Synergistic antibacterial activity of silver with antibiotics correlating with the upregulation of the ROS production

Thiol-dependent enzymes, including the thioredoxin (Trx) and glutathione (GSH) systems, have recently been found as promising bactericidal targets in multidrug-resistant (MDR) bacteria. We previously discovered that silver acted synergistically with ebselen in the inhibition of the Trx system and also resulted in a fast depletion of GSH in Gram-negative bacteria. Silver has been found by others to improve the sensitivity of bacteria to certain conventional antibiotics. Here, we found that the synergistic antibacterial effects of silver with four conventional antibiotics was correlated with the blockage of bacterial Trx system by silver. The synergistic antibacterial effect came along with the production of reactive oxygen species. All these results suggested that silver primarily enhanced the bactericidal activities of conventional antibiotics towards Gram-negative strains through the upregulation of ROS production.

and showed that antibiotics (ampicillin, kanamycin, and norfloxacin)-induced metabolic alterations elevated redox state, nucleotide oxidation, and increased oxidative stress by tightly regulated glutathione pools. Kohanski et al. 17 stated that all three classes of bactericidal drugs (aminoglycoside, quinolone, and beta-lactam) mediated hydroxyl radical damage. All these above reports pointed out that antibiotics can influence the redox balance in bacteria, but the exact mechanism is still not clear. Meanwhile, Collins and coworkers 17,18 proved that ROS production is one of the lethal factors of bactericidal antibiotics, and silver can enhance their antibacterial efficacy.
Antibiotic-mediated cell death is a complex, multi-faceted process that cannot be fully accounted for by the direct interactions of antibiotics with specific cellular targets. Here, we try to figure out whether antibiotics representing five different functional categories (beta-lactams, aminoglycosides, synthesis, tetracycline, and macrolides) can work synergistically with silver against Gram-negative bacteria through targeting redox system as ebselen and silver do 13 . Our results showed that silver could enhance certain antibiotics antibacterial effects against Gram-negative bacteria which are correlated to reactive oxygen species (ROS) production.

Results
Silver and certain antibiotics in combination exhibited synergistic antibacterial effects against E. coli. The antibacterial effects of silver nitrate (AgNO 3 ) and nine antibiotics representing five different functional categories (beta-lactams, aminoglycosides, synthesis, tetracycline, and macrolides) on the growth of a model Gram-negative bacterium, E. coli, which were detected in the 96 wells microplates. Overnight cultured E. coli DHB4 cells were diluted 1:1000 times in Luria Bertani (LB) medium, and incubated with serial dilutions of ionic silver (Ag + ) as a nitrate salt and 9 antibiotics in combinations, separately, for 24 h. Ebselen was used as the positive control, which acted synergistically with silver against Gram-negative bacteria in our recent report 13 . The results here showed that 4 (gentamicin, kanamycin, geneticin, tetracycline) out of 9 antibiotics had synergistic activity on E. coli DHB4 growth under the conditions we tested (Table S1). Further, the Bliss model was used to determine the nature of the therapeutic effects exhibited by the Ag + and antibiotics in combinations. We quantified the degree of synergy at 1 h and 4 h between Ag + and 4 antibiotics (gentamicin, kanamycin, geneticin, and tetracycline) in combinations, and the results showed that Ag + and 4 antibiotics indeed had synergistic antibacterial effects against E. coli (Fig. 1). All the results pointed out that Ag + could enhance the antibacterial effects of certain antibiotics against Gram-negative bacteria. In the following experiments, we used these 4 antibiotics for further studies.
ROS was one of the lethal factors for synergistic bactericidal effects of Ag + and antibiotics. Ag + and ebselen in combination could induce a high level of ROS production 13 , and the effects of Ag + and antibiotics in combinations need further studies. We, therefore, detected the production levels of ROS in Ag + and 4 antibiotics combinations treated cells, and Ag + and ebselen in combination was used as the positive control. The results showed that treatment with 5 μM Ag + alone could not cause a ROS response, and 5 μM Ag + and 80 μM antibiotics in combinations resulted in the upregulation of ROS (p < 0.0001) when compare with either antibiotics ( Fig. 2A) or silver (Fig. S1A). Furthermore, the increased levels of H 2 O 2 caused by the treatment with 5 μM Ag + and 80 μM antibiotics in combinations were also notarized by Amplex Red assay 15 (p < 0.0001) (Figs 2B and S1B). All the results demonstrated that ROS might be one of the determining factors for synergistic bactericidal effects of Ag + and antibiotics in combinations against E. coli. Ag + could disrupt bacterial Trx system. One of the main functions of thiol-dependent antioxidant systems is to scavenge ROS to keep cellular redox homeostasis and protect against oxidative stress. The disruption of the Trx and GSH systems may responsible for the upregulation of ROS. Ag + and ebselen in combination has been proven to target both bacterial Trx and GSH systems 13 , while the effects of Ag + and antibiotics in combinations need further studies. E. coli DHB4 cells grown to OD 600 nm of 0.4 were incubated with 5 µM Ag + and 80 µM antibiotics in combinations, and Ag + and ebselen in combination was used as the positive control. Results here showed that after 10 min treatment, the Trx activities in cell extracts treated by Ag + and antibiotics in combinations were dramatically inhibited compared with antibiotics or control group (Fig. 3A, p < 0.001); meanwhile, the TrxR activities in cell extracts treated by Ag + and antibiotics in combinations were also statistically lowered when compared with antibiotics or control group (Fig. 3B, p < 0.05). We also gained the same results when we prolonged the treatment time to 60 min (Fig. S2). But, there were no statistic difference between Ag + and antibiotics in combinations and silver (Fig. 3, p > 0.05). These results suggested that silver could disrupt bacterial Trx system; meanwhile, silver and antibiotics in combinations have direct influences on Trx1 when compared with antibiotics alone.
Bring into correspondence with above observation, the redox state of Trx1 verified by Redox Western Blot was also affected by Ag + and antibiotics in combinations after 60 min treatment when compared to antibiotics. Trx1 was in reduced form in untreated cells (negative control), and became oxidized upon the treatment by Ag + and antibiotics in combinations (Fig. 3C). In contrast, only 10 min treatment by Ag + and antibiotics in combinations could not cause Trx1 oxidization (Fig. S3). At the same time, the total protein levels of Trx1were not affected following a 10 min or 60 min treatment by Ag + and antibiotics in combinations (Figs 3C and S3). These results showed that when targeting Trx system, silver and antibiotics in combinations were not acting as fast as silver and ebselen do, which could effect Trx system in 10 min 13 .
Ag + and antibiotics could not directly disrupt bacterial GSH system. In our previous work, 5 µM Ag + and 80 μM ebselen in combination has been proven to deplete the GSH after 10 min treatment 13 . In our study, only 5 µM Ag + and 80 μM gentamicin or kanamycin in combinations could slightly deplete the total GSH amount in cell extracts when compared with antibiotics themselves (p < 0.05) (Fig. 4A), meanwhile other combinations showed no differences ( Fig. 4A) (p > 0.05). Further, the protein S-glutathionylation was decreased only in Ag + and ebselen in combination incubated bacteria 13 , but not in those bacteria cells treated with 5 µM Ag + and antibiotics in combinations for 10 min (Fig. 4B) or 60 min treatments (Fig. 5B).
All the results above suggested that silver and antibiotics had no direct effects on bacterial GSH system when acting against Gram-negative bacteria.
Five most difficult-to-treat MDR Gram-negative clinical isolates were highly sensitive to Ag + and ebselen in combination. We have isolated five clinically most difficult-to-treat MDR Gram-negative pathogen strains: Acinetobacter baumannii, Enterobacter cloacae, Escherichia coli, Klebsiella pneumonia and Pseudomonas aeruginosa. Overnight cultures of five isolates were diluted 1:1000 times in LB medium, and incubated with serial concentrations of Ag + and antibiotics in combinations for 24 h ( Table 1). The results showed that Ag + and antibiotics in combinations exhibited weak antibacterial effects on these MDR Gram-negative bacteria. Meanwhile, Ag + and ebselen in combination might be the only effective antibiotic against a range of resistant

Discussion
Recently a first new antibiotic in decades called Teixobactin was described inhibiting the peptidoglycan cell wall synthesis 3 . However, Teixobactin only inhibits Gram-positive bacteria like Staphylococcus aureus. Gram-positive bacteria in general have several differences to Gram-negative one which have a cell wall with an extra membrane. Most of Gram-positive bacteria lack GSH and associated enzymes like glutaredoxins 19 .
The current antibiotics principles are mainly based on disruption of protein synthesis, cell wall or cell membrane assembly, DNA or RNA replication 16 . We recently showed that targeting bacterial thiol-dependent redox enzymes is a distinct antibacterial mechanism compared with traditional antibiotics 4,13 .   There are two prime thiol-dependent enzyme systems, namely, thioredoxin (Trx) system, and glutathione (GSH)/glutaredoxin (Grx) system 19 . For the Trx system, electrons transferoccurs from NADPH to their substrates via thioredoxin reductases (TrxR) [20][21][22][23] ; meanwhile, for GSH/Grx system, electron transfer from NADPH to their substratesvia glutathione reductases (GR) 21,24,25 . Thiol-dependent enzymes are critical for ribonucleotide reductase (RNR) and DNA replication and repair, ordefense against oxidative stress via peroxiredoxins (Prxs) and methionine sulfoxide reductases (MSRs) 21,24,25 . Therefore, these two systems are critical for cell viability and proliferation.
Trx systems are ubiquitous in all bacteria, whereas the GSH system, as mentioned above, has been found to be absent in many pathogenic bacteria, including nearly all Gram-positive species 19,22,26 . We previously reported that ebselen, which is a substrate of mammalian TrxR but a competitive inhibitor of bacterial TrxR, executs selective antibacterial effect toward Gram-positive bacteria 4,12,27,28 . Further, we recently presented that silver and ebselen in synergistic combination had a strong selective bactericidal effect against Gram-negative bacterial infections. The combination directly inhibited both bacterial Trx and TrxR, and depleted intracellular GSH 13 . Based on these new findings, and combined with other published reports about antibiotics having influence on redox balance [15][16][17] , and that silver could sensitize Gram-negative bacteria to conventional antibiotics, thus we asked whether they also targeted the redox systems.
Drug-drug interactions can be classified as synergistic, antagonistic, or additive (no interaction). Nine antibiotics used in this work represent five different functional categories (beta-lactams, aminoglycosides, synthesis, tetracycline, and macrolides). Among them, total synthesis and tetracycline are broad-spectrum, aminoglycosides are particularly effective against Gram-negative bacteria, and beta-lactams and macrolides are particularly effective against Gram-positive bacteria. In the work reported here, 4 out of 9 antibiotics acted synergistically with silver against E. coli, a model Gram-negative bacterium (Fig. 1), which might occur through inducing ROS production as one of the lethal strategies (Figs 2 and 3).
E. coli has one TrxR, two Trxs (Trx1, and Trx2), and three major thiol peroxidases (bacterioferrit in comigratory protein (BCP), thiol peroxidase (Tpx), and alkyl hydroperoxide peroxidase subunit C (AhpC) 29,30 . Trx1 in E. coli is involved in protein repair by providing the electrons to E. coli methionine sulfoxide reductases (Msr) 20 , which participates in the protection of E. coli against oxidative damage from reactive nitrogen intermediates 31 . Trx1 in E. coli also acts as a specific reductase for homodimeric Tpx to scavenge ROS 32 . Thus, the decrease of Trx1 activity via its oxidization is at least partially responsible for the ROS production to cause E. coli cell death. Also the oxidative folding of disulfide containing membrane proteins is dependent on Trx and is possibly targeted 33 .
Meanwhile, although silve could inhibit Trx and TrxR (Fig. 3), yet the effect targeting GSH system is not universal (Figs 4 and 5). The presence of the GSH-Grx system in E. coli may be regarded as a backup for the Trx system. GSH/Grxs in E. coli participate in the antioxidant process by deglutathionylation and transfer electrons to ribonucleotide reductase. Silver and antibiotics in combinations showed no effects on GSH amount, and the S-glutathionylated proteins are not much different from that of in control group (Figs 4 and 5). Thus, this might be one of the reason to explain the anti-MDR-Gram-negative bacteria activities of silver and antibiotics in combinations were much weaker than silver and ebselen in combination (Table 1 and Fig. 6). All in all, silver enhanced the antibacterial effects of certain antibiotics against Gram-negative bacteria through upregulate the ROS production. In order to be more effective, the Trx and GSH systems must be targeted, as well, since both systems are particularly important for ribonucleotide reductase in bacteria which is essential for DNA replication and repair.

Materials and Methods
Bacterial strains. Escherichia coli (E. coli) DHB4 cells and multidrug-resistance (MDR) Gram-negative clinical isolatess (Tables 2 and 3 Table 2. Clinically isolated multidrug-resistant Gram-negative strains used in this work. Antibiotic KP-2 AB-1 PA-1 ECL-2 ECO-1  Quantifying synergy of ebselenand silver using the bliss model. Drug synergism was determined using the Bliss Independence Model, which calculates a degree of synergy using the following formula: S = (f X0 / f 00 )(f 0Y /f 00 ) − (f XY /f 00 ), where f XY refers to the wild-type growth rate in the presence of the combined drugs at a concentration X, for one of the drugs, and Y for the other; f X0 and f 0Y refer to the wild-type growth rates in the presence of the individual drugs at a concentration of X and Y, respectively; f 00 refers to the wild-type growth rate in the absence of drugs; and S corresponds to the degree of synergy, a parameter that determines a synergistic interaction for positive values and an antagonistic interaction for negative ones. Growth rates at different time points are determined by calculating the slope of the growth or kill curve being analyzed 18 .
Detection the production of ROS. The E. coli DHB4 were grown until the OD 600 nm of 0.4 in 5 ml LB medium, and the DHB4 cells were incubated with silver and antibiotics in combinations for 10 min. To measure the production level of ROS in the treated bacteria, DHB4 cells were harvested by centrifugation at 6,000 rpm for 5 min and thoroughly washed by PBS, which were further stained by 5 µM H 2 DCF-DA that protected from light for 20 min. After the incubation, treated cells were spun down and re-suspended in PBS, and the amount of ROS was detected by flow cytometry (CyAnadp, Beckman coulter).

Detection of Trx/TrxR activities and GSH amount in antibiotics and silver treated E. coli cell
lysates. E. coli DHB4 were grown until the OD 600 nm of 0.4 in LB medium, and the DHB4 cells were incubated with 80 µM antibiotics and 5 µM AgNO 3 for 10 and 60 min, respectively. The cultures that treated with 5 µM AgNO 3 and 80 µM ebselen were used as the positive control. Treated cells were harvested by centrifugation at 6,000 rpm for 5 min and thoroughly washed 3 times by PBS, and cells were further re-suspended in lysis buffer (100 mM NaCl, 20 mM NaF, 2.5 mM EDTA, 1 mM Na 3 VO 4 , 2.5 mM EGTA, 20 mM sodium ß-glycerophosphate, 0.5% Triton X-100, 10 mM sodium pyrophosphate, 25 mM Tris·HCl (pH 7.5)), which contains protease inhibitor cocktail, and further lysed by sonication. The treated cell lysates were collected by centrifugation at 13,000 rpm for 20 min and the protein concentrations were measured with the Lowry protein kit. The TrxR activity in E. coli DHB4 cell extracts was detected by the DTNB reduction activity assay 23 . These experiments were performed in the 96 microwell plates that contains 5 µM E. coli Trx1, 200 µM NADPH, 1 mM DTNB and 50 mM Tris·HCl (pH 7.5). The absorbance at 412 nm was detected for 5 min with a VERSA microwell plate reader and the slope of initial 2 min was set to verify TrxR activity in cell extracts. The Trx activity was detected by the same method which coupled with 100 nM E. coli TrxR instead of 5 µM E. coli Trx in the above discribed reaction solution. To measure the amount of GSH in cell extracts, 25 µg cell lysates were added into the reaction mixture containing 50 nM GR, 200 µM NADPH, 50 mM Tris·HCl (pH 7.5), 1 mM DTNB and 1 mM EDTA. The absorbance at 412 nm was detected for 5 min.

Redox state of Trx1 in E. coli treated by silver and antibiotics in combinations. E. coli DHB4
were grown until the OD 600 nm of 0.4 in LB medium, and the DHB4 cells were incubated with 80 µM antibiotics and 5 µM AgNO 3 for 10 and 60 min, respectively. The cultures treated with 5 µM AgNO 3 and 80 µM ebselen were used as the positive control. Redox Western blotting was performed to detect the redox state of Trx1 in the DHB4 cells. The treated cells werefurther harvested by centrifugation at 6,000 rpm for 5 min and thoroughly washed 3 times by PBS, and the protein was precipitated by 5% TCA in 1.0 ml. The precipitates were washed throughly 3 times with 1 ml pre-ice-cold acetone and dissolved in 50 mM Tris·HCl (pH 8.5) with 0.5% SDS containing 15 mM  MeO-PEG-Mal at 37 °C for 2 hours. Proteins were collected by centrifugation at 13,000 rpm for 20 min and the protein concentration was verified with the Lowry protein kit. Proteins were incubated with SDS-loading buffer at 90 °C for 10 min, and then separated on the 4-12% bolt Bis-Tris gel with MES running buffer (150 V, 40 min). The redox state of Trx1was detected withsheep anti-E. coli Trx1 antibody at 1:1000 dilution, followed by the detection of Chemiluminescence Reagent Plus.

Proteins S-glutathionylation in E. coli cells treated by silver and antibiotics in combinations.
Total protein S-glutathionylation of antibiotics and AgNO 3 in combinations treated E. coli DHB4 cells was verified by Western blot. E. coli DHB4 were grown until the OD 600 nm of 0.4 in LB medium, and the DHB4 cells were incubated with 80 µM antibiotics and 5 µM AgNO 3 for 10 and 60 min, respectively. The cells treated with 5 µM AgNO 3 and 80 µM ebselen were used as the positive control. Cells were washed 3 times, and re-suspended in lysis buffer (10 mM sodium pyrophosphate, 100 mM NaCl, 2.5 mM EDTA, 25 mM Tris·HCl (pH 7.5), 2.5 mM EGTA, 1 mM Na 3 VO 4 , 20 mM sodium ß-glycerophosphate, 0.5% Triton X-100, 20 mM NaF and 50 mM IAM) containing protease inhibitor cocktail. Afte sonication, the treated cell lysates were collected by centrifugation at 13,000 rpm for 20 min. Protein concentration was measured with Lowry protein kit, and Western blot was performed as described above with IgG2a mouse monoclonal antibody (VIROGEN, 101-A/D8) for S-glutathione-protein complexes. Statistical analysis. Mean, Standard Deviation (SD) and t-test (two tails, unpaired) significances were calculated in GrapPad Prism Software. *p < 0.05, **p < 0.01, ***p < 0.001.

Antibacterial activities of antibiotics
Data availability. The data sets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.