Citrate synthase condensation requires catalytic events in Corynebacterium glutamicum

Abstract

Protein condensation regulates cellular metabolism. Although many enzymes are highly structured proteins, the molecular mechanisms underlying their condensation remain poorly understood. Here, we report the discovery and characterization of Corynebacterium glutamicum citrate synthase (CgCS) condensates in relation to extracellular glutamate overproduction. CgCS, which does not possess typical intrinsically disordered sequences, formed droplet-like condensates in C. glutamicum. CgCS condensates were observed in growing cells and dispersed in stationary phase cells. Condensates were not formed for the CgCS variants that lacked a catalytic residue, and the degree of condensation was correlated with catalytic activity. Oxaloacetate, a substrate of CS, was a crucial factor affecting condensation. Both substrates (oxaloacetate and acetyl-CoA) affected the saturation concentration of CgCS in a catalytic residue-dependent manner. We suggest that conformational heterogeneity generated through catalytic events is important for condensation, i.e., catalysis-dependent condensation. This study provides insights into the mechanisms underlying condensation of enzymes lacking intrinsically disordered sequences.

The molecular mechanisms underlying protein condensation remain poorly understood. Here the authors discover and characterise the dynamic condensation of citrate synthase in C. glutamicum, showing condensation is dependent on catalytic activity and substrate binding.

Introduction

Protein condensates, also known as membraneless organelles, regulate cellular metabolism^1,2. Protein condensation plays various roles in regulating metabolism, such as enhancing enzymatic reactions and regulating metabolic nodes^3–8. Purinosomes, comprising at least 10 purine biosynthetic enzymes, are formed upon purine depletion for de novo purine biosynthesis⁹. Glycolytic enzymes form condensates called glycolytic bodies (G-bodies) in eukaryotic cells, which increase glycolytic flux under hypoxic conditions in yeast^10,11. In Escherichia coli, targeting of sugar metabolism-associated regulator (TmaR) forms condensates in growing cells and controls the subcellular localization and activity of phosphotransferase system Enzyme I, which plays a crucial role in sugar utilization and metabolism¹². These studies suggest the importance of protein condensation in metabolic regulation; however, the critical role of protein condensation in cellular metabolism, specifically, in regulating the metabolic pathways, remains poorly understood. Furthermore, the molecular mechanisms underlying metabolic enzyme condensation remain unclear.

Protein condensation is driven by multivalent interactions, which often arise from low-complexity or intrinsically disordered regions that increase effective valency¹³. Many enzymes are structured (folded) proteins and form condensates despite lacking intrinsically disordered regions, such as enolase from Saccharomyces cerevisiae¹⁴ and lipoate–protein ligase A (LplA) from Escherichia coli¹⁵. Although less studied, the condensation of structured proteins has recently attracted growing interests¹⁶.

Corynebacterium glutamicum, a gram-positive bacterium with high GC content, is generally recognized as safe in the industrial production of amino acids, particularly L-glutamate (an umami component) and L-lysine^17,18. Corynebacterium glutamicum overproduces extracellular glutamate in response to cell surface stimuli, such as Tween 40 addition, biotin limitation, and penicillin G addition¹⁹. Under glutamate-producing conditions, increases in the metabolic flux from glucose to glutamate occur without increased transcriptional and translational levels of corresponding enzymes^20–22, suggesting post-translational regulation of metabolism. OdhI, an intrinsic inhibitor of 2-oxoglutarate dehydrogenase (ODH), switches fluxes at the 2-oxoglutarate node toward glutamate synthesis^23–26. Metabolic switching at the phosphoenolpyruvate (PEP) node has also been reported²⁰, wherein the activation of phosphoenolpyruvate carboxylase (PEPC) by lysine deacetylase contributes to maintaining PEPC flux under glutamate-producing conditions despite the decreased protein level²⁷. However, these studies only account for the local reorientation of metabolic fluxes within large-scale metabolic flux redistribution. We focused on a possibility whether enzyme condensation, a post-translational mechanism, plays a role in metabolic changes leading to glutamate metabolism.

Herein, we report the discovery and characterization of citrate synthase (CgCS) condensates in C. glutamicum cells. We found that CgCS variants lacking a catalytic residue did not form condensates and that condensate formation was regulated by substrates. The degree of condensation was correlated with enzymatic activity. We propose that catalytic events generate conformational heterogeneity that is vital for the condensation of CgCS, such that the process of CgCS condensation is catalysis-dependent. This study provides insights into the mechanisms underlying condensation of enzymes lacking intrinsically disordered sequences.

Results

Citrate synthase forms condensates in C. glutamicum cells

Citrate synthase catalyzes the synthesis of citrate from oxaloacetate (OAA) and acetyl-CoA, the first rate-limiting step of the tricarboxylic acid cycle. Using green fluorescent protein (GFP) tagging, we found that CgCS (Cgl0829, encoded by gltA) formed condensates in C. glutamicum cells (Fig. 1a). The CgCS-mEGFP strain, which produced CgCS with a C-terminal monomeric enhanced GFP (mEGFP) tag, exhibited normal growth and glutamate production equivalent to that of the wild-type strain (Supplementary Fig. 1a), confirming that C-terminal mEGFP tagging did not affect CgCS function. Condensates were observed after 3–12 h of cultivation in the glutamate production medium, corresponding to the glucose-consuming growth phase, and were almost undetectable after 24 h (Fig. 1a, b). The abundances of CgCS produced at 9 and 24 h were almost identical (Fig. 1c), indicating that the loss of CgCS condensates was not ascribed to decreased CgCS production. mEGFP alone did not form condensates in C. glutamicum (Supplementary Fig. 1b).

Fig. 1. Citrate synthase forms condensates in Corynebacterium glutamicum cells.

a Representative images of CgCS-mEGFP in C. glutamicum cells (NMa43) cultured in glutamate-producing medium. “Non” and “Prod.” represent non-producing and glutamate-producing conditions, respectively. CS condensates are indicated by arrows. Scale bars, 1 µm. b Extracellular glucose consumption by the wild-type (WT) (ATCC 13032, black) and CgCS-mEGFP (NMa43, green) strains. Each data point represents one biologically independent experiment (n = 3). Open circles with dotted lines, non-producing conditions; closed circles with solid lines, glutamate-producing conditions. c Abundance of CgCS-mEGFP (77 kDa) in cell lysates (5 µg) was detected by western blotting using an anti-GFP antibody. d Size distribution of CgCS condensates. A total of 371 and 372 CgCS condensates were analyzed under non-producing and glutamate-producing conditions, respectively. Thick horizontal and vertical lines indicate the mean and SD, respectively. Means and p-values (p) obtained via two-sided unpaired Welch’s t-tests are shown. Exact p-value is presented in Supplementary Table 5. e Sucrose density gradient ultracentrifugation analysis using cell lysates (after 9 h) from the CgCS-Strep strain (YF1). CgCS-Strep (50 kDa) was detected via western blot using an anti-Strep antibody. Fraction numbers from top to bottom are shown. f FRAP analysis of the CgCS-mEGFP condensates in C. glutamicum. Data are presented as the mean +/− SD of four independent photobleaching experiments. Experiments were independently repeated twice (a, c, e), three times (f), or five times (d) with similar results, and a representative data set is shown. Source data are provided as a Source Data file.

CgCS condensates were observed under both glutamate-producing and non-producing conditions; however, they exhibited several characteristic differences. The CgCS condensates were larger under glutamate-producing conditions than under non-producing conditions (Fig. 1d). The aspect ratio (ratio of the major axis to the minor axis after ellipse approximation) was slightly higher in the condensates formed under glutamate-producing conditions, suggesting that the condensates adopted more rod-shaped forms (Supplementary Fig. 1c). Condensates with similar features were consistently observed in cells treated with Tween 40 addition, biotin limitation, or penicillin G addition, which induced glutamate production (Supplementary Fig. 1d).

Next, we performed an ultracentrifugation analysis of cell lysates from the C. glutamicum strain that produced C-terminal-Strep-tagged CgCS (CgCS-Strep). The CgCS-Strep protein from cells cultured under glutamate-producing conditions reproducibly migrated with smaller molecular-weight fractions compared to that from cells cultured under non-glutamate-producing conditions (Fig. 1e). This result suggested that the properties of CgCS depended on the culture conditions, which was consistent with the results for CgCS-mEGFP, revealing different characteristics between the two conditions. The amount of CgCS produced did not differ between the glutamate-producing and non-producing conditions (Fig. 1c). Most bacterial CS proteins assemble into dimers or hexamers²⁸. Size-exclusion chromatography revealed that CgCS-Strep purified from cultured C. glutamicum cells existed as a dimer (Supplementary Fig. 1e), as previously reported²⁹.

Bright-field microscopy showed that CgCS-mEGFP did not form aggregates in the cells (Supplementary Fig. 1f)³⁰. To further examine whether the condensates were liquid droplets or aggregates, fluorescence recovery after photobleaching (FRAP) analysis was performed³¹. After bleaching, the fluorescence intensity recovered, suggesting that the condensates were droplets formed via a phase separation mechanism (Fig. 1f). The fluorescence recovery was low, likely owing to the small volume of bacterial cells compared to the volume of the FRAP excitation, which was consistent with other FRAP analyses of bacterial cells³². 1,6-Hexandiol, which interferes with weak hydrophobic interactions, is often used to dissolve protein condensates³³. Treatment with 1,6-hexandiol did not affect CgCS condensates, suggesting that CgCS condensation does not depend on hydrophobic interactions.

Active CgCS is required for condensation

To further characterize CgCS condensation, we constructed mutant strains of C. glutamicum that produced an inactive CgCS variant as a mEGFP fusion protein. Based on the amino acid sequence and structural alignments among the CS proteins, His273, His316, and Asp372 were likely the catalytic residues of CgCS (Supplementary Fig. 2a)³⁴. We confirmed that recombinant CgCS^H273A, CgCS^H316A, and CgCS^D372A proteins did not exhibit CS activity (Supplementary Fig. 2b). AlphaFold modeling suggested that these variants possessed a structure similar to that of CgCS^WT (Supplementary Fig. 2c). All strains producing the CgCS^H273A, CgCS^H316A, and CgCS^D372A variants exhibited substantially impaired growth, which was partially restored by the dose-dependent addition of citrate up to 100 mM (Fig. 2a and Supplementary Fig. 2d, e), consistent with the results of a previous report³⁵.

Fig. 2. Active CgCS is required for condensation.

a Restoration of growth of the CgCS^H316A (NMa191) strain with 100 mM citrate. Cells were cultured in glutamate production medium. Each data point represents one biologically independent experiment (n = 3). WT, ATCC 13032 strain. b Representative images of mEGFP-fused CgCS variants (NMa43, NMa94, NMa183, and NMa381) in C. glutamicum cells cultured under glutamate-producing conditions with or without 100 mM citrate (after 9 h). Scale bars, 1 µm. c Population ratio of cells forming CgCS condensates. A total of 373, 372, and 371 cells were analyzed for CgCS^WT-mEGFP (NMa43) under glutamate-producing conditions and CgCS^H316A-mEGFP (NMa94) under glutamate-producing conditions with and without 100 mM citrate, respectively. d Amounts of CgCS^WT-mEGFP and CgCS^H316A-mEGFP in cell lysates (5 µg protein) were detected via western blotting using an anti-GFP antibody. e Fluorescent images of CgCS^WT-mEGFP and CgCS-mCherry variants coproduced in single cells: CgCS^WT-mEGFP and CgCS^WT-mCherry (NMa260), CgCS^WT-mEGFP and CgCS^H316A-mCherry (NMa263). Line-scan profiles of fluorescent intensity corresponding to the white dotted line is shown. The black triangle indicates the peak position of fluorescent intensity derived from CgCS^WT-mEGFP. Experiments were independently repeated twice (c) or three (b, d, e) with similar results, and a representative dataset is shown. Source data are provided as a Source Data file.

Inactive CgCS variants did not form condensates (Fig. 2b, c). The amount of CgCS^H316A-mEGFP was comparable to that of CgCS^WT-mEGFP, indicating that the loss of condensation was not due to a reduction in the amount of CgCS protein produced (Fig. 2d). These results suggested that condensation was linked to CS activity.

The loss of condensation in inactive CgCS variants may be caused by a change in the intracellular environment owing to the loss of CS activity. Therefore, we examined the localization of the wild-type CgCS (CgCS^WT) and CgCS^H316A in CgCS^WT-mEGFP-producing cells that coproduced plasmid-derived CgCS^H316A-mCherry and CgCS^WT-mCherry. Both genome-derived CgCS^WT-mEGFP and plasmid-derived CgCS^WT-mCherry formed condensates. Conversely, when genome-derived CgCS^WT-mEGFP was coproduced with plasmid-derived CgCS^H316A-mCherry, CgCS^WT-mEGFP formed condensates, but CgCS^H316A-mCherry did not (Fig. 2e). These results suggested that intrinsic properties of CgCS^H316A protein caused the loss of condensation.

Oxaloacetate affects CgCS condensation

We previously reported that PEPC (Cgl1585) functions as a major anaplerotic enzyme to supply OAA, and the PEPC gene-deleted strain (ΔPEPC strain) exhibits strongly reduced glutamate production in the studied conditions using the ATCC 13032 (Kyowa Hakko) strain²⁷. Notably, in the ΔPEPC strain, CgCS-mEGFP formed enlarged and rod-shaped condensates under glutamate-producing conditions (Fig. 3a), although the frequency of the condensations per cell between the ΔPEPC and wild-type strains was similar (Fig. 3b). Western blotting analysis confirmed that the amount of CgCS-mEGFP protein produced was not substantially different between the ΔPEPC and wild-type strains (Fig. 3c).

Fig. 3. Oxaloacetate affects CgCS condensation.

a Representative images of CgCS-mEGFP in the wild-type (WT, NMa43) and phosphoenolpyruvate carboxylase-deleted (ΔPEPC, NMa240) strains cultured under glutamate-producing conditions (left), and the areas and aspect ratios of CgCS condensates (right). A total of 391 (WT) and 546 (ΔPEPC) condensates were analyzed. b Population of cells forming CgCS condensates. A total of 907 (WT) and 1129 (ΔPEPC) cells were analyzed. c Abundance of CgCS-mEGFP in cell lysates (5 µg) of WT and ΔPEPC strains detected by western blotting using an anti-GFP antibody. d Representative images of CgCS-mEGFP in WT (NMa43), PEPC^K653R (NMa211), and PEPC^K653Q (NMa190) strains (upper), and the areas and aspect ratios of CgCS condensates (lower). A total of 424, 398, and 408 condensates were analyzed, respectively. e Biotin-bound pyruvate carboxylase (PC) in cell lysates (8 µg) of the ΔPEPC strain (MY100) detected by western blotting using a streptavidin-HRP conjugate. The ΔPC strain (MY66) served as a negative control. f Excess biotin restored CgCS condensate morphology. Representative images of CgCS-mEGFP in WT (NMa43) and ΔPEPC (NMa240) strains (upper), and the sizes and aspect ratios of the CS condensates (lower). A total of 437 (WT, 30 µg/L biotin), 502 (ΔPEPC, 30 µg/L biotin), and 604 (ΔPEPC, 300 µg/L biotin) condensates were analyzed. g Effect of exogeneous oxaloacetate (30 mM) on CgCS condensates. Representative images of CgCS-mEGFP in WT (NMa43) and ΔPEPC (NMa240) strains cultured under glutamate-producing conditions (upper), and the areas and aspect ratios of CgCS condensates (lower). A total of 250 condensates were analyzed for each strain. For (a, d, f, and g), data are presented as the mean and SD, indicated by thick horizontal and vertical lines, respectively. Statistical significance was determined by two-sided unpaired Welch’s t-tests (a), one-way ANOVA followed by two-sided Dunnett’s multiple-comparison test (d), or one-way ANOVA followed by two-sided Tukey’s honestly significant difference (HSD) test (f, g). p-values (p) and mean values are shown. Exact p-values are presented in Supplementary Table 5. Experiments were repeated independently twice (b, d, g), three times (c, e, f), or five times (a) with similar results, and a representative dataset is shown. Source data are provided as a Source Data file.

We identified two PEPC variants, PEPC^K653Q and PEPC^K653R, which exhibited severely and moderately decreased PEPC activity, respectively²⁷. Although both PEPC-variant strains of C. glutamicum exhibited growth equivalent to that of the wild-type strain, PEPC^K653Q exhibited severely impaired glutamate production, whereas PEPC^K653R produced moderate levels of glutamate²⁷. CgCS-mEGFP formed severely enlarged and more rod-shaped condensates in PEPC^K653Q, consistent with that in the ΔPEPC strain. FRAP analysis confirmed that the morphologically abnormal CgCS condensates in PEPC^K653Q showed a fluidity similar to that of the wild type (Supplementary Fig. 3a). The features of condensates in PEPC^K653R were similar to those in the wild type (Fig. 3d). Moreover, the area and aspect ratio of condensates were well correlated with the PEPC activity among WT, PEPC^K653R, and PEPC^K653Q variants (Supplementary Fig. 3b). These results indicated a correlation between PEPC activity and CS condensation. Because impaired PEPC may decrease the intracellular levels of OAA, we hypothesized that the abnormal condensates of CgCS in the PEPC mutants were correlated with reduced OAA availability.

Corynebacterium glutamicum contains pyruvate carboxylase (PC, Cgl0689), an anaplerotic enzyme that supplies OAA. Excess biotin (200 μg/L), a cofactor of PC, increases biotin-bound PC levels and restores glutamate production of the ΔPEPC strain, likely because of the recovery of OAA supply³⁶. We added a tenfold excess of biotin (300 μg/L) to the glutamate production medium to examine the effect on CgCS condensation in the ΔPEPC strain. Under these conditions, biotin-bound PC production increased compared to that under normal biotin (30 μg/L) conditions (Fig. 3e). The features (area and aspect ratio) of the CgCS condensates were partially restored to those of the wild type (Fig. 3f).

Based on these results, we hypothesized that OAA affects CgCS condensation. To examine this possibility, we attempted to determine whether an external supply of OAA could reverse abnormal condensation. ΔPEPC cells producing abnormal CgCS-mEGFP condensates, which were prepared under glutamate-producing conditions, were suspended in a buffer containing 30 mM OAA. Upon suspension with the OAA-containing buffer, the abnormal CgCS-mEGFP condensates in the ΔPEPC strain decreased to an area and an aspect ratio similar to that of the wild type (Fig. 3f). A similar OAA-dependent morphological change in condensates was observed for CgCS-mEGFP in PEPC^K653Q through biotin-dependent PC activation and external OAA supply (Supplementary Fig. 3c, d). CgCS-mEGFP in the wild-type strain exhibited decreased condensation in the OAA-containing buffer (Fig. 3g and Supplementary Fig. 3e).

Substrate-assisted conformational change is important for CgCS condensation

In general, protein condensation or aggregation occurs above the saturation concentration (c_sat) and can be easily assessed using turbidity measurements^31,37. Therefore, recombinant CgCS^WT and CgCS^H316A proteins were incubated in the presence of various OAA concentrations, and the change in turbidity (ΔOD₆₀₀) before and after incubation was measured. CgCS^H316A exhibited decreased turbidity compared with CgCS^WT at the same protein concentrations (50, 100, and 200 μM) in the absence of OAA (Fig. 4a, Supplementary Fig. 4a), suggesting that CgCS^H316A had a higher c_sat value than CgCS^WT. Staining with 1-anilinonaphthalene-8-sulphonate fluorescent dye revealed that the condensates resembled amorphous protein aggregates (Supplementary Fig. 4b).

Fig. 4. Oxaloacetate increases the c_sat of CgCS through binding to the catalytic site.

a, b Turbidity assays of recombinant CgCS^WT and CgCS^H316A. Changes in turbidity (ΔOD₆₀₀) after incubation for 24 h at 30 °C in the presence of indicated oxaloacetate (OAA) concentrations are shown as a color gradient (a). Turbidity of CgCS^WT protein (100 µM) after incubation with 1 mM OAA, pyruvate (PYR), malate (MAL), or 2-oxoglutarate (2OG) (b). Each data point represents one biologically independent experiment (n = 2). c Intrinsic fluorescence measurements. Fluorescence spectra of CgCS^WT and CgCS^H316A proteins (10 µM) in the presence of indicated concentrations of oxaloacetate (left). Fluorescence intensity plot at 330 nm in the presence of OAA (0.0001–5 mM) (right). Intensity was normalized to that without OAA. Dotted lines of magenta and turquoise represent the inflection point and tangent line at the inflection point (the slope parameter), respectively. d Effect of various organic acids on the fluorescence profiles of CgCS^WT. Experiments were repeated independently twice (a, d) or three times (c) with similar results, and a representative dataset is shown. Source data are provided as a Source Data file.

As the OAA concentration increased, the turbidity of CgCS^WT decreased (Fig. 4a). Such a turbidity-reducing effect of OAA was not observed with the CgCS^H316A solution (Fig. 4a), and neither pyruvate, malate, nor 2-oxoglutarate decreased the turbidity of the CgCS^WT solution (Fig. 4b and Supplementary Fig. 4c). His316 serves as a binding site for OAA in CS³⁸, and OAA likely increases the c_sat of CgCS by binding to the active site.

CS is generally considered to undergo large conformational changes during catalytic cycles; OAA binding to active-site residues, including His316, induces a conformational change from an open to a semi-closed form, and acetyl-CoA binding then leads to a closed form³⁹. OAA is the primary cause of the conformational changes of CS. To characterize the conformational changes, we measured intrinsic fluorescence⁴⁰ using CgCS^WT and CgCS^H316A. No difference was observed in the fluorescence profiles of CgCS^WT and CgCS^H316A-Strep in the absence of OAA (Fig. 4c), suggesting a lack of overall structural difference between them, which was also supported by AlphaFold modeling (Supplementary Fig. 2c). Upon addition of OAA to the CgCS^WT protein, a decrease in the fluorescence peak intensity was observed, which was dose-dependent, reflecting OAA-induced conformational changes (Fig. 4c and Supplementary Fig. 4d). Meanwhile, the change in peak intensity was smaller for CgCS^H316A than for CgCS^WT, suggesting that CgCS^H316A underwent a limited conformational change. To quantitatively evaluate conformational changes following OAA addition, the fluorescence intensity at 330 nm was plotted against varying OAA concentrations and fitted using a logistic function. We evaluated responsiveness to OAA using two parameters derived from the fitted curve: the inflection point, which reflects a K_d-like parameter, and the slope parameter, which represents the steepness of the transition and reflects the sensitivity in the conformational response (Fig. 4c). The inflection point values were 5.7 ± 4.6 × 10 and 2.5 ± 0.8 × 10³ µM for CgCS^WT and CgCS^H316A, respectively. OAA specifically induced conformational changes in CgCS^WT (Fig. 4d), consistent with the turbidity assay results (Fig. 4b). Collectively, these results suggested that CgCS^H316A had a lower binding affinity for OAA and showed limited conformational change compared to CgCS^WT, which may be related to reduced condensate formation in cells.

To reinforce the hypothesis that OAA-dependent conformational change or flexibility is involved in the formation of CgCS condensates, we attempted to isolate CgCS variants with impaired OAA binding. Using the dimer structural model of CgCS, we performed molecular dynamics (MD) simulation analysis in the presence of OAA molecules and found five residues that OAA most frequently contacts: R317, R418, R310, K415, and H316, the latter being a catalytic residue (Fig. 5a and Supplementary Fig. 5a). K415 and R418 are on the other chain of the dimer, and R418 corresponds to R421 in porcine CS that is involved in OAA/citrate binding³⁸. Enzyme activity assays using recombinant CgCS proteins revealed that the CgCS^R317A and CgCS^R418A variants exhibited negligible activity, similar to the catalytic site-deficient CgCS^H316A variant, and the CgCS^R310A and CgCS^K415A variants retained lower activity than CgCS^WT (Fig. 5b and Supplementary Fig. 5b).

Fig. 5. Correlation between the catalytic activity and condensation of CgCS.

a Residues frequently contacting OAA in MD simulations mapped onto an Alphafold3 model of the CgCS dimer. Two different chains are shown in blue and green. OAA in the catalytic pocket is shown in spherical mode. b CS activity of His-tagged recombinant variants. Each data point represents one biologically independent experiment (n = 3; n = 4 for R310A), averaged from three technical replicates. Bars represent the mean, and error bars indicate +/− SD. c OAA responsiveness of CgCS variants, represented by the inflection points and slope parameters of fitted curves of fluorescence intensity plots. Each square symbol represents one biologically independent experiment (n = 3 for H316A, K415A, and R418A; n = 4 for R310A; and n = 5 for WT and R317A). Closed circles represents the mean, and vertical and horizontal lines indicate +/− SD. d Condensate formation by CgCS^WT-mEGFP (NMa43), CgCS^R310A-mEGFP (NMa498), CgCS^R317A-mEGFP (NMa437), CgCS^K415A-mEGFP (NMa506), and CgCS^R418A-mEGFP (NMa503) strains under glutamate-producing conditions. Citrate (100 mM) was added where indicated. Population of condensate-forming cells and condensate areas (R310A and K415A) are shown. The number (n) of cells and condensates analyzed is shown in parentheses. Thick horizontal and vertical lines indicate the mean and +/− SD, respectively. Mean values and p-values (p) obtained via two-sided unpaired Welch t-tests are shown. Exact p-values are presented in Supplementary Table 5. e Correlation between CS activity and condensate area using data from (b and d). Condensate areas are shown relative to the WT value. r represents the Pearson’s correlation efficient value. Experiments were repeated independently twice (population of condensate-forming cells) or three times (condensate areas) with similar results for (d), and a representative dataset is shown. Source data are provided as a Source Data file.

We evaluated responsiveness of these variants to OAA using the inflection point and slope parameters (Fig. 5c and Supplementary Fig. 5c). The CgCS^R418A variant showed a substantially higher value of inflection point (4.9 ± 1.7  × 10³ µM) than CgCS^WT, similar to CgCS^H316A, suggesting the loss of enzymatic activity. CgCS^R310A (1.6 ± 1.1 × 10² µM), CgCS^R317A (2.2 ± 0.7 × 10² µM), and CgCS^K415A (1.4 ± 0.6 × 10² µM) showed slightly higher inflection point values than CgCS^WT. CgCS^R317A and CgCS^K415A further showed increased slope parameters, suggesting that these variants responded less to OAA at concentrations lower than the inflection point, but responded sharply around the inflection point. Thus, the response to OAA was considered to be substantially decreased for CgCS^H316A and CgCS^R418A and moderately decreased for CgCS^R310A, CgCS^R317A, and CgCS^K415A (Fig. 5c).

We constructed CgCS-mEGFP strains of these variants to determine the condensate formation. CgCS^R317A and CgCS^R418A variants that lost enzymatic activity exhibited almost no condensate formation under glutamate-producing conditions with or without citrate (Fig. 5d). CgCS^R310A and CgCS^K415A variants, which retained residual activity and had moderately decreased responses to OAA, formed smaller condensates (Fig. 5d). The amounts of CgCS^variant-mEGFP proteins produced were almost identical to the amounts of CgCS^WT-mEGFP (Supplementary Fig. 5d). We found a correlation between the catalytic activity and condensation in these CgCS variants (Fig. 5e). These results indicated that the catalytic activity and/or OAA-induced conformational change was important for CgCS condensation.

Acetyl-CoA may decrease the c_sat of CgCS

We attempted to examine the effect of acetyl-CoA, the other substrate of CS, on CgCS condensation. Acetyl-CoA, but not CoA, increased the turbidity of CgCS^WT but not CgCS^H316A protein, suggesting that the decrease in c_sat and effect of acetyl-CoA are not characteristic of the CgCS^H316A protein (Fig. 6a and Supplementary Fig. 6a). Acetyl-CoA and CoA affected the intrinsic fluorescent profiles of CgCS^WT to a similar extent (Fig. 6b), but more moderately than OAA. The inflection point value for acetyl-CoA was 1.1 ± 0.4 × 10³ µM, which was one order higher than that for OAA (0.57 × 10² µM). No apparent difference was observed between CgCS^WT and CgCS^H316A (1.0 ± 0.5 × 10³ µM) in the response to acetyl-CoA in fluorescent spectra (Fig. 6b and Supplementary Fig. 6b). We also examined CgCS^R310A, CgCS^R317A, and CgCS^K415A variants, which showed moderately decreased responses to OAA. Their responses to acetyl-CoA did not substantially differ compared to that of CgCS^WT (Supplementary Fig. 6b, c).

Fig. 6. Acetyl-CoA may decrease the c_sat of CgCS.

a Change in turbidity (ΔOD₆₀₀) after 24-h incubation of CgCS^WT and CS^H316A protein (100 µM) in the presence of 1 mM of acetyl-CoA or CoA. Each data point represents one biologically independent experiment (n = 3), and data from the same experiment are indicated by the same symbol (circles, triangles, and squares). Bars represent the mean, and error bars indicate +/− SD. b Fluorescence spectra of CgCS^WT and CgCS^H316A proteins (10 µM) in the presence of the indicated concentrations of acetyl-CoA (AcCoA) or CoA (left). Fluorescence intensity plot at 330 nm in the presence of various concentrations (0.001–10 mM) of acetyl-CoA or CoA (right). Intensity was normalized to that without acetyl-CoA or CoA. Dotted lines of magenta and turquoise represent the inflection point and tangent line at the inflection point, respectively. Two experiments were repeated independently with similar results for (b), and a representative dataset is shown. Source data are provided as a Source Data file.

Discussion

In this study, we discovered and characterized CS condensates in C. glutamicum. The condensates behaved as fluid-like droplets, which likely formed through a phase separation mechanism. CgCS formed larger condensates under glutamate-producing conditions, whereas it reproducibly migrated in smaller molecular-weight fractions, according to ultracentrifugation analysis. We speculate that under glutamate-producing conditions, the density of CgCS molecules in the condensate was decreased.

Protein condensation is driven by multivalent interactions. Intrinsically disordered domains (proteins) are characterized by low interaction specificity and high conformational heterogeneity, which increases effective valency^13,16. Structured proteins are characterized by high interaction specificity and low conformational heterogeneity, and high specific interactions, including protein–protein interactions and oligomerization, such as coiled-coil and head-to-tail polymerization, can affect the valency of protein and condensate formation¹⁶. CgCS does not possess typical intrinsically disordered sequences capable of phase separation and does not form higher-order assemblies (a simple homodimer in solution), similar to the cases of dimeric S. cerevisiae enolase¹⁴ and homomeric E. coli LplA¹⁵. Specific amino acid sequences intrinsic to the S. cerevisiae enolase induce condensation¹⁴. Escherichia coli LplA can form self-assembled condensates in cells through multivalent interactions with residues on the protein surface¹⁵. Similar mechanisms can be considered for CgCS condensation.

Although the detailed mechanism underlying CgCS condensation remains unclear, our in vivo and in vitro studies suggested that OAA is a crucial factor affecting condensation in C. glutamicum. CgCS^WT condensates in the wild-type cells were dispersed by an external supply of OAA. CgCS^WT protein formed amorphous aggregates as the concentration increased in vitro, and OAA reduced these aggregates. These results suggest that OAA negatively affected the condensation of CgCS, likely by increasing the c_sat of CgCS. The other substrate, acetyl-CoA, enhanced the aggregation. As neither OAA nor acetyl-CoA had any effect on the formation of CgCS^H316A aggregates, we assumed that both substrates affected the c_sat of CgCS by binding to the active site. We consider that the acetyl moiety of acetyl-CoA is involved in the binding.

Another interesting feature was that inactive CgCS variants with catalytic residue substitutions had markedly reduced abilities to form condensates. Metabolites affect protein condensation and substrate- or ligand-induced protein condensation occurs in several proteins. In the case of Plant Toll/interleukin-1 receptor (TIR) domain proteins, the binding of NAD⁺ and ATP, co-substrates of TIR-encoded NADase, induces head-to-tail dimerization of the domains, which are involved in the condensation of the TIR domain proteins in plant cells^41,42. In this case, a catalytic residue (Glu86 of RBA1 from Arabidopsis thaliana) is essential for TIR domain protein-mediated cell death but is not required for protein condensation⁴¹. Similarly, the condensate formation of E. coli LplA is negatively affected by binding of the substrate lipoic acid, although an LplA variant that loses lipoic acid-binding activity retains the ability to form condensates¹⁵. The condensation of CgCS is unique in that both catalytic residues and substrates are directly involved in condensate formation.

The conformational flexibility of CS is important for its overall catalytic efficiency^43,44. We hypothesized that substrate-induced conformational changes and flexibility are involved in CgCS condensation. This hypothesis may explain the reduced condensate formation of catalytic residue-substituted CgCS variants in cells. Analysis of additional CgCS variants strongly suggested a linkage of enzymatic activity to the condensation of CgCS. The degree of OAA-dependent conformational changes was mostly correlated with the degree of condensate formation of CgCS in cells. The characteristics of the CgCS^R317A variant, which retained OAA-induced conformational changes but exhibited no activity or condensate formation, were intriguing. The variant likely cannot undergo appropriate conformational changes during the catalytic reaction cycle or cannot undergo conformational changes under physiological concentrations of OAA owing to decreased responsiveness. This observation is consistent with our hypothesis that conformational changes and flexibility are important for CgCS condensation. Although the effect on conformation changes could not be detected by our intrinsic fluorescence measurement, acetyl-CoA binding may also have a minor effect on the conformation. Conformational changes linked to the catalytic cycle may be involved in generating multivalent interactions that can mediate phase separation. Therefore, we propose that CgCS condensation is catalysis-dependent. Our results suggest that conformational heterogeneity generated through catalytic events can be a novel property for condensation of structured proteins.

CgCS condensates were dispersed in stationary-phase cells when glucose was depleted, where the substrates OAA and acetyl-CoA were likely limited. This observation can be explained by the above hypothesis that CgCS with a low reaction turnover, i.e., a conformationally static CgCS, exhibits a reduced ability to form condensates. However, we cannot exclude the possible involvement of unknown positive or negative factors that promote CS dispersion or condensation, respectively. We assume that a balanced supply of OAA and acetyl-CoA, which may affect conformational heterogeneity during the catalytic cycle, is important for the appropriate formation of CgCS condensates. The abnormal formation of CgCS condensates observed in the PEPC-deficient strains may be due to an imbalanced supply of OAA and acetyl-CoA. Consistent with this hypothesis, normal condensate formation in the ∆PEPC strain was restored by an increased supply of OAA.

The effect of CgCS condensation on catalytic efficiency warrants further investigation. Enzyme condensation can increase the catalytic efficiency of single-step and multistep reactions^7,45. Recent studies have shown that substrate influx and product outflux peak at the condensate surface, and increased surface area of the condensate (correlated to the size) is beneficial for chemical reactions^46,47. Enlargement of CgCS condensates under glutamate-producing conditions may contribute to increased CS flux under glutamate-producing conditions²⁰. We attempted to determine the catalytic efficiency using permeabilized cells. However, this approach was unsuccessful because CgCS condensates were labile and dispersed during cell washing. Consequently, a suitable in vitro system will be required to directly assess the effect of condensation on the catalytic efficiency of CgCS.

Biologically important protein condensates, such as purinosomes and G-bodies, comprise various enzymes and proteins^9,11. These observations suggest that the CgCS condensates may contain additional components. However, our study did not comprehensively investigate the enzymes and proteins recruited to the condensate. Addressing this limitation will be important for further understanding the role of condensation in metabolic flux control and its potential physiological significance. Whether CgCS condensation contributes to metabolic changes associated with glutamate overproduction, and whether other enzyme condensates are involved, will be reported in the near future.

CgCS and most eukaryotic CSs belong to the dimeric type I CS and are not allosterically regulated, unlike the type II CS from E. coli, which is strongly inhibited by NADH via a mechanism that relies on the hexamer structure of CS^35,48. Condensate formation might be a novel regulatory mechanism of allosterically unregulated CSs. In the catalysis-dependent mechanism, substrate availability promotes enzyme condensation, thereby potentially regulating metabolic flux. As CS is one of the most conserved enzymes across domains and plays a critical role in controlling carbon flux through the tricarboxylic acid cycle, our study provides insights into metabolic flux regulation in other organisms, including industrial bioproduction hosts. These findings could guide the development of engineered condensate-based regulatory modules in microbial production strains.

Methods

Strain construction and culture conditions

Corynebacterium glutamicum ATCC 13032 (JCM 1318) was used as the wild-type strain. All strains, plasmids, and oligonucleotide primers used in this study are listed in Supplementary Tables 1–3, respectively. All oligonucleotide primers were synthesized commercially by Fasmac (Atsugi, Japan). Gene deletion, insertion of tag-coding sequences, and site-directed mutagenesis for amino acid substitution on the C. glutamicum genome were performed using a two-step homologous recombination method as described previously²² and details are provided in Supplementary Method 1. The monomeric EGFP (mEGFP, EGFP variant with A206K substitution) and mCherry were used as fluorescent protein tags. The CgCS-mEGFP and CgCS-mCherry constructs consisted of the CgCS (Cgl0829) sequence followed by the mEGFP and mCherry sequences, respectively, without the initial ATG codon. The CgCS-Strep construct consists of the CgCS (Cgl0829) sequence followed by a Gly–Gly linker and Strep-tag(II) sequence. The CgCS-mEGFP and CgCS-Strep constructs were genomically expressed under the original promoter. The pECt plasmid⁴⁹, containing a P_trc promoter and lacI^q gene, was used for expression of the CgCS-mCherry construct in C. glutamicum cells. The DNA sequences for CgCS-Strep, CgCS-mEGFP, and CgCS-mCherry are provided in Supplementary Table 4.

Corynebacterium glutamicum cells were grown in 25 mL of biotin-30 glutamate production medium²² in a 500 mL-baffled flask on a rotary shaker operating at 100 rpm and 31.5 °C. Biotin-30 medium contained 60 g/L D-glucose, 30 g/L (NH₄)₂SO₄, 1 g/L KH₂PO₄, 0.4 g/L MgSO₄·7H₂O, 0.01 g/L FeSO₄·7H₂O, 0.01 g/L MnSO₄·5H₂O, 0.2 mg/L thiamine, 0.48 g/L soybean protein hydrolysate, and 30 µg/L D-biotin. After autoclaving, 1 g of sterilized CaCO₃ was added into a flask to adjust the pH. Glutamate production was routinely induced through addition of 1.5 g/L Tween 40 after 3 h of the main cultivation. Alternatively, 0.4 U/mL of penicillin G was added after 3 h of the main cultivation or the preculture was inoculated into fresh glutamate production medium containing 3 µg/L D-biotin to induce glutamate production.

Fluorescence microscope imaging

Corynebacterium glutamicum cells were diluted in phosphate-buffered saline (pH 7.4) and adjusted to an OD₆₆₀ of 20. The cell-containing buffer was spotted on a 1%(w/v) agarose pad on a glass slide. Images were obtained using an all-in-one fluorescence microscope BZ-X810 (KEYENCE, Osaka, Japan) with a 100× objective lens (oil immersion numerical aperture [NA], 1.35). Green and red fluorescence was detected using GFP (ex: 470/40 nm, em: 525/50 nm, dichroic: 495 nm, OP-87763, KEYENCE, Japan) and TRITC (ex: 545/25 nm, em: 605/70 nm, dichroic: 565 nm, OP-87764, KEYENCE, Japan) filters, respectively. When counting the number of CS condensates per cell, the cell membrane was stained with 5%(v/v) ViVidFluor Neuro Red (Fujifilm Wako, Tokyo, Japan) via incubation for 30 min at 30 °C. When determining the effect of external OAA supply on CS condensation, a 10 µL culture of C. glutamicum was mixed with 35 µL of 100 mM HEPES (pH 8.0) buffer containing 30 mM OAA, and the samples were observed within 3 min. Images were analyzed using ImageJ Fiji software (version 1.54, NIH, Bethesda, MD, USA)⁵⁰.

Fluorescence recovery after photobleaching (FRAP) analysis

FRAP analysis was performed using an FV1200 confocal microscope (Evident, Japan) equipped with a 100× objective lens (oil immersion NA, 1.40). Green fluorescence was detected using a 488 nm laser and a GFP filter (505–540 nm, FV12-MHBY, Evident, Tokyo, Japan). Photobleaching was performed using a 488 nm laser at 25% intensity for 1 s. Images were analyzed using ImageJ Fiji software. The fluorescence intensity before bleaching was calculated as the average of five repeated measurements. The fluorescence intensity upon bleaching was set to zero and plotted relative to that before bleaching. Data are shown as the mean with standard deviations from four independent photobleaching experiments. FRAP experiments were performed independently at least twice, and a representative dataset is shown.

Western blotting analysis

Western blotting was performed as previously described in ref. ²². To detect mEGFP- and Strep-tagged proteins and biotinylated PC, rabbit polyclonal anti-GFP (A6455, MBL, Tokyo, Japan) antibody at 1:2000 dilution, mouse monoclonal anti-Strep-tag II antibody (M211-3, MBL) at 1:2000 dilution, and monoclonal anti-streptavidin (HRP) (AB7403, Abcam, Cambridge, UK) antibody at 1:5000 dilution were used, respectively. Horseradish peroxidase-conjugated anti-rabbit IgG (A0545-1ML, Sigma-Aldrich, St. Louis, MO, USA) and anti-mouse IgG (12-349, Merck Millipore, Billerica, MA, USA) were used as the secondary antibodies at 1:5000 dilution. Signals were detected using an Amersham ImageQuant 800 imaging system (Cytiva, Marlborough, MA, USA).

Sucrose gradient ultracentrifugation analysis

Corynebacterium glutamicum cells were lysed in buffer containing 50 mM TES-NaOH pH 7.6, 15%(w/v) glycerol, 1 mM phenylmethanesulfonylfluoride (PMSF), 0.1 mM dithiothreitol (DTT), and 10 µg/mL DNase at high-pressure (30,000 psi) using an EmulsiFlex-B15 high-pressure homogenizer (Avestin, Ottawa, Canada), and the cells were passed through the homogenizer twice. After cell debris removal via centrifugation (4800 × g, 10 min, 4 °C), the cleared lysate containing 10 µg of protein was layered onto a 5–30% (w/v) sucrose density gradient in 50 mM TES-NaOH pH 7.6, using a Gradient Station system (BioComp Instruments, Nova Scotia, Canada). The gradient was then centrifuged at 36,000 rpm (max. 230,000 × g) in a P40ST rotor (HITACHI, Japan) for 17 h at 4 °C. After centrifugation, the gradient was separated into 21 fractions (0.5 mL) and a bottom fraction using a Gradient Station system. Aliquots (10 µL) were separated using sodium dodecyl sulfate-polyacrylamide gel electrophoresis and subjected to western blotting analysis to detect CS-Strep.

Metabolite analysis

To measure extracellular metabolites, the culture was centrifuged at 20,000 × g for 10 min at 4 °C to remove the bacterial cells, and L-glutamate and D-glucose in the supernatant were analyzed using a L-glutamate KIT YAMASA NEO (Yamasa, Japan) kit and F-kit D-glucose or E-kit liquid D-glucose (Roche/R-Biopharm, Germany), respectively.

CS activity assay

Corynebacterium glutamicum cells were lysed in TESG10N buffer (0.1 M TES-NaOH pH 7.6, 10%(w/v) glycerol, 150 mM NaCl) containing 1 mM PMSF, 0.1 mM DTT, and 10 µg/mL DNase using a EmulsiFlex-B15 high-pressure homogenizer. After removal of cell debris via centrifugation (4800 × g, 10 min, 4 °C), the supernatant was subjected to buffer exchange against TESG10N buffer using a PD-10 column (Cytiva). CS activity assays were performed via colorimetric quantification of CoASH using 5,5-dithiobis-(2-nitrobenzoic acid) (DTNB)⁴⁴. The reaction mixtures (1 mL) contained either 50 mM Tris-HCl pH 7.5, 200 mM potassium glutamate, 0.1 mM DTNB, 0.15 mM acetyl-CoA, 0.2 mM OAA, or Milli-Q water (blank). The reaction commenced by adding 10 µg of crude proteins. The initial rate of 2-nitro-5-thiobenzoate generation was monitored using spectrometry at 412 nm and 30 °C. The extinction coefficient used was 14,150 M^–1cm^–1.

Turbidity assay

Recombinant CS-Strep proteins were produced and purified from E. coli BL21 (DE3) CodonPlus-RIL (Agilent Technologies, Santa Clara, CA, USA) cells using the pET21a vector (69740, Novagen, Merck Millipore). Escherichia coli cells were grown in 2× YT medium (16 g/L Bacto tryptone, 10 g/L Bacto yeast extract, 5 g/L NaCl) at 37 °C for 3 h and then cultivated at 25 °C for 21 h in the presence of 0.1 mM IPTG. The cells were lysed in buffer D (100 mM Tris-HCl pH 8.0 and 150 mM NaCl), and affinity purification was performed using a Strep-Tactin Superflow Plus resin (Qiagen, Venlo, The Netherlands). CS-Strep proteins were eluted with buffer D containing 2.5 mM D-desthiobiotin, followed by buffer exchange against buffer D using a PD-10 column. Purified proteins were concentrated using Amicon Ultra filter (3 K NMWL, Millipore), and assay mixtures containing purified proteins and compounds were prepared and incubated at 30 °C for 24 h. Turbidity was measured at 600 nm using a Spectra Max i3 multimode microplate reader (Molecular Devices, San Jose, CA, USA), and the differences before and after incubation were calculated.

Intrinsic fluorescence measurement

Recombinant CS-His₆ proteins were produced and purified from E. coli BL21(DE3) CodonPlus-RIL cells using the pET21a vector. Escherichia coli cells were cultured in 2 × YT medium at 37 °C for 3 h and then at 25 °C for 21 h in the presence of 0.1 mM IPTG. The cells were lysed in buffer A (20 mM Tris-HCl pH 8.0 and 150 mM NaCl), and affinity purification was performed using a TALON metal affinity resin (Takara Bio USA, San Jose, CA, USA). CS-His₆ proteins were eluted with buffer A containing 200 mM imidazole, followed by buffer exchange against buffer D using a PD-10 column. Purified proteins were concentrated using Amicon Ultra filter (3 K NMWL).

Recombinant CS-Strep or CS-His₆ proteins (10 µM) were excited at 280 nm in buffer D to characterize possible differences in the behavior of tryptophan and tyrosine residues⁴⁰. Fluorescence spectra were collected at 25 °C using a Spectra Max i3 multimode microplate reader between 305 and 450 nm at wavelength increments of 1 nm. Logistic curve fittings were performed using compound concentrations and fluorescence peak intensities at 330 nm, and the inflection point and scale (slope) parameter were obtained using R software (version 4.4.1, R Foundation for Statistical Computing, Austria).

AlphaFold modeling of protein structures

Modeling of CgCS monomer was performed using an AlphaFold3 server. Modeling of the complex structure of CgCS dimer with OAA was constructed using a locally installed version of AlphaFold3, using the protein sequence and OAA as inputs. The top-ranked predicted complex was used. UCSF Chimera software (ver. 1.19) was used for visualization and analysis of protein structures.

MD simulation

MD simulations were performed to explore the interactions between CgCS and OAA. The initial protein coordinates for MD simulations were developed via homology modeling using a SWISS MODEL server⁵¹. The structure of citrate synthase from Francisella tularensis (PDB ID:3MSU) was used as a template. Each protein was simulated within a water box. The system comprised a CgCS dimer molecule, 40 OAA molecules, and 79, 834 water molecules inside a dodecahedron box with 15.37-nm sides. Sodium ions were added to neutralize the charge of the system. The AMBER14SB force field⁵² and TIP3P⁵³ were used to describe protein and water molecules, respectively. OAA was modeled using a generalized amber force field⁵⁴. The charge and atom types for the OAA molecule are shown in Supplementary Method 2. Sodium ions were modelled by the Joung–Cheatham model⁵⁵. Following energy minimization, MD simulations under constant pressure and temperature (NPT) of 1 bar and 300 K, respectively, were run for 0.1 ns to equilibrate the system. The production runs lasted 100 ns, with eight runs conducted at each pressure, each with different initial velocities. Pressure during NPT simulations was maintained using the Parrinello–Rahman method⁵⁶, and temperature was controlled using the Langevin dynamics, with water viscosity of 2 ps^–1. The LINCS method⁵⁷ was used to constrain covalent bonds involving hydrogen atoms in the proteins, and the time step for integration was set to 2.0 fs. These simulations were performed using GROMACS v.2022.5⁵⁸. To analyze the contacts between CgCS and OAA, eight snapshots of structure were examined after carrying out 100-ns MD simulations. When an OAA molecule was present within 0.5 nm of each amino acid, it was considered in contact with the amino acid and the number of contacts for that amino acid increased by one (Supplementary Method 2).

Statistics and reproducibility

The normality of the data was assessed using the Shapiro–Wilk test in R software. Comparisons between two groups were conducted using a two-sided unpaired Welch’s t-test. Comparisons among three or more groups was performed using one-way analysis of variance (ANOVA). When ANOVA indicated a statistically significant difference, F values and corresponding p-values were examined, followed by post-hoc multiple comparisons using Dunnett’s test or Tukey’s honestly significant difference (HSD) test, as appropriate. The Pearson’s correlation coefficient was calculated using the cor.test() function in R. The Smirnov–Grubbs test was performed, and data identified as outliers at a 5% significance level were excluded. All experiments were performed independently at least twice, and the exact numbers of biological and technical replicates are presented in the figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

M.Na. contributed to the acquisition and analysis of most data and drafted the manuscript; T.K. and K.K. contributed to MD simulation analysis. A.Y. contributed to modeling of protein structures; M.Ni. contributed to the conception of the study and revised the manuscript. S.K. contributed to the conception and design of the study and substantively revised the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

S.K. discloses support for the research of this work from JSPS KAKENHI (grant no. 24K01662) and the Cooperative Research Program of the Network Joint Research Center for Material and Device. Additionally, partial support was provided by JST SPRING (grant no. JPMJSP2108) for M.Na. T.K., K.K., A.Y., and M.Ni. declare no relevant funding.

Data availability

All data obtained in this study are provided in this article and its Supplementary Information file. Source data are provided with this paper. The genomic sequence information of Corynebacterium glutamicum ATCC 13032 [Kyowa Hakko] (KEGG organism code: cgl) was obtained from the KEGG database [https://www.genome.jp/kegg]. Protein structures used in this work include PDB IDs 3MSU and 3ENJ. The bacterial strains and plasmids generated in this study are available from the corresponding author [S.K.] under a Material Transfer Agreement (MTA) with the University of Tokyo. The corresponding author will aim to respond to requests for these materials within two weeks of receiving a request. Provision of the materials is subject to completion of the MTA. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76904-7.