Synergistic effects of polymyxin and vancomycin combinations on carbapenem- and polymyxin-resistant Klebsiella pneumoniae and their molecular characteristics

ABSTRACT The emergence and spread of polymyxin resistance, especially among Klebsiella pneumoniae isolates, threaten the effective management of infections. This study profiled for polymyxin resistance mechanisms and investigated the activity of polymyxins plus vancomycin against carbapenem- and polymyxin-resistant K. pneumoniae. The entire genome sequences of seven isolates were profiled for resistance and virulence determinants. The effects of combination therapy were evaluated using the checkerboard technique, time-kill assay, and population profile analysis. Protein profiles of the isolates treated with monotherapy were compared to that of combination therapy. The whole-genome sequencing data revealed that the isolates harbored β-lactams, carbapenems, aminoglycoside, fluoroquinolones, macrolides, and tetracycline resistance genes, with several virulence-associated genes. The plasmids including the bla OXA-232-bearing ColKP3 plasmid were also identified in our isolates. Profiling for polymyxin resistance mechanism revealed a missense mutation in the crrB gene that resulted in a Q180L variant that conferred a deleterious effect on protein function. The combination assay indicated fractional inhibitory concentration index ranging from 0.31 to 1.13, whereas the time-kill assay demonstrated synergistic log reduction in colony-forming units per milliliter. Furthermore, population analysis profiling using dual antibiotics indicated enhancement in bacterial log reduction at lower antibiotics concentrations, compared to higher concentrations of single polymyxins. For protein profiling, 796 proteins were identified, and 56 and 94 of them were increased and decreased in the combined drug treatment groups, respectively, while other differentially produced proteins were detected in all treatment groups, except for the control group. The results demonstrated that the vancomycin combination might benefit the antimicrobial activities of polymyxins. IMPORTANCE This study provides insights into the mechanisms of polymyxin resistance in K. pneumoniae clinical isolates and demonstrates potential strategies of polymyxin and vancomycin combinations for combating this resistance. We also identified possible mechanisms that might be associated with the treatment of these combinations against carbapenem- and polymyxin-resistant K. pneumoniae clinical isolates. The findings have significant implications for the development of alternative therapies and the effective management of infections caused by these pathogens.

medical practitioners are without strict restrictions on antibiotic choice and at liberty to administer antibiotics deemed effective against MDR pathogens.Thus, permuta tions of antibiotic combinations as last-resort options are on the increase, and several in vitro and in vivo studies, as well as clinical trials, have demonstrated that antibiotic combination therapies ensure positive outcomes compared to monotherapies.
Although once discontinued for reported cases of toxicity, polymyxins, a group of polypeptide antibacterial compounds, are currently choice options for the management of infections caused by MDR Gram-negative bacteria, especially carbapenem-resistant pathogens (1)(2)(3).Among the polymyxin class, polymyxins B and E (colistin) are clini cally relevant and share structural similarities, except for an amino acid in the L-Dab peptide ring (4).Both also share a similar antimicrobial spectrum, with activity against most Gram-negative bacteria.Nephrotoxicity and neurotoxicity are common concerns associated with polymyxin B and colistin.However, studies have reported lower rates of nephrotoxicity with polymyxin B compared to colistin (5,6).
The recent emergence of polymyxin resistance among Gram-negative pathogens (7,8) constitutes a public health emergency and is proposed to mark an end to the antibiotic era (9).Colistin resistance in Gram-negative bacteria is mostly due to acquired mutations in the two-component systems including PhoPQ, CrrAB, and PmrAB (10)(11)(12).In addition, the emergence of the mcr-colistin resistance gene mobilized by transmissible plasmids (13)(14)(15)(16) promotes the spread and prevalence of polymyxin resistance.The use of multiple antimicrobial agents in combination therapies has been adopted as an interim viable solution to the management of MDR infections.Studies have demonstra ted that polymyxins in combination with other antibiotic classes act in synergy against MDR pathogens (7,17,18).Various studies have investigated the use of antimicrobial agents ineffective against Gram negative as adjunctive agents to enhance the inactiva tion of MDR Gram negatives or as resistance modifying agents to attenuate bacterial mechanisms of resistance.
Vancomycin, a bactericidal glycopeptide antibiotic indicated for Gram-positive bacteria, inhibits cell wall biosynthesis by preventing the incorporation of N-acetylmur amic acid-and N-acetylglucosamine-peptide subunits into the peptidoglycan matrix.In addition, vancomycin damages protoplasts by affecting the cytoplasmic membrane through the inhibition of RNA synthesis (19).Although vancomycin is not indicated for the treatment of Gram-negative bacteria, drug repurposing is proposed as a fast and inexpensive strategy for the management of MDR infections.Studies have evinced the in vitro synergistic effects of vancomycin as adjunctive therapy in combination with polymyxins for the management of MDR isolates of Acinetobacter baumannii and other clinically relevant bacterial strains (20)(21)(22).We then hypothesized that the combination of vancomycin and polymyxin (B and E) might be used to combat carbapenem-and polymyxin-resistant K. pneumoniae clinical isolates.Thus, this study aims to investi gate the antibacterial effects of vancomycin plus polymyxin combinations on these pathogens.Resistant mechanisms and possible mechanisms that occurred from these combinations were also studied by whole-genome sequencing and protein profiling, respectively.

Antibacterial activity of carbapenems, polymyxin, and vancomycin
Minimum inhibitory concentrations (MICs) of vancomycin after 2-h resazurin exposure ranged from 512 to 1024 µg/mL.With prolonged incubation of the cells, color changes were subsequently observed after 24 h in higher antibiotics concentrations of 2,048 and 4,096 µg/mL, suggesting bacteria re-growth.The MICs of imipenem and meropenem on the isolates were >64 and >128 µg/mL, respectively.The antibacterial effects of antibiotics on the isolates are presented in Table 1.According to Clinical and Labora tory Standards Institute (CLSI), the antimicrobial breakpoint average of Enterobacterales describes resistance to imipenem, meropenem, polymyxin B, and colistin as MIC values ≥4 µg/mL (23).

Detection of antimicrobial resistance genes, virulence genes, and plasmid types
The carbapenem-and polymyxin-resistant K. pneumoniae isolates were all found to belong to the sequence type 16 (ST16) with average nucleotide identity values of ≥99.99% (Table S1).The isolates harbored multiple antimicrobial resistance and virulence genes (Fig. 1), with a similar pattern of antimicrobial resistance (AMR) genes including the carbapenemase genes bla NDM-1 , bla OXA-9 , bla OXA-10 , and bla OXA-232 (except isolate NT10 which lacked the bla OXA-232 gene) and other beta-lactamase genes (bla CTX-M-15 , bla SHV-1 , bla TEM-1A , and bla VEB-1 ).In addition, genes mediating resist ance to aminoglycoside, fluoroquinolones, rifampicin, chloramphenicol, trimethoprim, macrolides, sulfisoxazole, and tetracycline were found in all isolates.Virulence genes harbored in the isolates included fimbriae and capsular genes, efflux pump genes AcrAB, genes involved in the regulation of capsular polysaccharide biosynthesis RcsAB, and genes of the type 6 secretion system involved in bacterial competitive advantage and associated with pathogenesis and biofilm formation.Plasmid profiling showed the presence of similar plasmids including the Col440I, Col440II, ColKP3, IncA/C2, IncFIA, and IncFIB(pQil).However, the plasmid ColKP3 was not found in isolate NT10, whereas the plasmid Col440II was not found in isolate NT23.The bla OXA-232 gene and ColKP3 plasmid were identified in the same contig; thus, ColKP3 plasmid might be harboring the bla OXA-232 gene.

Molecular characteristics of polymyxin resistance
Polymyxin resistance-related genes were compared to the complete sequence of a polymyxin-susceptible K. pneumoniae MGH 78578 strain (NC_009648.1)(24).The results demonstrated the presence of mutations resulting in amino acid substitutions in multiple genes.Table 2 presents nucleotide changes that resulted in an amino acid change (missense and nonsense mutations).However, nucleotide changes without amino acid change (silent mutation) were not reported in this study.The effects of protein variants were predicted by the PROVEAN score.Most of the amino acid substitutions with PROVEAN score greater than −2.5 indicated missense and nonsense mutations, which did not affect the functionality of the protein and were interpreted as neutral.Two substitutions were observed in the crrB of the CrrAB two-component regulatory system which normally modulates the PmrAB system and is involved in modifications of lipopolysaccharide.The substitution of glutamine with leucine at position 180 (Q180L) indicated a PROVEAN score of −6.578 and was interpreted as a deleterious effect, whereas the substitution of cysteine with serine at position 68 (C68S) yielded a PROVEAN score of 2.234 and was interpreted as a neutral effect.

Antibacterial effect of combinations of polymyxin and vancomycin
The fractional inhibitory concentration index (FICI) (Table 1) presents the effects of combinations of polymyxin (B and E) with vancomycin.The results revealed FICI ranges of 0.38-1.13for vancomycin with polymyxin B combinations and 0.31-1.13for vancomy cin with colistin combinations.Furthermore, the isobologram of fractional inhibitory concentrations (FICs) (Fig. 2) showed superadditive effects indicated by the downward concavity of the isoboles.Nonlinear isobolograms are used to indicate whether the effect produced by the combination of two drugs is synergistic (superadditive) or antagonistic (subadditive).From the isobole model, FICI = 1 represents an additive effect, FICI <1 represents a synergistic effect, whereas FICI >1 represents an antagonistic effect.The isobologram of combinations of vancomycin with polymyxin B and colistin showed that vancomycin in combination with either colistin or polymyxin B induced a synergistic effect against the isolates.

Population analysis profiling of carbapenem-and polymyxin-resistant K. pneumoniae
The population analysis profiling of the isolates showed resistance to both colistin and polymyxin B, with the growth of subpopulations at 2,048 µg/mL (Fig. 3).Isolates NT03 and NT09 indicated heteroresistance to colistin and polymyxin B, with subpopulations with MIC of 256 µg/mL, while isolates 07 showed heteroresistance to polymyxin B, with subpopulations with MIC of 8 µg/mL.The population analysis profiling of dual antibiotics (colistin and vancomycin or polymyxin B and vancomycin) (Fig. 4) showed that at concentrations <2,048 µg/mL, combinations at sub-inhibitory concentrations demonstrated enhanced antibacterial activities with higher reduction in the growth.a The variants of nucleotide and protein substitutions were equally found in all the isolates.b If the PROVEAN score is less than or equal to −2.5, the protein variant is predicted to have a "deleterious effect." If the PROVEAN score is greater than −2.5, the protein variant is predicted to have a "neutral effect." c We only reported nucleotide changes that resulted in the amino acid change (missense and nonsense mutations), while the nucleotide changes that did not lead to the amino acid change (silent mutation) were not reported in this study.d ND, not detected.

Time-kill kinetics of vancomycin combination with colistin or polymyxin B
The time-kill kinetics of polymyxins in combination with vancomycin is presented in Fig. 5.The results demonstrated enhanced synergistic activity for polymyxin B and vanco mycin combinations compared to colistin and vancomycin combinations.At 1/2MIC polymyxin B and 1/2MIC vancomycin, a >3 log reduction in colony-forming units per milliliter (CFU/mL) was recorded for isolate NT03 at 15 and 18 h, compared to cell count at 0 h.At 1/4MIC polymyxin B and 1/2MIC vancomycin, >2 log reduction in CFU/mL was observed.Similarly, isolate NT07 at 1/2MIC polymyxin B and 1/2MIC vancomycin displayed >2 log reduction, and at 1/4 MIC polymyxin B and 1/2 MIC vancomycin, showed >1 log reduction.At 12 h, a >3 log reduction in CFU/mL was observed for NT09; however, at 18 h, a regrowth was observed.At the tested concentrations, the combina tions did not show effects for isolate NT23.Also, for isolates NT03, NT07, and NT23, combinations of colistin with vancomycin at the tested concentration did not provoke enhanced activities.However, for isolates NT09, the combinations demonstrated >1 log reduction in CFU/mL at 8 h and with subsequent regrowth after 12 h.The overall result demonstrated that combinations of polymyxins with vancomycin were effective against the isolates.

Protein profiling analysis related to polymyxin and vancomycin activities
Isolate NT09 was used as a representative for this experiment because it was the only isolate that seems to show a response to colistin and polymyxin from the time-kill result.Protein profiling results of treatment and non-treatment groups are illustrated in Fig. 6 to 10 and Tables S2 to S5.In proteomes of six groups, the principal component analy sis showed obviously separated groups indicating that proteomes were significantly changed among the groups.Venn diagram in Fig. 7A with the data in Table S3 revealed that 39 (7.69%), 102 (20.12%), 81 (15.98%), and 54 (10.65%) out of 507 proteins were only observed in the groups of control, colistin, vancomycin, and colistin plus vancomycin, respectively.A total of 33 (6.51%) proteins were shared among these treatment groups, whereas they were not found in the control group.Thirty-two (6.31%) proteins were detected in both groups of single-drug treatments (colistin and vancomycin), which were not presented in a group of combined drug treatment.Meanwhile, a Venn diagram in Fig. 7B with the data in Table S4 showed that 37 (7.87%), 51 (10.85%), 113 (24.04%), and 62 (13.19%) out of 470 proteins were only observed in the groups of control, polymyxin B, vancomycin, and polymyxin B plus vancomycin, respectively.A total of 31 (6.60%) proteins were shared among these treatment groups, whereas they were not found in the control group.Twenty-five (5.32%) proteins were detected in both two groups of single-drug treatment (polymyxin B and vancomycin), which were not presented in a group of combined drug treatment.
In addition, volcano plots showing differentially produced proteins (DPPs) in all treatment groups (n = 5) and control groups are presented in Fig. 8 and 9; Table S5.Among the proportion of statistical significance of DPPs, four different proteins showed increased abundance in the colistin, polymyxin B, vancomycin, and polymyxin B plus vancomycin groups, while three proteins showed increased abundance in the colistin plus vancomycin treatment group, respectively.Similarly, eight proteins showed decreased abundance in the polymyxin B and the colistin plus vancomycin groups, and seven proteins, for the polymyxin B plus vancomycin, while no proteins showed decreased abundance in the colistin and the vancomycin treatment groups, respectively, compared to the control group.Remarkably, the MdtJ protein was significantly increased and the DsbD protein was significantly decreased in only groups of combined drug treatments.
Among the 796 proteins found among five treatment groups and the control group, 34 (4.27%) proteins involved in polymyxins and vancomycin activities were selected, and their fold changes are shown in Fig. 10.Hierarchical clustering of these 24 proteins revealed two groups (n = 3 and n = 31).The PhoR and WaaE proteins were increased in the colistin plus vancomycin treatment group, while the LplT, KdsB, and ArnB proteins were increased in the polymyxin B plus vancomycin treatment group, compared to the control group.Additionally, increase of the ArnT protein in polymyxin B and the PagP, Aas, and KdsA proteins in vancomycin was observed.

DISCUSSION
The resistance of K. pneumoniae to polymyxins limits treatment options for infections caused by carbapenem resistance isolates and has become a global health emer gency.The whole-genome sequencing showed that all the polymyxin-resistant isolates belonged to ST16 and shared similar genomic composition (≥99.99% identity) (36).This might be because all the isolates were sourced from an intensive care unit of the same hospital and might point to the local spread of isolates from the same clone within the facility.Antimicrobial resistance gene profiling revealed several resistant mediating genes responsible for multidrug resistance.A similar occurrence of multiple AMR genes especially of extended-spectrum beta-lactamases (ESBL) and carbapenemase genes in K. pneumoniae isolates has been rampantly reported (37,38).Due to the high prevalence and the intraspecies spread of AMR genes in K. pneumoniae, it has been described as a major worldwide source and shuttle for antibiotic resistance and a key trafficker of drug resistance genes from environmental to clinically important bacteria (39,40).The isolates carried genes encoding for carbapenems, aminoglycoside, fluoroquinolones, rifampicin, chloramphenicol, trimethoprim, macrolides, sulfisoxazole, and tetracycline resistance, in addition to beta-lactamase genes including bla NDM-1 , bla OXA-10 , bla OXA-232 , and bla OXA-9 .The isolates also harbored fimbriae and capsular genes, efflux pump genes, capsular polysaccharide biosynthesis genes RcsAB, and genes of the type 6 secretion system coding for virulent factors.Furthermore, plasmid types were identified in the isolates, especially the ColKP3 plasmid with the carriage of the class D carbapenemase gene, bla OXA-232 .Studies have previously reported that the bla OXA-232 gene is often harbored in the ColKP3 plasmid (41)(42)(43).Genomic profiling for polymyxin-resistant mediators identified the occurrence of multiple mutations resulting in amino acid substitutions.Amino acid substitutions in genes regulating lipid A biosynthesis, lipopolysaccharide biosynthesis, the two-component regulatory systems PhoQP, PmrAB, CrrAB, and the MgrB regulator are associated with polymyxin resistance.Although numerous mutations were present in the isolate, most of the mutations did not confer a deleterious effect on the gene function.However, an amino acid substitution in the CrrAB system resulted in a mutation in the crrB gene with a deleterious effect on the protein function.Deleterious effects resulting from mutations in the crrB gene have been reported as responsible for polymyxin resistance (29,(44)(45)(46).In addition, the presence of efflux pump genes including the sapABCDF operon associated with resistance to antimicrobial peptides (47), the kpnEF genes in a small multidrug resistance (SMR)-type efflux pump (31), and the acrB gene (25-27), which upon activation results in the increase of polymyxin resistance, might likewise contribute to the resistance of the isolates.
The use of combination therapies is an interim strategy for the effective management of polymyxin-resistant isolates.Studies have evinced the promising synergistic effects of polymyxins in combination with antibiotics of other classes.We investigated the effects of polymyxin B and polymyxin E (colistin) combinations with vancomycin, as a drug repurposing strategy for the management of infections caused by Gram-negative bacteria.The combinations demonstrated better in vitro effects when compared to individual antibiotic monotherapy.Previous studies have demonstrated that adjunctive vancomycin therapies potentiate enhanced antimicrobial effects of polymyxins on MDR A. baumannii (20,21,(48)(49)(50).However, the effects of these combinations on MDR K. pneumoniae isolates have not been elaborated.Although vancomycin is inactive against Gram-negative pathogens due to its large molecular size and inability to penetrate the outer bacterial membrane (51), compromise of bacterial outer membrane integrity by polymyxins might promote the penetration of vancomycin molecules through the outer membrane, resulting in synergistic bacteriostatic or bactericidal effects.Moreover, modification of vancomycin through conjugation with arginine conferred antimicrobial potentials against carbapenem-resistant E. coli, by targeting cell wall synthesis (52).The checkerboard assays indicated that the combination of polymyxins and vancomycin was not antagonistic but varied over a range of FICI, with synergistic, additive, or indifferent outcomes.Isobologram of FICs yielded a characteristic superadditive curve, with isobars below the additive line.Furthermore, the time-kill assay showed that the combinations exhibited higher bactericidal effects compared to individual monothera pies as demonstrated by the log reduction in CFU/mL.However, the combination of polymyxin B and vancomycin displayed better activities than combinations of colistin and vancomycin.It has been previously reported that polymyxin B exhibited better antimicrobial effects compared to colistin (53).Based on the pharmacokinetics of the drugs, polymyxin B achieves rapid therapeutic concentrations, provides more predict able serum concentrations, does not require renal dose adjustments, and might not necessitate a loading dose.In contrast, colistin is required to prolong time to achieve peak serum concentration, with unpredictable serum concentrations due to significant interpatient variability; renal dose adjustments must be considered; and a loading dose is required (54).Additionally, a recent systematic review and a prospective study on the nephrotoxicity of colistin and polymyxin B concluded that colistin is associated with significantly higher nephrotoxicity compared to polymyxin B (55,56).
Recently, heteroresistance which is defined as the occurrence of subpopulations with elevated antimicrobial tolerance within a population has become a critical issue in antimicrobial susceptibility testing and is often associated with false-positive or false-negative antimicrobial susceptibility results.Thus, we further performed a population analysis profiling of the isolates on individual polymyxin antibiotic and dual antibiotics (colistin + vancomycin OR polymyxin B + vancomycin).The results revealed that the isolates were entirely resistant to both polymyxins, with no susceptible subpopulation.Enhancement in bacterial log reduction (%) to dual antibiotics at reduced concentration was observed compared to single polymyxins at higher concentrations.Furthermore, the results indicated that the polymyxin B and vancomycin combinations were superior to colistin and vancomycin combinations.Previous studies have dem onstrated heteroresistant through population analysis profiling of resistant isolates to antibiotic combinations with additive outcomes (2,57).Most studies attribute hetero resistance to the co-occurrence of susceptible and resistant subpopulations within a population.However, heteroresistance includes the co-occurrence of subpopulations with various levels of resistance within a population, as well as the co-occurrence of subpopulations with various levels of susceptibility within a population (58).This study indicates that novel mutation in the CrrB protein and the presence of efflux pump genes might be responsible for high-level polymyxin resistance and highlights that repurposing of vancomycin might benefit the antimicrobial activities of polymyxins against MDR K. pneumoniae.
In the proteomic analysis, where the expected number of proteins was approximately 5,300, the actual liquid chromatography-tandem mass spectrometry (LC-MS/MS) results in our study showed only 796 proteins.Factors contributing to this phenomenon may include sample complexity, limited LC-MS/MS range, instrument sensitivity, post-trans lational modifications, and data analysis criteria (housekeeping protein exclusion).Furthermore, our findings revealed that the protein levels in the control group were lower compared to the treatment groups.It is possible that not all Klebsiella cells were eradicated within 6 h of antibiotic exposure.As a result, we postulate that the surviving cells could have been highly active in defending against the antibiotics, leading to higher protein levels observed in the treatment groups.Protein profiling also revealed possible mechanisms of action of polymyxin and vancomycin combination, compared to each single-drug treatment and non-treatment.The functions of all the studied proteins were retrieved from the UniProt database (https://www.uniprot.org/),as illustrated in Table S2.A total of three to four increased protein abundance and zero to seven decreased protein abundance was found among the treatment groups, compared to the control group.The DPPs might be associated with polymyxin and vancomycin activities, since they were only detected in the treatment groups.In the comparison between singledrug treatments and combined drug treatments, the increased MdtJ protein and the decreased DsbD protein were only observed in polymyxin plus vancomycin combination.MdtJ is a protein belonging to the SMR family efflux pumps, initially associated with resistance against deoxycholate and sodium dodecyl sulfate (SDS) in Escherichia coli (59).We then hypothesize that the increased MdtJ production in our Klebsiella pneu moniae isolates can enhance antibiotic efflux, reducing susceptibility and promoting multidrug resistance, cross-resistance, persistent infections, and horizontal gene transfer of resistance genes (60).DsbD (thiol:disulfide interchange protein DsbD) generally transports electrons from cytoplasmic thioredoxin to the periplasmic oxidized substrates, which are involved in the formation and reshuffling of disulfide bonds in proteins.Decrease of the DsbD protein may have consequences for bacterial cell physiology, including the proper folding of proteins (especially outer membrane porin and efflux pumps), susceptibility to oxidative stress, and virulence (61).Therefore, decreased DsbD protein probably compromises the ability of bacteria to resist antibiotics, leading to increased sensitivity to polymyxin plus vancomycin combination.
The heatmap with selected protein features (Fig. 10; Table S2) revealed some increased protein abundance, which were specifically observed in combined drug treatments.In the colistin plus vancomycin treatment, increased PhoR and WaaE proteins were found.PhoR (phosphate regulon sensor protein PhoR) is a member of the twocomponent regulatory system PhoR/PhoB involved in the phosphate regulon genes expression, enhancing bacterial survival in low-phosphate environments.Meanwhile, WaaE [lipopolysaccharide (LPS) core biosynthesis glycosyltransferase WaaE] is involved in the biosynthesis of the LPS core, which is a component of the outer membrane of Gram-negative bacteria.We hypothesize that increase of the WaaE protein may lead to alterations in the LPS structure, reducing the binding of colistin to the bacte rial cell surface and decreasing its effectiveness.In the polymyxin B plus vancomycin treatment, the LplT, KdsB, and ArnB proteins were increased.LplT [lysophospholipid (LPL) transporter LplT] generally catalyzes the facilitated diffusion of 2-acyl-glycero-3phosphoethanolamine (2-acyl-GPE) into the cell, leading to changes in the compo sition of the bacterial cell membrane.Therefore, increase of the LplT protein may contribute to bacterial resistance to antimicrobial agents that target the cell mem brane.KdsB (3-deoxy-manno-octulosonate cytidylyltransferase) generally catalyzes the transfer of a cytidine monophosphate group to 3-deoxy-D-manno-octulosonic acid, an important component of the LPS molecule.We then hypothesize that increase of the KdsB protein probably increases the production and the alteration of LPS, leading to antibiotic resistance in bacteria.ArnB (UDP-4-amino-4-deoxy-L-arabinose-oxoglutarate aminotransferase) is involved in the biosynthesis of LPS by attaching the modified arabinose to lipid A. Thus, we hypothesize that increase of ArnB may result in resistance to polymyxin B and cationic antimicrobial peptides.

Combination assay
The effects of vancomycin combination with either colistin or polymyxin were investiga ted using the checkerboard technique as previously modified (18).In brief, antibiotic dilutions containing 50 µL of serially diluted colistin or polymyxin B and 50 µL of serially diluted vancomycin were prepared in a 96-well plate.A 100 µL of 1 × 10 6 CFU/mL bacterial suspension was added to each well, and the plates were incubated for 18 h at 37°C.Inhibitory concentrations were determined as concentrations without visible color changes as indicated by the resazurin test.The antibacterial effects of single antibiotics were tested as a control.The experiment was performed for three independent repeats.The effects of the antimicrobial combination were defined according to the FICI as shown in the following equation: The FICI results for each combination were interpreted as follows: FICI ≤0.5, syner gism; 0.5 < FICI < 1, additive; 1 ≤ FICI < 2, indifference; and FICI ≥2, antagonism.

Population analysis profile of bacterial isolates
Population analysis profiling was used to investigate for colistin and polymyxin B heteroresistance (2), with slight modifications.Briefly, 100 mL of overnight bacterial culture in TSB was adjusted to 0.5 McFarland standard (10 8 CFU/mL) in phosphate buffer solution.The cultures were serially diluted from 10 8 to 10 1 , and dilutions were plated onto Mueller-Hinton agar containing a twofold gradient of various concentra tions of colistin or polymyxin B from 2 to 2,048 µg/mL, respectively.The drop plate technique was employed in this study.After 24 h of incubation at 37°C, the subpopula tions that grew on the plates were enumerated.Heteroresistance was defined as the presence of a subpopulation of cells capable of growing at a concentration of antibiot ics at least twofold higher than that of the antibiotic-susceptible parental strains.The population analysis profile of the bacterial isolates was further conducted with antibiotic combinations of polymyxin and vancomycin.Single polymyxin concentrations of 256-1,024 µg/mL and combinations with subinhibitory concentrations at 128-512 µg/mL of vancomycin were prepared on Mueller-Hinton agar (MHA) plates.Exponential phase bacterial cultures incubated for 5-6 h in Mueller-Hinton broth (MHB) were adjusted to 10 8 CFU/mL and diluted further to 10 7 , 10 6 , 10 5 , 10 4 , 10 3 , 10 2 , and 10 1 .Serial dilutions of bacterial culture were drop plated at each concentration of antibiotic and incubated at 37°C for 48 h.Antibiotic combination treatment was compared with treatment with polymyxin alone, and the results were presented as log reductions in CFU/mL.

Time-kill assay
The antibiotic efficacy of colistin, polymyxin B, vancomycin, and combinations of either colistin or polymyxin B with vancomycin against the polymyxin-hyper-resistant isolates was evaluated by an in vitro time-kill assay.In brief, overnight cultures of each of the four selected isolates were diluted to 10 6 CFU/mL and exposed to the MICs of each antibiotic and combinations of sub-inhibitory concentrations in MHB.Viable cells were enumerated at 0, 2, 4, 8, 12, 15, and 18 h by spot plating and expressed in CFU/mL.The limit of detection was set at 100 CFU/mL.

Protein profiling
To assess the possible mechanisms of the monotherapy and combination therapy against carbapenem-and polymyxin-resistant K. pneumoniae isolates, protein profiling was investigated using LC-MS/MS.Briefly, a representative isolate was selected and treated with colistin, polymyxin B, vancomycin, colistin plus vancomycin, and polymyxin B plus vancomycin at 37°C and 150 rpm for 6 h, while a non-treatment group was used as a control.The cultures were centrifuged at 4°C and 10,000 rpm for 10 min, and the supernatants were discarded.The pellets were resuspended in 800 µL of 1× phosphate-buffered saline and 200 µL of 10% SDS.They were incubated at 37°C for 1 h and then sonicated on ice with the condition of 30%, 9 pulse, 95 Aml, and 15 min.Then, they were centrifuged at 4°C and 12,000 rpm for 15 min, and the supernatants were collected.The protein concentration was measured using the Bradford assay, and the concentrations were normalized.Normalized protein concentrations were then sent to the National Center for Genetic Engineering and Biotechnology for performing the LC-MS/MS.
Protein quantitation was assessed by Maxquant 1.6.6.0.The analyzed LC-MS/MS data from Maxquant 1.6.6.0 were searched against the Uniprot Klebsiella pneumoniae database using the Andromeda software for protein identification.For all comparison, the data sets of statistically significant proteins (P < 0.05) from the LC-MS/MS results were statistically analyzed by analysis of variance (ANOVA) with Fisher's Least Significant Difference (LSD) test.Then, all DPPs were analyzed for their intersections among the different sample groups using jvenn (http://jvenn.toulouse.inra.fr/app/example.html)(70).To identify the fold changes of DPPs with statistical significance, the volcano plots were generated using MetaboAnalyst 5.0 (https://www.metaboanalyst.ca/).Generating volcano plots in MetaboAnalyst 5.0 involves uploading and preprocessing metabolomics data, normalizing by medium, applying statistical tests to detect DPPs between groups, and plotting proteins on a graph based on statistical significance (P-value) and fold change (effect size) (71).In addition, the heatmaps of the proteins related to polymyxin and vancomycin activities were also constructed using MetaboAnalyst 5.0.

Statistical analysis
All experiments were performed in triplicate for two independent repeats.Results were presented as mean ± standard deviation.Statistical analysis was performed using GraphPad Prism v8.Comparisons between means were carried out using analysis of variance and interpreted based on Tukey multiple comparisons at P < 0.0001.For protein profiling, the statistics in the comparison between treatment and control groups were performed by ANOVA with Fisher's LSD test at P < 0.05.

FIG 4
FIG 4 Result of population analysis profiling of the isolates on combinations of colistin and vancomycin (A) and polymyxin B and vancomycin (B).

FIG 5
FIG 5 Time-kill graph of polymyxin and vancomycin combinations on carbapenem-and polymyxin-resistant Klebsiella pneumoniae isolates.

FIG 6 AFIG 7
FIG 6 A 2D scores plot from a principal component analysis of all detected proteins demonstrating the global trends in differences across treatments.CST, colistin; PMB, polymyxin B; VAN, vancomycin; PC, principal component.

FIG 9
FIG 9 Volcano plots of differentially produced proteins (DPPs) in colistin plus vancomycin (A) and polymyxin B plus vancomycin (B) treatments, compared to the control group.

FIG 10
FIG 10 Heatmap of DPPs that might be associated with the activities of colistin, polymyxin B, vancomycin, colistin plus vancomycin, and polymyxin B plus vancomycin treatments, compared to the control group.CST, colistin; PMB, polymyxin B; VAN, vancomycin.

TABLE 1
Minimum inhibitory concentrations and combination effects of carbapenem, polymyxins, and vancomycin a

TABLE 2
Genomic alterations in polymyxin resistance-related genes c

Gene Role Mutation found in isolates a % Identity Nucleotide substitution c Amino acid substitution Prediction of protein function PROVEAN score Interpretation b
lpxMCatalyzes the transfer of myristate from myristoyl-acyl carrier protein to Kdo2-(lauroyl)-lipid IV(A) to form Kdo2-lipid A.