-regulated in CTX-resistant isolates and ACB-resistant isolates but not affected by AMI administration comparing with SEN strains, and 111 proteins on the contrary, as candidate proteins specifically for CTX-resistant isolates. Similarly, a total of 98 and 55 proteins exclusively up-regulated or down-regulated in AMI-resistant isolates, respectively, were identified as candidates specifically for AMI-resistant isolates (Figure 4A).

Next, to further identify the proteins closely correlated with mechanism of antibiotic resistance in KPN, we introduced the MICs of cotrimoxazole and amikacin on all the KPN hospital-isolates studied (Table 1). The Figures 4B,C showed the distribution of Pearson correlation coefficients between MICs and expression of protein sets above, respectively, for CTX-resistant isolates and AMI-resistant isolates. Finally, based on the results of gene enrichment analysis and correlation analysis, we identified nine proteins (Table 2), which were not only involved in the metabolic regulation of drug-resistant isolates, but also significantly correlated (positively or negatively) with the degree of drug resistance of KPN isolates (Figures 4D,E). Therefore, these enzymes possess important roles in reducing the ROS levels in cytoplasmic of cells. Further investigations need to be processed to determine the precise biological mechanism of the influence of these proteins on isolates resistance.