To explore pyruvate-stimulated antibiotic-mediated killing of E. tarda, viability of PPD200/87, a multidrug-resistant E. tarda (Fig. S2), was measured in the presence of pyruvate, antibiotics, or both. Five different classes of antibiotics were used, including β-lactams (amoxicillin and cefperazone-sulbactam), polypeptide (polymyxin B), quinolones (ofloxacin), tetracyclines (tetracycline), and aminoglycosides (gentamicin, micronomicin, and tobramycin). Aminoglycosides were most potentiated by pyruvate compared to other classes of antibiotics (Fig. 1A). Further, pyruvate-stimulated gentamicin killing occurred in a pyruvate dose-dependent manner, with 5 mM pyruvate being the most effective (Fig. 1B). Such a synergistic effect was also time and gentamicin dose dependent (Fig. 1C and D). The pyruvate-potentiated gentamicin killing was also effective against persisters and biofilms that could not be eradicated by pyruvate or gentamicin alone (Fig. 1E and F and Fig. S3).

To validate the synergistic effect of pyruvate and gentamicin, 2 infection models were investigated. In the E. tarda–tilapia infection model, pyruvate and gentamicin combination treatment elevated the tilapia survival rate from 20% to 100% compared to gentamicin treatment alone (Fig. 1G). In addition, combination treatment promoted mouse survival from 10% to 90% (Fig. 1H). Moreover, pyruvate and gentamicin treatment reduced bacterial loads by 293-, 198-, and 266-fold in mouse liver, spleen, and kidney, respectively (Fig. 1I). Furthermore, over 24 strains of clinically isolated E. tarda were tested to exclude strain-specific effects. All strains were more sensitive to killing by pyruvate and gentamicin combination treatment than by monotreatment with either, and the increased killing ranged from 4.2- to 8,130.8-fold (Fig. 1J). These results indicate that pyruvate potentiates aminoglycoside killing of multidrug-resistant E. tarda.

To investigate the mechanism by which pyruvate potentiates gentamicin killing, we performed metabolomic and transcriptomic analyses. Using gas chromatography–mass spectrometry (GC-MS)-based metabolomics, we identified a total of 106 metabolites from each sample (Fig. S4), including 30 up-regulated and 20 down-regulated metabolites (Fig. S5A and B). Pathway analysis indicated that the 50 differentially expressed metabolites were enriched to 11 metabolic pathways, with glycine, serine, and threonine metabolism, as well as cysteine and methionine metabolism being the top 2 pathways (Fig. 2A). Interestingly, the abundance of all and most of the metabolites from the top 1 and 2 enriched metabolic pathways, respectively, was up-regulated (Fig. S5C). Furthermore, orthogonal partial least square discriminant analysis (OPLS-DA) was conducted to identify sample patterns between the 2 groups (Fig. S5D). S-plot analysis identified metabolites with absolute values of the S-plot variation weight t and correlation coefficient P (corr) ≥ 0.05 and 0.5, respectively, as biomarkers. Among these biomarkers, serine, aspartate, glycine, and cysteine belong to the 2 most impacted metabolic pathways (Fig. 2B and Fig. S5E). Collectively, glycine, serine, and threonine metabolism, as well as cysteine and methionine metabolism are the 2 most critical metabolic pathways boosted by exogenous pyruvate.

Further, RNA-sequencing analysis was performed on bacteria treated with pyruvate, gentamicin, or combination treatment with both. Differentially expressed genes (DEGs) were identified with adjusted P values < 0.05 and an absolute |log₂ fold change| value > 1. In total, 1,198, 667, and 1,767 DEGs were identified in the pyruvate treatment group, gentamicin treatment group, and combination treatment group, respectively. Moreover, 506, 315, and 869 of those were up-regulated, while 692, 352, and 898 were down-regulated in the same groups, respectively (Fig. S6A). All DEGs were visualized using volcano plots, where the top 10 altered genes included cysk, uptG, metB, aspC, ETAE_3430, and thrA, which are implicated in cysteine and methionine metabolism, as well as glycine, serine, and threonine metabolism. Additionally, aceE, which plays a role in the P cycle, was also included (Fig. 2C and D and Table S1). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of the DEGs also enriched glycine, serine, and threonine metabolism, cysteine and methionine metabolism, and glutathione metabolism (Fig. S6B), rationalizing the shift of these pathways upon combined treatment with pyruvate and gentamicin. Following integration of the metabolomic and transcriptomic data, the 3 pathways were all up-regulated at the transcriptional and metabolite level (Fig. 2E). Pyruvate has 2 pathways (to cysteine by metB from pyruvate directly and to aspartate by aspC from oxaloacetate indirectly) to flux to cysteine and methionine metabolism, and to glycine, serine, and threonine metabolism, and then possibly to glutathione metabolism. It is important to note that the conversion of pyruvate to oxaloacetate has 3 metabolic pathways (i.e., pyruvate–PEP–oxaloacetate, pyruvate–oxaloacetate, and pyruvate–TCA cycle), which are included in the P cycle (Fig. 2E).

Thus, quantitative real-time polymerase chain reaction (qRT-PCR) was used to measure expression of genes encoding the 3 metabolic pathways (glycine, serine, and threonine metabolism; cysteine and methionine metabolism; and the P cycle on a holistic basis to fully understand this potentiation). For glycine, serine, and threonine metabolism, as well as cysteine and methionine metabolism, the expression of 10 genes was measured. We found that pyruvate increased the expression of all genes and gentamicin elevated the expression of 4 genes (aspC, thrA, ETAE_2812, and ETAE_1960), while the other 6 were unchanged. Distinctly higher expression was detected in the pyruvate group compared to the gentamicin group. More importantly, the synergistic use of pyruvate and gentamicin caused all genes to exhibit high expression compared to the control, where higher expression for most genes was found in the combined therapy treatment group compared to the pyruvate group (Fig. 3A). The enzymes, glutamic-oxaloacetic transaminase (GOT) and CGL, by which pyruvate enters the 2 pathways were also measured. GOT (encoded by aspC) reversibly converts oxaloacetate and glutamate into aspartate and α-ketoglutarate, while CGL (encoded by metB) reversibly converts pyruvate to cysteine. We observed increased GOT and CGL activity in the pyruvate treatment group as well as in the combination treatment group compared to the control (Fig. 3B and C). For the P cycle, pyruvate mono- and combination treatment promoted the expression of ace/F, icd, sucD, DCL27_RS10995, and maeB, but inhibited the expression of sucA, sdhB (both only the combination treatment group), and pckA. Gentamicin inhibited the expression of aceE, sucA, sucD, and pckA and promoted the expression of acnB, icd, and sucC (Fig. 3D). The activated P cycle was demonstrated by increased activity of enzymes in the P cycle (Fig. 3E). Furthermore, pyruvate was replaced with one of the intermediate metabolites in glycine, serine, and threonine metabolism; cysteine and methionine metabolism; and glutathione metabolism to determine whether they play a similar role to pyruvate. To achieve this, several intermediate metabolites, including aspartate, glycine, serine, homoserine, cysteine, and cystathionine, were quantified. All were higher in PPD200/87 in the presence than absence of pyruvate (Fig. 3F). Since sulfur-containing metabolites are sensitive to oxygen and heat, we directly quantified l-cysteine level to exclude such effect during GC-MS sample preparation [24]. This test validates that pyruvate increased cysteine content (Fig. 3G). Intermediate metabolites of the pyruvate to glutathione metabolism were individually synergized with gentamicin to kill PPD200/87, including oxaloacetate, aspartate, glycine, serine, cystathionine, cysteine, glutathione (GSH), and glutathione disulfide (GSSG). All of these metabolites synergized with gentamicin to kill E. tarda in a dose-dependent manner, with the exception of GSSG, which had no effect (Fig. 3H to O). GSH and GSSG are the principal intracellular antioxidant buffers against oxidative stress. GSH was the end effector, implying that the pyruvate-potentiated killing is related to GSH and the GSH/GSSG ratio. In addition, we tested the role of acetate, lactate, and formate, 3 products of other pyruvate metabolic pathways, in potentiating gentamicin. Formate, but not acetate and lactate, weakly potentiated the gentamicin killing (Fig. 3P), suggesting that pyruvate does not transform these 3 products to play the key role. These findings not only support the hypothesis that a pyruvate–cysteine/oxaloacetate–aspartate–serine–cysteine–glutathione pathway contributes to the pyruvate potentiation but also indicate the importance of understanding the balance between GSH and GSSG.

Cysteine and methionine metabolism and glutathione metabolism provide sulfur-containing amino acids and low molecular weight thiol compounds, respectively, which serve as potent antioxidants and redox signaling intermediates [25]. In glutathione metabolism, 2 molecules of GSH are oxidized by GSH peroxidase to one molecule of GSSG to directly scavenge ROS [26], while glutathione reductase (GR) converts GSSG to 2 molecules of GSH primarily using nicotinamide adenine dinucleotide phosphate (NADPH) or, in rare cases, NADH [reduced form of nicotinamide adenine dinucleotide (oxidized form) (NAD⁺)] [27]. Thus, GSH and GSSG and their ratio were measured in the absence and presence of pyruvate and/or gentamicin, in addition to the ROS scavenger, α-tocopherol, and the ROS promoter, H₂O₂ for positive and negative controls, respectively. The GSH level was reduced with the addition of pyruvate or H₂O₂, but not of α-tocopherol. In the combination treatment with gentamicin, GSH levels were considerably inhibited by pyruvate and then H₂O₂, but weakly by α-tocopherol. The action of H₂O₂ and α-tocopherol was enhanced by pyruvate (Fig. 4A). In contrast, we observed the opposite with GSSG levels (i.e., GSSG level was elevated with addition of pyruvate or H₂O₂). In the combination treatment with gentamicin, the GSSG level was greatly increased by pyruvate and then H₂O₂, which was enhanced by pyruvate. Importantly, α-tocopherol did not influence these effects (Fig. 4B). The GSH/GSSG ratio showed that pyruvate and H₂O₂ decreased the ratio, especially in the synergy with gentamicin and pyruvate, whereas the ROS scavenger and gentamicin treatment alone did not affect the ratio (Fig. 4C). Consistently, the second and first lowest viability was detected with the synergistic use of gentamicin and pyruvate, as well as with the addition of H₂O₂, respectively. Higher and lower viability was detected in the gentamicin/α-tocopherol combination treatment group and with H₂O₂, respectively, compared to gentamicin alone. Unchanged survival was measured in addition to α-tocopherol and gentamicin with and without pyruvate (Fig. 4D). The viability should be attributed to ROS. Specifically, α-tocopherol inhibited ROS, while H₂O₂ and pyruvate stimulated ROS, which was enhanced by gentamicin and pyruvate (Fig. 4E). Further, low concentrations of pyruvate promoted ROS, while high concentrations inhibited ROS in the absence of gentamicin (Fig. 4F), which is consistent with previous studies of pyruvate as a ROS scavenger [28]. Pyruvate-mediated ROS enhancement is positively proportional to pyruvate-mediated gentamicin killing within 5 mM pyruvate, but parallel when >5 mM pyruvate is used (Fig. 4G). These results indicate that pyruvate regulates GSH, GSSG, and GSH/GSSG ratios to promote ROS, which is needed for pyruvate-potentiated gentamicin killing. To further confirm this, GSH, GSSG, and GSH/GSSG ratios were measured in the presence of gentamicin with a pyruvate concentration gradient. Indeed, GSH and GSSG were reduced and elevated with the increasing pyruvate, respectively, and reached their maxima at approximately 5 mM pyruvate (Fig. 4H and I), which is a similar concentration observed to potentiate gentamicin killing (Fig. 1B). Consistently, we found that the GSH/GSSG ratio decreased in a dose-dependent manner (Fig. 4J). GSH, GSSG, GSH/GSSG ratio, and ROS were also measured in the presence of glycine and cysteine, the downstream metabolites of pyruvate. As compared to pyruvate (5 mM), which decreased GSH for 3.94-fold, increased GSSG for 7.17-fold, and reduced GSH/GSSG for 28.13-fold (Fig. 4A to C), glycine (5 mM) and cysteine (0.6 mM) decreased GSH level for 3.28- and 3.01-fold, increased GSSG for 6.82- and 5.42-fold, and decreased GSH/GSSG ratio for 22.43- and 16.34-fold, respectively (Fig. 4K). The folds of change for glycine and pyruvate were similar to pyruvate, but cysteine was less potent, which can be due to the lower dose of cysteine we used because of its poor solubility in water. Similarly, ROS was 121.3 ± 4.2 relative fluorescence units (RFU) and 110.7 ± 1.5 RFU in the presence of glycine and cysteine, respectively, as compared to 132.3 ± 4.9 RFU in the presence of pyruvate (Fig. 4F and L). Taken together, these experiments showed that pyruvate reaches its optimum effect at 5 mM. Moreover, these data suggest that regulation of GSH and GSSG ratio by exogenous pyruvate is responsible for ROS abundance and viability.

To further examine the pyruvate-mediated glutathione system regulation, qRT-PCR was used to measure the expression of genes encoding GSH generation and conversion. The expression of gshA and gshB, which convert cysteine and glycine to GSH, was higher in pyruvate and pyruvate/gentamicin combination treatment groups compared to the control and gentamicin monotreatment groups. However, the expression of genes that convert GSH to glycine (i.e., pepA, pepB, pepD, and pepN) was higher in the pyruvate, gentamicin, and pyruvate/gentamicin combination treatment groups compared to the control, with the highest expression observed in the combination group. Very interestingly, a reversal change was determined between higher gpx and low gor (no significance) expression in the pyruvate and pyruvate/gentamicin groups compared to the gentamicin only and control groups (Fig. 5A and Fig. S7). Note that gpx encodes glutathione peroxidase (GPX) that oxidizes GSH to GSSG, while gor encodes GR, which reduces GSSG back to GSH. Similar results (pyruvate caused the higher GPX activity and the lower GR activity, which were enhanced by gentamicin) were measured for the activity of GPX and GR (Fig. 5B and C). These results suggest that the reduction of GSSG to GSH is impaired by combination treatment of pyruvate and gentamicin. Since NADPH is required for reduction, NADP_total, NADPH, NADP⁺, and the NADHP/NADP⁺ ratio were measured. Pyruvate and pyruvate with gentamicin elevated NADP, NADPH, and reduced NADP⁺, leading to an increased NADHP/NADP⁺ ratio (Fig. 5D to G). These results indicate that the depressed GR is the most characteristic feature in the synergistic use of pyruvate and gentamicin, which is not related to the NADHP/NADP⁺ ratio. Because exogenous pyruvate alone caused a similar effect to the activity of GR, we speculated that pyruvate or its conversion products inhibited the activity of GR directly. To explore this, the GR activity was measured in the presence of aspartate, glycine, serine, cysteine, GSH, or GSSG. The activity was inhibited by glycine but not by the other metabolites (Fig. 5H). To validate these data, gor was cloned and the expressed recombinant protein was used to measure its activity in the presence of different concentrations of glycine. Our data show that its activity was reduced with increasing concentrations of glycine (Fig. 5I). Further, isothermal titration calorimetry (ITC) analysis revealed the binding of glycine with GR (Fig. 5J).

To further explain the relationship between pyruvate-boosted cysteine and pyruvate-depressed GSH, the expression of genes encoding the conversion from cysteine to GSH and from GSH to glycine was measured and the activity of GPX and GR was detected in the presence of different concentrations of pyruvate. The expression of these genes and GPX activity were elevated with increasing concentrations of pyruvate (Fig. 5K and L), while GR activity was reduced in a pyruvate dose-dependent manner (Fig. 5L). These results suggest that pyruvate metabolic flux promotes GSH and glycine, and subsequently, glycine in turn inhibits GR to depress the GSH source from GSSG, thereby resulting in GSH depression.

We further supposed that the depressed GSH impairs the clearance of ROS and thereby caused ROS elevation, which is responsible for the pyruvate-potentiated gentamicin killing. Thus, pyruvate was replaced with H₂O₂ to test whether H₂O₂ potentiated gentamicin killing. We found that the viability of PPD200/87 was reduced in a H₂O₂ dose-dependent manner (Fig. 5M), while ROS was elevated with the increasing H₂O₂ concentrations (Fig. 5N). When α-tocopherol was added, the viability and ROS were up- and down-regulated, respectively, in an α-tocopherol dose-dependent manner (Fig. 5O and P). Viability was also related to gentamicin dose under the same level of H₂O₂, the dose that had no killing effect (Fig. 5Q). To understand why pyruvate and H₂O₂ caused the elevated ROS but relatively higher and lower killing with the same dose of gentamicin, intracellular drug concentration was measured. Pyruvate elevated drug intracellular concentration, but H₂O₂ did not (Fig. 5R). These results support the conclusion that pyruvate regulates glutathione system to accumulate ROS that potentiates gentamicin killing and that gentamicin killing is ROS dependent. Therefore, the pyruvate–cysteine–glutathione system/glycine–ROS metabolic pathway is potentiated to synergize gentamicin killing.

The role of the pyruvate–cysteine–glutathione system/glycine–ROS metabolic pathway in antibiotic resistance was further evaluated using deletion mutants. Loss of metB, aspC, pepB, gpx, or gor elevated (metB, aspC, pepB) and decreased (gpx, gor) viability in the presence of gentamicin alone (Fig. 6A), suggesting the physiological existence of this metabolic resistance pathway. When exogenous pyruvate was added, the viability of ΔaspC, ΔaceE, ΔmetB, and ΔpepB was increased, while that of Δgpx and Δgor was reduced. Importantly, the absence of metB abrogated the pyruvate potentiation and the loss of gpx or Δgor distinctly promoted the effect (Fig. 6B). This metB-dependent abrogation implied that the flux of pyruvate to cysteine via CGL plays the most key role in the pyruvate potentiation, while gpx- or gor-dependent promotion suggested that the glutathione system dominates the potentiation. Thus, further experiments were performed in ΔmetB, Δgpx, and Δgor. Intermediate metabolites, glycine, cysteine, and glutathione of the pyruvate–cysteine–glutathione system/glycine–ROS metabolic pathway promoted gentamicin killing in ΔmetB, where glutathione caused similar killing to Δgpx as pyruvate did (Fig. 6C). Moreover, ROS, GSH, GSSG, and GSH/GSSG were measured. Compared to EIB202, ΔmetB exhibited a low abundance of ROS. When pyruvate was added, ROS was promoted, which was further boosted by the addition of gentamicin (Fig. 6D). To demonstrate that the killing in Δgpx and Δgor was attributed to the elevation of ROS, the ROS inhibitor α-tocopherol was used. Viability was elevated with increasing α-tocopherol dose in Δgpx and Δgor, when ROS was reduced (Fig. 6E and F). Absence of gpx increased GSH level but not GSSG and GSH/GSSG ratio, while absence of gor decreased GSH level, increased GSSG level, and decreased GSH/GSSG ratio. Deletion of metB caused down-regulation of GSH and GSSG, but increased the GSH/GSSG ratio. Similar results were detected in addition of gentamicin. When pyruvate or combination treatment with gentamicin was used, GSH was elevated, while GSSG expression and GSH/GSSG ratio were decreased in Δgpx and ΔmetB. However, a reversal result was measured in Δgor (Fig. 6G). The GSH/GSSG couple plays a crucial role in the 2-electron reduction of peroxides such as H₂O₂ using NADPH as an electron donor [29]. However, the GSH/GSSG ratio was not related to ROS expression level in Δgpx, which exhibited higher ROS and a higher GSH/GSSG ratio. This is because the absence of gpx causes GSH accumulation and limits the GSSG source; thus, the accumulated GSH did not reduce ROS. Finally, a mouse model was used to confirm that the loss of metB, gpx, or gor leads to the changes in the resistance/sensitivity to gentamicin and gentamicin + pyruvate. Further, we observed 20% survival in EIB202-infected mice that had been treated with gentamicin. Subsequently, we also observed 40%, 30%, and 10% survival in mice infected with Δgpx, Δgor, and ΔmetB mutants, respectively, following treatment with gentamicin alone (Fig. 6H). In the pyruvate combination treatment group, survival of Δgpx, Δgor, EIB202, and ΔmetB was 90%, 70%, 50%, and 0%, respectively (Fig. 6I). Therefore, the pyruvate–cysteine–glutathione system/glycine–ROS metabolic pathway physiologically contributes to antibiotic resistance that can be reversed by exogenous pyruvate in vitro and in vivo.

We further explored whether the pyruvate–cysteine–glutathione system/glycine–ROS metabolic pathway is not only promoted by exogenous pyruvate but also inhibited during drug resistance. To do this, LTB4 and the 8- and 16-fold minimum inhibitory concentration (MIC) of gentamicin isogenic strains (LTB4-S, LTB4-R_8MIC, and LTB4-R_16MIC) were collected. Further, qRT-PCR showed that the reduced expression of most genes encoding the pyruvate to glutathione system via glycine, serine, and threonine metabolism, and cysteine and methionine metabolism was associated with an increasing MIC in LTB4-R_8MIC and LTB4-R_16MIC compared to LTB4-S (Fig. 7A to C). The activity of GOT (catalyzing the reversible reaction of oxaloacetate and l-glutamate into l-aspartate and α-ketoglutarate) and CGL (reversibly converting pyruvate to cysteine) was reduced in a MIC-dependent manner (Fig. 7D and E). Aspartate, glycine, serine, homoserine, cystathionine, and cysteine were decreased with increasing MICs (Fig. 7F). The activity of GPX and GR was either unchanged or up-regulated (Fig. 7G and H). Consistently, GSH and GSSG were increased or unchanged (Fig. 7I and J). The GSH/GSSG ratio was elevated with increasing MICs (Fig. 7K). Furthermore, NADP_total, NAPDH, NAPD⁺, and NAPDH/NAPD⁺ ratios were measured. NADP_total and NAPD⁺ were unchanged, whereas NAPDH and NAPDH/NAPD⁺ ratios were reduced with the increasing MICs (Fig. 7L to O). Finally, pyruvate potentiation was demonstrated in these laboratory-evolved strains (Fig. 7P). These results indicate that the pyruvate–cysteine–glutathione system/glycine–ROS metabolic pathway contributes to antibiotic resistance.

To explore the role of the pyruvate–cysteine–glutathione system/glycine–ROS metabolic pathway in the resistance of other clinically isolated pathogens, clinically isolated multidrug-resistant and/or carbapenem-resistant P. auroginosa, E. coli, K. pneumonia, and MRSA were collected (Fig. S8). Metabolomics analysis revealed down-regulated aspartate, glycine, serine, homoserine, cysteine, and cystathionine in the pyruvate–cysteine–glutathione system/glycine–ROS metabolic pathway (Fig. 8A). Higher GSH, unchanged GSSG, and higher GSH/GSSG ratios were measured in 6 multidrug-resistant pathogens compared to 6 antibiotic-sensitive strains, which was reversed by exogenous pyruvate (Fig. 8B and Fig. S7). The activity of GPX was unchanged between these sensitive and resistant strains, which was promoted by exogenous pyruvate (Fig. 8C). However, GR activity was higher in these resistant strains compared to sensitive strains, and this activity was inhibited to low or normal levels by exogenous pyruvate compared to the sensitive strains (Fig. 8D). Moreover, lower ROS was consistently measured in resistant strains compared to sensitive pathogens, which was reversed by exogenous pyruvate (Fig. 8E). Exogenous pyruvate potentiated gentamicin to kill multidrug-resistant or carbapenem-resistant clinic isolates, P. auroginosa, E. coli, K. pneumonia, and MRSA (Fig. 8F). The pyruvate-potentiated gentamicin was effective in mouse models infected with these pathogens (Fig. 8G and H). Specifically, combination treatment with exogenous pyruvate and gentamicin increased survival in mice infected with MDR-ECO41, MDR-KPN68, CR-PAE17, and MRSA16 by 70%, 50%, 30%, and 50%, respectively, compared to gentamicin monotherapy (Fig. 8G). Plate counting in mouse livers, spleens, and kidneys showed that bacterial numbers were reduced by 241.9-, 268.8-, and 270-fold in mice infected with MDR-ECO41, 73.3-, 118.6-, and 43.1-fold in mice infected with MDR-KPN68, 23.5-, 31.3-, and 12.6-fold in mice infected with CR-PAE17, and 42.1-, 50.1-, and 12.2-fold in mice infected with MRSA16 (Fig. 8H). These results indicate that the pyruvate–cysteine–glutathione system/glycine–ROS metabolic pathway also confounds antibiotic sensitivity/resistance in these pathogens.

All bacterial strains came from the collection of our laboratory, where LTB4_-8MIC and LTB4_-16MIC originated from LTB4 that were sequentially propagated in tryptic soy broth (TSB) medium with ¹/₂ MIC of gentamicin. E. tarda strains were cultured in fresh TSB medium at 30 °C, with shaking at 200 rpm for 24 h, while the other bacteria were cultured in fresh Luria–Bertani medium (LB medium) at 37 °C, with shaking at 200 rpm for 16 h. The overnight bacterial cultures were collected, washed twice with saline, and resuspended in an M9 medium (containing 17.1 g/l Na₂HPO₄·12H₂O, 3 g/l KH₂HPO₄, 1 g/l NH₄Cl, 0.5 g/l NaCl) to an optical density at 600 nm (OD₆₀₀) of 0.2 when metabolites and/or antibiotics were added if desired; 10 mM sodium acetate, 2 mM 7H₂O·MgSO₄, and 0.1 M CaCl₂ were supplemented. Metabolites were added if desired. The bacterial cells were then cultured at 37 or 30 °C with shaking at 200 rpm for 6 h, excluding the incubation period experiment.

MICs were tested using Clinical and Laboratory Standards Institute (CLSI) methods. Overnight LB cultures were diluted and grown to OD₆₀₀ = 0.5, and then 10 μl of 5 × 10⁶ colony-forming units (CFU)/ml bacteria was mixed with 100 μl of diluted antibiotics in a 96-well plate and incubated for 16 h at 37 °C. The MIC was the lowest concentration preventing growth.

Sample preparation for the metabolomes of PPD200/87 and PPD200/87 plus pyruvate was carried out as previously described [60]. Bacterial cultures were grown in TSB medium at 30 °C and 200 rpm for 24 h, diluted 1:100 into fresh TSB, and cultured until OD₆₀₀ reached 1.0. The cells were harvested, washed with sterile saline, and quenched with precooled methanol. After sonication (650 W, 35% output, 2 s on, 3 s off) and addition of ribitol as an internal standard, the mixture was centrifuged to obtain supernatants, which were dried under vacuum at 37 °C. The dried samples were then derivatized with methoximation–pyridine hydrochloride for 3 h at 37 °C, followed by N-methyl-N-trimethylsilyltrifluoroacetamide (MSTFA) treatment for 30 min at the same temperature.

GC-MS was performed in Agilent 7890A GC equipped with an Agilent 5975C VL MSD detector (Agilent Technologies). The sample of 1 μl was injected into a 30 m × 250 μm internal diameter (i.d.) × 0.25 μm DB-1ms GC column. Initial temperature of the GC oven was kept at 85 °C for 5 min followed by an increase to 270 °C at a rate of 15 °C/min and then held for 5 min. Helium was used as the carrier gas, and flow was kept constant at 1 ml/min. The MS was operated in a range of 50 to 600 m/z. Each sample was analyzed in 4 biological repeats with 2 technical replicas. Identification of metabolites corresponding to the chromatographic peaks in the GC-MS analysis was performed using the National Institute of Standards and Technology (NIST) Mass Spectral Library. The data were normalized by applying total amount correction and generating standardized datasets consisting of metabolite information, retention times, and peak areas, which were further processed for metabolomics analysis. Significant differences in the standardized data were calculated and selected using IBM SPSS Statistics 19 software, considering a threshold of P < 0.01. Cluster analysis was performed using R software (R×64 3.6.1). The normalized areas of differential metabolites were analyzed using z score. Principal components and S-plot analyses were carried out using SIMCA-P+12.0 software (version 12; Umetrics, Umea, Sweden), while metabolic pathway analysis was performed using MetaboAnalyst 4.0 enrichment. Figures were created using GraphPad Prism 8.0 and Adobe Illustrator CS6.

Bacterial collection was performed according to the method described in metabolomic profiling. Total RNA was extracted from 3 ml of bacterial cultures using the TRIzol Reagent Kit (Thermo Fisher Scientific, MA, USA), as previously reported [61].

To investigate the effect of gentamicin and pyruvate on gene expression levels, qRT-PCR was conducted following previously described methods [62]. Primers used in this study are listed in Table S1.

Bacterial cells were cultured in M9 medium with/without metabolites at 30 °C and 200 rpm for 6 h. After the incubation, cells were centrifuged at 8,000g for 5 min, washed with phosphate-buffered saline (PBS), and adjusted to OD₆₀₀ = 1.0. The cells were sonicated in PBS for 10 min at 650 W × 35% power (2 s on, 3 s off) and centrifuged at 12,000g and 4 °C for 10 min to collect supernatants. Protein quantification was conducted using the BCA Protein Assay Kit. Enzyme activities were measured using specific assay kits (GOT: Solarbio BC1565, China, CGL: HalingBio HL50550.1v.A, China, GPX: Solarbio BC1195, China, and GR: Solarbio BC 1165, China) and colorimetric readings at designated wavelengths.

In accordance with previous reports [12], antibiotic killing assays involved centrifuging samples at 8,000 rpm for 3 min, washing with 50 ml of sterile saline, and resuspending in M9 medium with supplements to OD₆₀₀ = 0.2. Antibiotics were added, and samples were incubated at 30 °C for 6 h. Survival was assessed by periodic sampling, serial dilution, and CFU counting on TSB plates after 24 h at 30 °C. Survival rates were calculated relative to controls, with at least 3 biological replicates.

The biofilm cultivation experiment was conducted according to previously published methods [12]. In brief, PE50 catheters measuring 6 mm in length and 0.58 mm × 0.96 mm in dimeter were utilized. The catheters were inoculated with 1 ml of LB medium containing 10 μl of overnight cultures and incubated aerobically at 37 °C for 24 h. The medium was replaced by fresh LB every 24 h for 3 d. To remove loosely adherent cells, the PE50 catheters were washed 5 times with 1 ml of saline.

The persister cultivation experiment was conducted according to a previously published method [8]. PPD200/87 was cultured overnight to saturation, then washed with saline, and centrifuged at 8,000 rpm for 5 min. The samples were resuspended in TSB to OD₆₀₀ = 1.0 and exposed to escalating ofloxacin concentrations for 4 h to identify persisters. These persisters were then washed, resuspended in M9 to OD₆₀₀ = 0.2, and treated with gentamicin and sodium pyruvate. After 6 h at 30 °C and 200 rpm, colonies were counted as before.

The experiment was performed as previously described [63]. Balb/c mice (18 to 20 g) from Sun Yat-sen University's Animal Center were acclimated for a week before use in pathogen-killing and survival studies. Mice were infected intraperitoneally with multidrug-resistant strains and treated intramuscularly 1 h later with saline (group 1), sodium pyruvate (group 2), gentamicin (group 3), or both (group 4). Organs were harvested at 6 h for CFU/g analysis. For survival tests, 20 mice per group were infected and monitored for 7 d.

Three-centimeter-long, 0.4-g tilapia (Oreochromis mossambicus) from a Guangzhou breeder, confirmed infection-free, were acclimated for 2 weeks at 25 °C in a recirculating aquarium system with a basal diet. Eighty tilapia were split into 4 groups of 20 for intraperitoneal bacterial infection. One hour later, they were injected with 5 μl of saline (control), 30 μg of gentamicin, 90 μg of sodium pyruvate, or both. Survival was tracked for 7 d.

The GSSG/GSH Quantification Kit II (DOJINDO, Japan) was used for these experiments, where bacteria were cultured in M9 medium with various treatments at 30 °C for 6 h, adjusted to OD₆₀₀ = 1.0, and sonicated. Protein content was measured with a BCA kit (Beyotime, China), and absorbance at 405 nm was read using a microplate reader (BioTek Instruments, USA), with data normalized to protein levels.

Bacteria in M9 with additives were cultured at 30 °C for 6 h, mixed with 2′,7′-dichlorofluorescin diacetate (Sigma), and incubated at 37 °C for 30 min. Fluorescence was read at 485 nm/535 nm on a Victor X5 plate reader.

NADPH level was determined using a NADP⁺/NADPH Assay Kit with WST-8 (Beyotime, S0179, China) following the manufacturer's instructions.

Bacteria at OD₆₀₀ = 0.2 were cultured with pyruvate/H₂O₂ and gentamicin at 30 °C for 6 h, washed with PBS, and resuspended to OD₆₀₀ = 1.0. Gentamicin concentration was measured using a Gentamicin ELISA Kit (Shanghai Jianglai Industrial Limited By Share Ltd., JL17568, China) after sonication.

A gene deletion mutant was constructed from E. tarda EIB202 via sacB-based allelic exchange [54]. Briefly, using the genome of E. tarda EIB202 as the template, primers were set at about 500 base pairs (bp) up- and downstream of the gene to be knocked out and at appropriate positions within the gene (Table S3). A DNA fragment of ~1,000 bp in length with no target gene was obtained following 2 PCRs. This DNA fragment was seamlessly cloned into the linearized pDS132 vector, and the product was transferred into E. coli MC1061 to screen for positive clones (survival in kanamycin containing medium). The recombinant plasmid was further transformed into E. coli MFD-λpir (survived in medium containing 2,6-diaminopeptic acid) and screened for positive clones. All positive clones were verified by PCR (primers as above). E. coli MFD-λpir and E. tarda EIB202 were mixed at a ratio of 4:1 and cultured on TSB solid medium overnight at 30 °C for conjugation transfer. Then, the bacteria were washed with sterile saline and the mixture was collected, which was coated on TSB solid medium (containing 50 μg/ml kanamycin, 30 °C, 24 h). The positive clones were selected from the plate and cultured in 1ml of fresh TSB medium at 30 °C and 200 rpm. After about 24 h, the bacterial solution was coated on TSB solid medium (containing 12.5% w/v sucrose, 30 °C, 36 h). Then, the colonies without kanamycin resistance were selected from the plates and further verified by PCR and sequencing. Primers used to construct deletion mutants are listed in Table S2.

Cloning of GR gene and expression and purification of GR recombinant proteins were performed using a routine procedure as previously described [64]. Primers were designed for PCR amplification (Table S3). pET-32a and E. coli BL21 were used as expressional vector and cells.

ITC (Nano-ITC, TA Instruments, USA) was used to detect the thermodynamic parameters between GR with glycine at 25 °C. The detection procedure was carried out as the same as previously described [65].

Animal work was conducted in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Institutional Animal Care and Use Committee of Sun Yat-sen University (approval no. SYSU-IACUC-2020-B126716).