Metabolism-dependent succinylation governs resource allocation for antibiotic resistance

Abstract

The mechanisms that organisms allocate resources to sustain biological phenotypes remain largely unknown. Here, we use mobilized colistin resistance (mcr-1), which modifies lipopolysaccharide (LPS) to confer colistin resistance, as a model to explore how bacteria reallocate resources to support mcr-1–mediated resistance. We show that bacteria redirect resources from glycolysis, the pyruvate cycle, and LPS biosynthesis toward glycerophospholipid metabolism to produce phosphatidylethanolamine, the substrate for mcr-1 to modify LPS, while reducing LPS content to limit colistin binding. This reallocation down-regulates succinyl–coenzyme A (CoA) to diminish succinylation of proteins including triosephosphate isomerase (TPI), CpxR, and PdhR, thereby sustaining resistance. Exogenous succinate or α-ketoglutarate restores succinylation in a succinyl-CoA–dependent manner. Succinylation of TPI redirects metabolic flux to glycolysis and the pyruvate cycle, while succinylation of CpxR and PdhR up-regulates LPS biosynthesis, ultimately attenuating colistin resistance. Thus, we reveal a previously unrecognized mechanism by which bacteria regulate resource allocation through metabolism-driven posttranslational protein modification, offering strategies to combat antibiotic resistance.

Antibiotic resistance gene expression drives resource allocation, which can be reversed via metabolism-dependent succinylation.

INTRODUCTION

Biological phenotypes are supported by biological functions that are associated with resource cost (1, 2). Consequently, cells must optimize resource allocation to meet the demand for appropriate cellular functions that ultimately contribute to the phenotype (2). Antibiotic resistance, for example, is facilitated by the expression of resistance-conferring genes, which can originate from genetic mutations (either in antibiotic-related genes or their target sites) or be encoded on plasmids (3). While it is widely recognized that antibiotic-resistant genes are the principal drivers of the resistant phenotype, emerging evidence suggests that bacterial metabolism also plays a crucial role in linking gene products to resistance (4). Acquisition of antibiotic-resistant genes is always accompanied by metabolome reprogramming that allows the reallocation of resources. This is a pivotal step in establishing the resistant phenotype, as reverting the altered metabolism to its normal state and inhibiting the altered metabolic pathways without altering the antibiotic-resistant genes can reverse antibiotic resistance, presenting an alternative strategy for combating antibiotic resistance (5, 6). Therefore, it is a fundamental question of how bacteria modulate resource allocation in response to antibiotic-resistant gene expression, especially the connection between antibiotic-resistant gene expression, metabolic changes, and the resistant phenotype, which is largely unknown.

Various types of protein posttranslational modification (PTM) such as phosphorylation, acetylation, succinylation, pupylation, ubiquitin-like modification, glycosylation, lipidation, and crotonylation are identified in bacteria that regulate different biological processes including antibiotic persistence, antibiotic resistance, virulence, cell cycle, metabolism, dormancy, and sporulation (7, 8). Among these, certain PTMs have well-characterized modifying enzymes, and their roles in regulating biological phenotypes are well studied. YjgM, for example, a recently identified crotonyltransferase, modifies PmrA to enhance its binding affinity to eptA promoter, thereby increasing polymixin resistance (8). Similarly, HipA phosphorylates glutamate-tRNA ligase, causing a halt of translation, and ultimately antibiotic tolerance (9, 10). However, while PTMs such as succinylation are well known to regulate cellular metabolism, the specific succinyltransferase and desuccinylases involved remain unidentified in bacteria. Thus, how succinylation is controlled and its contribution to biological phenotypes is less explored in bacteria.

Mobilized colistin resistance (mcr-1), a plasmid-encoded antibiotic-resistant gene, confers resistance to colistin (CST), an antibiotic historically widely used in aquaculture and agriculture (11). The protein encoded by mcr-1 belongs to the phosphoethanolamine (PEtN) transferase family, which catalyzes the modification of lipid A with PEtN. This modification reduces the negative surface charge of the bacterial membrane, thereby preventing CST binding (12). However, overexpression of mcr-1 is toxic to bacteria, leading to membrane deformation and impaired growth (13). In addition, mcr-1 expression causes global metabolic change but the mechanism is unknown (14). Hence, mcr-1 represents an excellent model to explore how the expression of an antibiotic-resistant gene can trigger global metabolic change that contributes to the resistant phenotype. Unraveling the mechanisms behind this process could provide a fresh perspective on how bacteria reallocate resources to support resistance, from which to develop strategies to combat antibiotic resistance.

RESULT

Bacteria adopt metabolic strategies for resource reallocation in response to mcr-1 expression

The phenotype of mcr-1 expression is bacterial resistance to CST. To investigate how bacteria adjust resource reallocation to support mcr-1–mediated resistance, we cloned the mcr-1 gene from a plasmid isolated from Enterococcus faecalis (obtained from diseased fish) into a low-copy plasmid, pACYC184, generating pACYC184–mcr-1, which constitutively expresses MCR-1. As a control, a point mutation was introduced to disrupt MCR-1 catalytic activity by replacing glutamic acid (E) with alanine (A) at amino acid residue 246 (15). Both plasmids were individually transformed into Escherichia coli K12 BW25113, resulting in three strains: the wild-type– (Ec-K12), mcr-1–expressing strain (Ec-MCR-1), and the catalytically inactive mutant strain (Ec-E246A). The minimum inhibitory concentration (MIC) of CST for Ec-E246A was identical to that of Ec-K12 (0.5 μg/ml), confirming the loss of resistance in the mutant. In contrast, Ec-MCR-1 exhibited an eightfold increase in MIC (4 μg/ml; fig. S1A), consistent with levels observed in most clinical isolates (16).

Resource allocation was investigated at the metabolome level. Ec-MCR-1 and Ec-E246A were analyzed at three growth phases: exponential phase [i.e., at an optical density at 600 nm (OD₆₀₀ = 0.6)], late-exponential phase (OD₆₀₀ = 1.0), and stationary phase (OD₆₀₀ = 2.0). For each phase, four biological replicates were collected, with two technical replicates per sample, generating 48 data points. A total of 83 metabolites were identified in each sample. Metabolite profiles were visualized as a heatmap (fig. S1B). Ec-MCR-1 and Ec-E246A were clustered by growth phases, but the two strains were separately grouped (fig. S1B). Using a Mann-Whitney test, 69 metabolites were identified as differentially expressed in Ec-MCR-1 across all time points, with a false discovery rate of 0.20% (Fig. 1A and fig. S1, C to E). These metabolites were associated with 16 enriched pathways (Fig. 1B; P < 0.05). Metabolites of the tricarboxylic acid cycle (TCA cycle), which is part of the pyruvate cycle (17), and butanoate metabolism were shared across all three growth phases. Most metabolites in the two pathways were down-regulated in Ec-MCR-1 as compared to Ec-E246A (Fig. 1C and fig. S1F). In contrast, metabolites of sulfur metabolism, glutathione metabolism, pentose phosphate pathway (PPP), and fatty acid biosynthesis were up-regulated in Ec-MCR-1 (fig. S1F).

Fig. 1. Bacteria adopt metabolic strategies for resource reallocation in response to mcr-1 expression.

(A) Heatmap illustrating the relative abundances of differentially expressed metabolites between Ec-MCR-1 and Ec-E246A across the three growth phases. Yellow and blue indicate up- and down-regulated metabolites, respectively. (B) Pathway enrichment analyses of differentially expressed metabolites. Impacted pathways were identified by two criteria: P ≤ 0.05 (indicated by yellow bars) and a positive impact value (indicated by gray bars). The bottom x axis represents P value, the y axis lists the enriched pathways, and the top x axis shows the impact value. (C) Relative abundance of metabolites of pyruvate metabolism in the Ec-MCR-1 and Ec-E246A strain. Metabolites with increased and decreased abundance in Ec-MCR-1 relative to Ec-E246A are highlighted in yellow and blue, respectively. (D) Orthogonal partial least squares discriminant analysis (OPLS-DA). Each dot represents an individual sample. 0.6-E246A, 1.0-E246A, and 2.0-E246A refer to the Ec-E246A strain cultured at OD₆₀₀ of 0.6, 1.0, and 2.0, respectively. Similarly, 0.6-MCR-1, 1.0-MCR-1, and 2.0-MCR-1 refer to the Ec-MCR-1 strain cultured at OD₆₀₀ of 0.6, 1.0, and 2.0, respectively. (E and F) Scatter plots of succinate (E) and α-ketoglutarate (α-KG) (F) at different growth phases. The numbers in the x axis indicate the OD₆₀₀ values. Numbers in the y axis refer to the area quantified in gas chromatography–MS (GC-MS). (G) Enzymatic activity of pyruvate dehydrogenase (PDH), OGDH, succinate dehydrogenase (SDH), and malate dehydrogenase (MDH) in Ec-MCR-1 and Ec-E246A (n = 4). Specific enzyme activity was measured in whole-cell lysates and normalized to total protein abundance. n denotes biological replicates. prot, protein. Data are presented as mean ± SD from biological replicates. Statistical significance was determined using a Student’s t test.

Furthermore, orthogonal partial least squares discriminant analysis (OPLS-DA) and principal components analysis (PCA) were used to analyze sample patterns that allow for biomarker identification. Both methods distinguished Ec-MCR-1 from Ec-E246A at each growth phase (Fig. 1D and fig. S1G). In OPLS-DA analysis, the covariance and correlation between metabolites and class designation were visualized using S-plots (fig. S1, H to J). Five biomarkers were identified including succinate and α-ketoglutarate (α-KG) from the pyruvate cycle and octadecanoic acid from fatty acid biosynthesis (fig. S1, K and M). The abundance of both succinate and α-KG was lower in Ec-MCR-1 than Ec-E246A (Fig. 1, E and F). Down-regulation of the pyruvate cycle was confirmed with specific enzyme activity assays for pyruvate dehydrogenase (PDH), α-KG dehydrogenase (OGDH), succinate dehydrogenase (SDH), and malate dehydrogenase (MDH), all of which were down-regulated in Ec-MCR-1 but not in Ec-E246A (Fig. 1G). Collectively, the pathway enrichment and OPLS-DA analyses indicate that bacteria reallocate resources from the pyruvate cycle to other pathways, potentially to support CST resistance.

Bacteria reallocate resources to glycerophospholipid metabolism to support CST resistance

On the basis of metabolomic analysis revealing increased fatty acid levels, we hypothesized that bacteria may redirect resources from the pyruvate cycle toward fatty acid synthesis, which generates phosphatidylethanolamine (PE), the substrate for MCR-1 to modify lipopolysaccharide (LPS). To test this possibility, we inhibited MCR-1 activity by treating Ec-MCR-1 cells with ethanolamine (EA), a competitive inhibitor that blocks PEtN binding to MCR-1 (fig. S2A) (18). Exogenous EA restored CST sensitivity in Ec-MCR-1 in a dose-dependent manner (Fig. 2A). In contrast, exogenous PE, an alternative source of PEtN, exhibited a dose-dependent protective effect against CST killing in Ec-MCR-1 (Fig. 2B). Fractional inhibitory concentration index (FICI) further revealed that EA could be partially synergistic with CST (FICI = 0.625), while PE showed antagonism (FICI > 4) (table S1). Consistently, EA did not alter the MIC of Ec-MCR-1, while PE increased the MIC from 4 to 32 μg/ml (table S1). In addition, MCR-1–overexpressing cells showed slower proliferation compared to wild-type cells (13), suggesting that MCR-1 hyperactivation may be detrimental. Thus, we introduced the l-arabinose–inducible expression plasmid, pJN105-MCR-1, into Ec-K12. Induction with 1.25 and 10 mM l-arabinose increased the MICs of CST from 0.5 to 4 and 8 μg/ml, respectively (fig. S2B). Growth curves demonstrated that cells induced with 10 mM l-arabinose proliferated much slower than untreated cells and those induced with 1.25 mM l-arabinose (fig. S2C). However, exogenous EA rescued this growth defect (Fig. 2C). Moreover, EA, but not PE, restored the specific enzymatic activities of PDH, OGDH, SDH, and MDH in the pyruvate cycle in the Ec-MCR-1 strain (Fig. 2D). Together, these data suggest that MCR-1 reduces pyruvate cycle activity by its catalytic function.

Fig. 2. Bacteria allocate resources to glycerophospholipid metabolism to support CST resistance.

(A) Percent survival of Ec-MCR-1 cells in the presence of EA (0, 0.5, 1, and 2 mM) and CST (2 μg/ml) (n = 4) for 6 hours. Con, Control. (B) Percent survival of Ec-MCR-1 cells in the presence of EA (0 and 2 mM) and indicated concentrations of CST for 6 hours (n = 4). (C) Colony-forming units (CFUs) of Ec-MCR-1 cells in the presence of EA (1 mM) or l-arabinose (Ara) (10 mM) or both at indicated incubation times (n = 3). (D) Specific enzyme activities of PDH, OGDH, SDH, and MDH in Ec-MCR-1 cells in the presence of EA or PE (n = 4). (E) Schematic of glycolysis and glycerophospholipid metabolism connection. (F) mRNA levels of glycerophospholipid metabolism genes in Ec-K12, Ec-MCR-1, and Ec-E246A (n = 5). (G) Percent survival of Ec-MCR-1 and Ec-MCR-1::ΔpssA in the presence of PE (2 mM) and/or CST (4 μg/ml) for 6 hours (n = 4). (H) PDH, OGDH, SDH, and MDH activities in the indicated bacterial strains (n = 4). (I) Percent survival of Ec-MCR-1 cells in the presence of the indicated concentrations of Barlox 12 and CST for 6 hours (n = 4). (J) Glyceraldehyde-3-phosphate (G3P) and dihydroxyacetone phosphate (DHAP) abundance in Ec-MCR-1 and Ec-E246A cells (n = 4). (K) Specific enzyme activities of 6-phosphofructokinase (PFK) and pyruvate kinase (PK) in Ec-MCR-1 and Ec-E246A (n = 4). (L) Percent survival of indicated mutant bacteria in the presence of CST (2 μg/ml) (n = 4). Specific enzyme activity was measured in whole-cell lysates and normalized to total protein abundance [(D), (H), and (K)]. Percent survival was calculated by dividing the CFU of the treated group by that of the nontreated group. Data are presented as mean ± SD [(A) to (D) and (F) to (L)]. Statistical significance was assessed using Student’s t test, with a P value threshold of 0.05 (P < 0.05) considered statistically significant. n denotes biological replicates.

PE deficiency impairs bacterial cytokinesis, stress response, membrane protein folding, motility, and chemotaxis (19–23). It is highly possible that bacteria up-regulate PE biosynthesis upon MCR-1 expression to compensate for the PE consumed by active MCR-1, thereby maintaining normal growth. Gene expression analysis showed that specific genes involved in glycerophospholipid metabolism (i.e., cdsA, ynbB, pssA, and psd) were up-regulated in the Ec-MCR-1 strain but not in the Ec-E246A strain (Fig. 2, E and F, and fig. S2D). Because the genes of glycerophospholipid metabolism, from gpsA to psd, are essential except pssA, which encodes phosphatidylserine synthase (24), the susceptibility of ΔpssA::MCR-1 exhibited CST sensitivity comparable to that of Ec-E246A but was abolished by exogenous PE (Fig. 2G). Furthermore, the previously observed reductions in specific enzymatic activities of PDH, OGDH, SDH, and MDH were restored in the ΔpssA::MCR-1 strain (Fig. 2H). In addition, Barlox 12, an amine oxide inhibiting phosphatidylserine decarboxylase proenzyme (encoded by psd) (25), increased Ec-MCR-1 susceptibility to CST (Fig. 2I). Barlox 12 was synergistic with CST with an FICI value of 0.375 and reduced MIC by eightfold (table S1). Moreover, overexpression of glycerophospholipid metabolism genes (i.e., plsC, csdA, ynbB, pssA, and psd) in Ec-MCR-1 cells reduced bacterial sensitivity to CST (2 μg/ml), even at a higher concentration (8 μg/ml; fig. S2E). These data indicate that bacteria up-regulate PE biosynthesis to support MCR-1–mediated CST resistance.

Glycerophospholipid metabolism is a branch of glycolysis, while the pyruvate cycle operates downstream of glycolysis. The down-regulated pyruvate cycle and up-regulated glycerophospholipid metabolism suggest that bacteria reallocate resources from the pyruvate cycle to glycerophospholipid metabolism to support the resistance. Specifically, glycerophospholipid metabolism is connected to glycolysis via dihydroxyacetone phosphate (DHAP), which is interconverted to glyceraldehyde-3-phosphate (G3P). G3P, in turn, regulates metabolic flux into the pyruvate cycle (Fig. 2E). We observed a decrease in G3P abundance and an increase in DHAP abundance in Ec-MCR-1 cells as compared to Ec-E246A cells (Fig. 2J). These metabolic changes were further supported by increased specific enzymatic activity of adenosine 5′-triphosphate–dependent 6-phosphofructokinase (PFK) isozyme 1, which catalyzes the phosphorylation of d-fructose 6-phosphate to fructose 1,6-bisphosphate, and reduced activity of pyruvate kinase (PK), which catalyzes pyruvate formation, in Ec-MCR-1 cells (Fig. 2K). These findings confirm that metabolic flux is redirected toward glycerophospholipid metabolism but does not proceed through glycolysis.

Furthermore, genetic ablation of the glucose metabolism genes, ptsG and pgi, which mediate glucose catabolism, increased the sensitivity of Ec-MCR-1 to CST (fig. S2F). In contrast, Ec-MCR-1::ΔpfkA and Ec-MCR-1::ΔpfkB strains showed no change in CST sensitivity, likely due to compensatory mechanisms diverting metabolic flux into the PPP to generate G3P in the absence of pfkA or pfkB. As expected, the Ec-MCR-1::Δzwf strain, deficient in glucose-6-phosphate-1-dehydrogenase, disrupted the metabolic flux from glycolysis to PPP, although it had minimal effects on CST sensitivity. However, double knockout mutants, Ec-MCR-1::ΔpfkA Δzwf and Ec-MCR-1::ΔpfkB Δzwf, enhanced bacterial sensitivity to CST (Fig. 2L). Collectively, these results suggest that bacterial cells reallocate resources from glycolysis to glycerophospholipid metabolism upon MCR-1 expression to promote resistance.

Bacteria down-regulate LPS biosynthesis in response to mcr-1 expression

Phospholipid biosynthesis, including glycerophospholipids and LPS, shares a common precursor, β-hydroxymyristate–acyl carrier protein, enabling bacteria to balance lipid production (26). The up-regulation of phospholipid metabolism suggests a corresponding down-regulation of LPS biosynthesis. Given that LPS is the target of CST, we observed lower LPS abundance in Ec-MCR-1 compared to the Ec-E246A strain, while LPS levels were similar between Ec-K12 and Ec-E246A (Fig. 3A). LPS biosynthesis is tightly controlled by the biosynthetic pathway, including lipid A biosynthesis (lpxA, lpxC, lpxL, and lpxT), core oligosaccharide biosynthesis (waaA and waaO), and the LPS transporter system (lptG, lptB, and lptE) (27). Consistent with this, the expression of lpxA and lpxC was reduced in the Ec-MCR-1 strain (Fig. 3B and fig. S3A), indicating that LPS biosynthesis was down-regulated upon MCR-1 expression. To further address whether LPS abundance influences CST susceptibility, we treated bacteria with CHIR-090, a specific LpxC inhibitor (28, 29). CHIR-090 reduced LPS levels in a dose-dependent manner, which correspondingly decreased CST efficacy (Fig. 3, C and D), as confirmed by the checkerboard assay (FICI > 4) and increased the MIC of CST by fourfold (16 μg/ml) (table S1). Conversely, individual overexpression of LpxC and LpxA increased LPS abundance (fig. S3B) and enhanced CST efficacy (Fig. 3E).

Fig. 3. Bacteria down-regulate LPS biosynthesis in response to mcr-1 expression.

(A) LPS abundance in Ec-K12, Ec-MCR-1, and Ec-E246A strains (n = 4). Equal amount of bacterial cells was collected and used for LPS quantification. (B) Quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis of lpxA and lpxC expression in Ec-K12, Ec-MCR-1, and Ec-E246A strains (n = 5). (C) LPS abundance in Ec-MCR-1, which was treated with increasing concentrations of CHIR-090 (0, 0.25, 0.5, and 1 μg/liter) (n = 4). (D) Percent survival of Ec-MCR-1 treated with increasing concentrations of CHIR-090 under different concentrations of CST (0, 2, and 8 μg/ml) (n = 4). (E) Percent survival of Ec-MCR-1::lpxA or Ec-MCR-1::lpxC treated with increasing concentrations of isopropyl-β-d-thiogalactopyranoside (IPTG) (0, 10⁻⁴, 10⁻³, and 10⁻² mM) plus CST (2 μg/ml) (n = 4). (F) Quantification of the lpxA and lpxC expression in indicated mutant strains, control, ΔcpxR, and ΔpdhR (n = 5). Percent survival was calculated by dividing the CFU of the treated group by that of the nontreated group. Data are presented as mean ± SD. Statistical significance was assessed using Student’s t test, with a P value threshold of 0.05 (P < 0.05) considered statistically significant. n denotes biological replicates.

lpxA and lpxC are the most down-regulated genes in the Ec-MCR-1 strain. CpxR and PdhR are transcription factors that regulate lpxA and lpxC, respectively. Specifically, CpxR binds to the promoter region of lpxA at hlpAp (30), while PdhR binds to the promoter region, mraZp of the dcw gene cluster, where lpxC is located (31). We found that the expression of lpxA and lpxC was reduced in ΔcpxR and ΔpdhR mutants, respectively (Fig. 3F). These findings suggest that bacteria reallocate resources from the pyruvate cycle and LPS biosynthesis to support PE generation. This metabolic shift indicates that bacteria simultaneously enhance PE production and reduce LPS levels to facilitate MCR-1–mediated CST resistance.

Exogenous succinate restores resource allocation in Ec-MCR-1 cells

Previous studies have demonstrated that exogenous metabolites can reverse phenotypic traits, e.g., reversing the bacterial antibiotic resistance phenotype to a sensitive phenotype (4, 32). On the basis of this, we hypothesized that exogenous succinate or α-KG, the two biomarkers found at reduced levels in Ec-MCR-1, may restore the normal resource allocation. Using untargeted metabolomics, succinate reprogrammed the metabolome of Ec-MCR-1 cells, inducing a metabolic shift in Ec-MCR-1 (fig. S4A). This shift involved 60 differentially abundant metabolites, with 33 increasing and 27 decreasing in abundance (fig. S4, B and C). Specifically, exogenous succinate decreased fatty acid abundance while replenishing metabolites in the pyruvate cycle (Fig. 4A). Furthermore, exogenous succinate down-regulated the transcription of genes in glycerophospholipid metabolism in Ec-MCR-1 (gpsA, glpA, glpB, cdsA, ynbB, pssA, psd, cdh, and plsC) (Fig. 4B). These data together suggest that exogenous succinate partially restores resource allocation in Ec-MCR-1.

Fig. 4. Exogenous succinate restores resource allocation in Ec-MCR-1 cells.

(A) Heatmap showing relative abundances of differentially expressed metabolites in fatty acids and the pyruvate cycle in Ec-MCR-1 treated with or without succinate (20 mM) for 6 hours. Suc, succinate. (B) qRT-PCR analysis of glycerophospholipid metabolism genes in Ec-MCR-1 treated with or without 20 mM succinate for 6 hours (n = 5). (C) qRT-PCR analysis of LPS biosynthetic genes in Ec-MCR-1 treated with or without 20 mM succinate for 6 hours (n = 5). (D) LPS abundance in Ec-K12, Ec-MCR-1, and Ec-E246A treated with or without 20 mM succinate for 6 hours (n = 4). (E) Relative percentage of fluorescence-labeled PEtN (NBD-PEtN)–modified lipid A in Ec-MCR-1 treated with succinate (n = 4). RFU, relative fluorescence units. (F) Zeta potential of Ec-K12, Ec-MCR-1, and Ec-E246A in the presence of 20 mM succinate for 6 hours (n = 4). (G) CST binding to purified LPS from Ec-K12, Ec-MCR-1, and Ec-E246A treated with or without 20 mM succinate (n = 4). Percent survival was calculated by dividing the CFU of the treated group by that of the nontreated group. Data are presented as mean ± SD [(B) to (G)]. Statistical significance was assessed using Student’s t test, with a P value threshold of 0.05 (P < 0.05) considered statistically significant. n denotes biological replicates.

Meanwhile, exogenous succinate up-regulated gene expression for lipid A biosynthesis (lpxA, lpxC, lpxL, and lpxT), core oligosaccharide biosynthesis (waaA and waaO), and LPS transport system (lptG, lptB, and lptE), but not LPS modification genes (phoP, phoQ, pmrA, pmrB, and eptA) (Fig. 4C). Treatment with succinate increased LPS abundance by 1.52-fold, bringing it closer to levels observed in Ec-K12. This effect was also observed in Ec-K12 and Ec-E246A (Fig. 4D). The increased LPS was succinate concentration dependent in Ec-MCR-1 (fig. S4D). MCR-1 is known to mediate lipid A modification with PEtN (15), which can be detected by mass spectrometry (MS). PEtN modification of lipid A was only observed in Ec-MCR-1 but not in Ec-K12 or Ec-E246A (fig. S4E). However, exogenous succinate reduced the relative percentage of PEtN modification, as quantified by the relative intensity (fig. S4F). To further validate this, we developed a fluorescence-based assay leveraging the fact that MCR-1 uses PEtN as a substrate. Fluorescence-labeled PEtN (NBD-PEtN) was incorporated into LPS, serving as an indicator of modification levels. When NBD-PEtN was incubated with Ec-K12, Ec-MCR-1, or Ec-E246A, Ec-MCR-1 exhibited stronger fluorescence than Ec-K12 and Ec-E246A, confirming the assay’s validity (fig. S4G). In the presence of succinate, the amount of NBD-PEtN–modified LPS in Ec-MCR-1 decreased (Fig. 4E). CST binds to LPS via electrostatic interactions, but lipid A modification by MCR-1 reduces the negative charge, diminishing CST affinity (33). Surface charge changes were measured using zeta potential (34). Ec-K12 and Ec-E246A had similar levels of zeta potential (−20.39 ± 0.31 and −21.01 ± 0.86 mV, respectively), but Ec-MCR-1 showed a reduced negative charge (−3.05 ± 0.56 mV) (Fig. 4F). Exogenous succinate decreased the zeta potential of Ec-MCR-1 to −17.50 ± 0.86 mV, comparable to Ec-K12 and Ec-E246A (Fig. 4F). This change was succinate dose dependent (fig. S4H). The restored zeta potential correlated with increased LPS levels, facilitating CST binding. To directly assess the effect of succinate on LPS, we purified LPS from succinate-treated bacteria and measured CST binding. As expected, the amount of CST bound to LPS was increased (Fig. 4G). These data together suggest that succinate suppresses fatty acid biosynthesis, reduces glycerophospholipid metabolism, and enhances LPS biosynthesis. This reduces the relative percentage of PEtN modification, restores zeta potential, and increases CST binding.

Succinylation of CpxR and PdhR up-regulates LPS biosynthesis

To dissect the mechanism of how succinate promotes LPS biosynthesis, we traced succinate metabolism. Succinate is metabolized to fumarate via SDH, encoded by sdhA/B/C/D, and interconverted to succinyl–coenzyme A (CoA) via sucC and sucD and then to α-KG via sucA and sucB (Fig. 5A). We transformed pACYC184-mcr-1 into the deletion mutants of ΔsdhA, ΔsdhB, ΔsdhC, ΔsdhD, ΔsucA, ΔsucB, ΔsucC, and ΔsucD. Only ΔsucC and ΔsucD reduced LPS biosynthesis and increased zeta potential compared to controls and other mutants (fig. S5A), suggesting that succinate promotes LPS biosynthesis via SucC and SucD, which are critical for succinyl-CoA formation. Exogenous succinate increased succinyl-CoA levels in Ec-K12, Ec-MCR-1, and Ec-E246A (Fig. 5B), but this effect was abolished in ΔsucC and ΔsucD mutants (Fig. 5C).

Fig. 5. Succinylation of CpxR and PdhR up-regulates LPS biosynthesis.

(A) Succinate metabolism. (B and C) Abundance of succinyl-CoA in (B) Ec-K12, Ec-MCR-1, and Ec-E246A or (C) in indicated mutants transformed with mcr-1 treated with or without 20 mM succinate for 6 hours (n = 4). (D and E) (D) Western blot analysis of succinylation of CpxR and PdhR or (E) their point mutations treated with succinate (20 mM) for 6 hours. Proteins were purified using Flag-tag antibody. Positive bands were detected by rabbit anti–succinyl lysine antibody or rabbit anti-Flag antibody. (F) Fluorescence intensity of lpxA or lpxC promoters in Ec-K12, ΔcpxR, and ΔpdhR (n = 4). LpxA-Pro and LpxC-Pro refer to lpxA and lpxC promoters, respectively. WT, wild type. (G) Relative fluorescence intensity of lpxA or lpxC promoters in the ΔcpxR::cpxR, ΔcpxR::K188R, ΔcpxR::K218R, and ΔpdhR::pdhR, ΔpdhR::K61R treated with or without succinate for 6 hours (n = 4). (H) LPS abundance in indicated mutant bacteria treated with or without 20 mM succinate for 6 hours (n = 4). cpxR, K188R, and K219R refer to wild-type cpxR, cpxR K188R mutation, and cpxR K219R mutation expressed in ΔcpxR; pdhR and K61R refer to the wild-type pdhR and pdhR K61R mutation expressed in ΔpdhR. (I and J) (I) Zeta potential or (J) percent survival of indicated mutant bacteria treated with succinate (20 mM) for 6 hours (n = 4). (K) Western blot showing nonenzymatic succinylation of CpxR and PdhR. (L) Percent survival of Ec-MCR-1::CobB exposed to increasing concentrations of CST for 6 hours (n = 4). Percent survival was calculated by dividing the CFU of the treated group by that of the nontreated group. Data are presented as mean ± SD [(B) and (C), (F) to (J), and (L)]. Statistical significance was assessed using Student’s t test, with a P value threshold of 0.05 (P < 0.05) considered statistically significant. n denotes biological replicates.

Succinyl-CoA serves as a succinyl-group donor for protein succinylation (35). We hypothesized that succinate enhances LPS biosynthesis by succinylating the transcription factors CpxR and PdhR. Succinate increased global shift in succinylation in Ec-K12, Ec-MCR-1, and Ec-E246A (fig. S5B). To test whether CpxR and PdhR undergo succinylation, we ectopically expressed these two genes with a 3× Flag tag. Recombinant proteins were purified using anti-Flag antibody and probed with anti-Flag as input control or anti–succinyl lysine antibodies. Exogenous succinate increased succinylation of both CpxR and PdhR (Fig. 5D), indicating they are potential targets. Isocitrate dehydrogenase and glyceraldehyde-3-phosphate dehydrogenase served as positive controls for succinylation (fig. S5C), while SoxS was the negative control (36). Potential succinylation sites were predicted using the Compendium of Protein Lysine Modifications (CPLM) database (37). CpxR and PdhR were predicted to have succinylation sites at K61, K213, and K219, respectively (fig. S5D). MS confirmed these sites, with a 100.02 mass/charge ratio (m/z) shift corresponding to succinylation (fig. S5E). Then, a point mutation was introduced individually to these three sites, K61R, K213R, and K219R, which decreased succinylation of the two transcription factors upon succinate treatment (Fig. 5E). Transcriptional activity of CpxR and PdhR was assessed using a promoter activity assay. The promoter region for lpxA and lpxC was inserted into the promoterless plasmid, pMW82, which contains an unstable green fluorescent protein (38). Fluorescence intensity for the lpxA promoter decreased only in ΔcpxR, while the lpxC promoter decreased only in ΔpdhR (Fig. 5F). Succinate increased fluorescence intensity in ΔcpxR::cpxR but not in ΔcpxR::cpxR K188R and ΔcpxR::cpxR K219R (Fig. 5G). Similar results were observed for ΔpdhR::pdhR and ΔpdhR::pdhR K61R. Furthermore, deletion mutants failed to increase LPS biosynthesis when complemented with mutated CpxR or PdhR (Fig. 5H). Similar results were obtained for the zeta potential (Fig. 5I), indicating impaired surface charge restoration and reduced CST binding. Consistently, the mutants were resistant to the synergistic killing by CST and succinate (Fig. 5J). Thus, CpxR and PdhR were the targets of succinate in promoting LPS biosynthesis.

Succinylation can occur via succinyltransferase-dependent or -independent mechanisms (39). Because succinyltransferases are rarely reported in prokaryotes, we explored alternative mechanisms. KAT2A is the succinyltransferase in eukaryotes (40), but we failed to identify any ortholog in prokaryotes. Lysine (K) residues are common targets for succinylation, and GCN5-related N-acetyltransferases (GNATs), a family of lysine acetylases in E. coli, have been implicated in other posttranslational modifications (41, 42). We tested 12 potential GNATs including ypeA, yhbS, yhhK, yhhY, yjaM, ypfI, yfjQ, rffC, rimI, astA, argA, and atoB. Overexpression of argA, astA, rimI, yfjQ, yhhY, yhbS, and ypeA altered global succinylation patterns (fig. S5F), but none affected CpxR or PdhR succinylation (fig. S5G). This suggests that succinate-triggered succinylation of CpxR and PdhR may occur via nonenzymatic mechanisms, which are common in prokaryotes and eukaryotes (43, 44).

An in vitro succinylation assay confirmed that succinyl-CoA directly succinylates CpxR and PdhR (Fig. 5K). In addition, overexpression of CobB, a desuccinylase in E. coli (45), abolished succinate-enhanced CST killing (Fig. 5L). These data together suggest that succinate promotes nonenzymatic succinylation of CpxR and PdhR, up-regulating LPS biosynthesis.

Succinylation of triosephosphate isomerase redirects metabolic flow from glycerophospholipid metabolism to glycolysis

Previous reports showed that triosephosphate isomerase (TPI), encoded by tpiA, is a succinylated protein (36). TPI catalyzes the interconversion of G3P and DHAP, with DHAP serving as a key node connecting glycolysis and glycerophospholipid metabolism for PE biosynthesis. We first assessed the specific enzymatic activity of TPI in Ec-MCR-1. The specific enzymatic activity of TPI converting G3P to DHAP was increased, while conversion of DHAP to G3P was reduced compared to Ec-E246A (Fig. 6A). This indicates that MCR-1 expression favors DHAP production, consistent with the redirection of metabolic flux from glycolysis to glycerophospholipid metabolism. Deletion of tpiA increased sensitivity to CST, while overexpression of tpiA enhanced CST resistance in Ec-MCR-1 (Fig. 6B). TPI expression is negatively regulated by ribonuclease G, encoded by rng, and PK, encoded by pykF, at the mRNA and/or protein level (46, 47). Accordingly, Δrng and ΔpykF had increased TPI expression (46, 47) and were more resistant to CST (Fig. 6C). These findings support the hypothesis that MCR-1 redirects metabolic flux toward glycerophospholipid metabolism to confer CST resistance.

Fig. 6. Succinylation of TPI redirects metabolic flow from glycerophospholipid metabolism to glycolysis.

(A) Specific enzyme activity of TPI for the interconversion of G3P and DHAP in Ec-MCR-1 or Ec-E246A (n = 4). (B) Percent survival of MCR-1, MCR-1::ΔtpiA, and MCR-1::OE-ΔtpiA, which correspond to Ec-MCR-1 strain, tpiA deletion mutant, and tpiA deletion mutant overexpressing wild-type tpiA. The strains were treated with CST for 6 hours (n = 4). (C) Percent survival of deletion mutants MCR-1, MCR-1::ΔpykF, and MCR-1::Δrng treated with CST at 2 or 8 μg/ml (n = 4). (D) Specific enzyme activity of TPI for the interconversion of G3P and DHAP in Ec-MCR-1 after incubation with 20 mM succinate for 6 hours (n = 4). (E) Abundance of DHAP and G3P in Ec-MCR-1 following succinate treatment for 6 hours (n = 4). (F) Western blot analysis of TPI succinylation levels after incubation with 20 mM succinate for 6 hours. (G) Specific enzyme activity of TPI for the interconversion of G3P and DHAP in indicated bacterial strains (n = 4). (H) Percent survival of indicated bacterial strains under combined succinate and CST treatment (n = 4). (I) Specific enzyme activities of PDH, OGDH, SDH, and MDH in indicated bacterial strains (n = 4). Specific enzyme activity was measured in whole-cell lysates and normalized to total protein abundance [(A), (D), (G), and (I)]. Percent survival was calculated by dividing the CFU of the treated group by that of the nontreated group. Data are presented as mean ± SD [(A) to (E) and (G) to (I)]. Statistical significance was assessed using Student’s t test, with a P value threshold of 0.05 (P < 0.05) considered statistically significant. n denotes biological replicates.

Given the global impact of succinate on protein succinylation, we attempted to see whether the succinate-mediated succinylation alters TPI activity. Exogenous succinate restored the enzymatic activity that converts DHAP to G3P but decreased the conversion of G3P to DHAP (Fig. 6, D and E). MS confirmed succinylation of TPI at K188 (fig. S6A), consistent with predictions (fig. S6B). Succinylation of TPI was increased in succinate-treated Ec-MCR-1 but remained lower than in Ec-E246A (Fig. 6F). In Ec-MCR-1::TPI K188R, the enzymatic activity of G3P-to-DHAP conversion increased, while DHAP-to-G3P conversion remained unchanged (Fig. 6G). Consistently, Ec-MCR-1::TPI K188R had similar CST susceptibility to Ec-MCR-1 but was resistant to succinate-enhanced CST killing (Fig. 6H). Last, Ec-MCR-1::ΔtpiA showed higher pyruvate cycle activity than Ec-MCR-1 and Ec-MCR-1::tpiA-K188R (Fig. 6I). These data together suggest that succinate restores normal metabolic flow, counteracting MCR-1–induced redirection through succinylation of TPI.

Succinylation-dependent restoration of resource allocation ameliorates MCR-1–induced CST resistance

The above findings suggest that restoring succinylation may ameliorate MCR-1–induced CST resistance. Both succinate and α-KG increased global succinylation levels (figs. S5B and S7A) and enhanced bactericidal activity of CST in a concentration- and time-dependent manner (Fig. 7, A to D). To further validate these results, we tested mcr-1–positive strains from diverse sources, including animal husbandry, aquaculture, and human settings. These strains included E. coli (LD22, LD39, LD67, LD93, Gga4, Gga5, and Gga6), Klebsiella pneumoniae (KPN52, KPN54, KPN60, and KPN62), Pseudomonas aeruginosa (PA1 and PA2), and Edwardsiella tarda (ET20-MCR1), with CST MICs ranging from 4 to 160 μg/ml. Because of the variability in MICs, the CST dose was adjusted accordingly for each strain (table S2). The combination of succinate and CST effectively reversed MCR-1–mediated polymyxin resistance and reduced bacterial viability (Fig. 7E). Notably, we were unable to calculate the FICI value for succinate or α-KG in combination with CST, as these metabolites are nontoxic to bacteria (table S1). This is consistent with the behavior of metabolism-based bactericidal agents, which often function differently from conventional antibiotic potentiators, enzyme inhibitors, or active-site competitors. Their efficacy may vary depending on culture conditions and involve distinct mechanisms (6, 48).

Fig. 7. Succinylation-dependent restoration of resource allocation ameliorates MCR-1–induced CST resistance.

(A to D) Percent survival of Ec-MCR-1 treated with indicated concentrations of (A) succinate or (C) α-KG plus CST for 6 hours and (B) succinate (20 mM) or (D) α-KG (20 mM) plus CST (2 μg/ml) at indicated incubation time (n = 4). (E) Percent survival of mcr-1–positive colistin-resistant pathogens treated with succinate (20 mM) plus CST (n = 4). CST concentration used for each strain was summarized in table S2. (F and G) Survival rates of mice infected with E. coli LD67 and treated with saline, CST (10 mg/kg), (F) succinate (100 mg/kg) or (G) α-KG (100 mg/kg), or their combinations (n = 10 per group). Survival rates were monitored over 1 week. (H and I) CFUs in the liver, spleen, and kidney of mice treated with saline, CST (10 mg/kg), (H) succinate (100 mg/kg) or (I) α-KG (100 mg/kg), or their combinations (n = 6 per group). Tissues were collected and homogenized 24 hours posttreatment. (J and K) (J) Percent survival (n = 20 per group) or (K) CFU (n = 6 per group) of zebrafish infected with ET20-MCR1 and treated with saline, CST (10 mg/kg), succinate (100 mg/kg), or both. Survival rates were monitored over 1 week (J). Tissues were collected and homogenized at 24 or 48 hours posttreatment (K). Data are presented as mean ± SD [(A) to (E), (H), (I), and (K)]. Statistical significance was assessed using Student’s t test [(A) and (C)] or one-way analysis of variance (ANOVA) [(B), (D), (E), (H), (I), and (K)] or two-sided Mann-Whitney U test [(F), (G), and (J)], with a P value threshold of 0.05 (P < 0.05) considered statistically significant. n denotes biological replicates [(A) to (E)] or the number of animals [(F) to (K)].

When mice were challenged with E. coli LD67, an mcr-1–positive strain that is virulent to mice, the survival rate was 0% (fig. S7B). However, the combination of succinate (100 mg/kg) or α-KG (100 mg/kg) with CST (10 mg/kg) increased survival rates by 50%, representing a fivefold improvement compared to CST alone (Fig. 7, F and G). Correspondingly, bacterial colony-forming units (CFUs) in the liver, kidney, and spleen decreased by 48.2- to 59.1-fold (Fig. 7, H and I). We also evaluated the efficacy of this combination in a clinical isolate from the environment, specifically ET20-MCR-1, obtained from diseased fish. The addition of succinate to CST increased survival rates from 20 to 80%, while CFUs were reduced by 89.6- to 136.2-fold (Fig. 7, J and K).

To assess whether compensatory up-regulation of mcr-1 expression might occur during treatment, we isolated E. coli LD67 that survived succinate-CST treatment in mice. These bacteria were subjected to five reinfection cycles, with mcr-1 expression monitored at each stage. No changes in mcr-1 expression were observed (fig. S7C). Furthermore, bacteria from reinfected mice remained susceptible to CST in the presence of succinate (fig. S7D). Together, these findings demonstrate that succinate combined with CST ameliorates MCR-1–mediated resistance in vivo without increasing CST toxicity or resistance risk.

DISCUSSION

Organisms must finely tune resource allocation to enhance fitness, as imbalances can disrupt normal biological processes, leading to altered phenotypes or even harm to the host. A long-standing question in biology is how organisms optimize resource allocation in response to stress. Extensive research has explored how cells respond to environmental cues and regulate growth, fate, motility, and secretion through genome-scale modeling, whole-cell modeling, Pareto analysis, and cell circuit studies (49–54). These efforts have illuminated how resource allocation influences population-level behavior and adaptation to microenvironments. However, the molecular mechanisms underlying resource allocation are less explored at the molecular level.

In this study, we provide a system-level perspective on how resource allocation governs MCR-1–mediated CST resistance. We found that bacteria reroute metabolic flux from glycolysis and fatty acid synthesis to glycerophospholipid metabolism to produce sufficient PE for MCR-1 function. This shift reduces pyruvate cycle activity and LPS biogenesis [Fig. 8; which was created in the Generic Diagramming Platform (55)]. This resource reallocation helps bacteria cope with the stress imposed by MCR-1, as its high-level expression, requiring substantial PE, could otherwise be detrimental to the host (13). The diversion of metabolic flux toward glycerophospholipid metabolism to support PE production is consistent with prior studies showing PE accumulation in response to MCR-1 expression (56). This study, as well as others, has shown that mcr-1–expressing cells increase fatty acid biosynthesis (14, 56). Notably, inhibitors of fatty acid biosynthesis (e.g., cerulenin, triclosan, and Debio1452-NH3) and biotin metabolism (e.g., MAC13772) have been shown to synergize with CST, effectively countering both chromosomal- and plasmid-mediated CST resistance (57). Consistently, disruption of arginine degradation alters membrane lipid composition that reduces bacterial interaction with polymyxins in Acinetobacter baumannii (58). Beyond glycolysis, the pyruvate cycle, and fatty acid synthesis, we also observed metabolite accumulation in sulfur and glutathione metabolism. These pathways are critical for oxidative stress defense (59, 60), although their specific roles in MCR-1–induced resistance remain unclear. The alterations in these pathways suggest that MCR-1 expression causes more profound resource allocation perturbations than anticipated. Bacteria appear to use feedback mechanisms to stabilize cellular homeostasis, mitigating the stress induced by MCR-1.

Fig. 8. Proposed model.

Upon MCR-1 expression, bacteria reduce glycolysis and the pyruvate cycle while enhancing glycerophospholipid metabolism. This shift produces more PE, enabling MCR-1 to transfer PEtN to modify LPS. This modification reduces bacterial surface negative charge, blocking colistin binding. In addition, the reduced P cycle lowers succinyl-CoA levels, decreasing protein succinylation. This sustains glycolysis-to-glycerophospholipid flow and reduces LPS biosynthesis by lowering CpxR and PdhR succinylation, further decreasing colistin targets. OM, outer membrane; IM, inner membrane. This figure was created in Generic Diagramming Platform (55).

In this study, we also explore how metabolomic changes regulate protein posttranslational modifications (PTMs). For instance, a reduction in the pyruvate cycle decreases the levels of succinate and α-KG, which, in turn, lowers succinyl-CoA availability and affects protein succinylation. PTMs play a critical role in modulating protein activity, and extensive efforts have been dedicated to identifying additional modifications, their substrates, and the enzymes responsible for these modifications in bacteria (36, 39). While these studies have shed light on the metabolic consequences of PTMs, how bacterial metabolism influences PTMs remains poorly understood. Here, we establish a connection between bacterial metabolism, PTMs, and phenotypic outcomes. We demonstrate that key metabolic pathways not only support antibiotic resistance but also influence downstream PTMs, which ultimately redirect metabolic flux and drive resource reallocation.

We highlight two roles of succinylation in resource allocation. First, succinylation regulates LPS biosynthesis, providing insights into the posttranslational control of this critical bacterial molecule. Current knowledge of LPS biosynthesis is largely limited to the enzymatic pathways involving hundreds of proteins distributed across the three compartments of Gram-negative bacteria (61). We show that metabolic signals modulate LPS biosynthesis through the succinylation of key transcriptional regulators, CpxR and PdhR, which directly control the expression of LpxA and LpxC. These enzymes catalyze the first two steps of LPS biosynthesis (27, 62), with LpxC acting as the rate-limiting step of LPS biosynthesis (62). Reduced LpxC expression contributes to CST resistance (63, 64), and succinate supplementation restores LpxC levels, rescuing the mcr-1–induced decline in LPS abundance. Increased LPS alters bacterial surface charge, enhancing CST binding through electrostatic interactions. This is consistent with observations that mid-logarithmic phase bacteria, which carry a more negative surface charge, are more susceptible to CST than resistant strains (65). Second, succinylation of TPI acts as a metabolic switch, directing flux from glycolysis toward either glycerophospholipid metabolism or the pyruvate cycle. Although TPI succinylation was found a decade ago, its biological significance has remained unclear. Our findings reveal its role in metabolic regulation, further underscoring the broad impact of metabolome changes on phenotypic outcomes.

In summary, we demonstrate that bacteria use metabolic strategies to mitigate stress caused by the expression of exogenous antibiotic resistance genes, reallocating resources to support the MCR-1–induced resistance phenotype. We show that succinylation plays a central role in this process, balancing decreased glycerophospholipid metabolism with increased LPS biosynthesis, ultimately reducing CST resistance. These findings not only provide insights into the mechanisms of bacterial antibiotic resistance but also suggest potential targets for developing strategies to control bacterial infections.

MATERIALS AND METHODS

Ethic statement

This study was conducted in compliance with the Guide for the Care and Use of Laboratory Animals (National Institutes of Health). Animals were housed under standard protocols, maintained on a 12:12 light/dark cycle with ad libitum access to food and water. The mice were kept in a sterile, independent air supply isolation cage within a laminar airflow purification room, maintained at a constant temperature (26° to 28°C) and humidity (relative humidity 40 to 60%). All experimental procedures were approved by the Institutional Animal Care and Use Committee of Sun Yat-sen University (approval no. SYSU-IACUC-2020-B126716).

Bacterial strains and growth conditions

E. coli strains (LD22, LD39, LD67, and LD93) were gifts from Y. Liu, Yangzhou University (66). E. coli (gag1, gag2, and gag3) were isolated from diseased chickens. E. tarda ET20 was isolated from diseased fish. Clinical isolates of Pseudomonas aeruginosa and A. baumannii were obtained from patients at Zhongshan Hospital, Xiamen. All clinical isolates are listed in table S2. To propagate bacterial growth, a single colony was picked up and inoculated into LB medium, shaken at 37°C except for E. tarda that were grown in Tryptic Soy Broth (TSB) medium, and shaken at 30°C.

Plasmids and genetic manipulations

The wild-type mcr-1 gene and its mutant variant mcr-1(E246A) were cloned into the pACYC184 expression vector, generating the plasmids pACYC184::mcr-1 and pACYC184::mcr-1(E246A), respectively. Similarly, the E. coli genes cpxR and pdhR were cloned into pACYC184 with a 3× Flag tag to create recombinant plasmids. All constructs were verified by polymerase chain reaction (PCR) and direct DNA sequencing. A complete list of plasmids used in this study is provided in table S3.

Metabolomic profiling and gas chromatography analysis (gas chromatography–MS)

Metabolomic analysis was performed using gas chromatography–MS (GC-MS) as previously described (67). For untargeted metabolomic analysis comparing Ec-MCR-1 and Ec-E246A strains, overnight bacteria were diluted 1:100 in 50 ml of fresh LB medium and cultured at 37°C to an OD₆₀₀ of 1.0. Cells were quenched with 60% methanol (aqueous) at −80°C and collected by centrifugation at 8000g for 5 min. Metabolites were extracted using 1 ml of cold methanol containing 5 μl of ribitol (0.1 mg/ml) as an analytical internal standard.

To analyze the role of succinate in reprogramming the metabolome of Ec-MCR-1 cells, overnight cultures were harvested, washed, and resuspended in M9 minimal medium to an OD₆₀₀ of 0.2. Succinate was added to a final concentration of 20 mM, followed by incubation for 6 hours. After incubation, bacteria were collected, washed, and resuspended to an OD₆₀₀ of 0.2. A 10-ml aliquot of the bacterial suspension was rapidly quenched in prechilled methanol to halt metabolic activity.

Cells were lysed by sonication for 3 min at a 10-W power setting and were centrifuged at 12,000g for 10 min at 4°C. Supernatant was transferred to a 1.5-ml centrifuge tube and then dried in a vacuum concentrator system (Labconco) for GC-MS analysis.

GC-MS analysis was carried out using an Agilent 5975 GC-MSD system coupled to an Agilent 7890A GC (Agilent Technologies, USA). Before GC-MS analysis, samples were derived with a two-stage technique. Briefly, dried samples were suspended in 40 μl of methoxyamine hydrochloride (20 mg/ml in pyridine) and incubated at 37°C for 90 min. Then, 80 μl of N-methyl-N-trimethylsilyltrifluoroacetamide containing 1% trimethylchlorosilane was added and incubated at 37°C for 30 min. After incubation, 1 μl of the derivatized sample was injected into a DB-5MS capillary column (30 m by 0.25 mm by 0.25 μm) with helium as the carrier gas (1.1 ml/min). The GC oven temperature was kept at 60°C for 1 min, programmed to 300°C at a rate of 10°C/min, and maintained for 10 min. The injector temperature was set to 250°C, and mass spectra were recorded at 70 eV. The mass range was adjusted from 50 to 600 m/z.

Multivariate analysis and sample pattern recognition

Multivariate analysis was performed as described previously (68). Initial peak detection and mass spectral deconvolution were performed using Agilent Technologies MSD Productivity ChemStation software (E.02.02.1431; 2011). Metabolites were identified using spectral matching and retention indexes from the National Institute of Standards and Technology (NIST) library using the NIST MS search 2.0. The peak intensities were normalized to form a single matrix with Rt-m/z pairs (retention time–mass charge ratio pairs) for each file in the dataset. The data matrix was normalized to the internal standard (ribitol) and then to the total intensity. Normalized peak intensities formed a single matrix with Rt-m/z pairs for each file in the dataset. The matrix can be used for further analysis. According to a reference distribution, z-score analysis scaled each metabolite. Statistical difference was obtained by Kruskal-Wallis test and Mann-Whitney test using SPSS 26.0. P < 0.05 was considered significant. Hierarchical clustering was completed in the R platform (https://cran.r-project.org/) with the function “heatmap.2” of “gplots library.” Multivariate statistical analysis included PCA and OPLS-DA implemented with SIMCA 14.0 (Umetrics, Umeå, Sweden). Control scaling was selected before fitting. All variables were mean cantered and scaled to Pareto variance of each variable. PCA was used to reduce the high dimensionality of the dataset. We analyzed the differential metabolites to their respective biochemical pathways as outlined in the MetaboAnalyst 6.0 (www.metaboanalyst.ca/).

Minimal inhibitory concentration and checkboard assay

Minimal inhibitory concentration (MIC) was measured with the standard broth microdilution method (69). All antibiotics were twofold diluted in LB and equally mixed with bacterial suspensions in a 96-well microtiter plate (Corning, New York, USA). MIC values were defined as the lowest concentrations of drugs with no visible growth of bacteria after 18-hour incubation at 37°C.

Checkboard assay was conducted as previously described (70) in a clear, flat-bottom 96-well assay plate containing twofold dilutions of each compound in an 8 × 8 dose-point matrix. The plates were incubated at 37°C without shaking overnight (~18 hours), and then the absorbance at OD₆₀₀ was measured using a Tecan Infinite M1000 Pro plate reader. The synergistic effect was evaluated by calculating the FICI using the formula: FICI = MIC_AB/MIC_A + MIC_BA/MIC_B = FIC_A + FIC_B. MIC_A represents the MIC of compound A alone; MIC_AB denotes the MIC of compound A in the presence of compound B; MIC_B is the MIC of compound B alone; MIC_BA is the MIC of compound B in the presence of compound A; FIC_A and FIC_B represent the FIC of compounds A and B, respectively. FICI ≤ 0.5 indicates synergism, 0.5 < FICI ≤ 1 indicates an additive effect, 1 < FICI ≤ 4 indicates indifference, and FICI > 4 indicates antagonism.

In vitro bactericidal assay

A single colony was grown in 50 ml of LB broth in 250-ml flasks for 18 hours at 37°C, except that E. tarda were grown at 30°C. After centrifugation at ×8000 rpm for 3 min, bacteria were washed three times with 20 ml of sterile saline. E. coli (gag1, gag2, and gag3) were resuspended in M9 minimal medium supplemented with 10 mM acetate, 1 mM MgSO₄, and 100 μM CaCl₂; P. aeruginosa and A. baumannii were resuspended in M9 minimal medium supplemented with 10 mM acetate, 1 mM MgSO₄, and 100 μM CaCl₂; and E. coli (LD22, LD39, LD67, and LD93) were resuspended in 0.85% NaCl supplemented with 10 mM citrate, 1 mM MgSO₄, and 100 μM CaCl₂. CST and/or metabolite(s) were added and incubated at 37° or 30°C for 6 hours. After incubation, aliquot samples (100 μl) were serially diluted, and 5 μl of the diluted sample was plated onto LB agar plates. The plates were cultured at 37° or 30°C for 12 hours. Only those dilutions yielding 20 to 200 colonies were enumerated to calculate CFUs. Percent survival was determined by dividing the CFU obtained from a treated sample by the CFU obtained from the control.

Investigation of immune protection in a model of zebrafish

Zebrafish, Danio rerio, with an average length of 2.5 to 3 cm and an average weight of 0.2 ± 0.05 g, were purchased from the Guangzhou Zebrafish Breeding Base. These animals were acclimatized to laboratory conditions in fish tanks (80 cm by 75 cm by 90 cm) for 7 days before experiments. Zebrafish were randomly divided into control and experimental groups, with 20 fish in each group. Fish were challenged with 5 μl of 3 × 10⁸ CFUs. At 1-hour postinfection, fish were treated with a single dose of CST (10 mg/kg), succinate (100 mg/kg), α-KG (100 mg/kg), or combinations of CST plus metabolite through muscle injection. Survival rates of fish were recorded for 7 days. Alternatively, whole fish were collected, disinfected, and homogenized for plating at 12, 24, and 48 hours postinfection.

Mouse peritonitis-sepsis infection model

Six-week-old male Kunming (KM) mice were obtained from the Laboratory Animal Center of Sun Yat-sen University (Guangzhou, China) and were housed for 1 week before experiments (71, 72). Mice (n = 10 per group) were intraperitoneally infected with a dose of 3.0 × 10⁸ CFU E. coli LD67 and treated with CST (10 mg/kg), succinate (100 mg/kg), α-KG (100 mg/kg), or combinations of CST plus metabolite via intraperitoneal injection 1 hour postinfection. Survival rates of mice were monitored for at least 7 days. For mouse peritonitis-sepsis infection models, another group of mice was treated in parallel. However, they were euthanized by cervical dislocation at 24 hours postinfection. The liver, kidney, and spleen were aseptically removed, homogenized, and serially diluted for plating.

Quantification of LPS/succinyl-CoA

Bacteria were washed three times with 15 ml of 0.85% NaCl and resuspended in M9 minimal medium supplemented with 10 mM acetate, 1 mM MgSO₄, and 100 μM CaCl2 to an OD₆₀₀ of 0.2. Succinate or α-KG was added and incubated at 37°C for 6 hours. Bacterial cells were collected by centrifugation at 8000 rpm for 3 min at 4°C and resuspended in 1 ml of double-distilled water (ddH₂O). The precipitate was sonicated in 200 μl of ddH₂O as a pellet group. The amounts of LPS and succinyl-CoA were quantified using enzyme-linked immunosorbent assay (ELISA) test kits (MSKbio, Wuhan, China). LPS content was calculated on the basis of the OD value of trimethylboron catalyzed by antibody horseradish peroxidase (HRP) after binding with the LPS antibody (LPS ELISA test kit). Succinyl-CoA was detected after binding with the capture antibody (succinyl-CoA ELISA test kit). Final data were analyzed and processed according to the standard curve.

Quantification of LPS-bound CST

To quantify the LPS-bound CST, LPS was purified from bacteria as previously described (73). Excessive LPS (200 ng) was added to the LPS quantification ELISA kit, which recognizes the core polysaccharides of LPS, and incubated for 30 min to saturate the antibody with LPS. Then, 100 μl of CST (200 ng/liter) was added and incubated for an additional 2 hours. Unbound CST was quantified by a CST detection kit as previously described (74). LPS-bound CST was then calculated by subtracting the free CST from the total CST added.

Quantification of PEtN-LPS

To assess PEtN transfer by MCR-1, the NBD-PE [N-(7-Nitrobenz-2-oxa-1,3-diazol-4-yl)-1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine, triethylammonium salt; N131159] with NBD-PEtN modification was used to quantify PEtN-LPS. MCR-1 transfers NBD-PEtN from NBD-PE to lipid A in LPS on the cell membrane. Ec-K12 or Ec-MCR-1 cells treated with succinate were mixed with freshly prepared NBD-PE in ddH₂O to a final concentration of 5 μg/ml. Reactions were incubated at 37°C for 1 hour. Bacterial cells (10⁶ CFU/ml) were collected and then analyzed by flow cytometry (CytoFLEX, Beckman Coulter Ltd., USA) with a record count of 10,000 cells. Excitation and emission wavelengths of 490 and 517 nm, respectively, were used to detect NBD-PEtN fluorescence (75).

Measurement of specific enzymatic activities

The specific enzymatic activities of SDH, PDH, MDH, OGDH, PFK, PK, and triose phosphate isomerase (DHAP→G3P) were measured by colorimetric assay kits (Genmed Scientifics Inc.). Given that transcriptomic and proteomic analyses conducted by two independent research groups demonstrated that the expression levels of the encoding genes remained unchanged, we measured the specific enzyme activities in total cell lysates (56, 76). Briefly, Ec-MCR-1 cultures were collected, washed, and resuspended in lysis buffer. Cells were disrupted by sonication. After removing debris, supernatants were collected and processed for enzyme activity assay according to the manufacturer’s instructions. Specific enzyme activity was normalized to the total amount of proteins, which was determined by Bradford assay.

Measurement of the enzymatic activity of TPI (G3P→DHAP)

The enzymatic activity of TPI was measured using G3P as the substrate in the presence of the reduced form of nicotinamide adenine dinucleotide (oxidized form) (NADH) (β-NADH; extinction coefficient at 340 nm = 6.2 mM/cm) in a coupled enzyme assay (77). The reaction was initiated by the addition of 0.2 mM G3P to a reaction mixture containing 20 mM Hepes (pH 7.4), 5 U of α-glycerol-3-phosphate dehydrogenase, 0.1 mM β-NADH, and 25 to 100 μl of protein extract (50 μg of total protein). Production of β-NAD⁺ was monitored at 340 nm in a PerkinElmer LS 55 Fluorescence Spectrophotometer (PerkinElmer). Specific enzyme activity was normalized to the total amount of proteins, which was determined by Bradford assay.

Quantification of DHAP and G3P

The content of DHAP was measured according to the manufacturer’s protocol (Yuduo Biological Science and Technology Co. Ltd., Shanghai, China). Bacteria (1 × 10¹⁰ CFU) were collected, washed three times with 0.85% NaCl solution, and resuspended in assay buffer. Cells were disrupted by sonication for 3 min at a 10-W power setting on ice. After removing debris, supernatants were collected. The quantification of DHAP was achieved through a colorimetric assay. Specifically, DHAP was converted into hydrogen peroxide—the generated hydrogen peroxide reacted with 4-aminoantipyrine to form a red quinoneimine dye. The concentration of DHAP in the samples was then quantitatively determined using a spectrophotometer at a wavelength of 510 nm. The concentration was determined via a standard curve.

The content of G3P was measured using a colorimetric assay (78). Bacteria (1 × 10¹⁰ CFU) were collected, and lysis and collection of the supernatant were carried out according to the conditions mentioned above. Briefly, the assay measured the G3P release of NADH under the catalytic action of G3P enzyme (4 U/ml) in 50 mM Hepes buffer (pH 7.4). The increase in NADH concentration can be measured spectrophotometrically at 340 nm. The concentration of G3P was determined via standard curve.

Nonenzymatic succinylation and liquid chromatography–tandem MS analysis of succinylation site(s)

A total of 1 mg of CpxR, PdhR, and TPI (0.1 ml) was mixed with ¹/₁₀ volume of freshly prepared succinyl-CoA in ddH₂O to a final concentrations of 10 mM. Reactions were incubated at 30°C for 3 hours and stopped by addition of five volumes of prechilled acetone to precipitate the protein. Protein precipitates were resuspended in 0.2 ml of 100 mM triethylammonium bicarbonate and digested for 16 hours by adding 1/100 (w/w) trypsin protease (Sigma-Aldrich, USA). A 2-μl aliquot of the digest was loaded into a nanoViper C18 trap column (Acclaim PepMap 100; 75 μm by 2 cm). Online chromatographic separation was performed using an Easy-nLC 1200 system (Thermo Fisher Scientific, USA). Data-dependent acquisition MS was performed on a Thermo Fisher Scientific Q Exactive mass spectrometer (Thermo Fisher Scientific, USA) fitted with a Nano Flex ion source. Data were acquired using an ion spray voltage of 1.9 kV and an interface heater temperature of 275°C. Tandem MS data were analyzed for protein identification and quantification using PEAKS Studio 8.5 (Bioinformatics Solutions Inc., USA). The local false discovery rate at the peptide-spectrum match level was set to 1.0% after searching against the K12 database, allowing a maximum of two missed cleavages (39).

Measurement of zeta potential

Zeta potential (in millivolts) of bacterial cells was measured using a ZetaPALS Zeta Potential Analyzer (Brookhaven Instruments Corporation, Holtsville, NY). Before analysis, bacterial suspensions grown in the presence or absence of metabolites were centrifuged (5000 rpm for 5 min at 5°C), and pellets were washed twice and resuspended in 2 ml of sterile 0.85% saline, adjusted to an OD₆₀₀ of 0.5. Each bacterial sample was further diluted 1:200 in 0.85% saline, and zeta potential measurements were performed on 2-ml aliquots.

Quantitative reverse transcription-PCR

The expression level of mRNA in bacterial cells was determined by quantitative reverse transcription–PCR (qRT-PCR). Briefly, total RNA was extracted from cells using TRIzol Reagent. Equal amount of each RNA sample was reverse transcribed into cDNA using the PrimeScript RT reagent kit (Accurate Biology, Hunan, China) according to the manufacturer’s instructions. cDNA templates were quantified using SYBR Green reagent (Accurate Biology, Hunan, China) on a QuantStudio 6 Flex system (Life Technologies Holding Pte. Ltd., Singapore). The amplification conditions were as follows: 95°C for 30 s, 40 cycles of 95°C for 10 s, and 60°C for 30 s. Fluorescence measurements were taken at 70°C for 1 s during each cycle. Cycling was terminated at 95°C with a heating rate of 0.1°C/s to obtain a melting curve. Data are presented as relative mRNA expression compared to the control group and normalized to the endogenous reference gene 16S ribosomal RNA. Primers used in this study are listed in tables S4 and S5.

Effect of succinate on mcr-1 expression and bacterial resistance to CST in vivo

Six-week-old KM mice were obtained from the Laboratory Animal Center of Sun Yat-sen University (Guangzhou, China) and acclimatized for 1 week. Seventy-two infection-free mice were infected with Ec-MCR-1 (3.0 × 10⁸ CFU per mouse). One hour later, the peritonitis-sepsis model mice were treated with CST (10 mg/kg) or succinate (100 mg/kg) plus CST (10 mg/kg). At 24 hours, the spleen and liver were collected and homogenized. The supernatant was used for bacterial cell enumeration via plate counting or for mRNA extraction to quantify mcr-1 expression. Bacteria were grown on plate and used for another round of infection. Five cycles were performed.

Western blot

Bacterial cells were lysed directly in 5× SDS loading buffer and boiled for 10 min. After centrifugation, 50 μg of total protein extracts were separated by 12% SDS–polyacrylamide gel electrophoresis and transferred to a polyvinylidene difluoride membrane (Merck Millipore and Sigma-Aldrich, USA). The membrane was blocked with 3% bovine serum albumin in tris-buffered saline containing 0.05% Tween-20 for 1 hour at room temperature. The membrane was then incubated with a rabbit anti-FLAG monoclonal antibody (catalog no. R24091; lot no. KK0517) (Zen BioScience, Sichuan, China) or a rabbit anti–succinyl lysine monoclonal antibody (catalog no. PTM-401; lot no. ZDJ32M51P1) (PTM BIO, Zhengjiang, China) at a 1:1000 dilution, followed by a goat anti-rabbit immunoglobulin G H&L secondary antibody conjugated with HRP (catalog no. 511203; lot no. HH0518) (Zen BioScience, Sichuan, China). Positive bands were detected using a gel documentation system (LAS-3000 Fujifilm Medical Systems, Stamford, CT, USA).

Statistical analysis

Statistical analysis was performed using GraphPad Prism 8 and SPSS 26.0 software. All data are presented as mean ± SD. For the in vitro studies, Student’s t tests (for normally distributed data) or one-way analysis of variance (ANOVA) (for multiple groups) were used to calculate P values. For the in vivo studies, n represents the number of animals per group, and statistical significance was determined by a two-sided Mann-Whitney U test unless otherwise indicated. Differences with P < 0.05 were considered significant. The presentation of P values was formatted according to the guidelines of APA style (Numbers and Statistics Guild, APA style seventh edition): A P value larger than 0.01 was reported to two decimal places, P values between 0.01 and 0.001 to three decimal places, and P values less than 0.001 as P < 0.001.

Supplementary Materials

This PDF file includes:

Figs. S1 to S8

Tables S1 to S5