Engineering of a LysG-derived arginine-specific biosensor for high-throughput screening of arginine overproducers in Corynebacterium glutamicum Authors: Ningyun Cai, Ye Wang, Pengcheng Ren, Deming Rao, Tuo Shi, Ziyao Wang, Jie Shen, Wenjuan Zhou, Linlin Qi, Lixian Wang, Ning Chen, Jiuzhou Chen, Ping Zheng, Jibin Sun Published: 2026 ABSTRACT Arginine is a semi-essential amino acid and holds significant value in the pharmaceutical and nutraceutical industries. Corynebacterium glutamicum is a promising host for arginine production, yet its industrial titers remain limited. Biosensor-based high-throughput screening enables rapid isolation of overproducers. However, the native LysG-based biosensor in C. glutamicum lacks arginine specificity, limiting its efficiency for screening high arginine-producing strains. Here, we developed an arginine-specific biosensor pLysGF222I in C. glutamicum, by combining semi-rational design and directed evolution of LysG, an endogenous regulator responsive to basic amino acids. Structural and interaction analyses revealed that the F222I substitution in LysG retained high affinity for arginine while significantly decreasing its responsiveness to histidine and lysine. Subsequently, this arginine-specific biosensor was successfully applied to screen a library of N-acetylglutamate kinase (NAGK) mutants, identifying 18 novel NAGK mutants with enhanced arginine production compared with previously reported variants. The strain integrated with the optimal NAGK mutant was further subjected to mutagenesis and screening, yielding an engineered strain that achieved an arginine titer of 56.07 g/L in a 5-L bioreactor, a 42% increase relative to the original starting strain. Overall, this study provides a powerful tool for accelerating the metabolic engineering of C. glutamicum for high-level arginine production.Graphical abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s13068-026-02776-0. Introduction As a semi-essential amino acid, arginine plays a pivotal role in supporting regular physiological activities, such as enhancing the secretion of growth hormones, promoting wound healing, and facilitating endogenous nitric oxide synthesis for vascular relaxation [1]. It also acts as an essential precursor during the enzymatic production of ornithine and citrulline [2, 3]. These diverse applications of arginine in the food, pharmaceutical, feed, and industrial sectors underscore its considerable market potential, evidenced by the rapid growth of its global market. Currently, microbial fermentation stands as the primary method for industrial arginine production, with Escherichia coli and Corynebacterium glutamicum serving as the predominant chassis strains [1, 4]. To improve arginine productivity, multiple studies have adopted systematic metabolic engineering strategies, encompassing pathway design, key enzyme engineering, modular optimization, fermentation process optimization, as well as mutagenesis and screening [1, 5–8]. Via extensive strain engineering efforts, arginine production has been increased to 132 g/L in E. coli and 112 g/L in C. glutamicum [9, 10]. Nevertheless, the maximum conversion rate of arginine remains merely 50%, highlighting significant room for further improvement compared to the much higher conversion rate achieved during the production of its precursor glutamate. Non-rational strategies, such as mutagenesis screening, have emerged as an effective approach to develop strains with favorable phenotypes without the need for a comprehensive understanding of intricate intracellular interaction networks and metabolic regulation mechanisms [11–13]. The major challenge in these strategies is establishing an efficient high-throughput screening method. In this regard, biosensor-based screening has proven to be a pivotal solution, enabling both the isolation of high-yielding producers and the identification of key genetic traits [14–18]. Over the past decades, a variety of biosensors based on transcriptional regulators, riboswitches, protein–protein interactions, and translation machinery have been developed for screening in biomanufacturing [14, 15, 19–23]. And in arginine biosynthesis, biosensors based on the activator-type transcription factors ArgP from E. coli and LysG from C. glutamicum have been developed and applied in high-throughput screening of enzymes and strains [9, 24, 25]. For instance, an ArgP-based biosensor pABIO10, which exhibits high arginine specificity, has been successfully employed to screen arginine-producing strains in E. coli. This screening facilitated the isolation and construction of an E. coli arginine overproducer with a titer enhanced up to 132 g/L [9]. The LysG-based biosensor has also been developed in C. glutamicum and utilized for high-throughput screening of the N-acetylglutamate kinase (NAGK) mutants. Due to its inherent responsiveness to all three basic amino acids, a high false-positive rate is encountered in such screening [24]. To date, there have been no reports on the use of LysG-based biosensors for screening arginine-producing strains. Given C. glutamicum’s inherent advantages as a glutamate producer and industrial chassis, the development of efficient, arginine-specific biosensors is imperative for obtaining high-producing strains and for the subsequent identification of novel targets and genetic elements to further enhance arginine biosynthesis [6, 26]. As a typical LysR-type transcriptional regulator (LTTR), LysG adopts the conserved LTTR structure, where a DNA-binding domain (DBD) and an effector-binding domain (EBD) are connected by a flexible linker helix (LH) [27]. In recent years, engineering of LysG has yielded a series of functionally diversified biosensors. For instance, directed evolution of LysG has led to the development of biosensors with improved detection performance and dynamic range, which have been successfully applied to screen strains with a 1.22-fold increase in lysine production [28]. Consistently, our previous research demonstrated that localized mutations in the LH region are effective in extending the operational range without affecting effector specificity [29]. A pLysGE58V biosensor showed a 4.5-fold higher induced fluorescence intensity than the wild-type biosensor, enabling efficient high-throughput screening of phosphoenolpyruvate carboxylase (a key enzyme for lysine production) mutants. In contrast, protein engineering has also been employed to alter the effector specificity of wild-type LysG. Based on structural insights, semi-rational design targeting key amino acid residues in the EBD of LysG led to the identification of the lysine-insensitive mutant LysGA219L, which was successfully applied to screen histidine-producing strains and identify novel genomic targets [27, 30]. Using a similar strategy, the lysine-specific mutant LysGE123Y,E125A was obtained, enabling the screening of high lysine-producing strains [31]. However, LysG-based biosensors with specific responsiveness to arginine have not been reported. In this study, we integrated a semi-rational strategy with directed evolution to engineer LysG and developed an arginine-specific LysG-based biosensor in C. glutamicum. Structural modeling and molecular interaction analyses elucidated the molecular basis of arginine-specific recognition by LysGF222I. With high specificity, it was successfully applied to fluorescence-activated cell sorting (FACS)-based high-throughput screening for both the key enzyme N-acetylglutamate kinase (NAGK) and high arginine-producing strains in C. glutamicum. This study broadens the utility of LysG and provides a valuable tool for arginine-producing strain screening in C. glutamicum. Materials and methods Chemicals and enzymes All chemicals were purchased from Sangon Biotech (Shanghai, China) and Solarbio Life Science (Beijing, China) unless otherwise specified. Restriction enzymes were purchased from New England Biolabs (Beijing, China). The DNA purification kit and plasmid extraction kit were purchased from TIANGEN BIOTECH (Beijing, China). TransStart® FastPfu DNA Polymerase was purchased from TransGen Biotech (Beijing, China). DpnI, T4 PNK, and T4 DNA ligase were purchased from TaKaRa (Beijing, China). ClonExpress MultiS One Step Cloning Kit was purchased from Vazyme (Nanjing, China). Primer synthesis and sequencing services were provided by GENEWIZ (Suzhou, China). All dipeptides were purchased from Hangzhou Allpeptide Biotechnology Co., Ltd. (Hangzhou, China). Bacterial strains and culture conditions The main bacterial strains used in this study are listed in Table S1. E. coli Trans1-T1 (TransGen Biotech Co. Ltd., Beijing, China) was used for the plasmid construction and was cultivated aerobically in Luria–Bertani (LB) medium (5 g/L NaCl, 10 g/L tryptone, 5 g/L yeast extract, pH 7.0) at 37 °C. E. coli BL21 (DE3) (TransGen Biotech Co. Ltd., Beijing, China) used for heterologous gene expression was cultivated in 2 × YT medium (5 g/L NaCl, 16 g/L tryptone, 10 g/L yeast extract, pH 7.0). Kanamycin (Km, 50 μg/mL), chloramphenicol (Cm, 20 μg/mL), or isopropyl-β-d-thiogalactopyranoside (IPTG, 0.1 mM or 0.5 mM) was added to the medium as required. C. glutamicum ATCC 13869 and its derivatives were cultivated aerobically at 30 °C in TSB medium or CGXIIY medium (2 g/L yeast extract, 20 g/L NH4SO4, 5 g/L urea, 1 g/L KH2PO4, 1.3 g/L K2HPO4·3H2O, 42 g/L MOPS, 0.25 g/L MgSO4·7H2O, 0.01 g/L FeSO4·7H2O, 0.01 g/L MnSO4·H2O, 0.001 g/L ZnSO4·7H2O, 0.2 mg/L CuSO4, 0.02 mg/L NiCl2·6H2O, 0.01 g/L CaCl2, 0.03 g/L protocatechuic acid, 0.2 mg/L biotin, and 0.1 mg/L vitamin B1, pH 7.2). The composition of TSB medium is as previously reported [12]. Where appropriate, Km (25 μg/mL), Cm (5 μg/mL), or IPTG (0.1 mM) was added in the medium. Plasmid construction All the plasmids and primers used in this study are listed in Tables S1 and S2, respectively. Plasmids were constructed via recombination using the ClonExpress MultiS One Step Cloning Kit (Vazyme, Nanjing, China). Primers used for plasmid construction and strain verification are listed in Table S2. For the construction of plasmid pET-28a-LysGWT, a fragment containing the lysG gene was amplified from the genomic DNA of C. glutamicum ATCC 13032 using the primer pair LysG-1/LysG-2. The backbone of pET-28a was amplified by PCR with the primer pair pET-28a-1/pET-28a-2. Then lysG was cloned into the E. coli expression vector pET-28a through recombination ligation, generating plasmid pET-28a-LysGWT. On the basis of the pET-28a-LysGWT plasmid, the F222I mutation was introduced by PCR using the primer pair F222I-1/F222I-2. Terminal phosphorylation and ligation of the PCR product were conducted with T4 PNK and T4 ligase. E. coli BL21 (DE3) was transformed with the ligation product for isolation of the correct transformant with pET-28a-LysGF222I. According to the method described above, E. coli BL21 (DE3) transformants harboring pET-28a-LysGE58V and pET-28a-LysGE58V,F222I were constructed. For the construction of plasmid pK18-∆LysEG, the left and right homologous flanks of the lysG and lysE genes were amplified from the genomic DNA of C. glutamicum ATCC 13869 using the primer pairs lysEG-UH-F/lysEG-UH-R and lysEG-DH-F/lysEG-DH-R, respectively. The two fragments were ligated with the linearized pK18mobsacB that was amplified from pK18mobsacB using the primer pair pK-F/pK-R to construct pK18-∆LysEG. For the construction of plasmid pXM-NAGK, a fragment containing NAGK was amplified from genomic DNA of C. glutamicum ATCC 13869 using the primer pair NAGK-1/NAGK-2. The backbones of pXMJ19 were amplified by PCR with the primer pairs p19-1/p19-2 and p19-3/p19-4. Then NAGK was cloned into the E. coli–C. glutamicum shuttle expression vector pXMJ19 through recombination ligation, generating plasmid pXM-NAGK. Genetic manipulation of C. glutamicum Deletion of the lysG and lysE genes in the C. glutamicum ATCC 13869 chromosome was performed by using the suicide plasmid pK18mobsacB via allele exchange [32]. The plasmid pK18-∆LysEG was transformed into C. glutamicum ATCC 13869 via electroporation for allelic exchange to generate 13869∆LysEG, following the procedure described previously [32, 33]. The same method was employed to generate the engineered strains, including the argR and farR double-deletion mutant 13869∆ArgR∆FarR, as well as strains in which the native chromosomal argB (the gene encoding NAGK) was replaced with mutant alleles, yielding 13869∆ArgR∆FarR NAGKA49V,M54V, 13869∆ArgR∆FarR NAGKH268E, R1, and R2. Generation of mutation library To construct a saturation mutagenesis library targeting 12 putative ligand-binding sites on LysG, site-saturation mutations were introduced into pLysGWT [29] using NNK codons covering all 20 amino acids. The targeted residues [27] were N95, D97, D124, E125, R143, H161, F189, G190, P191, D193, A219, and F222, and mutagenesis was performed using the corresponding primer pairs N95-F/N95-R, D97-F/D97-R, D124-F/D124-R, E125-F/E125-R, R143-F/R143-R, H161-F/H161-R, F189-F/F189-R, G190-F/G190-R, P191-F/P191-R, D193-F/D193-R, A219-F/A219-R, and F222-F/F222-R. Terminal phosphorylation and ligation of the PCR products were conducted with T4 PNK and T4 ligase, respectively. E. coli Trans1-T1 was transformed with the ligation products for isolation of correct transformants. The plasmid libraries were then extracted and transferred into 13869∆LysEG for selection. To construct the random mutation library of NAGK, error-prone PCR was employed to produce different mutation rates of NAGK with the primer pair NAGK-M-1/NAGK-M-2 using the genomic DNA of C. glutamicum ATCC 13869 as the template by adding 0.01 mM of MnCl2. Then the error-prone PCR products were purified and recombined with the plasmid pXM-NAGK, linearized by PCR with the primer pair pXM-F/pXM-R. Ligation products were transferred into E. coli Trans1-T1 competent cells. These plasmids, containing the NAGK mutation library and named pXM-NAGKmut, were extracted and then transformed into the competent cells of 13869∆ArgR∆FarR/pLysGF222I by electroporation. Atmospheric and room temperature plasma mutagenesis The R2 strains harboring either the pLysGWT or pLysGF222I biosensor were used for subsequent mutagenesis and screening. These strains were cultivated in TSB medium at 30 °C for 6 h. The active cells were washed twice using PBS and diluted to an OD600nm of 0.6. Then, a 10 μL cell suspension was transferred into a sterilized stainless-steel plate and treated with the atmospheric and room temperature plasma (ARTP) system (ARTP-IIIS, Wuxi Tmaxtree Biotechnology Co., ltd., China) to construct the mutagenesis library. The operating parameters were as follows: the radio frequency power input was 120 W, and the flow of pure helium was 10 L/min. The ARTP treatment time was set to 20, 60, or 100 s. After ARTP treatment, the cells were inoculated into TSB medium supplemented with 25 µg/mL Km and cultivated at 30 °C. Then the cells were diluted and spread on TSB agar plates. After 2 days of incubation at 30 °C, the total number of clones, control clones (A), and survival clones (B) were counted. The lethality rates were calculated as (1 − B/A) × 100%. FACS-based screening For FACS-based sorting of the LysG mutation library, the strain 13869ΔLysEG harboring the LysG saturation mutagenesis library was cultivated in 800 µL TSB medium with 25 μg/mL Km in a 24-well plate at 800 rpm and 30 °C for 10 h, with 13869ΔLysEG/pLysGWT used as the control strain. Cells were used as seed cultures to inoculate 800 µL of modified CGXIIY medium by adding 0.8 M arginine in 24-well plates to an optical density at 600 nm (OD600nm) of 0.5. After further cultivation for 6 h, the cells were diluted to an OD600nm below 0.05 with sterile phosphate buffered saline (PBS, 37 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4, pH 7.4) and used immediately for FACS analysis. eYFP fluorescence was analyzed by flow cytometry (BD FACSAria™ Fusion, BD Biosciences, USA) with the following parameters: excitation at 488 nm, emission fluorescence at 520 ± 14 nm, sample pressure of 45 psi. The nozzle diameter was 85 μm. Cells were captured on the signal channels of GFP (voltage 478 V), FSC (voltage 165 V), and SSC (voltage 269 V). In the first-round screening, cells maintaining arginine responsiveness were sorted (positive sorting). The sorted cells were cultured overnight in the TSB medium, and then the cells were transferred into 24-well plates containing CGXIIY medium supplemented with 0.6 M lysine and 50 mM histidine for the next round of screening. In the second-round screening, cells showing no response to lysine and histidine were sorted (negative sorting). Similarly, the third-round screening (positive sorting) was performed. After the three rounds of sorting, 200 cells with the top 0.05% of fluorescence intensities were sorted by FACS on TSB agar plates supplemented with 25 µg/mL Km, which were subsequently incubated for 20 h at 30 °C. For FACS-based sorting of the NAGK mutation library, the strain 13869∆ArgR∆FarR harboring the NAGK mutagenesis library was first cultivated in 800 µL TSB medium with 15 μg/mL Km and 5 μg/mL Cm in a 24-well plate at 800 rpm and 30 °C for 10 h. Cells were used as seed cultures to inoculate 800 µL of modified CGXIIY medium with 15 μg/mL Km and 5 μg/mL Cm in 24-well plates to an optical density at 600 nm (OD600nm) of 0.5. After further cultivation for 24 h, the cells were diluted to an OD600nm below 0.05 with sterile PBS and used immediately for FACS analysis. FACS sorting was performed following the previously described procedure. Gates in the eYFP channel were set to exclude low fluorescent cells. 100 cells with the top 0.05% of fluorescence intensities were sorted by FACS on TSB agar plates with 15 μg/mL Km and 5 μg/mL Cm, which were subsequently incubated for 20 h at 30 °C. For FACS-based sorting of the arginine overproducers, the ARTP-derived mutant strains obtained from R2/pLysGWT and R2/pLysGF222I were first cultivated in 800 µL TSB medium with 25 μg/mL Km in a 24-well plate at 800 rpm and 30 °C for 10 h. Cells were used as seed cultures to inoculate 800 µL of modified CGXIIY medium with 25 μg/mL Km in 24-well plates to an optical density at 600 nm (OD600nm) of 0.5. After further cultivation for 6 h, the cells were diluted to an OD600nm below 0.05 with sterile PBS and used immediately for FACS analysis. FACS sorting was performed following the previously described procedure. 100 cells with the top 0.02% of fluorescence intensities were sorted by FACS on TSB agar plates with 25 μg/mL Km, which were subsequently incubated for 20 h at 30 °C. Secondary screening in 96-well plates For secondary screening of LysG mutants, the 96-well plates were utilized for high-throughput cultivation of the strains obtained from FACS. C. glutamicum strains pre-grown in TSB medium were transferred into 96-well plates with 200 µL modified CGXIIY medium supplemented with either 0.8 M arginine, 50 mM histidine, or 0.6 M lysine to an OD600nm of 0.5. After cultivation at 30 °C and shaking at 800 rpm for 9 h, cells were harvested and diluted properly with PBS. eYFP fluorescence intensities were determined using a microplate reader (SpectraMax M5, Molecular Devices, λ excitation = 488 nm, λ emission = 520 nm). The fluorescence intensities were normalized with OD600nm. Fold induction was defined as the ratio of fluorescence output of a biosensor in the presence of 0.8 M arginine to that in the absence of effector. To characterize the specificity of LysG mutants, fluorescence assays were also performed in the presence of 4 mM dipeptides. The experimental procedure was identical to that described previously, except that the cultivation time was extended to 21 h. For secondary screening of NAGK mutants, the 96-well plates were utilized for high-throughput cultivation of the strains obtained from FACS. C. glutamicum strains pre-grown in TSB medium were transferred into 96-well plates with 200 µL modified CGXIIY medium. After cultivation at 30 °C and shaking at 800 rpm for 24 h, cells were harvested and diluted properly with PBS. eYFP fluorescence intensities were determined using a microplate reader (SpectraMax M5, Molecular Devices, λ excitation = 488 nm, λ emission = 520 nm). The fluorescence intensities were normalized to OD600nm. Strains pre-grown in TSB medium were transferred into 96-well plates with 200 µL modified ARG medium. Simultaneously, FACS-sorted strains were cultured in arginine seed medium for 7 h and then transferred to 96-well plates containing 200 µL of arginine fermentation medium, with an inoculation volume of 5% (v/v). After cultivation for another 24 h, arginine concentrations in the cultures were measured using high-performance liquid chromatography (HPLC) [9]. For secondary screening of the arginine overproducers, cells pre-grown in arginine seed medium were transferred into 24-well plates with 800 µL arginine fermentation medium with an inoculation size of 5% (v/v). After cultivation at 30 °C and shaking at 800 rpm for 20 h, arginine concentrations in the cultures were measured using HPLC. Molecular docking analysis of LysG The LysGF222I mutant model was constructed via homology modeling using SWISS-MODEL [34]. Ligands (arginine, histidine, lysine) were obtained from PubChem and converted to SYBYL Mol2 format with Open Babel. Wild-type and mutant receptors were prepared in LeDock’s [35] LePro module by removing heteroatoms/water, protonating residues at pH 7.0, and defining a 20 × 20 × 20 Å3 docking box centered on the co-crystallized arginine site (PDB:6XTV). LeDock generated 20 poses per ligand, followed by clustering (1.0 Å RMSD cutoff) to retain top-scoring conformations for analysis. Microscale thermophoresis (MST) assay LysG protein was expressed and purified as previously described [29]. All binding experiments were carried out on a Monolith Label Free/Red system (NanoTemper Technologies) and were performed in standard label-free capillaries (MO-Z022, NanoTemper Technologies). Enzyme or mutant concentration was kept constant (4 µM) and mixed at 1:1 with corresponding amino acids, which were serially diluted in dialysis buffer (20 mM Tris, 50 mM NaCl, 200 mM KCl, 50 mM MgCl2, 1 mM ethylenediaminetetraacetic acid (EDTA), 1 mM dithiothreitol (DTT), 0.05% (v/v) Tween 20, pH 7.4) from 25 mM to 1.53 µM. After 5 min incubation at room temperature, all the samples were loaded into glass capillaries and measurement was performed at 5% excitation power with medium MST power using the MO control software. Three independent experiments were repeated and the data were analyzed by MO affinity analysis software (v.2.3). Fnorm was calculated as the ratio of fluorescence after thermodiffusion (5 s) to the initial fluorescence. ΔFnorm was obtained by subsequently subtracting the Fnorm of the unbound state. Data were fitted to a dissociation constant (Kd) 1:1 binding model and the confidence interval of the Kd was calculated from the variance of the fitted parameter. Construction of plasmid-cured strain C3N To cure the pLysGF222I plasmid from the variant C3, cells were cultivated in TSB medium without Km at 220 rpm for 12 h. The culture was then serially diluted and spread onto TSB agar plates with or without Km. Colonies that grew on plates without Km but failed to grow on Km-containing plates were identified as the plasmid-cured strain C3N. Arginine fermentation For 24-well plate fermentations, strains were cultivated in 24-well plates containing 800 µL arginine seed medium in each well at 30 °C, 800 rpm for 8 h. The seed culture was then used to inoculate 24-well plates containing 800 µL arginine fermentation medium in each well to an initial OD600nm of 0.3. The 24-well plates were cultivated at 30 °C, 800 rpm for 21 h and cultivation was stopped for glucose and arginine analysis. The arginine seed medium contains 20 g/L glucose, 20 g/L tryptone, 0.5 g/L KH2PO4, 1.5 g/L K2HPO4·3H2O, 0.5 g/L MgSO4·7H2O, 10 g/L yeast extract, 5 mg/L biotin, and 20 g/L MOPS (pH 7.2). The arginine fermentation medium contains 80 g/L glucose, 20 g/L soy flour, 10 g/L urea, 0.7 g/L K2HPO4·3H2O, 0.5 g/L MgSO4·7H2O, and 40 g/L MOPS (pH 7.5). For 5-L bioreactor fermentation, strains were incubated and spread on TSB plates. A colony was transferred into 500-mL shake flasks containing 100 mL of RSG seed medium and cultivated at 30 °C and 240 rpm for 12 h. The seed culture (10% v/v) was inoculated into a 5-L bioreactor with 2 L of RPG fermentation medium and cultivated at 36 °C. The composition of RSG and RPG media is as previously reported [10]. During fermentation, the pH value (6.8 ± 0.05) was automatically maintained using aqueous ammonia (25%, v/v). Dissolved oxygen (DO) was maintained at 20–30% by controlling both the aeration rate and agitation speed. After the sugar initially added was consumed, 2 L of fresh medium containing 600 g/L glucose, 1.2 g/L MgSO4·7H2O, 8 g/L KCl, 30 g/L (NH4)2SO4, 1 g/L NH4H2PO4, 150 mg/L CaCl2, 20 mg/L β-alanine, 20 mg/L VB5, 20 mg/L VB1, and 0.2 mg/L biotin was continuously fed. The feeding rate of the solution was controlled to maintain the glucose concentration in the medium at a low concentration (below 0.5%). Analytical methods The cell growth was monitored by measuring the absorbance at 600 nm (OD600nm) with an ultraviolet spectrophotometer. The SBA-40C biosensor (Shandong Province Academy of Sciences, China) was used to measure the residual glucose concentration. The arginine concentration was detected by HPLC (1260 Infinity II, Agilent Technologies, China) using derivatization with an acetonitrile/water mixture (50:50, v/v) and 50 mM sodium acetate as the mobile phase at a flow rate of 1 mL/min, and quantified by UV detection at 360 nm. RNA-seq transcriptome analysis Cells of strains R2 and C3N were harvested at the mid-exponential phase during fed-batch fermentation and used for RNA extraction. RNA preparation, library construction, and transcriptome sequencing were performed by Novogene Co., Ltd., yielding a total of 4.96 Gbp of data [2.48 ± 0.45 Gbp per sample, mean ± standard deviation (SD)]. After quality filtering, reads were aligned to the ATCC 13869 reference genome (GenBank: CP016335.1). Differentially expressed genes were identified using DESeq2 [36] with |log2FoldChange|≥ 1 and false discovery rate (FDR) < 0.05. Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis were conducted in R. Results and discussion Construction of arginine-specific biosensor To develop an arginine-specific biosensor in C. glutamicum, we initially evaluated a heterologous system based on the E. coli ArgP regulator and its cognate PargO promoter. This system has been proven to exhibit high arginine specificity and has been employed for screening arginine-producing strains in E. coli [9]. Unfortunately, it proved nonfunctional in C. glutamicum (Fig. S1), which might be attributed to the impact of interhost differences in protein expression and recognition of genetic elements. Consequently, an alternative strategy involving the engineering of the native broad-spectrum transcriptional regulator LysG was pursued. First, the lysG and lysE genes were deleted in the industrial glutamate-producing chassis strain C. glutamicum ATCC 13869, generating the strain 13869ΔLysEG with eliminated endogenous regulation. Subsequently, based on the biosensor plasmid pLysGWT [29], site-saturation mutagenesis libraries were constructed targeting the effector-associated sites in LysG’s EBD region, including eight residues directly contacting the amino acid ligand (N95, D97, D124, E125, H161, F189, D193, and F222) and four second-shell residues (R143, G190, P191, and A219) [27]. For each site-saturation mutation library, about 1,000 colonies were collected for plasmid extraction. The 12 plasmid libraries were then mixed and transformed into 13869∆LysEG. To identify arginine-specific LysG variants, the cell population was subjected to a FACS-based three-step positive–negative–positive screening. The first positive screening was conducted in the presence of 0.8 M arginine to identify variants maintaining arginine responsiveness (isolation of the top 15% fluorescing cells) and the subsequent negative screening was deployed in the presence of 0.6 M lysine and 50 mM histidine to exclude the variants that still responded to lysine and histidine (isolation of the bottom 10% non-fluorescing cells). Subsequently, the final round of positive screening with 0.8 M arginine was applied to separate and individually collect the high-fluorescence variants (top 0.05%). The sorted cells were recultured on agar plates (Fig. 1a). Following these screening processes, 137 colonies were selected for mutation enrichment analysis. DNA sequencing of these variants revealed that 4 kinds of LysG mutants were obtained with different amino acid sequences and the LysGF222I was significantly enriched, accounting for 91.24% of the total (Table S3). Owing to extensive sample sequencing, relatively rare LysG variants, including LysGF222L, LysGF189W, and LysGA219I, were also identified.Fig. 1Screening for arginine-specific biosensors through directed evolution of LysG. a Schematic illustration of the screening procedures for LysG variants specifically responsive to arginine. The cell population underwent iterative selection to enrich arginine-specific biosensors, including a positive screen with the target ligand (0.8 M arginine), a subsequent negative screen with non-target ligands (0.6 M lysine and 50 mM histidine), and a final round of positive screening with 0.8 M arginine to eliminate false positives. Flow cytometric analysis of strain 13869∆LysEG cultured with arginine (green), histidine (blue), or lysine (pink). Flow cytometric analysis of LysG libraries cultured with arginine (grey) or with histidine and lysine (black). The sorting gate is indicated by a purple box in the histogram. b Fluorescence response intensities of different pLysG variants and the control strain harboring the empty pEC-XK99E plasmid under conditions supplemented with 0.8 M arginine, 0.6 M lysine, or 50 mM histidine. c Fluorescence response intensities of the pLysGWT and pLysGF222I biosensors in the presence of 4 mM dipeptides. Values and error bars reflect the mean ± SD of three biological replicates (n = 3) Screening for arginine-specific biosensors through directed evolution of LysG. a Schematic illustration of the screening procedures for LysG variants specifically responsive to arginine. The cell population underwent iterative selection to enrich arginine-specific biosensors, including a positive screen with the target ligand (0.8 M arginine), a subsequent negative screen with non-target ligands (0.6 M lysine and 50 mM histidine), and a final round of positive screening with 0.8 M arginine to eliminate false positives. Flow cytometric analysis of strain 13869∆LysEG cultured with arginine (green), histidine (blue), or lysine (pink). Flow cytometric analysis of LysG libraries cultured with arginine (grey) or with histidine and lysine (black). The sorting gate is indicated by a purple box in the histogram. b Fluorescence response intensities of different pLysG variants and the control strain harboring the empty pEC-XK99E plasmid under conditions supplemented with 0.8 M arginine, 0.6 M lysine, or 50 mM histidine. c Fluorescence response intensities of the pLysGWT and pLysGF222I biosensors in the presence of 4 mM dipeptides. Values and error bars reflect the mean ± SD of three biological replicates (n = 3) To characterize the specificity of the screened LysG variants, the four mutants were cultivated in 96-well plates supplemented separately with arginine, histidine, or lysine, and their fluorescence response was quantitatively assessed. All strains carrying the different biosensor exhibited fluorescence intensities similar to those of the control strain harboring the empty pEC-XK99E plasmid in the absence of effector addition (Fig. S2). Compared with the LysGWT, all mutants maintained their responsiveness to arginine. However, most exhibited significantly reduced or even abolished sensitivity to lysine, while displaying variable responses to histidine (Fig. 1b). For instance, LysGA219I retained a response to histidine and the LysGF189W mutant showed a markedly reduced histidine response. The phenotype of LysGA219I is broadly similar to that of the previously reported LysGA219L variant, likely due to the similar hydrophobic properties of isoleucine and leucine [27]. In contrast, LysGF222I and LysGF222L completely lost responsiveness to histidine and displayed a distinct shift toward arginine specificity. Furthermore, these two mutants (LysGF222I and LysGF222L) also exhibited higher fold induction values toward arginine (6.69 and 5.84, respectively) than LysGWT (fold induction = 4.44) (Fig. 1b). Given its higher fold induction activity and arginine specificity, LysGF222I was selected for subsequent characterization. Dose–response curves were firstly measured for strains carrying LysGWT, the LysGF222I mutant, or the empty pEC-XK99E plasmid across a range of arginine concentrations (0 to 1.2 M) (Fig. S3). As expected, elevated fluorescence signals from both pLysGWT and pLysGF222I were observed upon the addition of arginine. While both pLysGWT and pLysGF222I biosensors exhibited an excellent dynamic range, pLysGF222I showed a significantly higher fluorescence response at the same arginine concentration. To further assess the specificity of the pLysGF222I biosensor, dipeptides comprising basic amino acids (Lys-Ala, Arg-Ala, and His-Ala) were used at the same concentration to avoid the growth inhibition caused by high amino acid concentrations [29]. Compared with pLysGWT, pLysGF222I exhibited significantly attenuated responses to Lys-Ala and His-Ala, with decreases of 91.74% and 93.00%, respectively, while maintaining a specific response to Arg-Ala (Fig. 1c). In summary, these results demonstrate that pLysGF222I functions as a highly arginine-specific biosensor, successfully developed through the engineering LysG in C. glutamicum. Characterization of the mutant LysGF222I Residue F222, one of eight key residues in the ligand-binding pocket of the reported [LysG + Arg] co-crystal structure, is predicted to interact with arginine and the other two basic amino acids [27]. To investigate the mechanism underlying arginine selectivity over lysine and histidine in the LysGF222I mutant, comparative structural analyses of LysGWT and LysGF222I were performed using molecular docking. Docking analysis suggests that arginine binding in LysGWT is stabilized by a triple hydrogen-bond network anchored by the D193 carboxylate and the L234 backbone carbonyl. In contrast, lysine and histidine form only a single hydrogen bond with D97. F222 engages in a cation–π interaction with all three amino acids, potentially contributing to their stabilization (Fig. 2a). The substitution of phenylalanine (F) with isoleucine (I) at position 222 would eliminate the aromatic side chain required for cation-π binding. Consequently, arginine might retain its binding pose through its extensive hydrogen-bonding framework, whereas lysine and histidine—deprived of the crucial cation–π anchor and deficient in compensatory hydrogen bonds—may exhibit reduced binding stability (Fig. 2b). These findings suggest that the LysGF222I mutant might achieve arginine-specific responsiveness by reducing its binding affinity for histidine and lysine.Fig. 2Molecular docking and MST analyses of interactions between LysG variants and effectors. a LysGWT docked with arginine, histidine, and lysine (cyan sticks), showing conserved hydrogen bonds (red dashed lines) and cation–π interactions (blue dashed lines) between ligands and residue F222 (green sticks). b The LysGF222I mutant docking demonstrates disrupted cation–π interactions with Ile222. Critical binding site residues are explicitly labeled. Orange arrows indicate no binding. c Dose–response curves for the binding interactions between LysGE58V mutant and basic amino acids. d Dose–response curves for the binding interactions between LysGE58V,F222I mutant and basic amino acids. ΔFnorm, normalized fluorescence from which the relative fluorescence of the unbound state was subtracted (see Methods, MST). Data represent mean ± SEM of three independent measurements Molecular docking and MST analyses of interactions between LysG variants and effectors. a LysGWT docked with arginine, histidine, and lysine (cyan sticks), showing conserved hydrogen bonds (red dashed lines) and cation–π interactions (blue dashed lines) between ligands and residue F222 (green sticks). b The LysGF222I mutant docking demonstrates disrupted cation–π interactions with Ile222. Critical binding site residues are explicitly labeled. Orange arrows indicate no binding. c Dose–response curves for the binding interactions between LysGE58V mutant and basic amino acids. d Dose–response curves for the binding interactions between LysGE58V,F222I mutant and basic amino acids. ΔFnorm, normalized fluorescence from which the relative fluorescence of the unbound state was subtracted (see Methods, MST). Data represent mean ± SEM of three independent measurements To verify this hypothesis, the binding affinities of LysGWT and LysGF222I for basic amino acids were assessed by microscale thermophoresis (MST). However, no detectable signals were observed for either protein, presumably because the conformational changes induced by these effectors are too subtle to be resolved by MST. To enhance LysG’s conformational flexibility and promote structural rearrangements that improve protein–DNA binding dynamics and signal sensitivity, the previously reported E58V mutation which exerts these effects was introduced into the LysGF222I mutant [29]. This modification did not alter the mutant’s effector specificity, while enhancing its induction activity (Fig. S4). MST results confirmed that LysGE58V exhibited relatively comparable binding affinities for arginine, histidine, and lysine, with Kd of 7.3 mM, 6.7 mM, and 12.6 mM, respectively (Fig. 2c). In contrast, the LysGE58V,F222I mutant retained arginine-binding capacity with enhanced affinity (Kd = 3.4 mM) but showed substantially reduced affinity for histidine and lysine (Fig. 2d). These findings are consistent with our structural analysis and corroborate the previous specificity characterization results, supporting the conclusion that the F222I mutant confers arginine-specific responsiveness by retaining enhanced arginine-binding capacity while displaying reduced affinity for histidine and lysine. To date, no studies have investigated the specific role of F222 in effector binding of LysG. Moreover, mutations at this position have not yet been identified to modulate the specificity of LysG. According to previous studies, the LysGE123Y mutant exhibited a diminished histidine response, whereas the LysGA219L mutation abolished the response to lysine [27, 31]. These findings demonstrate that residues E123 and A219 are critical for the binding of histidine and lysine, respectively. However, each of these mutations abrogates the binding of only one target amino acid, rendering them unsuitable for the development of biosensors with single-ligand specificity. In contrast, residue F222 is the first to be identified to be critical for the binding of both histidine and lysine, which broadens our understanding of the molecular mechanism underlying LysG–effector interactions. Furthermore, the LysGF222I mutant identified in our study eliminated responses to both amino acids, thus enabling the development of an arginine-specific biosensor that is well suited for high-throughput screening applications. Directed evolution of NAGK based on pLysGF222I NAGK catalyzes a critical step in the arginine biosynthesis pathway and is subject to strong feedback inhibition by arginine in C. glutamicum [19, 37]. Relieving this inhibition is therefore a primary objective for arginine overproduction. To evaluate the applicability of pLysGF222I for high-throughput screening, NAGK was first selected as the target for directed evolution. The NAGKA49V,M54V mutant, previously reported to relieve feedback inhibition and promote arginine accumulation [38], was used as a positive control to evaluate the performance of different LysG-derived biosensors. In C. glutamicum, arginine biosynthesis is tightly controlled by the transcriptional repressors ArgR and FarR, which inhibit the expression of the arg operon (e.g., NAGK) [39, 40]. To eliminate endogenous regulatory interference, the genes argR and farR were deleted in C. glutamicum ATCC 13869, generating the host strain 13869∆ArgR∆FarR (Fig. 3a). Overexpression of NAGKA49V,M54V in the 13869ΔArgRΔFarR strain increased arginine production from 0.19 g/L to 2.01 g/L (Fig. 3b). Subsequently, fluorescence intensities were measured in the 13869ΔArgRΔFarR strain overexpressing either NAGKWT (wild-type) or the NAGKA49V,M54V mutant using the biosensors pLysGWT, pLysGF222I, pLysGE58V,F222I (Fig. 3c). The results showed that across all tested biosensors, strains overexpressing the NAGKA49V,M54V mutant exhibited higher fluorescence intensity than those overexpressing NAGKWT. Among the three biosensors, pLysGF222I displayed the strongest discriminatory ability, the mutant-overexpressing strain showed a 3.08-fold higher fluorescence intensity than the wild-type NAGK overexpressing strain, in contrast to a 1.61-fold increase observed under the pLysGWT. Although pLysGE58V,F222I exhibited the highest overall fluorescence intensity, the fold change (mutant vs. wild-type NAGK) was only 1.89-fold. These results confirm that the pLysGF222I biosensor offers an enhanced response and superior discrimination of arginine-producing strains, establishing it as a robust tool for the high-throughput screening of NAGK variants and arginine overproducers.Fig. 3High-throughput screening system testing. a Arginine biosynthesis pathway in C. glutamicum. Red dashed lines indicate feedback inhibition, orange arrows indicate FarR-mediated repression, purple arrows indicate ArgR-mediated repression, and the symbol “ × ” indicates gene deletion. G6P glucose 6-phosphate, F6P fructose-6-phosphate, α-KG a-ketoglutarate, Ac-Glu acetylglutamate, Ac-Glu-P acetylglutamyl phosphate, Ac-Glu-SA acetylglutamate semialdehyde, Ac-Orn acetylornithine, Orn ornithine, Cit citrulline, Arg-Suc arginosuccinate. b Arginine production by strains overexpressing wild-type NAGK (NAGKWT) and the mutant NAGKA49V,M54V cultivated in CGXIIY medium. c Fluorescence intensities of strains overexpressing NAGKWT and NAGKA49V,M54V measured in CGXIIY medium. Values and error bars represent the mean ± SD of three biological replicates (n = 3) High-throughput screening system testing. a Arginine biosynthesis pathway in C. glutamicum. Red dashed lines indicate feedback inhibition, orange arrows indicate FarR-mediated repression, purple arrows indicate ArgR-mediated repression, and the symbol “ × ” indicates gene deletion. G6P glucose 6-phosphate, F6P fructose-6-phosphate, α-KG a-ketoglutarate, Ac-Glu acetylglutamate, Ac-Glu-P acetylglutamyl phosphate, Ac-Glu-SA acetylglutamate semialdehyde, Ac-Orn acetylornithine, Orn ornithine, Cit citrulline, Arg-Suc arginosuccinate. b Arginine production by strains overexpressing wild-type NAGK (NAGKWT) and the mutant NAGKA49V,M54V cultivated in CGXIIY medium. c Fluorescence intensities of strains overexpressing NAGKWT and NAGKA49V,M54V measured in CGXIIY medium. Values and error bars represent the mean ± SD of three biological replicates (n = 3) Next, a random mutation library of the full-length NAGK was constructed by error-prone PCR, yielding approximately 5 × 104 colonies. Sequence analysis indicated an average of 3.25 nucleotide mutations per gene, providing sufficient sequence diversity for subsequent screening. This plasmid library was then transformed into 13869∆ArgR∆FarR containing pLysGF222I and subjected to primary screening via FACS (Fig. 4a). Approximately 200 cells representing the top 0.05% of fluorescence intensities were sorted and recovered on agar plates. From these, 152 individual colonies were picked and cultivated in 96-well plates for a second round of screening (Fig. S5). Although three variants showed lower fluorescence intensities, most selected mutants exhibited higher fluorescence intensities than the strain overexpressed NAGKWT. Notably, 114 out of 152 mutants (76.97%) displayed higher fluorescence than the positive control strain, which overexpressed the NAGKA49V,M54V mutant. Sequence analysis of the top 25 revealed 18 kinds of NAGK mutants harboring various amino acid substitutions. Among these, five mutants were recurrently identified, appearing 2–3 occurrences each (Table S4). Then, strains overexpressing these NAGK variants and the positive control strain overexpressing NAGKA49V,M54V were selected for 24-well plate-based fermentation in ARG2 medium. Expression of these mutants resulted in much higher arginine titers than the control strain. The best two mutants NAGK1-67 and NAGK1-48 showed over a 4-fold improvement of arginine titers compared with the positive control NAGKA49V,M54V (Fig. 4b).Fig. 4Directed evolution of NAGK using pLysGF222I. a Schematic illustration of the directed evolution of NAGK in C. glutamicum and FACS-based screening of desired mutants. The sorting gate is indicated by an orange box in the histogram. b Arginine production of strains overexpressing wild-type NAGK and its mutants after 24 h of cultivation in ARG2 medium. c Arginine production and yield of strains in which the chromosomal NAGK was replaced with a mutant allele after 20 h of cultivation in a 24-well plate. Values and error bars represent the mean ± SD of three biological replicates (n = 3) Directed evolution of NAGK using pLysGF222I. a Schematic illustration of the directed evolution of NAGK in C. glutamicum and FACS-based screening of desired mutants. The sorting gate is indicated by an orange box in the histogram. b Arginine production of strains overexpressing wild-type NAGK and its mutants after 24 h of cultivation in ARG2 medium. c Arginine production and yield of strains in which the chromosomal NAGK was replaced with a mutant allele after 20 h of cultivation in a 24-well plate. Values and error bars represent the mean ± SD of three biological replicates (n = 3) Among these NAGK variants, several mutation sites align with previously reported key sites or regions. For instance, mutations in the N-terminal region and substitutions at specific residues (e.g., K47E, A49V) have been proven to be critical for relieving arginine feedback inhibition [19, 24, 38]. Consistent with these reports, our screened mutants include variants harboring these key mutations, including NAGK1-13 and NAGK2-1 (both including K47E), NAGK2-62 (including A49V), and NAGK2-70, NAGK1-48, NAGK1-72, NAGK2-65 (all including N-terminal region mutations). Moreover, our biosensor enabled the discovery of novel mutants with previously unreported mutation profiles (e.g., NAGK1-43, NAGK1-67) exhibiting superior arginine-producing performance. Notably, these mutation sites are distinct from characterized key regions, suggesting that NAGK feedback inhibition is governed by a complex regulatory mechanism and can be alleviated through multiple routes. This outcome highlights a key advantage of the pLysGF222I biosensor: its enhanced response intensity enables robust discrimination of arginine-producing variants. This ability expands accessible sequence space, facilitates high-throughput screening of key enzymes, and minimizes false-positive enrichment. To evaluate the applicability of the novel NAGK mutants for arginine production, the native NAGK in 13869∆ArgR∆FarR was replaced with NAGK1-48, NAGK1-67, NAGKA49V,M54V and NAGKH268E. The last two mutants were generated by introducing previously reported high-performing NAGK variants [41]. Both novel mutants outperformed the previously reported mutants (Fig. 4c) and the best-performing strain R2 (NAGK1-67) achieved an arginine titer of 6.05 g/L and a yield of 0.08 g/g glucose. Taken together, by employing the highly specific and robustly responsive biosensor, we successfully conducted efficient high-throughput screening of the key rate-limiting enzyme NAGK and obtained novel NAGK mutants with enhanced arginine synthesis capability, providing new genetic elements for arginine biosynthesis and facilitating strain-level improvement. Furthermore, we also constructed high-arginine-producing strains based on superior variants, laying a solid foundation for subsequent high-throughput screening of arginine overproducers. High-throughput screening of arginine overproducers based on pLysGF222I Biosensors with high specificity and robust response intensity are critical for efficiently identifying and isolating high-producing strains. To evaluate the utility of pLysGF222I in the screening of arginine overproducers, an untargeted mutagenesis and high-throughput screening workflow was conducted. The R2 strain with its chromosomal wild-type NAGK replaced by the optimal mutant NAGK1-67 was used as the starting strain. First, a series of mutant libraries was generated by subjecting R2/pLysGF222I to ARTP treatment for 20, 60, or 100 s, which produced different lethality rates (Table S5) and thus yielded libraries with distinct mutation rates. These libraries were subsequently mixed, cultivated in CGXIIY medium for 6 h, and subjected to FACS analysis to select variants with improved arginine production (Fig. 5a). Meanwhile, the same workflow was applied to R2 harboring the pLysGWT biosensor. Approximately 100 cells displaying the top 0.02% of fluorescence intensities in each biosensor were sorted by FACS and cultivated in 96-well plates for fermentation assessment.Fig. 5High-throughput screening of arginine overproducers using the pLysGF222I biosensor. a Schematic illustration of ARTP mutagenesis coupled with FACS sorting for evolutionary screening of arginine overproducers. The sorting gate is indicated by an orange box in the histogram. b Deep-well plate fermentation results of arginine-producing colonies sorted using the pLysGF222I biosensor. The red box indicates the R2/pLysGF222I strain used as the reference for normalization, with its value set as 1. c Deep-well plate fermentation results of arginine-producing colonies sorted using the pLysGWT biosensor. The red box indicates the R2/pLysGWT strain used as the reference for normalization, with its value set as 1. Values and error bars represent the mean ± SD of three biological replicates (n = 3). *P < 0.05, **P < 0.01, ***P < 0.001 High-throughput screening of arginine overproducers using the pLysGF222I biosensor. a Schematic illustration of ARTP mutagenesis coupled with FACS sorting for evolutionary screening of arginine overproducers. The sorting gate is indicated by an orange box in the histogram. b Deep-well plate fermentation results of arginine-producing colonies sorted using the pLysGF222I biosensor. The red box indicates the R2/pLysGF222I strain used as the reference for normalization, with its value set as 1. c Deep-well plate fermentation results of arginine-producing colonies sorted using the pLysGWT biosensor. The red box indicates the R2/pLysGWT strain used as the reference for normalization, with its value set as 1. Values and error bars represent the mean ± SD of three biological replicates (n = 3). *P < 0.05, **P < 0.01, ***P < 0.001 Among colonies screened by pLysGF222I, 23.21% (13 of 56) exhibited a statistically significant increase of at least 10% in arginine titer compared with the control strain R2/pLysGF222I (Fig. 5b). The variant C3 screened by pLysGF222I displayed the highest arginine production, with a 38.33% increase relative to R2/pLysGF222I. In contrast, parallel screening using pLysGWT identified only a single colony with a comparable increase. The variant D6, isolated using pLysGWT, exhibited merely an 18.47% increase in arginine production. Instead, 50.91% of the colonies (28 of 55) showed a more than 10% decrease in arginine production (Fig. 5c). This discrepancy is likely due to the broader effector specificity of pLysGWT, which might enrich strains overproducing any basic amino acids and thereby reduce the selection efficiency for the desired phenotype. These results confirm that the pLysGF222I biosensor significantly improves the efficiency of isolating arginine-overproducing mutants. Combining the ARTP-generated large-scale mutant library (up to 10⁷ clones) with FACS, the arginine-specific pLysGF222I biosensor enables ultra-high-throughput screening that can be completed within several days. Compared with traditional methods such as analog-based screening, this approach exhibits significantly higher efficiency and accuracy, providing a powerful tool for rapid isolation of arginine overproducers. Fed-batch fermentation of the evolved strains for arginine production To assess the potential production performance of the evolved strains, fed-batch fermentations were conducted. Firstly, pLysGF222I was cured from C3, generating strain C3N. Then, C3N and the parental strain R2 were subjected to fed-batch fermentation in a 5-L bioreactor with a total glucose of 1320 g (Fig. 6). The parental strain R2 exhibited a rapid glucose consumption rate, achieving an arginine titer of 39.49 g/L with a glucose-to-arginine yield of 0.11 g/g. In contrast, the evolved strain C3N showed significantly faster growth and arginine production rates. At the same glucose consumption level, C3N produced an arginine titer of 56.07 g/L, representing a 42% improvement compared to R2. Furthermore, its glucose-to-arginine yield reached 0.15 g/g, a 36% increase relative to the parental strain. These results indicate that the evolved strain C3N, which carries additional beneficial ARTP-induced mutations, can significantly enhance the arginine synthesis efficiency.Fig. 6Fed-batch fermentation for arginine production using strains R2 (black lines) and C3N (red lines). Strain R2 is derived from 13869∆ArgR∆FarR, in which the chromosomal NAGKWT has been replaced with NAGK1-67. Strain C3N is a derivative of R2 obtained through mutagenesis. Cultivation was performed in 5-L bioreactors with RPG medium Fed-batch fermentation for arginine production using strains R2 (black lines) and C3N (red lines). Strain R2 is derived from 13869∆ArgR∆FarR, in which the chromosomal NAGKWT has been replaced with NAGK1-67. Strain C3N is a derivative of R2 obtained through mutagenesis. Cultivation was performed in 5-L bioreactors with RPG medium To elucidate mechanisms underlying the enhanced arginine production in the C3N strain, a comparative transcriptomic analysis was conducted between C3N and its parental strain R2. Cells were harvested at 20 h of fed-batch fermentation for analysis, a time point corresponding to the mid-exponential growth phase and the onset of rapid arginine accumulation (Fig. 6). Transcriptomic profiling revealed that 61 genes were significantly upregulated and 71 genes were significantly downregulated in C3N relative to R2 (Fig. S6a). Notably, no differentially expressed genes were directly mapped to the arginine biosynthesis pathway. However, the expression of lysE, which encodes the basic amino acid exporter, was upregulated 2.50-fold, likely facilitating arginine secretion and contributing to the enhanced production phenotype [42, 43]. Additionally, aceA (encoding isocitrate lyase) was downregulated to 0.35-fold, potentially suggesting a redirection of metabolic flux from the glyoxylate cycle toward α-ketoglutarate, a key precursor for arginine biosynthesis. KEGG enrichment analysis further revealed that sulfur metabolism was significantly downregulated in C3N (Fig. S6b). As a fundamental physiological process, sulfur metabolism plays a critical role in energy conversion, redox homeostasis, and signal transduction [44–46]; however, its specific impact on arginine biosynthesis remains to be elucidated. Overall, these results highlight that the evolved strain C3N, obtained via biosensor-assisted screening, serves as a promising chassis for arginine production in C. glutamicum. Further in-depth investigation into the genetic disparities between C3N and R2 will be instrumental in developing strategies to further enhance arginine titers. Conclusions In summary, this study generated LysG mutants with diminished responsiveness to lysine and histidine while retaining arginine specificity through a semi-rational engineering approach. An efficient arginine-specific biosensor pLysGF222I was subsequently developed and successfully applied for the high-throughput screening of high-performance NAGK mutants and the subsequent isolation of arginine-overproducing strains. These efforts yielded multiple novel NAGK mutants and high-yield arginine-producing strains, thus providing valuable chassis strains for further metabolic engineering optimization and laying a solid foundation for discovering novel genetic elements that boost arginine biosynthesis. Collectively, this biosensor-assisted optimization strategy for key biosynthetic enzymes and producer strains offers a representative paradigm for the development of industrial-grade arginine-producing strains. Supplementary Information Additional file 1: Fig. S1. Testing of the ArgP-based biosensor in C. glutamicum. Fig. S2. Fluorescence response intensities of different pLysG variants and the control strain harboring the empty pEC-XK99E plasmid under conditions without effector addition. Fig. S3. Dose–response curves of the biosensors based on LysGWT, LysGF222I and the empty-plasmid control strain. Fig. S4. Fluorescence response intensities of different pLysG variants carrying the E58V mutation and the control strain harboring the empty pEC-XK99E plasmid under conditions without effector addition and with supplementation of 0.8 M arginine, 0.6 M lysine, or 50 mM histidine. Fig. S5. Preliminary evaluation of fluorescence intensity of NAGK variants cultured in CGXIIY medium for 24 h. Fig. S6. Comparative transcriptome analysis of C3N and R2. Table S1. Strains and plasmids used in this study. Table S2. Primers used in this study. Table S3. The amino acid substitutions in the LysG variants screened from the site-saturation mutation library. Table S4. The amino acid substitutions in the NAGK variants screened from the random mutation library. Table S5. Lethality rates at different ARTP treatment times. (DOCX 1907 KB)Additional file 2: GenBank sequence of pLysGF222I used in this study. (GB 13 KB) Additional file 1: Fig. S1. Testing of the ArgP-based biosensor in C. glutamicum. Fig. S2. Fluorescence response intensities of different pLysG variants and the control strain harboring the empty pEC-XK99E plasmid under conditions without effector addition. Fig. S3. Dose–response curves of the biosensors based on LysGWT, LysGF222I and the empty-plasmid control strain. Fig. S4. Fluorescence response intensities of different pLysG variants carrying the E58V mutation and the control strain harboring the empty pEC-XK99E plasmid under conditions without effector addition and with supplementation of 0.8 M arginine, 0.6 M lysine, or 50 mM histidine. Fig. S5. Preliminary evaluation of fluorescence intensity of NAGK variants cultured in CGXIIY medium for 24 h. Fig. S6. Comparative transcriptome analysis of C3N and R2. Table S1. Strains and plasmids used in this study. Table S2. Primers used in this study. Table S3. The amino acid substitutions in the LysG variants screened from the site-saturation mutation library. Table S4. The amino acid substitutions in the NAGK variants screened from the random mutation library. Table S5. Lethality rates at different ARTP treatment times. (DOCX 1907 KB) Additional file 2: GenBank sequence of pLysGF222I used in this study. (GB 13 KB)