Analysis of Regulatory Mechanism of AcrB and CpxR on Colistin Susceptibility Based on Transcriptome and Metabolome of Salmonella Typhimurium

ABSTRACT With the increasing and inappropriate use of colistin, the emerging colistin-resistant isolates have been frequently reported during the last few decades. Therefore, new potential targets and adjuvants to reverse colistin resistance are urgently needed. Our previous study has confirmed a marked increase of colistin susceptibility (16-fold compared to the wild-type Salmonella strain) of cpxR overexpression strain JSΔacrBΔcpxR::kan/pcpxR (simplified as JSΔΔ/pR). To searching for potential new drug targets, the transcriptome and metabolome analysis were carried out in this study. We found that the more susceptible strain JSΔΔ/pR displayed striking perturbations at both the transcriptomics and metabolomics levels. The virulence-related genes and colistin resistance-related genes (CRRGs) were significantly downregulated in JSΔΔ/pR. There were significant accumulation of citrate, α-ketoglutaric acid, and agmatine sulfate in JSΔΔ/pR, and exogenous supplement of them could synergistically enhance the bactericidal effect of colistin, indicating that these metabolites may serve as potential adjuvants for colistin therapy. Additionally, we also demonstrated that AcrB and CpxR could target the ATP and reactive oxygen species (ROS) generation, but not proton motive force (PMF) production pathway to potentiate antibacterial activity of colistin. Collectively, these findings have revealed several previously unknown mechanisms contributing to increased colistin susceptibility and identified potential targets and adjuvants for potentiating colistin treatment of Salmonella infections. IMPORTANCE Emergence of multidrug-resistant (MDR) Gram-negative (G-) bacteria have led to the reconsideration of colistin as the last-resort therapeutic option for health care-associated infections. Finding new drug targets and strategies against the spread of MDR G- bacteria are global challenges for the life sciences community and public health. In this paper, we demonstrated the more susceptibility strain JSΔΔ/pR displayed striking perturbations at both the transcriptomics and metabolomics levels and revealed several previously unknown regulatory mechanisms of AcrB and CpxR on the colistin susceptibility. Importantly, we found that exogenous supplement of citrate, α-ketoglutaric acid, and agmatine sulfate could synergistically enhance the bactericidal effect of colistin, indicating that these metabolites may serve as potential adjuvants for colistin therapy. These results provide a theoretical basis for finding potential new drug targets and adjuvants.

The virulence-related genes were mostly downregulated in JSDD/pR. Herein, we listed the top 20 significantly up-or downregulated genes of the three groups according to the log 2 FC in the Table S1. Generally, compared to JS, the top 20 SDEGs in JSDD were metabolism-related genes, for example the propanediol utilization and ethanolamine utilization-related genes. However, compared to JS or JSDD, transcriptional changes in JSDD/pR occurred not only in metabolism-related genes, but also in virulence-related genes (mostly downregulated), such as invF, invE, spaM, fimW, etc. The virulence-related genes included pathogenicity island genes (invG, invE, invF, sipA, sipB, sipC, sipD, prgJ, prgI), cell invasion protein gene (spaM), and fimbrial genes (fimW, fimC, fimA). The pathogenicity island genes were required for Salmonella invasion of host cell and aid in intracellular survival and replication (19). Mutations in invA, invC, and invG rendered Salmonella defective in their ability to invade cultured epithelial cells (20,21). Gene spaM encode protein involved in cell invasion (22). Type I fimbriae is known to play a role in Salmonella pathogenicity by facilitating adhering to and invasion of intestinal epithelial cells (23). FimW has been reported as a negative regulator of type I fimbriae genes (24). Genes fimC and fimA are required for the biogenesis of type I fimbriae, and FimA is found to be the major subunit of the type I fimbriae (25). Thus, decreased fimW expression may result in enhaced fimC, fimA transcription, and fimW mutant Salmonella strain were demonstrated to upregulated the fimA expression (24). However, in this paper, we detected simultaneous reduction of fimW, fimC, and  fimA, which may result from the cross-regulation role of FimY and FimZ (26). Hence, these data indicated that the more susceptible strain JSDD/pR may exhibit impaired adhesion and invasiveness.
Transcriptional changes occured in both colistin resistance-related genes and numerous other TCSs pathways. We summarized and drawn the networks of the SDEGs included in TCSs pathways, cationic antimicrobial peptide (CAMP) resistance pathway and lipopolysaccharide biosynthesis pathway in JSDD/pR versus JSDD and JSDD/pR versus JS groups. As shown in Fig. 3, the SDEGs were annotated and divided into three categories: colistin susceptibility, virulence, and metabolism. In terms of colistin susceptibility, the transcription levels of colistin resistance-related genes (CRRGs), such as TCSs genes basSR, phoPQ, pmrAB and the downstream genes arnABCDEFT operon (arnB also known as pmrH), pagP, pmrC, mdtABCD, were all significantly downregulated, which was reported to play important roles in the process of lipid modification especially the modification of pEtN and L-Ara4N (6). These results were in conformity with the electrospray ionization mass spectrometry (ESI-MS) analysis, which also demonstrated a reduction of pEtN and L-Ara4N modification of lipid A in the colistin susceptible Salmonella Typhimurium strain JSDD/pR (Fig. S1). Therefore, the detection of a markedly MIC reduction (16-fold) of colistin for JSDD/pR in our previous study were closely related to the reduction of L-Ara4N and pEtN modification of lipid A species (12). Besides, we also figured out the transcriptional changes of TCSs pathways. QseCB cascade have functions associated virulence as well as in flagella biosynthesis (27). The TCS MCP-CheA-CheW consists of methyl-accepting chemotaxis proteins (MCPs), histidine kinase (CheA), and scaffolding protein (CheW). Phosphorylated CheY diffuses to the flagellar motor and results in a change in the direction of cell movement (28). FIG 3 Networks of the SDEGs included in colistin susceptibility, virulence, and metabolism pathways in JSDD/pR versus JSDD and JSDD/pR versus JS groups. OM, PS, and CM represents outer membrane, periplasmic space, and cytoplasmic membrane, respectively. The red font, blue font, or black font of gene names indicated significantly downregulation, upregulation, or no significant change, respectively. Red dotted arrows between cpxR and colistin resistance-related genes (phoPQ, pmrC, pmrD, arnB) indicated the results in our previous paper that CpxR could markedly downregulate the transcriptional levels of these genes (12). The MICs of colistin for JS, JSDD and JSDD/pR were 0.8, 0.8, and 0.05 mg/L, respectively.
The metabolism-related TCSs NarQL, KdpDE, CitAB, and AauSR were associated with the fumarate reductase, potassium transport, citrate fermentation, acidic amino acid uptake, and metabolism (29)(30)(31)(32). These transcriptional changes of above metabolism-related genes and virulence-related genes may be a compensatory reaction. The results remind us to take a broader, holistic view of our research on antibiotic resistance, which should not only take into account the changes of resistance or susceptibility itself, but also the changes in metabolism and virulence of bacteria. There may be a new approach to the treatment of MDR bacteria in these compensatory changes.
More significantly differential metabolites were detected in JSDD/pR. In view of the obvious perturbation of metabolic pathways-related genes in JSDD/pR, and meanwhile, in consideration of reports that deletion or inhibition of acrB may cause the accumulation of antibiotics and harmful metabolites (14), we are curious that whether acrB deletion and cpxR overexpression in JSDD/pR could also increase the accumulation of harmful substances and ultimately led to the increased colistin susceptibility of JSDD/pR. We first detected whether the intracellular accumulation concentration of colistin was increased by liquid chromatography tandem mass spectrometry (LC-MS/MS). As shown in the Fig. S2, compared to JS or JSDD, colistin as well as the reported high/low-accumulating drugs, ciprofloxacin and rifampicin, were all significantly accumulated in JSDD/pR. Consisting with the mechanism of "direct antibacterial colistin activity" reviewed by El-Sayed Ahmed et al. (33), we considered that the increased uptake of colistin in JSDD/pR was responsible for the increased bactericidal effect of colistin.
The metabolome analysis were subsequently performed between JSDD versus JS, JSDD/pR versus JSDD, and JSDD/pR versus JS groups. Overall, only 40% of the detected metabolites in these strains were identifiable, and there were more significantly differential metabolites (SDMs) in JSDD/pR than JS and JSDD (Table S2). The principal-component analysis (PCA) were carried out and these strains formed two separate clusters, in which JSDD clustered tightly with the wild-type strain JS (Fig. 4). These results indicated that JSDD/pR was metabolically different form JS and JSDD.
We then performed a KEGG pathway analysis among these SDMs. A total of 45 metabolic pathways were significantly perturbed (P , 0.05) in JSDD/pR compared with JS or JSDD, embodying in carbon metabolism, pentose phosphate pathway, amino acid metabolism (cysteine, methionine, arginine, histidine metabolism, etc.), TCA cycle, nucleotide (purine and proline) metabolism, and some other metabolism (Table S3). But it should be noted that many SDMs were involved in multiple metabolic networks. For example, pyruvic acid was simultaneously included in pentose phosphate pathway, TCA cycle, biosynthesis of amino acids, and so on. Therefore, by comprehensive analysis of metabolomics, transcriptomics, and previous reports, we concentrated specifically on eight pathways that changed both at transcriptional and metabolic levels ( Table 1).
Exogenous citrate and a-ketoglutaric acid potentiated the colistin-mediated killing of Salmonella. The TCA cycle, as the common and hub metabolic pathway of carbohydrate lipid and amino acid metabolism, is an integral part for efficient bacterial metabolism in changing environment (18). Recently, numerous studies revealed that metabolic slowdown were associated with the drug-tolerant state of bacteria, and boosting the TCA cycle could alter the metabolic state, which thereby restoring antibiotic sensitivity (34,35). Surprisingly, we found that the more susceptible strain JSDD/pR exhibited a significant increase in several key metabolites of TCA cycle, compared to that of JS or JSDD (Fig. 5). The pyruvic acid, citrate, cis-aconitic acid, a-ketoglutarate, and (s)-malate level exhibited a substantial increase, being elevated 8 to 38-fold in JSDD/pR strain (Fig. 5).
Other studies have demonstrated that excess carbon sources such as fumarate, succinate, a-ketoglutarate, oxaloacetate, and pyruvate could restore susceptibility of pseudomonas aeruginosa, Edwardsiella tarda, or Vibrio alginolyticus to antibiotics (36)(37)(38)(39). Mechanistic investigations showed that supplementation of these carbon sources sensitized cells to antibiotic killing by stimulating the TCA cycle, which may result in activation of respiratory metabolism, more generation of PMF. By contrast, Goossens et al. had reviewed that metabolic slowdown with reduced TCA cycle conferred a drug-tolerant phenotype in Mycobacterium tuberculosis (40). Thus, exogenous pyruvic acid, citrate, cis-aconitic acid, a-ketoglutaric acid, and (s)-malate were used in this study to test whether they could promote the colistin-mediated killing of JS and JSDD strains. The survival of JS and JSDD were both significantly reduced in the presence of colistin plus pyruvic acid, citrate, and a-ketoglutarate, and their antibacterial effects were enhanced successively by treatment of citrate, a-ketoglutarate, and pyruvic acid (Fig. 6), which were proportional to their fold changes in JSDD/pR. While cis-aconitic acid and (s)-malate supplementation did not significantly alter the survival of JS (Fig. S3). The synergistic effects of these metabolites would be more significant, as the surviving cells may be re-growth after 6 h and 12 h of incubation. Therefore, we proposed that the significant accumulation of TCA cycle-associated metabolites, including citrate and a-ketoglutaric acid, in the more susceptible strain JSDD/pR make contributions to the increased antibiotic lethality.
Oxidative phosphorylation contributed little to enhance the colistin-mediated bacterial lethality. Oxidative phosphorylation can be a major source of ATP for the normal function of most cells. Three types of respiratory NADH dehydrogenases (NDH-1, NDH-2, and NQR) involving in oxidative phosphorylation have been identified in bacteria (41). NDH-1 is encoded by the nuoABCDEFGHIJKLMN operon, which using flavin mononucleotide (FMN) to transport electrons during NADH oxidation and PMF generation (42). As prior mentioned, the decline of PMF and ATP have been of great importance for the increased tolerance of antibiotic treatment (43,44), and enhanced PMF is known to drive the uptake of aminoglycoside antibiotics (45). Unexpectedly, we detected opposite changes in the level of FMN (0.19-fold) and phosphoric acid (Pi, 0.24-fold)  (TCA) cycle  map00020  3  3  2  3  0  0  20  Oxidative phosphorylation  map00190  0  2  0  0  0  0  16  Arginine and proline metabolism  map00330  0  7  0  4  0  0  78  Cysteine and methionine metabolism  map00270  2  6  2  7  0  0  61  Biosynthesis of amino acids  map01230  4  11  4  10  0  0  128  Histidine metabolism  map00340  0  5  2  5  1  0  47  Purine metabolism  map00230  0  7  3  7  0  2  95  Pyrimidine metabolism  map00240  2  5  3  3  in JSDD/pR compared to that of JSDD (Fig. 5a). Additionally, we also found that the expression level of nuoABCDEFGHIJKLMN operon were downregulated by 1.2-to 2.8-fold in JSDD/pR compared to that of JSDD and JS. These results indicated that there was decreased oxidative phosphorylation and may also in consequence PMF and ATP deficiency in JSDD/pR. We subsequently measured the PMF (DpH and Dc ) and intracellular ATP levels in JS, JSDD and JSDD/pR. Unexpectedly, we found no dramatic change of Dc and DpH, but significant increase of ATP production in JSDD/pR (Fig. 7a to c). These phenotypes may be explained by the fact that ATP generated by oxidative phosphorylation contributes little to the colistin susceptibility of Salmonella. As suggested earlier, ATP formed via oxidative phosphorylation is dispensable for growth of Salmonella, because the strains lacking nuo, ndh, or atpB-encoded subunits of ATP synthase recovered from nitrosative stress as efficiently as wild-type controls (46). The NADH-driven ATP synthesis through the transfer of electrons always coupled with ROS production. ROS play an important role in the physiology and pathology of cells, and usually associated with antimicrobial-mediated lethality (47). We discovered a marked increase of ROS production in JSDD/pR after examining the ROS levels in JS, JSDD, JSDD/pR (Fig. 7d). We suggest that AcrB deletion and CpxR overexpression may act synergistically to increase the ROS level. As it has been shown that activated AcrAB-TolC could suppress antibiotic-mediated intracellular ROS accumulation, and ROS levels were significantly increased when the DacrB strain was exposed to external stressful events (48,49). Furthermore, there were reports that CpxR could bind directly or work with additional regulatory pathways, e.g., ArcA, to repress the transcription of nuo operon, and thereby trigger ROS based cell death (50,51). Overall, we suggested that AcrB and CpxR could target the ATP and ROS generation, but not PMF production to potentiate antibiotic activity of colistin.  (a and b), citrate (c and d) and a-ketoglutaric acid (e and f) in the presence or absence of colistin, respectively. Purple or blue color indicates that sampling time is 6 h or12 h, respectively, and color deepens with an increase in the concentration of metabolites. "1" and "-" represent that colistin (2 mg/L) or a metabolite (1.2, 2.5, 5, and 10 mM) were added or not. CFU were counted to estimate the number of viable bacteria. Statistically significant differences were indicated by asterisk * P , 0.05, ** P , 0.01, *** P , 0.001, **** P , 0.0001, and ns, no significant difference. Values in brackets represent the MICs of colistin for JS (0.8 mg/L) and JSDD (0.8 mg/L).
Striking accumulations of agmatine sulfate and perturbation of nucleotide metabolism were considered to play roles in the increased colistin susceptibility as well decreased virulence of JSDD/pR. Further alterations in the metabolic profile of JSDD/pR were observed for amino acid concentrations (Fig. 8a). These include multiple pathways devoted to biosynthesis and metabolism of arginine, proline, cysteine, methionine, and histidine. Surprisingly, we observed an abnormally high agmatine sulfate content (104.53fold) in JSDD/pR. Meanwhile, exogenous agmatine sulfate could also drastically enhanced colistin bactericidal activity against JS and JSDD in a concentration dependent manner (Fig. 9). Agmatine sulfate can be produced through the decarboxylation of arginine, and then converted to putrescine by agmatinase (52). Putrescine is ultimately metabolized by polyamine aminopropyltransferase, resulting in the production of polyamines, which have been reported previously to interact with anionic complexes such as DNA, RNA, ATP, and phospholipids, and then stimulate cell proliferation, gene expression for the survival of cells. Recent studies have also confirmed polyamines as potent antioxidants by blocking free radicals from binding DNA and inhibiting lipid peroxidation in cell membranes (53,54). Thus, we considered that the accumulation of agmatine sulfate may be resulted from the downstream inhibition of polyamine production, which did not conducive for cell survival and enhanced the colistin-mediated killing of JSDD/pR. Nucleotide biosynthesis function in a variety of vital cellular and metabolic processes, such as synthesis of nucleic acids, energy production, delivering drugs (55). We have found in this study that both the purine and pyrimidine metabolism pathways were markedly changed in JSDD/pR (Fig. 8b), and the intermediates involved in purine metabolism pathway (adenine, 29-deoxyadenosine 59-monophosphate, AMP) or pyrimidine metabolism pathway (29, were upregulated by approximately 15-to 40-fold. Yang et al. (56) have found that adenine supplementation in E. coli decreased ATP demand, which elevating metabolic activity of central carbon, and thus reduced the lethality of antibiotics. In contrast, pyrimidine supplementation, including uracil or cytosine, could inhibit pyrimidine biosynthesis, promote purine biosynthesis activity and increase the antibiotic lethality. However, the  relationship between nucleotide biosynthesis and antibiotic was still controversial, since daptomycin-nonsusceptible Staphylococcus aureus showed increased levels of both purines and pyrimidines, while mutations in these pathways showed contrarily decreased formation of persister cells after treatment with rifampicin (57,58). Nucleotide synthesis inhibitors (e.g., 6-thioguanine) have been developed for treatment of Staphylococcus aureus or Mycoplasma pneumoniae (59,60). Of note, nucleotide biosynthesis pathways have strong links with full virulence of bacterial pathogens, and researchers have discovered that mutations in different pathway steps of nucleotide biosynthesis, including carA, carB, or pyrD, could consequently lead to decreased expression of virulence factors in Pseudomonas aeruginosa (61,62). Hence, we suggested that the disruption of nucleotide metabolism after acrB deletion and cpxR overexpression played a role in the increased colistin susceptibility as well decreased virulence of JSDD/pR in this paper.
Conclusions. Taken together, we have demonstrated that acrB deletion and cpxR overexpression have caused striking perturbations at both the transcriptomics and metabolomics levels in JSDD/pR. The virulence-related genes and CRRGs were markedly downregulated in the more susceptible strain JSDD/pR. The significant accumulation of metabolites in the TCA cycle and amino acid metabolism, including citrate, a-ketoglutaric acid, and agmatine sulfate, could potentiate the colistin-mediated killing of Salmonella, indicating that these metabolites may serve as potential adjuvants for colistin therapy. Additionally, we also suggested that AcrB and CpxR could synergistically target the ATP and ROS, but not PMF production to potentiate antibiotic activity of colistin. These results revealed several previously unknown regulatory mechanisms of AcrB and CpxR on the colistin susceptibility and provides a theoretical basis for finding potential new drug targets and adjuvants.

MATERIALS AND METHODS
Bacterial strains and growth conditions. The acrB and cpxR double-deleted mutant JSDacrBDcpxR:: kan (simplified as JSDD) and corresponding cpxR complementary strain JSDacrBDcpxR::kan/pcpxR (simplified as JSDD/pR) were generated from a multidrug-susceptible standard strain of Salmonella Typhimurium CVCC541 (named as JS) in our previous paper (63). The MICs of colistin for these strains were 0.8 mg/L (JS), 0.8 mg/L (JSDD), and 0.05 mg/L (JSDD/pR), respectively. For bacterial growth, single clones were propagated in LB medium and cultured to reach an OD 600 of 0.6, then cells were washed 3 times and collected for further use.
Transcriptome analysis. Transcriptome sequencing and functional annotation in JS, JSDD, and JSDD/pR were carried out by Sangon Biotech (China). The clean reads were mapped to the Salmonella  Typhimurium CT18 genome (NC003198) from NCBI using Bowtie2. We identified significantly differentially expressed genes (SDEGs) that with a qValue , 0.05 and jFold Changej .2 (log 2 FC $ 1or # 21). We further analyzed and compared the SDEGs between each pair of the three groups, i.e., JSDD compared to JS (JSDD versus JS), JSDD/pR compared to JSDD (JSDD/pR versus JSDD), or JSDD/pR compared to JS (JSDD/pR versus JS). The functional enrichment of SDEGs were annotated and classified by Gene Ontology (GO: http://geneontology.org) and Kyoto Encyclopedia of Genes and Genomes pathways (KEGG: http://www.kegg.jp). The relative transcription levels described in this paper referred to Log 2 FC of all transcripts.
Lipid A characterization by ESI-MS. Lipid A samples were prepared by the Bligh-Dyer method as previously described (64). For ESI-MS analysis, samples were dissolved with 1 mL solvent mixture of chloroform/methanol (4:1, vol/vol). The ESI-MS analysis were performed with the negative-ion mode of a Q Exactive Plus MALDI-TOF mass spectrometer (Thermo Fisher Scientific, USA).
Colistin accumulation measured by LC-MS/MS. Overnight cultures of JS, JSDD, and JSDD/pR were washed twice with 40 mL PBS, and pellets were resuspended in PBS and 900 mL each tube for three repetitions for each compound. Meanwhile, CFU were determined by the 10-fold serial broth microdilution method. Samples were equilibrated at 37°C with shaking for 5 min, and then 100 mL ciprofloxacin, novobiocin, and colistin were added (final concentrations were 1.5, 0.6, and 0.4 mmol/mL, respectively), with continually incubated for another 10 min. After incubation, 800 mL cultures were gingerly layered on 700 mL silicone oil (AR20/Sigma High Temperature), then centrifuged (9,400 rcf, 5 min) and removed the above supernatant and oil. After the resuspension in 200 mL water, each pellet were subjected to three freeze-thaw cycle of liquid nitrogen (3 min) in followed by 65°C water bath (3 min). Samples were pelleted at 9,400 rcf for 5 min and 180 mL supernatant was collected. Each pellet was resuspended in 100 mL methanol, and supernatant was combined with the previous supernatant collected after centrifuging (9,400 rcf, 5 min). Supernatants were 100 or 1,000 times diluted and analyzed by LC-MS/MS.
Bactericidal assay in vitro. Bacteria cells were prepared as described above and resuspended in LB media, adjusted OD 600 to 0.5. Colistin (2 mg/L) or/and a metabolite (1.2, 2.5, 5, and 10 mM) were added and incubated at 37°C for 6 h or 12 h. After incubation, CFU were determined by the 10-fold serial broth microdilution method. The metabolites, including pyruvic acid, citrate, a-ketoglutaric acid, cis-aconitic acid, and (s)-malate and agmatine sulfate, were all purchased from Sigma-Aldrich and dissolved in water.
Intracellular ATP and ROS levels assay. Intracellular ATP and ROS were assessed with the Enhanced ATP assay kit (Beyotime, Shanghai, China) and Intracellular ROS assay kit (Beyotime, Shanghai, China), respectively. Bacterial precipitates were lysed using lysozyme, and centrifuged. Subsequently, the supernatant liquids were prepared for intracellular ATP and ROS measurement according to the manufacturer's instructions.
Data availability. Transcriptome data have been submitted to the Sequence Read Archive (SRA) database under the BioProject accession number PRJNA625887 (SRA accession numbers SAMN14610720, SAMN14610710, and SAMN14610595). The raw data for the metabolomics have been uploaded to the MetaboLights (www.ebi.ac.uk/metabolights) under accession number MTBLS7745. It is anticipated that this accession number will be released by 2 July 2023; until that time, the data will be available from the corresponding author upon request.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, DOCX file, 1.1 MB.

ACKNOWLEDGMENTS
No potential conflict of interest was reported by the authors. This study was financed by the National Natural Science Foundation of China (NO.32072913 and 32102716).