Optimization of an improved chemically defined medium for high-efficiency expression of glucokinase in Escherichia coli

Abstract

Background

Recombinant glucokinase (GLK), an important enzyme for blood glucose diagnostics, is commonly prepared in Escherichia coli (E.coli )for laboratory research and related biosynthetic applications. In small-scale laboratory cultivation, complex media are often used because of their convenience and strong growth-supporting capacity, but their ill-defined composition and batch-to-batch variability can compromise reproducibility and increase medium-related costs. Although minimal defined medium (e.g., M9 medium) are generally preferred in scalable E. coli bioprocesses because of their compositional consistency and process controllability, their relatively simple nutrient composition may limit recombinant protein expression performance under laboratory-scale conditions. Therefore, developing an improved chemically defined medium represents a practical strategy to balance expression efficiency, compositional control, and medium cost for recombinant GLK production.

Results

To address this, we developed a chemically defined medium including amino acids using high-throughput respiration activity monitoring system (RAMOS) and a high-throughput growth curve analysis system (Growth Profiler 960, GP960), alongside a machine learning-driven iterative optimization system (Machine-learning guided Experimental Trials for Improvement of Systems, METIS). The optimized medium enhanced soluble GLK expression by 33 fold compared to the pre-optimization medium, reaching 74% of the titer obtained in complex media, while reducing medium costs by 74%. Moreover, soluble protein expression per cell increased by 80% compared to the complex medium, lowering the medium usage cost per unit of bacterial protein by 85%. Validation in a 7.5 L bioreactor demonstrated that the optimized medium supported continuous growth to an optical density at 600 nm (OD₆₀₀) of 103, achieving 4-fold and 5.4-fold higher biomass than Luria-Bertina (LB) and M9 media, respectively, with soluble GLK titer reaching 920 mg L^− 1, which exceeded conventional media by more than 10-fold.

Conclusion

This study tackles the challenge of suboptimal media in GLK fermentation, offering a cost-effective alternative and a transferable strategy for optimizing fermentation processes for other high-value enzymes.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12934-026-03027-5.

Background

Efficient preparation of recombinant proteins is essential for laboratory research in biodetection, enzyme engineering, and related biosynthetic applications. In such studies, reducing medium-related costs can improve the overall efficiency of experimental work, especially when repeated small-scale cultivation is required. Glucokinase (GLK), a representative tool enzyme widely used in biodetection applications such as blood glucose test kits [1], is commonly prepared through recombinant expression in Escherichia coli (E. coli). Consequently, improving the economy and reproducibility of GLK preparation is of practical importance for laboratory research and related biosynthetic technology development.

For recombinant GLK expression in E. coli, complex media such as Luria-Bertani (LB) [2], Terrific Broth (TB) [2, 3], and Super Optimal Broth (SOB) [2] are commonly used in laboratory-scale cultivation because of their convenience and strong growth-supporting capacity. However, these media contain ill-defined components and may exhibit substantial batch-to-batch variability [4], which can compromise expression reproducibility, increase background impurities during culture, and elevate medium-related costs [5, 6]. Therefore, although complex media are effective for routine laboratory protein expression, their limited compositional controllability can restrict their utility when reproducibility and cost efficiency are important considerations.

In scalable E. coli bioprocesses, traditional minimal defined media (e.g., M9 medium) are generally preferred due to their compositional consistency, reproducibility, and process controllability, as well as their ability to support high-cell-density cultivation [7]. However, under laboratory-scale recombinant protein expression conditions, conventional defined media often show lower expression performance than complex media, partly because of their relatively simple nutrient composition [8]. This creates a practical gap between process controllability and expression efficiency. A promising strategy to bridge this gap is the development of improved chemically defined medium [9, 10], in which selected nutrients are introduced into a defined basal formulation to improve protein production while retaining better compositional control and lower cost than complex media. Among such nutrients, amino acids are particularly attractive supplements because they can enhance biosynthetic capacity and reduce metabolic burden associated with de novo synthesis.

Despite this potential, an improved chemically defined medium specifically optimized for recombinant GLK expression remain insufficiently explored. In particular, the identities and concentration ranges of key nutrients that can simultaneously support efficient GLK production and reduce medium-related costs have not been systematically established. This knowledge gap limits the rational balancing of protein expression performance, medium composition, and economic efficiency, highlighting the need for an efficient medium-optimization strategy.

To facilitate such optimization, a previously reported machine learning-driven iterative optimization framework, Machine-learning guided Experimental Trials for Improvement of Systems (METIS), was adopted as an auxiliary engineering tool under constrained experimental conditions [11]. By iteratively incorporating experimental feedback, METIS enables the identification of improved medium compositions without imposing restrictive assumptions regarding factor interactions. In parallel, recent advances in high-throughput phenotyping technologies provide a practical experimental basis for systematic medium optimization. For example, the Respiration Activity Monitoring System (RAMOS) allows online respiration analysis [12], while the Growth Profiler 960 (GP960) enables automated biomass monitoring [13]. Together, these tools make it possible to rapidly and quantitatively evaluate cellular responses to medium variations, thereby enhancing the feasibility and efficiency of medium optimization.

In this study, we aimed to develop an improved chemically defined medium with relatively clear composition and reduced medium cost for recombinant GLK production in E. coli by combining experimental screening with METIS-assisted iterative optimization (Fig. 1). Key nutrients affecting cell growth and metabolic activity were first systematically screened using RAMOS and GP960. Subsequently, METIS was employed together with GP960 to iteratively refine the concentration ranges of these key components. Through this integrated strategy, an optimized chemically defined medium was obtained that improved GLK production under aerobic cultivation conditions while reducing medium cost. In addition, bioreactor experiments were performed to preliminarily evaluate its performance under scaled cultivation conditions. This work provides a practical framework for medium optimization in recombinant protein expression and offers a useful basis for further evaluation of GLK production under scalable bioprocess conditions.

Fig. 1.

Schematic workflow of the iterative optimization of a chemically defined medium for glucokinase expression in Escherichia coli

Materials and methods

Strains and media

In this study, E. coli BL21(DE3) containing the recombinant plasmid pET-32a(+)-Trx-His-GLK was used. The GLK gene was synthesized and cloned into the pET-32a(+) expression vector by GenScript (China). The experimental media included LB medium, a non-optimized defined minimal medium, and an optimized chemically defined medium, with detailed formulations detailed in Table 1. Salts, amino acids, vitamins, trace elements, and nucleotides were prepared as concentrated stock solutions (The detailed formulation of these components is provided in Table S1, with the component information sourced from Battling S. et al. [10]). The chemically defined medium used for industrial-scale aerobic cultivation was formulated by adding 5 mL L^− 1 trace element stock solution to the optimized medium obtained from shake-flask screening. The M9 medium contained KH₂PO₄ 3000 µg mL^− 1, MgSO₄·7H₂O 241 µg mL^− 1, Na₂HPO₄ 6780 µg mL^− 1, NaCl 500 µg mL^− 1, NH₄Cl 100 µg mL^− 1, CaCl₂ 11 µg mL^− 1, and 5 mL L^− 1 trace elements. Except when noted otherwise, buffers, chemicals and reagents were of analytical grade from Sigma Chemical Co. (St Louis, MO) or Fisher Scientific (Nepean, ON, Canada). All media were adjusted to pH 7.0 using 1 M HCl or NaOH, filter-sterilized through a 0.22 μm membrane, and stored at 4 ℃. Each medium was supplemented with 100 µg mL^− 1 ampicillin before use.

Table 1.

Composition of different culture media used for Escherichia coli

Ingredients	LB medium (µg mL− 1)	Non-optimized defined minimal medium (µg mL− 1)	Optimized chemically defined medium (µg mL− 1)
Yeast extract	5000
Tryptone	10,000
NaCl	10,000
Inorganic salt
(NH4)2HPO4		5000	2500
K2HPO4		5000	2500
MgSO4·7H2O		2000	2500
Citric Acid		1000	1000
Amino acids
Histidine			187.5
Proline
Glutamate			750
Arginine			250
Aspartate
Methionine			250
Isoleucine			168.75
Threonine
Cysteine
Serine
Glycine
Phenylalanine			275
Tyrosine
Tryptophan			12.5
Alanine
Leucine
Lysine			125
Valine			1000
Vitamins
Ascorbic acid
Biotin
Folic acid
Nicotinic acid
Ortoric acid
P-aminobenzoic acid
Pantothenic acid
Pyridoxine ∙ HCl
Riboflavin
Thiamine ∙ HCl
Vitamin B12
Nucleotides
Adenine
Cytosine
Guanine
Inosine
Thymidine
Uracil
Xanthine
Trace elements
FeSO4 ∙ 7H2O
ZnSO4 ∙ 7H2O
CuSO4 ∙ 5H2O
MnSO4 ∙ 5H2O
Na2B4O7·10H2O
CaCl2 ∙ 2H2O
(NH4)6Mo7O24

Aerobic cultivation

Luria-Bertani (LB) medium was originally formulated by American microbiologist Giuseppe Bertani [53]

Non-optimized defined minimal medium was provided by the research institute

Optimized chemically defined medium is derived from the optimization results obtained in this study

pDNA transformation

Chemically competent E. coli BL21(DE3) cells were used for transformation. The plasmid pET-32a(+)-Trx-His-GLK was added to the competent cells and incubated on ice for 30 min. Heat shock was then performed at 42 ℃ for 90 s, followed by immediate cooling on ice for 2–3 min. Subsequently, 1 mL of sterile LB medium was added, and the cells were recovered at 37 ℃ with shaking at 220 rpm for 1 h. Finally, the culture was plated on LB agar containing 100 µg mL^− 1 ampicillin and incubated overnight at 37 ℃ for selection of positive clones.

Strain aerobic culture

Seed culture preparation

A cryopreserved glycerol stock of the strain was inoculated into 100 mL of LB medium supplemented with 100 µg mL^− 1 ampicillin and cultured at 37 ℃ with shaking at 220 rpm for 7 h. The primary seed culture was examined under a microscope to confirm the absence of contamination and uniform cell morphology, ensuring that it reached the mid-logarithmic growth phase (OD₆₀₀ = 0.6–0.8). Before inoculation into the secondary seed medium, the primary seed culture was harvested by centrifugation at 5000 rpm for 5 min at 4 ℃. The harvested cell pellet was washed twice with 0.9% (w/w) NaCl solution to completely remove residual LB medium and avoid interference with subsequent aerobic cultivation. The washed cells were then resuspended in 100 mL of medium (identical in composition to the fermentation medium) containing 100 µg mL^− 1 ampicillin, and cultured at 37 ℃ with shaking at 220 rpm until reaching an OD₆₀₀ of 0.6–0.8. This secondary seed-culture step allowed the strain to fully adapt to the minimal-medium environment. The secondary seed culture was collected, washed once more with 0.9% (w/w) NaCl solution, and then inoculated at an appropriate volume into the fermentation medium to initiate the formal fermentation experiment.

Aerobic cultivation in RAMOS shake flasks

The RAMOS (HiTec Zang GmbH, Herzogenrath, Germany) is an online off-gas analysis system designed for shake flask fermentation with living cells. It enables real-time monitoring of key gaseous components, such as oxygen and carbon dioxide [14], reflecting the physiological and metabolic state of cells during aerobic cultivation.

Prior to aerobic cultivation, the washed seed culture was inoculated into RAMOS shake flasks at 1% or 5% (v/v) of the seed culture. Each flask contained a working volume of 30 mL. Aerobic cultivation was performed at 37 ℃ with shaking at 250 rpm on an orbital shaker with a 100 mm shaking diameter. For all media, 0.2 M S3-(N-morpholino) propanesulfonic Acid (MOPS) buffer was added to maintain the pH within 6.0–7.0 throughout the fermentation. After 2 h of aerobic cultivation, isopropyl β-D-thiogalactopyranoside (IPTG) was added to a final concentration of 0.1 mM to induce target protein expression. Aerobic cultivation continued under the same conditions, and oxygen transfer rate (OTR) was monitored in real time via the RAMOS system. Cultivation was terminated when OTR showed a sustained decline after reaching its maximum.

Aerobic cultivation in GP960-well plates

The GP960 (Enzyscreen, Heemstede, The Netherlands) is a high-throughput automated system capable of simultaneously culturing up to ten microplates under continuous shaking conditions [15]. It monitors microbial growth in real time by analyzing changes in green pixel intensity (G-value) [16] from images captured at regular intervals, facilitating efficient high-throughput screening.

Seed culture was added to each well at 1% (v/v), resulting in a working volume of 250 µL per well. Aerobic cultivation was carried out at 37 ℃ with orbital shaking at 220 rpm. After 2 h, IPTG was added to a final concentration of 0.1 mM to induce GLK expression. Except for the LB medium group, 0.2 M MOPS buffer was added to maintain the pH between 6.0 and 7.0. Images were captured automatically every 15 min, and G-value of each well was recorded continuously. Cell growth dynamics were quantified by analyzing time-dependent changes in G-values, and aerobic cultivation was stopped when the G-value plateaued, indicating growth cessation.

Iterative culture optimization using the METIS system

The METIS system is an active-learning-based optimization platform that iteratively predicts optimized medium compositions from limited experimental data [15]. It employs machine learning to capture complex, nonlinear interactions among medium components, enabling rational exploration of high-dimensional formulation spaces without exhaustive experimentation [17]. Initial concentration ranges for thirteen medium components—K₂HPO₄, (NH₄)₂HPO₄, MgSO₄·7H₂O, citric acid, Histidine (His), Glutamic (Glu), Arginine (Arg), Methionine (Met), Isoleucine (Ile), Phenylalanine (Phe), Tryptophan (Try), Valine (Val) and Lysine (Lys)—were defined as the starting input for the system, with multiple concentration levels manually assigned to establish the initial concentration gradients for iterative optimization (Table 2).

Table 2.

Initial concentration ranges of medium components before METIS optimization

Component	Concentration Levels
	µg mL− 1
(NH4)2HPO4	1000	2500	5000	10,000
K2HPO4	25,000	20,000	15,000	10,000	5000	2500
Citric Acid	5000	4000	3000	2000	1000
MgSO4·7H2O	10,000	5000	2500	1000
Histidine	750	375	187.5	93.7
Glutamate	1500	750	375	187.5	93.7
Arginine	500	250	125	62.5
Methionine	500	250	125	62.5
Isoleucine	675	337.5	168.7	84.3
Phenylalanine	2200	1100	550	275	137.5
Tryptophan	100	50	25	12.5
Lysine	1000	500	250	125	50
Valine	1000	500	250	125	50

Each row represents a different concentration level

Based on these input ranges, the system generated 48 distinct medium formulations for the first experimental round. Each formulation was cultured in GP960 plates, with images captured automatically every 15 min to record the G-value, representing cell growth. Growth curves were derived by analyzing time-dependent changes in G-values [16]. After aerobic cultivation, the final G-values and corresponding component concentrations were fed back into METIS to update the prediction model. The system then recommended a new set of 48 formulations predicted to support the highest biomass for the next round of aerobic cultivation.

Fermentation in 7.5 L bioreactor

The 7.5 L NBS bioreactor (BioFlo 120, Eppendorf, Germany) equipped with online monitoring and automatic control systems for pH, dissolved oxygen (DO), and temperature was used for high-cell-density fermentation. The bioreactor was integrated with constant-flow feed pumps, acid-base addition pumps, sterile aeration system, and exhaust gas collection system to achieve precise parameter control and automated operation throughout the fermentation process.

The optimized chemically defined medium was prepared according to Table 1, and the feed medium consisted of 50% glycerol. The prepared medium was loaded into the vessel and sterilized at 115 ℃, 75 kPa for 30 min. After cooling, air was sparged to calibrate DO to 100%, with agitation at 800 rpm and aeration at 1 vvm to verify system functionality. The secondary seed culture was inoculated at 2.5% (v/v) via flame inoculation. Fermentation was initiated with aeration at 1 vvm and agitation at 800 rpm.

The fermentation process was divided into cell growth phase (0–9 h) and induction phase (9–40 h). During the growth phase, temperature was maintained at 37 ℃, pH at 7.0 ± 0.1 (controlled by automatic addition of 25% ammonia solution and phosphoric acid), with DO maintained above 30% by dynamic regulation of agitation speed. Upon DO rebound, temperature was decreased to 25 ℃, glycerol feeding was initiated at 20 g h^− 1, and 0.1 mM IPTG was added for protein induction after temperature stabilization.

Analytical methods

Biomass determination

Cell concentration was determined by measuring the optical density at 600 nm (OD₆₀₀) using a spectrophotometer (UV-1900i, Shimadzu Instruments (Suzhou) Company Limited, Suzhou, China) at room temperature. Since a linear correlation for OD₆₀₀ and cell mass according to the Lambert-Beer law is only viable for an OD₆₀₀ between 0.1 and 0.3 [10, 18], samples were diluted using deionized water, if necessary.

pH measurement

The pH of samples was measured using a pH meter (S400-B, Thermo Fisher Scientific, USA) at room temperature. Prior to measurement, the instrument was calibrated with standard buffer solutions at pH 4.01 and pH 6.86.

Protein quantified by SDS-PAGE

Following expression, recombinant E. coli cells were harvested by centrifugation at 7,000 rpm for 5 min. The resulting cell pellet was washed twice with 100 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) buffer (pH 7.4) to remove residual medium components and resuspended to an OD₆₀₀ of 5.0. A 600 µL aliquot of the resuspended cells was subjected to mechanical disruption using a cryogenic grinder (JXFSTPRP-CL, Shanghai, China) under the following parameters: 35 s run time, 0 s pause, 20 cycles, at -50 ℃. After disruption, 100 µL of the lysate was centrifuged at 12,000 rpm for 5 min to separate the supernatant (soluble protein fraction) from the pellet (insoluble inclusion body fraction). The pellet was resuspended in 100 µL of 100 mM HEPES buffer (pH 7.4) for further analysis.

Aliquots of 40 µL from each fraction (whole-cell lysate, supernatant, and resuspended pellet) were separately transferred to 1.5 mL microcentrifuge tubes. To each aliquot, 10 µL of 5 × SDS-PAGE protein loading buffer was added, mixed thoroughly, and denatured by boiling at 100 ℃ for 5 min. The samples were then separated on a 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gel at a constant voltage of 120 V for 60 min. After electrophoresis, the gel was stained with Coomassie Brilliant Blue G-250 for 40 min and washed twice with deionized water. Bovine serum albumin (BSA, purity ≥ 98%) was used as a standard protein, and the relative expression level of the target protein Glk was quantified based on the grayscale intensity of the corresponding bands using ImageJ (Version ImageJ 1.53t) software according to Eq. 1.

1

Here, represents the protein concentration (µg mL^− 1) of the BSA standard dissolved in HEPES buffer, and denotes the concentration (µg mL^− 1) of Glk. and correspond to the volumes (µL) of BSA and Glk samples loaded on the SDS-PAGE gel, respectively. is the integrated density (grayscale value) of the Glk band, and is the integrated density of the BSA band. The soluble expression level of Glk reported in this study corresponds to the concentration of Glk in the lysate obtained from recombinant E. coli cells at an OD₆₀₀ of 5.0.

Data analysis

Experimental data were processed using GraphPad Prism (Version 8.0.2 (263)) and Origin (OriginPro 2021 (64-bit) 9.8.0.200). Aerobic cultivations in microplates and shake flasks were performed in three biological replicates.

Results

Determination of the optimal carbon source

To identify a suitable carbon source for the development of a cost-effective chemically defined medium for E. coli growth, the bacterium was cultivated in a defined minimal medium (supplied by the company, Table 1) supplemented with glucose, sucrose, or glycerol as the sole carbon source. All carbon sources were supplied at an equal mass concentration of 20 g L^− 1. Cellular respiratory activity under each condition was dynamically monitored using the RAMOS. The results revealed distinct differences in metabolic behavior depending on the carbon source (Fig. 2). The peak OTR in glycerol-based medium (0.021 mol L^− 1 h^− 1) was higher than that observed in glucose (5.51 × 10^− 5 mol L^− 1 h^− 1) and sucrose (4.93 × 10^− 5 mol L^− 1 h^− 1) media, and was slightly lower than the respiratory level measured in LB medium (0.0226 mol L^− 1 h^− 1). The corresponding OD₆₀₀ value (2.5) was also significantly higher than those measured in glucose (0.285) and sucrose (0.187) conditions. These results indicate that glycerol supported markedly higher metabolic activity compared with the other carbon sources tested. However, the onset of metabolic activity in the glycerol medium was delayed by approximately 20 h relative to LB medium, suggesting that while glycerol is a favorable carbon source for respiratory metabolism, further optimization of other medium components is required to improve metabolic initiation and overall aerobic cultivation performance.

Fig. 2.

Oxygen uptake rate and biomass of Escherichia coli cultivated in media containing different carbon sources. a Time-resolved oxygen transfer rate (OTR) profiles under different carbon source conditions (black: glucose, red: glycerol, blue: sucrose, green: Luria-Bertani (LB) medium). b Biomass under different carbon source conditions. Aerobic cultivation were performed using a Respiration Activity Monitoring System (RAMOS) with an inoculate size of 1% (v/v)

Determination of the optimal nitrogen source

Following the identification of the optimal carbon source, the nitrogen source composition in the synthetic medium was further optimized. E. coli was aerobic cultivated in media containing equimolar concentrations (0.1 M) of three inorganic nitrogen sources: (NH₄)₂HPO₄, (NH₄)₂SO₄, and NH₄Cl. Metabolic responses were dynamically monitored using the RAMOS. The results showed that while the overall respiratory capacity was comparable under the different nitrogen sources, significant differences in lag phase duration were observed (Fig. 3). When cultivated with (NH₄)₂HPO₄, the strain reached its peak respiration rate at 15 h, corresponding to an OD₆₀₀ value of 2.89. In contrast, cultures supplied with (NH₄)₂SO₄ (OD₆₀₀ 2.67) and NH₄Cl (OD₆₀₀ 2.3) exhibited delayed peak OTR. The most pronounced delay observed under NH₄Cl conditions, which also showed lower respiratory activity throughout the aerobic cultivation. Overall, (NH₄)₂HPO₄ exhibited the most favorable performance in supporting rapid metabolic activation and respiratory activity, and was therefore selected as the sole nitrogen source for the synthetic medium. Notably, the time to reach peak OTR in all inorganic nitrogen source groups was delayed by 10–15 h compared with that observed in LB medium, suggesting that additional growth factors may be required to further enhance metabolic efficiency under defined medium conditions [10, 19, 20].

Fig. 3.

Oxygen uptake rate and biomass of Escherichia coli in media containing different nitrogen sources. a Time-resolved oxygen transfer rate (OTR) profiles under different nitrogen source conditions. b Biomass under different nitrogen source conditions. Aerobic cultivation were performed using a Respiration Activity Monitoring System (RAMOS) with an inoculate size of 1% (v/v)

Effects of growth factors supplementation on respiration activity of E. coli

To further optimize the synthetic medium and improve its respiratory performance, the effects of various growth factor categories, including trace elements, nucleotides, vitamins, and amino acids, were systematically evaluated. Among all tested supplements, the addition of amino acids led to a pronounced enhancement in cellular respiratory activity (Fig. 4), resulting in a peak OTR of 0.0203 mol L^− 1 h^− 1. This value was comparable to the peak OTR observed in LB medium (0.0209 mol L^− 1 h^− 1). In contrast, supplementation with trace elements, nucleotides, or vitamins did not lead to a noticeable improvement in respiratory performance (Fig. 4). These results indicate that amino acids are key growth factors for supporting efficient respiratory metabolism in E. coli under minimal medium conditions. Accordingly, amino acid supplementation was identified as a key priority for enhancing the respiratory efficiency of the synthetic medium.

Fig. 4.

Time-resolved oxygen transfer rate (OTR) profiles of Escherichia coli cultivated in media containing 20 g L^− 1 glycerol supplemented with different categories of growth factors. The inoculate size was 5% (v/v). LB medium (black) and minimal salts medium (red) served as controls. Test conditions included supplementation with trace elements (blue), nucleotides (green), vitamins (purple), and amino acids (brown). Aerobic cultivations were performed using a Respiration Activity Monitoring System (RAMOS). Data represent mean values from three independent aerobic cultivations, with shaded areas indicating the standard deviation

Identification of essential amino acid subsets supporting growth and respiration in E. coli

Although supplementation with amino acids can enhance respiratory activity, the inclusion of all amino acid species would impose substantial cost constraints for large-scale production. Therefore, a systematic screening strategy was implemented to identify a minimal yet functionally essential amino acid composition. To reduce the combinatorial complexity associated with one-by-one omission, amino acids were grouped into five families according to their biosynthetic origins: the glutamate, the aspartate, the serine, the aromatic amino acid, and the pyruvate-derived families (Table 3). Each family was omitted individually, and growth dynamics were monitored in real time using GP960. The results revealed distinct effects on both exponential-phase growth and stationary-phase biomass upon omission of different amino acid families (Fig. 5). Removal of the glutamate, aspartate, or pyruvate-derived families resulted in significantly reduced biomass between 2 h and 5 h of aerobic cultivation compared with the synthetic medium containing all amino acids and LB medium (Fig. 5A and B, and 5E). In contrast, omission of the aromatic amino acid family led to a noticeable biomass reduction only after 7 h (Fig. 5D), whereas omission of the serine family produced no significant differences throughout the aerobic cultivation process (Fig. 5C). Collectively, these results indicate that, with the exception of the serine family, the remaining four amino acid families are critical for supporting efficient cell growth.

Table 3.

Classification of amino acids based on their biosynthetic pathways

Group		Amino acids
1	Glutamate	Histidine	Proline	Glutamate	Arginine
2	Aspartate	Aspartate	Methionine	Isoleucine	Threonine
3	Serine	Cysteine	Serine	Glycine
4	Aromatic	Phenylalamine	Tyrosine	Tryptophan
5	Pyruvate	Alanine	Leucine	Lysine	Valine

Fig. 5.

Effects of omitting specific amino acid families on biomass accumulation of Escherichia coli cultivated in a fully chemically defined medium. Amino acids were classified into five biosynthetic families and omitted individually: Group 1 (Histidine, Proline, Glutamate, Arginine), Group 2 (Aspartate, Methionine, Isoleucine, Threonine), Group 3 (Cysteine, Serine, Glycine), Group 4 (Phenylalanine, Tyrosine, Tryptophan), and Group 5 (Alanine, Leucine, Lysine, Valine). Aerobic cultivations were performed in 96-well round-bottom microtiter plates with a working volume of 250 µL at 37 ℃ and 250 rpm. The fully supplemented synthetic medium (containing salts and all amino acids), LB medium, and the defined minimal medium served as controls. a–e correspond to omission of Groups 1–5, respectively. Relative biomass values were normalized to those obtained with the fully supplemented synthetic medium. Data points represent mean values from at least three independent replicates, and shaded areas indicate standard deviation

To further resolve the contribution of individual amino acids within the glutamate, aspartate, aromatic, and pyruvate-derived families, selected amino acids were supplemented back into the corresponding omission media in various combinations. Within the glutamate family, individual supplementation with Arginine (Arg) or glutamate (Glu) increased exponential-phase biomass, with Arg addition resulting in a 40%–50% increase relative to the complete omission condition (Fig. S1A). Co-supplementation of Arg and Glu did not yield additional improvement, whereas the inclusion of Histidine (His) alongside Arg and Glu further enhanced biomass by 15%–25% during exponential growth (Fig. 6A; Fig. S1B).

Fig. 6.

Effects of individual and combinatorial supplementation of selected amino acids on biomass accumulation of Escherichia coli in a chemically defined medium. Four amino acid families identified as growth-relevant were further dissected: Group 1 (Histidine, Proline, Glutamate, Arginine), Group 2 (Aspartate, Methionine, Isoleucine, Threonine), Group 4 (Phenylalanine, Tyrosine, Tryptophan), and Group 5 (Alanine, Leucine, Lysine, Valine). Aerobic cultivations were performed in 96-well round-bottom microtiter plates with a working volume of 250 µL at 37 ℃ and 250 rpm. The chemically defined medium, LB medium, and the defined minimal medium served as controls. a omission of Group 1 and supplementation with Arg, Glu, and His. b omission of Group 2 and supplementation with Met and Ile. c omission of Group 4 and supplementation with Phe and Trp. d omission of Group 5 and supplementation with Lys and Val. Biomass values were normalized to those obtained with the chemically defined medium. Data points represent mean values from at least three independent replicates, and shaded areas indicate standard deviation

For the aspartate family, Methionine (Met) or Isoleucine (Ile) supplementation individually increased exponential-phase biomass by 10%–20%, while threonine (Thr) primarily contributed to higher stationary-phase biomass (Fig. S1C). Combined supplementation of Met and Ile resulted in the highest final biomass, outperforming other combinations and the complete omission group (Fig. 6B; Fig. S1D). In the aromatic family, no significant growth enhancement was observed with single-amino-acid supplementation (Fig. S1F); however, combined addition of Phenylalanine (Phe) and Tryptophan (Trp) increased stationary-phase biomass by approximately 10%, approaching that of the chemically defined medium containing all amino acids (Fig. 6C; Fig. S1G). For the pyruvate-derived family, Valine (Val) alone strongly promoted exponential-phase growth, and co-supplementation with Lysine (Lys) further elevated biomass to levels nearly equivalent to those achieved with the complete amino acid formulation (Supplementary Fig. S1H-I).

Based on these screening results, nine amino acids—His, Glu, Arg, Met, Ile, Phe, Trp, Lys, and Val—were identified as the most influential contributors to cell growth. Validation experiments demonstrated that the medium containing this nine-amino-acid combination significantly enhanced biomass accumulation during both exponential and stationary phases compared with the defined minimal medium, while approaching the performance of the fully supplemented amino acid medium and LB control (Fig. S2). During the early growth phase (2–5 h), biomass in the optimized formulation was slightly lower than that in the fully supplemented medium, indicating that the reduced formulation retained high growth efficiency despite partial amino acid omission.

Optimization of component concentrations in the medium

To further identify the optimal concentration combination supporting efficient biomass accumulation of E. coli, multiple rounds of concentration optimization were performed for the nine selected amino acids (His, Glu, Arg, Met, Ile, Phe, Trp, Lys, Val) and key inorganic components (K₂HPO₄, (NH₄)₂HPO₄, MgSO₄·7H₂O, and citric acid). The optimization was conducted using the METIS system, integrated with the high-throughput GP960 for automated screening (details provided in the Materials and Methods section). After three rounds of iterative optimization, the distribution of experimentally measured biomass outcomes progressively converged, accompanied by an increase and stabilization of the median biomass value (Fig. 7). This trend indicates that the optimized medium formulations recommended by the METIS had reached a stable performance level. Comparison of the concentration intervals before and after optimization revealed substantial differences in the degree of adjustment among individual components (Fig. S3). For example, the optimized concentration range of Glu (750–1500 µg mL^− 1) was narrowed by approximately 87% relative to its initial range (93.7–1500 µg mL^− 1), while the range for Val (250–1000 µg mL^− 1) was reduced by 80% compared to the initial interval (50–1000 µg mL^− 1). Additionally, the concentration ranges of six other components, (NH₄)₂HPO₄, K₂HPO₄, MgSO₄·7H₂O, Ile, Phe, and Lys, were each reduced by more than 50%. These results indicate that biomass formation is highly sensitive to the concentrations of these components, necessitating relatively precise control during medium formulation. In contrast, the concentration ranges of citric acid, His, Arg, Met, and Trp changed only modestly throughout the optimization process, suggesting lower sensitivity of growth to their exact concentrations and allowing greater formulation flexibility in practical applications.

Fig. 7.

Distribution of relative biomass of Escherichia coli across three iterative rounds of chemically defined medium optimization. In each round, 48 medium formulations were tested, with each data point representing the relative biomass obtained for one formulation. Box plots summarize the outcomes per round, with whiskers extending to 1.5 times the interquartile range, the box bounds representing the 25th and 75th percentiles, and the center line indicating the median

Evaluation of biomass and recombinant protein production in optimized chemically defined medium and LB media

Although the optimized medium formulations all enabled attainment of comparable biomass levels (Fig. S4A), notable differences remained in both medium cost and recombinant protein expression. Therefore, further optimization was performed to identify a formulation that maintains similar biomass while maximizing protein expression at minimal production cost. Based on the results from the third round of METIS optimization, a fourth round of experimental groups was generated. Representative formulations were selected: three low-cost and three high-cost groups (The medium formulations are listed in Table. S2). Fermentation was carried out using the Ramos strain, and protein expression was induced by IPTG. LB medium and basal salts medium were used as controls.

The results showed that biomass levels were similar across formulations with different costs (Fig. 8A). Soluble protein expression in the low-cost group was higher than that in the high-cost group (Fig. 8A), with Complete Medium 1 achieving the highest expression of 36.44 µg mL^− 1, which represents a 33 fold relative to the basal salts medium (1.11 µg mL^− 1) and corresponds to 74% of the expression obtained with LB medium (49.57 µg mL^− 1). The soluble expression of GLK was verified by SDS-PAGE (Fig. S5). Although the optimized medium yielded lower absolute protein expression compared to the complex medium, the soluble protein content per unit biomass (16.23 µg mL^− 1 OD^− 1) was 80% higher than that of LB medium (9.04 µg mL^− 1 OD^− 1) (Fig. 8B). Moreover, the production cost per unit biomass and the medium formulation cost were reduced by 85% and 74%, respectively, compared to LB (Table 4; Fig. 8B, C), demonstrating a clear economic advantage. Collectively, these analyses indicate that the optimized medium formulation maintains comparable biomass levels, enhances protein production efficiency per cell, and substantially lowers production costs.

Fig. 8.

Aerobic cultivation of Escherichia coli under aerobic conditions in the Respiration Activity Monitoring System (RAMOS) and Growth Profiler 960 (GP960) using 20 g L^− 1 glycerol in different media. a Relationship between cell optical density (OD₆₀₀) and soluble protein per unit biomass. b Total soluble protein concentration versus soluble protein per unit biomass. c Relationship between soluble protein production cost per unit biomass and soluble protein per unit biomass. Media tested include Complete Media 1–6 (optimized synthetic media) and a defined minimal medium, as well as Luria-Bertani (LB). Colors correspond to the labels shown in the figure. Data represent the mean of three biological replicates, with shaded areas indicating standard deviation (SD)

Table 4.

Cost comparison between Luria-Bertani (LB) medium and optimized chemically defined medium

Component	Unit Price	LB Medium	Chemically Defined Medium 1	Chemically Defined Medium 2	Chemically Defined Medium 3	Chemically Defined Medium 4	Chemically Defined Medium 5	Chemically Defined Medium 6
	(¥/g)	(¥/L)	(¥/L)	(¥/L)	(¥/L)	(¥/L)	(¥/L)	(¥/L)
(NH4)2HPO4	0.06		0.16	0.03	0.16	0.16	0.16	0.16
K2HPO4	0.03		0.08	0.16	0.08	0.08	0.08	0.08
Citric Acid	0.06		0.06	0.11	0.06	0.11	0.06	0.11
MgSO4·7H2O	0.01		0.02	0.01	0.02	0.01	0.01	0.01
His	0.38		0.07	0.07	0.04	0.07	0.04	0.07
Glu	0.18		0.14	0.28	0.14	0.28	0.28	0.28
Arg	0.24		0.06	0.06	0.06	0.06	0.06	0.06
Met	0.18		0.05	0.01	0.03	0.01	0.02	0.09
Ile	0.51		0.09	0.04	0.09	0.04	0.09	0.04
Phe	0.23		0.06	0.03	0.06	0.51	0.51	0.51
Trp	0.20		0.01	0.01	0.01	0.01	0.01	0.01
Lys	0.18		0.02	0.02	0.02	0.02	0.04	0.18
Val	0.24		0.24	0.12	0.12	0.24	0.24	0.12
Glycerol	0.06		1.15	1.15	1.15	1.15	1.15	1.15
Yeast Extract	0.36	1.8	0.00	0.00	0.00	0.00	0.00	0.00
Tryptone	0.64	6.4	0.00	0.00	0.00	0.00	0.00	0.00
NaCl	0.04	0.40	0.00	0.00	0.00	0.00	0.00	0.00

These chemicals were of analytical grade and purchased from Sigma Chemical Co. (St Louis, MO, USA) or Fisher Scientific (Nepean, ON, Canada)

Fermentation in a 7.5 L-NBS bioreactor using LB, M9, and chemically defined medium

The optimized chemically defined medium was validated in a 7.5 L bioreactor for GLK production. During batch fermentation with the base chemically defined medium, severe respiratory and growth inhibition was observed, with a maximum OD₆₀₀ of only 13 (Fig. S6). At approximately 24 h, a constant feed of 50% glycerol at 20 g h^− 1 was initiated; however, neither respiration nor growth improved significantly, indicating that carbon source limitation was not the primary constraint. At 31 h, supplementation with 5 mL L^− 1 trace element solution markedly enhanced both respiratory activity and biomass accumulation, accompanied by a decrease in glycerol accumulation. These findings demonstrated that trace elements constitute a critical limiting factor for high-cell-density fermentation in this system.

Accordingly, trace element supplementation was incorporated into the chemically defined medium formulation. Given that excessive trace element concentrations may exert cytotoxic effects on cells, the optimal dosage was optimized (Fig. S7). No significant differences in cell growth or GLK expression were observed between 5 mL L^− 1 and 10 mL L^− 1 supplementation levels. To minimize production costs, 5 mL L^− 1 was selected as the optimal concentration for subsequent experiments.

The performance of the optimized chemically defined medium was benchmarked against LB medium (conventional laboratory complex medium) and M9 medium (industrially relevant defined medium), with an initial glycerol concentration of 20 g L^− 1 maintained across all conditions. Marked differences were observed in respiratory activity, biomass accumulation, and protein productivity among the three media (Fig. 9). The optimized chemically defined medium achieved a peak OUR of 172 mmol L^− 1 h^− 1 at 9 h, representing 39% and 75% improvements over LB (123.9 mmol L^− 1 h^− 1 at 7.5 h) and M9 (98.2 mmol L^− 1 h^− 1 at 11.5 h), respectively (Fig. 9a). While the OD₆₀₀ of LB and M9 cultures plateaued at 20.7 and 16.1, respectively, the optimized chemically defined medium supported continuous growth, reaching a maximum OD₆₀₀ of 103 at 34 h—approximately 4-fold and 5.4-fold higher than LB and M9 (Fig. 9b). Furthermore, the soluble protein titer in the optimized chemically defined medium reached 920 mg L^− 1, exceeding those of LB (82.5 mg L^− 1) and M9 (83.3 mg L^− 1) by more than 10-fold (SDS-PAGE is shown in Fig. S8). In conclusion, the optimized chemically defined medium outperformed conventional media in the 7.5 L bioreactor, with 4-5-fold higher biomass and an order-of-magnitude increase in protein titer. These results establish a solid foundation for the industrial production of GLK.

Fig. 9.

Profiles of E. coli fermentation in LB, Industrially-Applied M9, and chemically defined medium supplemented with 5 mL L^− 1 trace elements in the 7.5 L-NBS bioreactor. a shows the oxygen uptake rate of E. coli in three different media. b shows the biomass (OD₆₀₀) of E. coli cultured in three different media. c showsthe soluble GLK protein titer of E. coli cultured in three different media. Batch fermentation was initiated with an initial working volume of 4 L. The inoculation volume was 2.5%. The green line represents LB medium, the red line represents M9 medium, and the blue line represents the chemically defined medium. All three culture media contained 20 g L^− 1 glycerol in the initial batch. Feeding was performed at a constant rate of 20 g h ^− 1 with 50% glycerol

Discussion

This study aimed to develop an improved chemically defined medium to overcome the limitations of complex media in recombinant GLK production, including unstable expression and high production costs due to ill-defined composition and batch-to-batch variability. By combining high-throughput analytical platforms with the METIS, concentrations of carbon and nitrogen sources, salts, and key amino acids were systematically refined. The optimized medium maintained comparable biomass to LB while significantly enhancing soluble GLK titer per cell. Laboratory-scale screening showed a 74% reduction in medium cost compared to complex media, and 7.5 L bioreactor validation demonstrated its suitability for high-cell-density cultivation, achieving OD₆₀₀ up to 103 and soluble protein titer of 920 mg L^− 1—substantially higher than LB or M9 medium. These results indicate that the optimized chemically defined medium provides a cost-effective and efficient solution for laboratory GLK production.

Our findings demonstrate that glycerol as the optimal carbon source and diammonium phosphate as the optimal nitrogen source play a critical role in enhancing respiratory metabolism in E. coli. Specifically, the substitution of glucose or sucrose with glycerol significantly improved cellular respiratory activity and recombinant protein expression, consistent with previous reports that glycerol enhances respiratory metabolism and recombinant protein yield in E. coli [21–23]. The low respiration rate observed in minimal medium with glucose as the sole carbon source may be attributed to the "glucose effect" [24]. Glucose is transported into the cell via the phosphoenolpyruvate (PEP)-dependent phosphotransferase system (PTS), leading to dephosphorylation of enzyme IIA (EIIA), which subsequently inhibits adenylate cyclase activity and lowers intracellular cyclic adenosine monophosphate (cAMP) levels [23, 25]. The cAMP-CRP complex functions as a transcriptional activator for multiple genes involved in the tricarboxylic acid (TCA) cycle and respiratory chain. Consequently, reduced cAMP levels suppress the transcription of these associated genes [26]. In contrast, glycerol enters the cell via facilitated diffusion through the glycerol facilitator (GlpF) membrane protein and, without affecting the phosphorylation state of EIIA, is phosphorylated by glycerol kinase (GlpK) to glycerol-3-phosphate (G3P) [27]. G3P then enters glycolysis, generating substantial nicotinamide adenine dinucleotide (NADH) that provides ample electron donors for the respiratory chain [28, 29]. As a result, glycerol uptake and metabolic regulation do not trigger strong carbon catabolite repression.

Differences in the metabolic initiation time of E. coli under aerobic cultivation with different nitrogen sources may be associated with variations in the chemical structure of the nitrogen source, complexity of the metabolic pathway, and additional metabolic demands. Ammonium sulfate metabolism initially releases ammonium ions for nitrogen assimilation; however, the accompanying sulfate ions (SO₄²⁻) must undergo a series of energy-intensive activation and reduction steps before entering sulfur metabolism [30]. Sulfate activation is catalyzed by adenosine triphosphate (ATP) sulfurylase, which converts SO₄²⁻ to activated adenosine 5’-phosphosulfate (APS). Subsequent reactions, mediated by APS reductase and sulfite reductase, consume additional ATP and substantial amounts of nicotinamide adenine dinucleotide phosphate hydrogen (NADPH) to ultimately reduce APS to sulfide for the synthesis of sulfur-containing amino acids such as Cysteine (Cys) and Met [31, 32]. Although essential, these processes compete for reducing equivalents that would otherwise be used for ammonia assimilation, potentially creating metabolic competition between nitrogen and sulfur assimilation pathways and thereby delaying overall metabolic initiation to some extent. In contrast, the metabolic pathway of (NH₄)₂HPO₄ (phosphite, HPO₃²⁻) is fundamentally different. Phosphite is oxidized to phosphate via phosphite dehydrogenase with the concurrent reduction of nicotinamide adenine dinucleotide (NAD⁺) to NADH, and adenosine monophosphate (AMP) is phosphorylated to adenosine diphosphate (ADP) [33, 34]. The electron donor in this reaction operates at a substantially lower standard redox potential (-690 mV) compared to the NAD⁺/NADH couple (-320 mV) [33]. As a result, the free energy released per mole of substrate is significantly higher than that of sulfate reduction, and the process does not consume ATP, instead providing additional reducing power and substrate-level phosphorylation benefits to the cell. The delayed metabolic initiation observed with ammonium chloride as the nitrogen source may be associated with the presence of chloride ions (Cl⁻) [35]. Studies indicate that ammonium chloride can alter intracellular electrolyte balance, affecting energy metabolism related to proton pumps and membrane potential [35, 36]. Moreover, chloride ions may inhibit the regulation of amino acid transport systems [37], thereby indirectly reducing the efficiency of initial nitrogen metabolism.

Through growth factor screening and amino acid omission experiments, this study found that, with the exception of the serine family, deletion of the other four amino acid biosynthetic pathways significantly inhibited the growth of E. coli. In contrast, synergistic supplementation of nine key amino acids—His, Glu, Arg, Met, Ile, Phe, Trp, Lys, and Val—can influence cellular efficiency in converting nutrients into biomass [38] and effectively restore cell density to levels comparable to those achieved with complex medium (LB). Analysis indicates that the biosynthesis of these key amino acids typically involves multistep, irreversible reactions, and their key enzymes are subject to strong end-product feedback inhibition. For example, the synthesis of Phe and Trp relies on 3-Deoxy-D-arabino-heptulosonic acid 7-phosphate (DAHP) synthase, an enzyme strongly inhibited by its end products Phe and Trp [39, 40]. The synthesis of Ile and Val begins with threonine deaminase (IlvA), whose activity is feedback-inhibited by Ile itself through binding to a specific regulatory site, inducing conformational changes; recent studies have elucidated the role of key residues such as Phe352 in Ile binding and signal transmission [41]. His biosynthesis entails ten enzymatic steps, with the first enzyme, ATP-phosphoribosyltransferase (ATP-PRT, HisG), being dually regulated by His and ATP, a feedback-inhibition mechanism validated in multiple studies [42–44]. Inhibition of these key enzymes blocks metabolic flux at the initiation steps of synthesis, and cells cannot compensate via transamination, thereby severely restricting growth. In contrast, Ala and Asp can be synthesized directly from abundant precursors in single-step transamination reactions [45, 46]. Serine (Ser) and Glycine (Gly) are interconvertible through serine hydroxymethyltransferase (SHMT) [47]. Under carbon-limiting conditions, Pro can also provide additional reducing power and energy via a decarboxylation pathway [48]. Consequently, the absence of these amino acids exhibits a comparatively minor impact on cell growth.

Furthermore, this study employed an artificial intelligence-based iterative medium optimization system, which demonstrated that the optimized medium significantly enhanced both cell growth and GLK expression while alleviating the high-cost issue associated with complex media. The AI-driven approach exhibited notable advantages, particularly in efficiently handling multi-component nonlinear interactions with limited sample sizes. Using this method, we successfully elucidated the nonlinear relationship between biomass concentration and medium composition, and identified synergistic effects among different nutrients. A limitation of the current method, however, is the lack of a high-throughput protein concentration assay; optimization relied solely on cell density rather than soluble protein titer. Rapid cell growth does not necessarily match the rates of intracellular protein synthesis and folding, which may explain the relatively low soluble protein expression observed. Moreover, the data used for model training were derived from short-cycle, high-throughput screening, leaving gaps in the analysis of long-term aerobic cultivation processes such as protein folding and the accumulation of metabolic by-products.

The significant difference in cell growth and metabolic phenotypes between shake-flask and 7.5 L bioreactor systems observed in this study essentially stems from the distinct demand for trace elements at different aerobic cultivation scales [49, 50]. Shake flasks, as a low-cell-density open system, require relatively low total trace elements and show no obvious growth limitation. In contrast, the high-cell-density closed environment in the bioreactor leads to a sharp increase in biomass and trace element consumption, causing serious metabolic blockage when using the shake-flask-optimized medium without additional trace element supplementation. As essential cofactors for respiratory enzymes, ribosome assembly and biosynthetic pathways, insufficient trace elements directly restrict energy metabolism and anabolic efficiency, forcing cells to divert carbon flux toward byproducts rather than protein synthesis [49]. This explains why shake-flask optimization results are often difficult to scale up directly.

In industrial recombinant protein production, M9 is the most common chemically defined medium [51], and many defined media formulations have been developed for E. coli, particularly for high-cell-density aerobic cultivation and recombinant protein expression through optimization of carbon sources, nitrogen supply, and trace elements. However, these media are typically designed for specific production contexts and are not necessarily optimal for all target proteins. In this study, both M9 and our optimized chemically defined medium were uniformly supplemented with trace elements in the initial batch to support high-cell-density fermentation. In conventional fed-batch processes using M9 medium, extra magnesium is usually added into the feed to maintain normal cell metabolism, since long-term fermentation would deplete magnesium and cause metabolic restrictions. In this work, to ensure a strict and fair comparison, both media were fed with glycerol only during the fed-batch phase, without any extra magnesium or nutrients added to the feed. Despite the same feeding strategy and trace element supplementation, M9 still performed poorly in 7.5 L bioreactor fermentation, which may result from the lack of rational nutrient balance required for efficient GLK expression [52]. In contrast, our optimized chemically defined medium achieved excellent growth, high cell density, and significantly higher GLK production even with glycerol as the only feed supplement, demonstrating that its initial formulation provides sufficient and balanced magnesium, trace elements, and nutrients to support continuous cell metabolism and protein synthesis throughout fed-batch fermentation.

On top of nutritional discrepancies between small-scale and large-scale cultivations, operational parameters including induction conditions also exert vital impacts on fermentation performance. Consistent with the scale-up comparison above, it is worth noting that induction timing also impose remarkable effects on cellular physiology and heterologous protein synthesis. In order to maintain uniform conditions for medium performance comparison, relatively early induction was arranged in shake-flask screening assays. In high-density fermentation systems, premature induction tends to impose excessive metabolic burden on host cells, accelerate physiological decline, and impair overall production capacity. For this reason, delayed induction combined with low-temperature cultivation was adopted in 7.5 L scale-up fermentation. This strategy effectively alleviates cellular stress caused by heterologous protein expression, reduces protein misfolding and aggregation, and maintains stable cell viability throughout the fermentation process.

Overall, nutrient balance represents the key factor determining medium performance and scale-up feasibility in this fermentation system. Although the optimized medium developed in this study outperforms conventional minimal media, our screening strategy remains limited by its reliance on a single evaluation indicator and insufficient systematic optimization of multiple nutritional components. For follow-up studies, DoE methods can be applied to optimize the combined supply of amino acids, inorganic salts and trace elements. Revealing the synergistic interactions among these nutrients will achieve finer nutrient matching and relieve scale-up bottlenecks. In addition, future optimization should integrate growth performance, respiratory activity and target protein expression for comprehensive evaluation, which helps to detect hidden nutritional restrictions more effectively. Finally, larger-scale bioreactor trials are needed to verify the stability and industrial potential of the optimized medium for high-density recombinant protein fermentation.

Conclusion

This study developed a low-cost chemically defined medium optimized for recombinant GLK expression, comprising 20 g L^-1 glycerol as the carbon source, 2.5 g L^-1 diammonium phosphate as the nitrogen source, and optimized concentrations of salts and nine key amino acids (His, Glu, Arg, Met, Ile, Phe, Trp, Lys, and Val). Laboratory-scale experiments demonstrated that this medium enhanced respiratory metabolism, increased soluble protein expression per unit biomass by 80%, and reduced medium usage costs by 74% compared with LB medium. Validation in a 7.5 L bioreactor further confirmed its suitability for high-cell-density cultivation, achieving an OD₆₀₀ of 103 and a soluble GLK titer of 920 mg L^-1, substantially outperforming LB and M9 media under the same conditions. These results indicate that the optimized chemically defined medium provides an effective, cost-efficient solution for laboratory- and pilot-scale GLK production.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

Ala

Alanine

Arg

Arginine

Asp

Aspartate

cAMP

Cyclic adenosine monophosphate

Cys

Cysteine

DAHP

3-Deoxy-D-arabino-heptulosonic acid 7-phosphate

EIIA

Enzyme IIA

G3P

Glycerol-3-Phosphate

GFP

Green fluorescent protein

GLK

Glucokinase

GlpF

Glycerol facilitator

GlpK

Glycerol kinase

Glu

Glutamate

Gly

Glycine

GP960

Growth profiler 960

HEPES

4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid

His

Histidine

HisG

ATP-phosphoribosyltransferase

Ile

Isoleucine

IlvA

Threonine Deaminase

IPTG

Isopropyl β-D-1-thiogalactopyranoside

Leu

Leucine

Lys

Lysine

Met

Methionine

METIS

Machine-learning guided experimental trials for improvement of systems

MOPS

S3-(N-morpholino) propanesulfonic acid

NADH

Nicotinamide Adenine Dinucleotide

OD₆₀₀

Optical density at 600 nm

OTR

Oxygen Transfer Rate

PEP

Phosphoenolpyruvate

Phe

Phenylalamine

Pro

Proline

PTS

Phosphotransferase system

RAMOS

Respiration activity monitoring system

Ser

Serine

SHMT

Serine Hydroxymethyltransferase

Thr

Threonine

Trp

Tryptophan

Tyr

Tyrosine

Val

Valine

Author contributions

Yuhan Ma and Jiahao Kong designed and was responsible for medium selection, sample collection, and offline data analysis. Min Chen and Jianye Xia jointly contributed to the conceptualization of the study, project supervision, and funding acquisition. Zhenguo Wen provided guidance on experimental design, assisted in data interpretation, and participated in the drafting and revision of the manuscript. Zongjin Wang, Sitong Chen and Xu Yang assisted in conducting experiments, contributed to the conceptualization of the study, and participated in data collection and analysis.

Funding

This work was financially supported by the Tianjin Major Science and Techonology Projects and Engineering Programs (25ZXWCSY00190) and the Director’s Leading Scientific Innovation Fund (SZJJ2024-009).

Data availability

Data will be made available on request.

Declarations

Competing interests

The authors declare no competing interests.