https://orcid.org/0000-0002-4989-7598
Figures
Abstract
Comparison of Streptococcus zooepidemicus (S. zooepidemicus) as a natural strain for hyaluronic acid (HA) production with Corynebacterium glutamicum (C. glutamicum) as a recombinant host is a valuable tool to identify key metabolic targets for enhanced HA production. For this purpose, a comparative analysis based on the genome-scale metabolic networks of native and recombinant HA producers was conducted. In this study, the first genome-scale metabolic model (GEM) of S. zooepidemicus, was reconstructed and named iZN522. This systematically curated model was validated using various biochemical data, consisting of growth rate prediction, amino acids auxotrophy, and different carbon sources consumption. Following the validation, key genes for HA overproduction in both bacteria were identified through iterative single-gene deletion analysis. Comparing optimal flux distributions predicted by models of S. zooepidemicus strain with recombinant C. glutamicum revealed that removing the oxidative pentose-phosphate pathway in C. glutamicum effectively increases the hyaluronan synthase flux. Additionally, improvement strategies in C. glutamicum primarily rely on up- and down-regulation, while in S. zooepidemicus, gene deletions can optimize the HA production rate. By comparing the two metabolic models, it can be inferred that eliminating of glucose-6-phosphate dehydrogenase enzymatic reaction in C. glutamicum results in greater similarity in the central carbon metabolisms of C. glutamicum and S. zooepidemicus, thereby enhancing HA production.
Citation: Nikuiyan Z, Tabandeh F, Motamedian E, Salehi S, Talebi M, Ghale RA (2025) Reconstruction of a genome-scale metabolic model for Streptococcus zooepidemicus: Comparison with Corynebacterium glutamicum to study hyaluronic acid production. PLoS One 20(12): e0335509. https://doi.org/10.1371/journal.pone.0335509
Editor: Muhammad Afzal, Universitat fur Bodenkultur Wien, AUSTRIA
Received: July 21, 2025; Accepted: October 13, 2025; Published: December 31, 2025
Copyright: © 2025 Nikuiyan et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting information files.
Funding: This work is based upon research funded by Iran National Science Foundation (INSF) under project No. 4024643, and was supported by the National Institute of Genetic Engineering and Biotechnology (NIGEB project number 769). Both grants were received by Dr. Fatemeh Tabandeh.
Competing interests: The authors have declared that no competing interests exist.
1. Introduction
A genome-scale metabolic model (GEM) contains curated gene-protein-reaction (GPR) associations based on annotated genome and experimental observations. These models provide a quantitative format to predict the phenotypes of the target organism in a specific steady-state condition [1–3]. GEMs are becoming increasingly powerful tools for understanding and manipulating microbial metabolism and enable in-depth understanding of biological big data, with recent reviews highlighting their applications in investigating microbial metabolic adaptations [4,5]. Moreover, GEMs allow us to identify optimized flux distributions with computational tools such as flux balance analysis (FBA), which utilize linear programming solvers [3].
Hyaluronic acid is a non-immunogenic linear glycosaminoglycan composed of repeating disaccharide units of D-glucuronic acid and N-acetyl-D-glucosamine [6–8]. Due to its unique physicochemical properties, including high viscosity, biocompatibility, and hygroscopicity, HA is a biomaterial with a broad range of applications in various fields, including pharmaceuticals, cosmetics, biomedicine and the food industry [9–12]. The global HA market size was valued at United States dollar (USD) 8.3 billion in 2018 and is anticipated to reach USD 15.25 billion in 2026 [10]. Indeed, its expanding use across diverse sectors including ophthalmology, rheumatology for joint lubrication, advanced wound care, and tissue engineering, alongside significant demand from the pharmaceutical industry for drug delivery systems and the cosmetics sector for sophisticated skincare formulations, underpins this market growth [12,13]. Consequently, efficient and sustainable biotechnological production routes are increasingly critical to meet these varied and growing demands [7,12,14]. Recent market analyses affirm this trajectory, indicating sustained high demand and significant economic value, further motivating research into optimized bioproduction routes [15]. In recent years, to meet the market demands and circumvent issues with animal-derived HA, such as potential viral contamination and ethical concerns, HA is mainly provided by microbial fermentation [7,8]. Native and recombinant bacterial systems are used for HA production [16]. Streptococcus equi subsp. zooepidemicus strains are commonly used as native HA producers, capable of producing 6–7 g/L HA, under appropriate culture conditions [6,17,18]. However, as a biosafety level 2 opportunistic pathogen, its industrial use necessitates stringent safety and purification protocols [18]. Recently, heterologous HA biosynthesis by GRAS (generally recognized as safe) recombinant producers such as Corynebacterium glutamicum has come into consideration as a promising alternative [19–22]. While significant progress has been made in engineering these GRAS hosts, HA yield produced by recombinant C. glutamicum (as a recombinant HA producer) is lower than the HA produced by S. zooepidemicus (as a native HA producer), presenting ongoing challenges for metabolic engineering [16,20,23]. The HA capsule, a virulence factor for the animal pathogenic bacterium S. zooepidemicus, plays a vital role in evading the host immune system [18]. It seems that the metabolic pathways of S. zooepidemicus are adopted for HA biosynthesis in an efficient way. S. zooepidemicus uses approximately 5% of the carbon source for HA biosynthesis, but 10% and 80% of carbon sources were consumed for biomass generation and lactic acid production, respectively [24]; thus, growth rate competes with HA production for carbon source and energy [25], and experimental observations has confirmed this statement [24,26]. Many attempts have focused on improving the fermentation conditions [17,27–29] and medium composition [30–32] to overproduce HA in S. zooepidemicus. Overexpression of NADH oxidase as a metabolic engineering approach in S. zooepidemicus led to a 15% and 33% increase of biomass and ATP, respectively. Nevertheless, the HA yield remained constant, often because other factors like precursor availability or synthase activity subsequently become limiting [8,18]. Experimental observations have shown that redirecting carbon flux from glycolysis to the HA synthesis pathway improves HA productivity. For instance, the strategy of expressing polyhydroxybutyrate (PHB) synthesis genes (phbCAB) in S. zooepidemicus has been investigated as a means to create an alternative carbon sink. This approach, particularly when combined with the inhibition of lactic acid synthesis, reportedly increased the HA production from 5.5 g/L to 7.5 g/L [8,18]. Recombinant C. glutamicum produces lactic acid as the main by-product in the HA fermentation process. Cheng et al. (2017) [33] have reported that knockout of the lactate dehydrogenase gene in recombinant C. glutamicum increased HA titer significantly. The evaluation of C. glutamicum GEM, named iCW773, as detailed by Cheng et al. (2019) [20], indicated that glycolysis pathway attenuation, PPP knockout, lactate/acetate pathway deletion, and PDH activity attenuation enhanced HA titer to 27.8 g/L in fed-batch culture.
While extensive research has been conducted to optimize HA production in S. zooepidemicus and recombinant C. glutamicum [18,21] including recent efforts to control HA molecular weight through precursor flux engineering in S. zooepidemicus [34], no studies have investigated the optimization of the whole-cell metabolic networks of S. zooepidemicus for HA production using a newly reconstructed and comprehensively validated GEM for this specific purpose. In addition, there is no report comparing these bacteria via detailed computational systems biology and comparative GEM analysis to identify target genes for increasing HA production rate. Because of the HA pathway’s complexity and its relationship with biomass and ATP, along with the recognized influence of transcriptional regulatory networks on HA synthesis [35], a comprehensive metabolic model can help answer these questions. Such models, especially when standardized [36] and built upon current genomic and biochemical knowledge [37,38], provide a robust platform for in silico exploration.
In this study, a GEM was employed to explore novel strategies for enhancing the HA production rate. To accomplish this goal, the GEM of S. zooepidemicus ATCC 35246 was reconstructed for the first time and denoted as iZN522. Subsequently, the differences in optimal flux distributions between the native and recombinant producers were analyzed by the FBA method. Moreover, single- and multiple- gene knockout for HA optimization in both strains were systematically identified by FBA and iterative minimization of metabolic adjustment (MOMA) methods. Additionally, recombinant C. glutamicum GEM was evaluated computationally and strategies were introduced for HA production rate enhancement. This work aims to bridge the gap in comparative systems-level understanding of HA production in these key microorganisms.
2. Materials and methods
2.1. Genome-scale metabolic model reconstruction
The S. zooepidemicus GEM was reconstructed based on genome annotation data and experimental data from the literature according to the procedure presented by Thiele and Palsson [39]. While also incorporating best practices and considerations for model quality outlined in recent GEM reconstruction guidelines [5,38,40]. The reconstruction of the model was carried out in three steps: generation of the draft reconstruction, manual refinement of the draft reconstruction, and determining the pseudo reactions of biomass and HA.
The first step consisted of collecting and organizing the metabolic reactions of S. zooepidemicus, including specific pathways for HA biosynthesis, using biochemical data for the ATCC 35246 strain, based on its genome sequence (GenBank access code: CP002904.1) published by Ma et al. [41] provided by the KEGG database [42], transport reactions listed in transportDB [43], and corresponding metabolites for each reaction elicited from PubChem [44] and ChEBI [45] in an EXCEL file. Automated tools like CarveMe can assist in generating initial draft reconstructions from annotated genomes, which then require extensive curation [46]. For manual curation in the second step, initially, the names, formation, and direction of reactions were checked using the BIGG database [47]. Then, using Marvin View software 14.7.7, 2014, developed by ChemAxon (chemaxon.com), the metabolite’s charges were estimated and following that, the reactions were charge-balanced. Finally, the biochemical literature of S. zooepidemicus was reviewed, and new reactions were added to the draft file.
In the third step, experimental data for biomass composition were not available for S. zooepidemicus, and hence, the molar ratios for deoxyribonucleotides of DNA and ribonucleotides of RNA were calculated using the genome and the whole-cell RNA sequence of S. zooepidemicus provided by (microbedb.jp) and (bacteria.ensembl.org). Protein composition was calculated using codon preference data of S. zooepidemicus (dnahive.fda.gov). A detailed calculation of DNA, RNA, and protein composition is presented in the Supplementary File S1 in S1 File. Other data for the biomass reaction were obtained from the closest strain, Lactococcus lactis [48]. Formulating an accurate biomass objective function is crucial for GEM predictions, and guidelines for its systematic construction are available [37–39].
Fig 1 represents the HA biosynthetic pathway in S. zooepidemicus, showing two precursors consisting of UDP-glucuronic acid (UDP-GA) and UDP-acetylglucosamine (UDP-GlcNac) linked by a polymerase called hyaluronan synthase (encoded by hasA) [49,50]. The biosynthesis of the HA disaccharide unit consumes five ATP molecules, one acetyl-CoA, and two NAD cofactors [8,18]. The pseudo reaction of the HA formation and the HA exchange reaction were added to the draft file to simulate HA production by this model. The growth-associated maintenance (GAM) energy described as the amount of ATP consumed for biomass formation, was taken from the L. lactis model (39.4 mmol ATP.gDCW-1.h-1) [48]. Non-growth associated maintenance (NGAM) energy was calculated based on the experimental glucose consumption rate presented by Blank et al. and Hamilton et al. [51,52]. For more detailed NGAM calculation, see the Supplementary File S1 in S1 File.
This schematic illustrates the biosynthetic pathway of hyaluronic acid (HA) emphasizing the metabolic precursors and enzymatic steps. The pathway starts from glucose, which is metabolized to form two activated sugar precursors: UDP-glucuronic acid (UDP-GA) and UDP-N-acetylglucosamine (UDP-GlcNAc). These precursors are polymerized by the enzyme hyaluronan synthase, encoded by the hasA gene, to form HA. Arrows indicate the direction of enzymatic reactions. Key energy inputs and cofactors are labeled. This figure contextualizes the metabolic investment required for HA synthesis, highlighting the critical nodes for regulation and genetic manipulation. Abbreviations: ATP, adenosine triphosphate; NAD, nicotinamide adenine dinucleotide; UTP, uridine triphosphate.
2.2. Growth simulation and validation
Model simulations were carried out using the Systems Biology Markup Language (SBML) Level 3 [53], Constraint-Based Reconstruction and Analysis (COBRA) toolbox version 3.0 [54], and the GNU Linear Programming Kit (GLPK) package in MATLAB software. FBA [55] was applied to predict optimal flux distributions in different media and conditions. The intracellular reversible reaction fluxes were limited to between −1000–1000 mmol.gDCW-1.h-1, and the intracellular irreversible reaction fluxes were set between 0 and 1000 mmol.gDCW-1.h-1. The lower and upper bounds of other exchange reactions were calculated using Eq. 1, based on previous experimental data [51]. The glucose uptake rate was set to 18.56 mmol.gDCW-1.h-1 in all simulated conditions, and the exchange reaction of oxygen was set to zero to simulate anaerobic conditions. Other constraints of exchange reactions are presented in S1 Table in S1 File.
The qi is the flux of the corresponding metabolite in mmol.gDCW-1.h-1, Ci presents concentration of metabolite i in mmol per liter, Xbiomass is the biomass concentration in gDCW/L, and D is the dilution rate (h-1).
For the growth simulation, two chemically defined media named CDM and CDM1, and a complex medium were used. The compositions of these culture media are provided in S2 and S3 Tables in S1 File. For the growth prediction on complex media including yeast extract, the chemical composition was estimated according to the data presented by Oh et al. [56]. The objective function was set to biomass reaction to simulate the growth of S. zooepidemicus in all in silico experiments.
The growth rate prediction using a model is a crucial stage for model validation. Standardized model testing suites like MEMOTE [36] offer comprehensive quality control checks beyond basic growth prediction. To evaluate the predicted growth rate by our model, growth simulation on CDM medium in anaerobic conditions was performed using FBA based on the experimental data presented by Blank et al. [51]. Considering CDM1 and anaerobic conditions, the model was also evaluated by comparison of the predicted essential amino acids with experimental data provided by Armstrong et al. [57]. For this purpose, the lower bound for the particular amino acid’s exchange reaction was set to zero individually, and the essentiality of each amino acid was investigated using FBA by setting the biomass reaction as the objective function. To check the organism’s capability to consume different carbon sources consisting of sucrose, fructose, lactose, and mannose on a complex medium containing yeast extract in aerobic conditions, the organism’s growth was simulated on different carbon sources using FBA. The predicted results were compared with experimental data reported by Pan et al. [58]. The effect of the presence of oxygen on the HA production rate was also studies qualitatively by robustness analysis for HA’s exchange reaction with oxygen’s exchange reaction in comparison with experimental data from Liu et al. [29]. Since the information on the effect of oxygen exchange rate on HA production for a culture medium containing yeast extract and sucrose was available, first, the conditions of this culture medium were simulated. Under these conditions, the oxygen exchange reaction was set to a non-zero value. Then, the HA exchange reaction was set as the objective function and the robustness analysis was performed for the mentioned reaction versus the oxygen exchange reaction.
2.3. Comparison of S. zooepidemicus and C. glutamicum metabolic models for HA biosynthesis
The flux distributions for S. zooepidemicus and recombinant C. glutamicum metabolic models were calculated to compare central carbon metabolism and HA biosynthetic pathways in these GEMs. For this purpose, first, recombinant C. glutamicum was simulated by adding the reaction that simulates HA condensation, and the HA exchange reactions to iCW773 (as a well-established version of C. glutamicum ATCC13032 GEM) [59]. To model an HA-producing C. glutamicum, strategies and pathway modifications detailed in engineering studies, such as Cheng et al. (2019) [20] which utilized iCW773, were considered to define the heterologous system. Then, their flux distributions were predicted at a growth rate of 95% of the optimal and for the HA exchange reaction as an objective function using FBA in aerobic conditions. In these simulations, the CDM1 medium was used for the S. zooepidemicus model and the CGXII minimal medium was used for recombinant C. glutamicum model by setting glucose uptake rates to 18.56 mmol.gDCW-1.h-1 and 4.67 mmol.gDCW-1.h-1 respectively. Then, flux distributions were estimated for 100 mmol.gDCW-1.h-1 glucose consumption rate in each model according to the predicted data by the models. Finally, the flux distributions diagrams for central carbon metabolism and HA biosynthetic pathways for each model were presented for comparison purposes.
Also, in silico single- and double-gene deletions for S. zooepidemicus in CDM1 medium, and recombinant C. glutamicum in CGXII medium were carried out using FBA in aerobic conditions to compare gene interactions in the two networks. Such gene deletion analyses are standard applications of GEMs to understand network robustness and identify essential genes or synthetic lethals [5,60]. Therefore, the percentages of lethal, sick, synthetic lethal, and synthetic sick genes were determined. In single-gene deletion, the relative growth rate (Rgr) was calculated for each gene, which is the ratio of the mutant strain growth rate to the wild strain growth rate. Then, percentages of lethal (Rgr = 0) and sick genes (Rgr < 1) for the two networks were determined. In double-gene deletions, if the predicted growth rate of in silico double-mutant strains was less than the growth rate of either of the single-mutant strains, the two genes were considered to have interactions. When the Rgr of double-gene deletions was equal to zero, the two genes were designated as synthetic lethal. Two genes are synthetic sick if their Rgr in double-gene deletion is less than one.
2.4. Prediction of strategies for overproduction of HA for the two metabolic models
To predict gene-deletion strategies for overproduction of HA using the S. zooepidemicus GEM, an iterative computational approach based on the Minimization of Metabolic Adjustment (MOMA) principle was employed. The application of this principle, particularly for understanding metabolic rerouting and identifying impactful genetic interventions through iterative single-gene deletion analysis, is supported and contextualized by recent advancements in constraint-based modeling [60]. In this method, the biomass reaction is chosen as the objective function and this MOMA-based analysis is used for iterative single-gene deletion analysis to identify significant genes for increasing HA production rates. The CDM medium in aerobic conditions was considered in these calculations and the glucose consumption rate was set to 18.56 mmol.gDCW-1.h-1. Single, double and triple gene knockout mutant strains were investigated using the iterative approach: A gene was recognized in round one whose elimination improved the HA production rate and double-mutants were identified by repetition of single-gene deletion analysis in the genetic background of the suitable single-gene mutant. This process was continued until identifying triple-mutant strains with improved HA production rate. If the deletion of a gene in round one represented more than a 40% decrease in the growth rate, it was not allowed to proceed to the next round.
For the recombinant C. glutamicum model, MOMA was used for iterative single-gene deletions to determine simulated mutant phenotypes that produced optimal HA production rates. As the method suggested only one gene-deletion, it was used based on multiple optimal solutions [61] to predict candidate reactions that could improve the HA production rate by up- and down-regulation. This approach of leveraging multiple optimal solutions is important as metabolic networks can exhibit flux variability. To this end, 5000 multiple optimal flux distributions using LAMOS [62] of maximized and minimized HA production rate at a constant growth rate of 95% optimal were calculated. Finally, a comparison between multiple optimal solutions of maximized and minimized HA production rates identified the critical reactions to improve the HA production rate in C. glutamicum. Advanced algorithms continue to be developed for exploring solution spaces and identifying robust engineering targets [63].
3. Results and discussion
3.1. The genome-scale metabolic network of S. zooepidemicus
S. zooepidemicus is a fastidious and facultative anaerobic bacterium. This organism uses the Embden- Meyerhof pathway (EMP) to metabolize glucose for maintenance and cell growth [18]. The complete tricarboxylic acid (TCA) cycle is absent in S. zooepidemicus, and this bacterium, in order to grow, needs most amino acids and nucleotides to be present in the growth medium [64]. This auxotrophy is a common characteristic of highly host-adapted bacteria which often lose biosynthetic pathways for compounds readily available from their environment. HA is polymerised from UDP-N-acetylglucosamine (UDP-GlcNAC) and UDP-glucuronic acid (UDP-GlcA) dimers by hyaluronic acid synthase (HAS). The HA glycosyl skeleton originates from glucose. Fructose-6-phosphate is produced from glucose by EMP, then it is converted to glucosamine-6-phosphate, and finally to UDP-GlcNAC. UDP-GlcA, as the other HAS substrate, can also be produced from glucose via the glucuronic acid pathway. The balanced and efficient supply of these activated sugar precursors, UDP-GlcNAc and UDP-GlcA, is a critical control point for both HA yield and molecular weight, as demonstrated in recent metabolic engineering efforts in S. zooepidemicus [18,34]. A genome-scale metabolic network of Streptococcus equi subsp. zooepidemicus ATCC 35246 was reconstructed based on genome annotation and literature data. As the first functional GEM for S. zooepidemicus, it was named iZN522. The present model consisted of 677 reactions, 697 metabolites, and 522 genes out of 2049 whole genes (more than 25% coverage). This gene coverage is typical for curated GEMs, focusing on metabolic functions [40]. Out of the 677 chemical reactions in the model, 425 reactions had known genes, 20 reactions were added to the model based on physiological evidence and experimental reports, and 34 reactions were added to the model during the gap-filling process. The model’s features and a comparison of iZN522 with other Streptococci’s available metabolic models are presented in Tables 1 and 2, respectively. Notably, when comparing iZN522 with models like that for S. pyogenes [65], differences in pathways related to virulence and specific nutrient requirements can become apparent, reflecting their distinct ecological niches.
3.2. Evaluation of iZN522 predictions
Quantitative and qualitative validation approaches were applied to evaluate the accuracy of iZN522 predictions. Such validation against experimental data is a cornerstone of establishing confidence in GEM-based simulations [5,38]. One of the quantitative validation steps was predicting the growth rate using a metabolic model in a specific condition. The predicted growth rate by the model in anaerobic conditions and CDM culture medium is close to the reported growth rate by Blank et al. [51]. Table 3 shows the ability of iZN522 to predict the organism’s growth rate in these conditions correctly.
Furthermore, the prediction of essential amino acids for biomass generation was performed using FBA in CDM1 culture, qualitatively. The results obtained from the metabolic model were compared with experimental amino acid auxotrophy provided by Armstrong et al. [57]. Table 4 shows that the model predicts the essential amino acids for the organism’s growth with acceptable accuracy, except for the two amino acids lysine and glutamine.
Indeed, iZN522 forecasts the organism’s growth in the absence of lysine and glutamine in contrast to the experimental reports. As metabolic models cannot predict gene-expression data, it may be possible to correct this discordance with experimental data by adding gene-expression data to the model. The integration of omics data, including transcriptomics, is an ongoing effort in the GEM community to enhance model predictivity and address such discrepancies [4,5]. Because the available articles about amino acid consumption in defined culture media are limited, no other evidence was found to explain these results.
For more evaluation, the organism’s growth was qualitatively simulated on various carbon sources such as sucrose, fructose, lactose, and maltose in addition to glucose using FBA to investigate the capability of the model in prediction of the organism’s growth on different carbon sources. The modeling results are presented in Table 5 and were compared with experimental data reported by Pan et al. [58].
The effect of oxygen presence on the HA production rate was also studied using robustness analysis, qualitatively. Fig 2 shows that oxygen presence leads to an increase in HA production rate following the experimental data presented by Armstrong et al. and Liu et al. [57,68].
Graph showing the predicted effect of varying oxygen exchange flux on hyaluronic acid production rate in S. zooepidemicus culture media containing yeast extract and sucrose. The robustness curve demonstrates increasing HA flux with increasing oxygen availability, consistent with experimental observations. The x-axis represents oxygen exchange rate (mmol/gDCW/h), and the y-axis represents predicted HA synthesis flux (mmol/gDCW/h). The figure illustrates oxygen’s role as a critical environmental factor enhancing HA biosynthesis in aerobic conditions.
3.3. Strategies for enhancement of HA production rate in S. zooepidemicus using iZN522
Fig 2, in accordance with experimental data, shows that aerobic conditions are more suitable for the improvement of HA production rate. Thus, aerobic conditions were selected for the prediction of suitable knockout strategies. Then, the maximum growth rate was selected as an objective function, and single-gene deletion analysis was performed using MOMA. The application of MOMA helps identify gene deletions that cause minimal disruption to the overall metabolic flux distribution while achieving a desired phenotypic change [60]. The results of a single genome-wide knockout simulation are illustrated in Fig 3.
The stoichiometric model suggested that the deletion of eight genes, including trpG, ppc, aceF, fbaA, eno, asnA, asd, and cmk, led to a higher yield of HA production by improving the HA biosynthesis pathway directly and, by decreasing the growth rate, indirectly. Among the eight predicted gene targets, six were removed from further examination: the knockout of trpG and ppc decreased the growth rate intensively, and deletion of cmk did not significantly improve the HA production rate. Furthermore, aceF, fbaA, and eno were vital for the organism’s survival [69].
AceF functions as the E2 component of the pyruvate dehydrogenase complex connecting glycolysis to the TCA cycle, playing a critical role in central carbon metabolism. Mutation or deletion of aceF has been shown to significantly impact bacterial cytotoxicity and energy metabolism with direct effects on critical secretion systems. The essentiality of aceF for cell viability and metabolic regulation has been demonstrated in multiple bacterial species reinforcing its importance in fundamental metabolic flux control. This supports our in silico observation of aceF essentiality in S. zooepidemicus viability and validates its exclusion from knockout candidates [70].
Although our GEM simulations did not predict a significant growth reduction following the in silico deletion of eno, fbaA, and aceF, experimental evidence from other bacteria highlights their critical roles in metabolism, stress tolerance, immune evasion, and virulence. Accordingly, these genes were treated as high-risk knockout targets and excluded from further analysis. Nonetheless, direct validation of their essentiality in S. zooepidemicus remains necessary. We propose follow-up experiments including Δ or conditional knockdown strains, in vitro growth assays, oxidative stress and neutrophil killing tests, and complement binding/opsonization/invasion assays for fbaA [71,72].
Therefore, we consider only asd and asnA to predict in silico double- and triple-mutant strains to improve HA production. First, each of the asnA and asd genes was removed from the stoichiometric matrix separately. Then, a single-gene deletion analysis was performed again to identify a suitable double- and triple-mutant strain. In silico multiple gene-knockout strategies and relative in silico HA production yield of mutant strains are presented in Fig 4.
According to Fig 4, the ilvE, ppc, pyrD, and asnA genes encode valine transaminase, phosphoenolpyruvate carboxylase, dihydroorotate dehydrogenase, and asparagine synthase, respectively. The omission of these genes as double- and triple-gene deletions can be useful to optimize HA yield.
Table 6 summarizes the in silico iterative single-gene deletion results and their HA production rate. The mutant strain growth rate ratio to wild strain growth rate for each simulated mutant strain is presented in Table 6.
All mutant strains increase the HA production rate by decreasing the growth rate and increasing hyaluronan synthase availability to their precursors. This inverse relationship between growth and production of secondary metabolites or complex polymers like HA is a common challenge in metabolic engineering, often requiring strategies that decouple growth from production [8].
Finally, the triple-mutant strain Δasd ΔilvE Δpyr was selected as the most efficient mutant strain due to its suitable growth rate and high HA production rate. The comparison of flux distribution for the wild strain with the mentioned triple-mutant strain has demonstrated the effectiveness of eliminating three genes, including asd, ilvE, and pyrD on HA production metabolic pathway. Fig 5 illustrates the flux distributions of metabolic pathways after asd gene deletion. The asd encodes aspartate semi-aldehyde dehydrogenase and is involved in the transformation of aspartate to threonine and lysine.
Black values indicate the reaction fluxes of the wild strain, and red values indicate reaction fluxes of ∆asd strain. The unit of fluxes is mmol.gDCW-1.h-1 The multiplication sign indicates the deletion of a gene. The equivalent abbreviations for important metabolites are given below (Udpg: UDP-glucose, Uacgam: UDP-N-acetyl glucosamine, PG: Peptidoglycan, Clpn-lla: Cardiolipin, Lyspg-lla: Lysyl phosphatidylglycerol, LTAalaGal: Lipoteichoic acid derivatives, HA: Hyaluronic acid).
The elimination of the asd decreases the flux of reactions that produce aspartate-tRNA, threonine-tRNA, glycine-tRNA, lysine-tRNA, and asparagine-tRNA and led to reducing the flux of protein formation as part of the biomass from 1.64 in wild-type to 1.18 in ∆asd. Lysine is involved in the formation of cardiolipin, lipoic acid, lysyl phosphatidylglycerol, and peptidoglycan (as parts of biomass composition). When the aspartate to lysine conversion pathway is blocked, the flux passing through the enzymatic reactions mentioned is limited, and the growth rate decreases from 0.39 to 0.28 per hour. ilvE was identified as the second knockout for enhancement of HA production rate, which encodes the valine transaminase enzyme. Valine transaminase produces 3-methyl-2-oxobutanate by consuming the valine and converting one alpha-ketoglutarate to glutamine. Alanine is converted to pyruvate simultaneously. Finally, the pyruvate regenerates alanine by alanine transaminase. According to Fig 6, a cycle is formed in the ∆asd strain, which continuously produces alanine. Elimination of ilvE from the stoichiometric matrix blocks the cycle transforming pyruvate to alanine. Thus, the alanine-transaminase flux decreases from 999.29 mmol.gDCW-1.h-1 in ∆asd strain to 1.42 mmol.gDCW-1.h-1 in ∆asd ∆ilvE strain. Whereas alanine is involved in the biosynthesis of cardiolipin, peptidoglycan, and lipoteichoic acid, which generate a major part of biomass, the growth rate decreases from 0.28 h-1 in ∆asd to 0.14 h-1 in ∆asd ∆ilvE by reduced alanine production.
Red values refer to the reaction fluxes of the ilvE ∆asd strain. The fluxes are reported in mmol.gCDW-1.h-1. (Udpg: UDP-glucose, Uacgam: UDP-N-acetylglucosamine, PG: peptidoglycan, Clpn-lla: cardiolipin, Lyspg-lla: lysyl phosphatidylglycerol, LTAalaGal: lipoteichoic acid derivatives, HA Hyaluronic acid, 3 mob: 3-methyl-2-oxobutanate, akg: alpha-ketoglutarate, glu: glutamate).
Hyaluronic acid is synthesized in the two main branches of the intermediate compounds of the glycolysis pathway. The first branch originates from glucose-6-phosphate, which forms UDP-glucuronic acid by consuming one UTP and NAD molecules. Fructose-6-phosphate is the origin of the second branch, which is converted to UDP-N-acetylglucosamine by consuming one molecule of UTP, one molecule of glutamine, and acetyl CoA. Finally, the precursors UDP-N-acetylglucosamine and UDP- glucuronic acid bind together to form the HA chain by the hyaluronan synthase enzyme. UDP-glucose and UDP-N-acetylglucosamine are precursors of HA production. Furthermore, they are involved in the synthesis of cardiolipin, peptidoglycan, and lipoteichoic acid (Fig 5). Thus, when asd and ilvE genes are eliminated from the metabolic model, growth rate decreases, and more HA precursors are consumed for the HA biosynthesis pathway. As a result, hyaluronan synthase flux increases from zero in the wild strain to 5.5 mmol.gDCW-1.h-1 in ∆asd ∆ilvE.pyrD, which as the third identified knockout for improvement of HA biosynthesis, encodes dihydroorotate dehydrogenase, participates in oxidation of (S)-dihydroorotate to orotate and in the de novo biosynthesis of pyrimidine nucleotides. The deletion of pyrD increases both growth rate and HA production rates, simultaneously. On the other hand, in the ∆asd ∆ilvE ∆pyrD triple-mutant strain, HA production and growth rate are coupled together. The omission of pyrD blocks the biosynthesis of UMP from aspartate. Thus, UMP is provided by nucleotide salvage pathways in the triple-mutant strain. In the triple-mutant strain, the flux of reaction related to phosphoribosyl pyrophosphate synthetase (PRPP synthetase) increases compared with ∆asd ∆ilvE.
The improved flux of PRPP synthetase activates the associated reaction with the uracil phosphoribosyltransferase (UPRT) enzyme in the ∆asd ∆ilvE ∆pyrD strain. UPRT consumes one unit of uracil and 5-phosphoribosyl-1-pyrophosphate to produce a unit of UMP. Also, when the reaction of dihydroorotate dehydrogenase is inactivated, a cycle is created that continuously provides the UTP required to produce UDP-N-acetylglucosamine and UDP-glucose (similar precursors for HA and biomass production). Therefore, biomass and HA production are not limited. Furthermore, when the flux of reactions associated with the production of ribonucleotides and deoxyribonucleotides increases, the RNA and DNA production fluxes which form part of the biomass, enhance compared to the double-mutant strain. Thus, the growth rate and HA production rate increase from 0.15 h-1 and 5.5 mmol.gDCW-1.h-1 in ∆asd ∆ilvE strain to 0.20 h-1 and 7.33 mmol.gDCW-1.h-1 in the ∆asd ∆ilvE ∆pyrD, respectively. These predicted improvements highlight the potential of systematic multi-gene knockout strategies identified via GEMs, although experimental validation of such complex mutant phenotypes is essential [5].
3.4. Strategies for enhancement of HA production rate in recombinant C. glutamicum using iCW773
At first, an HA disaccharide binding reaction and an HA exchange reaction were added to iCW773. Then single-gene deletion analysis by MOMA was performed in aerobic conditions for CGXII minimal culture medium with a glucose consumption rate of 4.67 mmol.gDCW-1.h-1. Three genes, including eno, nadhd, and zwf, were suggested as genes whose knockouts improved the HA production rate, as shown in Fig 7.
The black values represent the reaction fluxes of the ∆asd ∆ilvE strain. Red values refer to the reaction fluxes of the ∆asd ∆ilvE ∆pyrD strain. The fluxes are reported in mmol.gCDW-1.h-1. The cross indicates the deletion of the gene. (Udpg: UDP-glucose, Uacgam: UDP-N-acetyl glucosamine, PG: peptidoglycan, CPS-lla: polysaccharide units, Ura: uracil, Uri: uridine, GUA: guanine).
The nadhd, eno, and zwf genes encode the NADH dehydrogenase, enolase, and glucose-6-phosphate dehydrogenase, respectively. As the lack of eno and nadhd genes, decrease the growth rate by less than 40 percent, the only suitable gene deletion is that of the zwf gene, which increases the HA production rate to 0.05 mmol.gDCW-1.h-1. The zwf gene encodes the enzyme glucose-6-phosphate dehydrogenase, which initiates the oxidative pentose phosphate pathway [73]. The elimination of this gene leads to more glucose-6-phosphate entering the HA biosynthetic pathway. This strategy of blocking the initial step of the PPP to channel carbon towards a desired product is a common approach in metabolic engineering of various microorganisms [21]. In addition, the growth rate is reduced by 0.78 percent of the wild strain’s growth rate by attenuation of the generation of NADPH in the oxidative pentose phosphate pathway, which is used for anabolic reactions. Thus, zwf knock out amplifies the flux of hyaluronan synthase by redirecting carbon flux from biomass to HA generation (Fig 8). This result is in accordance to experimental data presented by Cheng et al. [20].
In our in silico gene deletion experiments, we classified genes based on their impact on growth. ‘Lethal’ genes denote those whose deletion completely halts growth, indicating essentiality under modeled conditions. ‘Sick’ genes cause a significant but not fatal growth decrease, reflecting partial impairment. ‘Synthetic lethal’ pairs describe gene combinations where simultaneous deletion is lethal even though single deletions are not, highlighting compensatory pathways. ‘Synthetic sick’ pairs involve combined deletions leading to more severe growth defects than isolated knockouts. These categorizations enable precise characterization of network robustness and pinpoint promising targets for metabolic engineering.
We identified efficient double- and triple-gene knockouts in S. zooepidemicus to optimize HA production rate by iterative MOMA single-gene deletion analysis approach. However, our attempts at the above approach to identify efficient in silico multiple-mutant strain in the C. glutamicum model (iCW773) to produce an optimal HA production rate were not successful. It seems that the difference in gene interactions in the two metabolic networks has caused this difference. To prove this assumption, single-gene deletion analysis, double-gene deletion analysis with one, and at least five gene interactions were performed. Thus, the percentage of lethal and sick genes for the two metabolic networks iCW773 and iZN522 were obtained. In addition, synthetic lethal and synthetic sick genes with one and at least five interactions for the two models were calculated and compared. According to Table 7, the percentage of lethal genes, synthetic lethal genes with one interaction, and at least five interactions for the iZN522 model are much higher than the amount calculated for the iCW773 model. This suggests a potentially higher degree of metabolic network robustness or redundancy in the iCW773 model of C. glutamicum compared to iZN522 for S. zooepidemicus, which could explain the differential success of iterative single-gene deletion strategies for complex product optimization [60].
The results indicate that there are more target points in iZN522 than iCW773 to reduce the growth rate. Thus, the iterative single-gene deletions strategy will not be sufficient for HA optimization in this strain. Furthermore, a method based on multiple optimal solutions was used to improve HA production using the iCW773 metabolic network. To identify critical reactions that play an influential role in the enhancement of HA production under recombinant C. glutamicum, 5000 multiple optimal flux distributions using LAMOS in two conditions at a constant growth rate of 95% optimal were calculated. Then, the comparison of multiple optimal solutions of maximized and minimized HA production rate under recombinant C. glutamicum was carried out to identify target reactions that should be up- or down-regulated to optimize the HA production rate. According to model predictions, the up-regulation of reactions of phosphoribosylglycinamide formyltransferase, Mg antiporter, adenylate kinase, deoxy guanylate kinase, glutamate N-acetyltransferase, carbamate kinase, uridylate kinase, cytidylate kinase, hyaluronan synthase and down-regulation of reactions of GAR transformylase, Mg2+ ABC transporter, nucleoside diphosphate kinase, and nucleoside triphosphate pyrophosphorylase increase HA production rate from zero to 0.22 mmol.gDCW-1.h-1 (Fig 9).
The blue arrows represent reactions that carry a normal flux; the green arrows should up-regulate, while the flux of reactions with the red arrows should down-regulate.
The model predicts that upregulation of anabolic reactions responsible for precursor biosynthesis coupled with downregulation of ATP-consuming processes may collectively promote increased hyaluronic acid production, suggesting focused metabolic flux control as a promising strategy [8].
For the production of each unit of disaccharide in the HA chain, five molecules of ATP are consumed [33]. Therefore, the above changes reduce the energy cost of producing the metabolites which are essential to generate biomass, and the excess ATP is used to increase the HA production rate. In addition, for each entry of an Mg2+ ion, two proton ions are transferred to the extracellular environment by the Mg2+ antiporter. Assuming the transient flux of the Mg2+ antiporter enhances, the proton gradient in the extracellular environment increases, consequently increasing the flux of ATP synthase and ATP production. The method based on multiple optimal solutions suggested that the up-regulation of the glutamate-N-acetyltransferase reaction can also enhance the HA production rate. Acetylglucosamine is a precursor required for the formation of HA and biomass. Therefore, saving acetyl-CoA can also increase HA. In fact, increasing the reaction flux of glutamate-N-acetyltransferase, which converts a unit of acetyl ornithine to ornithine, provides the acetyl that is needed to produce acetyl glutamate. Thus, the acetyl-CoA required to increase HA production is maintained and improves the HA production rate.
3.5. The comparison of HA production pathway in iZN522 and iCW773
For this purpose, the flux distribution for HA production, especially in central carbon metabolism (glycolysis, pentose-phosphate, and Krebs cycle), was calculated for both strains using FBA with a glucose consumption rate of 100 mmol.gDCW-1.h-1. These flux distributions were compared to identify further strategies for improving HA production in recombinant C. glutamicum. In contrast to S. zooepidemicus, the oxidative pentose-phosphate pathway is active in C. glutamicum, as shown in Fig 10.
S. zooepidemicus fluxes were reported in green, and recombinant C. glutamicum fluxes were shown in red. The dot line also indicates reactions that are present only in C. glutamicum, and S. zooepidemicus lacks them. Negative fluxes also indicate the reaction flux in the opposite direction.
During this process, glucose-6-phosphate is first converted to 6-phosphogluconate and produces 1 mole of NADPH by the glucose-6-phosphate dehydrogenase. The enzyme 6-phosphogluconate dehydrogenase then produces ribulose-5-phosphate and releases 1 unit of CO2 and NADPH from the 6-phosphogluconate. As C. glutamicum has the potential to grow in a minimal culture medium, it needs higher volumes of NADPH to produce the amino acids, cofactors, and nucleotide compounds that are necessary for its growth. This inherent metabolic capability for NADPH generation is a hallmark of C. glutamicum’s robust central metabolism, often exploited in metabolic engineering [21,22]. S. zooepidemicus is a fastidious bacterium that can only grow in complex culture media, so it requires less NADPH than C. glutamicum due to its defective amino acid biosynthesis pathways. The oxidative pentose-phosphate pathway in C. glutamicum causes a large volume of glucose-6-phosphate to enter this pathway. Thus, less glucose-6-phosphate enters the HA production pathway. Hence the enzyme hyaluronan synthase in C. glutamicum carries a lower flux than in S. zooepidemicus under baseline non-engineered conditions for HA production. The non-oxidative pentose-phosphate pathway contains transketolase and transaldolase activity. In S. zooepidemicus, these enzymes use fructose-6-phosphate and glyceraldehyde-3-phosphate to provide ribose-5-phosphate needed for nucleotides production. In S. zooepidemicus, a part of the fructose-6-phosphate and glyceraldehyde-3-phosphate produced in the glycolysis pathway is used to produce ribose-5-phosphate, so the flux passing through both reactions of phosphofructokinase and glyceraldehyde-3-phosphate dehydrogenase is reduced. Whereas, in C. glutamicum, non-oxidative reactions of pentose-phosphate pathway are carried out to convert ribose-5-phosphate and xylulose-5-phosphate to fructose-6-phosphate and glyceraldehyde-3-phosphate, which increases the fluxes of phosphofructokinase and glyceraldehyde-3-phosphate dehydrogenase. This difference in the pentose-phosphate pathway depends on the activation or inactivation of glucose-6-phosphate dehydrogenase. Thus, the omission of the glucose-6-phosphate dehydrogenase gene can improve the HA production rate in recombinant C. glutamicum. Most of the Krebs cycle reactions in S. zooepidemicus were absent. In contrast, recombinant C. glutamicum has a complete Krebs cycle and electron transport chain. Therefore, in recombinant C. glutamicum, more ATP is produced per mole of glucose than in S. zooepidemicus. According to model results, the amount of ATP produced per mole of glucose in S. zooepidemicus is equivalent to 20 molecules of ATP, but in C. glutamicum it is equal to 26.58 molecules of ATP. The HA biosynthesis pathway consumes ATP; therefore, considering the ATP yield in both models, recombinant C. glutamicum appears theoretically more suitable for HA production from an energetic standpoint, provided carbon flux can be efficiently channeled [8].
This comprehensive comparison using GEMs reveals distinct metabolic architectures and energetic efficiencies that underpin the different natural and engineered capacities for HA production in these two organisms. Such insights are crucial for selecting appropriate host chassis and devising rational engineering strategies for improved biopolymer synthesis [5,37]. Further developments, such as incorporating kinetic information or using hybrid modeling approaches [74], could offer even deeper understanding and more precise predictions for optimizing HA production.
3.6. Comparative discussion on key essential genes across organisms
We elaborate on critical enzymatic functions shared by S. zooepidemicus and C. glutamicum, emphasizing genes such as asd (aspartate semi-aldehyde dehydrogenase) and asnA (asparagine synthase), which are central to amino acid biosynthesis. These enzymes influence the pool of metabolites that serve dual roles in biomass formation and hyaluronic acid precursor synthesis. The strategic modulation of these genes markedly affects carbon flux toward HA biosynthesis pathways, demonstrating their high value as engineering candidates across both models.
In this study, the iZN522 metabolic model was reconstructed as the first metabolic model for S. zooepidemicus. These flux perturbations support a balanced allocation of resources between biomass synthesis and hyaluronic acid production, thereby optimizing overall cellular productivity under engineering conditions. This model was utilized to identify the parameters influencing HA production and to explore potential solutions for enhancing HA production through targeted genetic modifications. Furthermore, the metabolic model of recombinant C. glutamicum was simulated as a suitable recombinant strain for HA production. Strategies involving up- and down-regulation of key reactions, guided by computational analysis, were also proposed to enhance HA production in this strain were also proposed. Lastly, a comparison of the central carbon metabolic pathways in both strains elucidated the factors contributing to the higher HA production potential and different metabolic efficiencies in S. zooepidemicus compared to the baseline engineered recombinant C. glutamicum.
The development of iZN522 offers a significant advancement for understanding and engineering S. zooepidemicus, a key industrial HA producer. The model-driven identification of gene knockout targets provides specific, actionable hypotheses for experimental validation aimed at increasing HA titers and productivity [8,18]. The insights gained from the comparative analysis with C. glutamicum not only explain observed differences in HA production capabilities but also inform broader strategies for host selection and optimization in metabolic engineering [37].
Looking forward, the iZN522 model can serve as a foundational tool for more advanced studies. Future efforts could focus on its expansion by integrating multi-omics data (e.g., transcriptomics, proteomics) to capture regulatory dynamics and further refine predictions [4,5]. Investigating the metabolic determinants of HA molecular weight control, a critical quality attribute, represents another important avenue where iZN522 could provide valuable insights, potentially guiding engineering strategies similar to those explored recently [34]. Moreover, the development of dynamic or hybrid models that incorporate kinetic parameters and fermentation process variables could lead to more accurate simulations of industrial bioprocesses [63,74]. Continued experimental validation of the targets identified in this study will be paramount. Ultimately, this work contributes to the growing body of knowledge in systems metabolic engineering, paving the way for more efficient and sustainable biotechnological production of HA and other valuable biopolymers [12,15].
4. Conclusion
In this study, we report the first comprehensive genome-scale metabolic model of Streptococcus zooepidemicus (iZN522), providing strong quantitative validation against experimental growth data and substrate utilization profiles. Comparative model-based analyses with the recombinant Corynebacterium glutamicum model (iCW773) reveal distinct metabolic engineering strategies for increased hyaluronic acid production: targeted gene knockouts in S. zooepidemicus versus flux modulation in C. glutamicum. These insights offer a systems-level understanding that bridges native and recombinant hosts, guiding rational strain improvement for industrial HA biosynthesis.
Acknowledgments
During the preparation of this article the authors used Perplexity in order to assist with English language. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.
References
- 1. Bordbar A, Monk JM, King ZA, Palsson BO. Constraint-based models predict metabolic and associated cellular functions. Nat Rev Genet. 2014;15(2):107–20. pmid:24430943
- 2.
Castillo S, Patil KR, Jouhten P. Yeast genome-scale metabolic models for simulating genotype–phenotype relations. Yeasts in Biotechnology and Human Health. Springer; 2019. pp. 111–33.
- 3. Gu C, Kim GB, Kim WJ, Kim HU, Lee SY. Current status and applications of genome-scale metabolic models. Genome Biol. 2019;20(1):121. pmid:31196170
- 4. Passi A, Tibocha-Bonilla JD, Kumar M, Tec-Campos D, Zengler K, Zuniga C. Genome-scale metabolic modeling enables in-depth understanding of big data. Metabolites. 2022;12(1):14. pmid:35050136
- 5. Carter EL, Constantinidou C, Alam MT. Applications of genome-scale metabolic models to investigate microbial metabolic adaptations in response to genetic or environmental perturbations. Briefings Bioinform. 2024;25(1):bbad439. pmid:38048080
- 6. Liu L, Liu Y, Li J, Du G, Chen J. Microbial production of hyaluronic acid: current state, challenges, and perspectives. Microbial Cell Fact. 2011;10:1–9. pmid:22088095
- 7. Serra M, Casas A, Toubarro D, Barros AN, Teixeira JA. Microbial hyaluronic acid production: a review. Molecules. 2023;28(5):2084. pmid:36903332
- 8. Marwan-Abdelbaset E, Samy-Kamal M, Tan D, Lu X. Microbial production of hyaluronic acid: the current advances, engineering strategies and trends. J Biotechnol. 2025;403:52–72. pmid:40154620
- 9. Chen WY, Marcellin E, Steen JA, Nielsen LK. The role of hyaluronic acid precursor concentrations in molecular weight control in Streptococcus zooepidemicus. Mol Biotechnol. 2014;56(2):147–56. pmid:23903961
- 10. Gunasekaran V, Gowdhaman D, Ponnusami V. Role of membrane proteins in bacterial synthesis of hyaluronic acid and their potential in industrial production. Int J Biol Macromol. 2020;164:1916–26. pmid:32791275
- 11. Ciriminna R, Scurria A, Pagliaro M. Microbial production of hyaluronic acid: the case of an emergent technology in the bioeconomy. Biofuels Bioprod Bioref. 2021;15(6):1604–10.
- 12. Saadati F, Bahrulolum H, Talebi M, Karimi M, Bozorgchami N, Ghale RA, et al. Advances and principles of hyaluronic acid production, extraction, purification, and its applications: a review. Int J Biol Macromol. 2025;312:143839. pmid:40318723
- 13. Talebi M, Ghale RA, Asl RM, Tabandeh F. Advancements in characterization and preclinical applications of hyaluronic acid-based biomaterials for wound healing: a review. Carbohydrate Polymer Technol Appl. 2025;9:100706.
- 14. Ucm R, Aem M, Lhb Z, Kumar V, Taherzadeh MJ, Garlapati VK, et al. Comprehensive review on biotechnological production of hyaluronic acid: status, innovation, market and applications. Bioengineered. 2022;13(4):9645–61. pmid:35436410
- 15. Ferreira RG, Azzoni AR, Santana MHA, Petrides D. Techno-economic analysis of a hyaluronic acid production process utilizing streptococcal fermentation. Processes. 2021;9(2):241.
- 16. de Oliveira JD, Carvalho LS, Gomes AMV, Queiroz LR, Magalhães BS, Parachin NS. Genetic basis for hyper production of hyaluronic acid in natural and engineered microorganisms. Microbial Cell Fact. 2016;15(1):1–19. pmid:27370777
- 17. Zhang Y, Dong J, Xu G, Han R, Zhou J, Ni Y. Efficient production of hyaluronic acid by Streptococcus zooepidemicus using two-stage semi-continuous fermentation. Bioresour Technol. 2023;377:128896. pmid:36933576
- 18. Wei M, Huang Y, Zhu J, Qiao Y, Xiao N, Jin M, et al. Advances in hyaluronic acid production: Biosynthesis and genetic engineering strategies based on Streptococcus—a review. Int J Biol Macromol. 2024;270(Pt 2):132334. pmid:38744368
- 19. Hoffmann J, Altenbuchner J. Hyaluronic acid production with Corynebacterium glutamicum: effect of media composition on yield and molecular weight. J Appl Microbiol. 2014;117(3):663–78. pmid:24863652
- 20. Cheng F, Yu H, Stephanopoulos G. Engineering Corynebacterium glutamicum for high-titer biosynthesis of hyaluronic acid. Metabolic Eng. 2019;55:276–89. pmid:31301358
- 21. Zha J, Zhao Z, Xiao Z, Eng T, Mukhopadhyay A, Koffas MA, et al. Biosystem design of Corynebacterium glutamicum for bioproduction. Curr Opin Biotechnol. 2023;79:102870.
- 22. Wolf S, Becker J, Tsuge Y, Kawaguchi H, Kondo A, Marienhagen J, et al. Advances in metabolic engineering of Corynebacterium glutamicum to produce high-value active ingredients for food, feed, human health, and well-being. Essays Biochem. 2021;65(2):197–212. pmid:34096577
- 23. Du Y, Cheng F, Wang M, Xu C, Yu H. Indirect pathway metabolic engineering strategies for enhanced biosynthesis of hyaluronic acid in engineered Corynebacterium glutamicum. Front Bioeng Biotechnol. 2021;9:768490. pmid:34988066
- 24. Liu L, Wang M, Du G, Chen J. Enhanced hyaluronic acid production of Streptococcus zooepidemicus by an intermittent alkaline‐stress strategy. Lett Appl Microbiol. 2008;46(3):383–8. pmid:18221275
- 25. Zhang X, Wang M, Li T, Fu L, Cao W, Liu H. Construction of efficient Streptococcus zooepidemicus strains for hyaluoronic acid production based on identification of key genes involved in sucrose metabolism. AMB Express. 2016;6(1):121. pmid:27896786
- 26. Jagannath S, Ramachandran KB. Influence of competing metabolic processes on the molecular weight of hyaluronic acid synthesized by Streptococcus zooepidemicus. Biochem Eng J. 2010;48(2):148–58.
- 27. Huang W-C, Chen S-J, Chen T-L. The role of dissolved oxygen and function of agitation in hyaluronic acid fermentation. Biochem Eng J. 2006;32(3):239–43.
- 28. Johns MR, Goh L-T, Oeggerli A. Effect of pH, agitation and aeration on hyaluronic acid production byStreptococcus zooepidemicus. Biotechnology Lett. 1994;16(5):507–12.
- 29. Liu L, Yang H, Du G, Chen J, Wang M. Enhanced hyaluronic acid production of streptococcus zooepidemicus by shifting dissolved oxygen level based on broth rheology and oxygen mass transfer characteristics. Food Biotechnol. 2009;23(2):148–61.
- 30. Patil KP, Kamalja KK, Chaudhari BL. Optimization of medium components for hyaluronic acid production by Streptococcus zooepidemicus MTCC 3523 using a statistical approach. Carbohydrate Polymers. 2011;86(4):1573–7.
- 31. Pires AMB, Eguchi SY, Santana MHA. The influence of mineral ions on the microbial production and molecular weight of hyaluronic acid. Appl Biochem Biotechnol. 2010;162(8):2125–35. pmid:20495890
- 32. Pires AMB, Santana MHA. Metabolic effects of the initial glucose concentration on microbial production of hyaluronic acid. Appl Biochem Biotechnol. 2010;162(6):1751–61. pmid:20411440
- 33. Cheng F, Luozhong S, Guo Z, Yu H, Stephanopoulos G. Enhanced biosynthesis of hyaluronic acid using engineered Corynebacterium glutamicum via metabolic pathway regulation. Biotechnology J. 2017;12(10):1700191. pmid:28869338
- 34. Zhao R, Li J, Li Y, Pei X, Di J, Xie Z, et al. Inducible engineering precursor metabolic flux for synthesizing hyaluronic acid of customized molecular weight in Streptococcus zooepidemicus. Microbial Cell Factories. 2025;24(1):1–13. pmid:39825423
- 35. Gao W, Zhang X, Zhang G, Zuo M, Cao W, Xie Z, et al. Is hyaluronic acid production transcriptionally regulated? A transcriptional repressor gene deletion study in Streptococcus zooepidemicus. Appl Microbiol Biotechnol. 2021;105:8495–504. pmid:34661707
- 36. Lieven C, Beber ME, Olivier BG, Bergmann FT, Ataman M, Babaei P, et al. MEMOTE for standardized genome-scale metabolic model testing. Nature Biotechnol. 2020;38(3):272–6. pmid:32123384
- 37. Fang X, Lloyd CJ, Palsson BO. Reconstructing organisms in silico: genome-scale models and their emerging applications. Nat Rev Microbiol. 2020;18(12):731–43. pmid:32958892
- 38. Bernstein DB, Sulheim S, Almaas E, Segrè D. Addressing uncertainty in genome-scale metabolic model reconstruction and analysis. Genome Biol. 2021;22:1–22. pmid:33602294
- 39. Thiele I, Palsson BØ. A protocol for generating a high-quality genome-scale metabolic reconstruction. Nature protocols. 2010;5(1):93. pmid:20057383
- 40. Gu C, Kim GB, Kim WJ, Kim HU, Lee SY. Current status and applications of genome-scale metabolic models. Genome Biol. 2019;20(1):121. pmid:31196170
- 41. Ma Z, Geng J, Zhang H, Yu H, Yi L, Lei M, et al. Complete genome sequence of Streptococcus equi subsp. zooepidemicus strain ATCC 35246. Am Soc Microbiol. 2011.
- 42. Kanehisa M, Furumichi M, Sato Y, Matsuura Y, Ishiguro-Watanabe M. KEGG: biological systems database as a model of the real world. Nucleic Acids Res. 2025;53(D1):D672–D7. pmid:39417505
- 43. Ren Q, Chen K, Paulsen IT. TransportDB: a comprehensive database resource for cytoplasmic membrane transport systems and outer membrane channels. Nucleic acids research. 2007;35(suppl_1):D274–D9. pmid:17135193
- 44. Kim S, Chen J, Cheng T, Gindulyte A, He J, He S, et al. PubChem 2025 update. Nucleic Acids Res. 2025;53(D1):D1516–D25. pmid:39558165
- 45. Hastings J, Owen G, Dekker A, Ennis M, Kale N, Muthukrishnan V, et al. ChEBI in 2016: Improved services and an expanding collection of metabolites. Nucleic Acids Res. 2016;44(D1):D1214–D9. pmid:26467479
- 46. Machado D, Andrejev S, Tramontano M, Patil KR. Fast automated reconstruction of genome-scale metabolic models for microbial species and communities. Nucleic Acids Res. 2018;46(15):7542–53. pmid:30192979
- 47. Norsigian CJ, Pusarla N, McConn JL, Yurkovich JT, Dräger A, Palsson BO, et al. BiGG Models 2020: multi-strain genome-scale models and expansion across the phylogenetic tree. Nucleic Acids Res. 2020;48(D1):D402–6. pmid:31696234
- 48. Flahaut NAL, Wiersma A, van de Bunt B, Martens DE, Schaap PJ, Sijtsma L, et al. Genome-scale metabolic model for Lactococcus lactis MG1363 and its application to the analysis of flavor formation. Appl Microbiol Biotechnol. 2013;97(19):8729–39. pmid:23974365
- 49. Ruffing A, Chen RR. Metabolic engineering of microbes for oligosaccharide and polysaccharide synthesis. Microb Cell Fact. 2006;5:25. pmid:16859553
- 50. Sze JH, Brownlie JC, Love CA. Biotechnological production of hyaluronic acid: a mini review. 3 Biotech. 2016;6:1–9. pmid:28330137
- 51. Blank LM, McLaughlin RL, Nielsen LK. Stable production of hyaluronic acid in Streptococcus zooepidemicus chemostats operated at high dilution rate. Biotechnol Bioeng. 2005;90(6):685–93. pmid:15803469
- 52. Hamilton I, Phipps P, Ellwood D. Effect of growth rate and glucose concentration on the biochemical properties of Streptococcus mutans Ingbritt in continuous culture. Infect Immun. 1979;26(3):861–9. pmid:43291
- 53. Keating SM, Waltemath D, König M, Zhang F, Dräger A, Chaouiya C, et al. SBML Level 3: an extensible format for the exchange and reuse of biological models. Mol Syst Biol. 2020;16(8):e9110. pmid:32845085
- 54. Heirendt L, Arreckx S, Pfau T, Mendoza SN, Richelle A, Heinken A, et al. Creation and analysis of biochemical constraint-based models using the COBRA Toolbox v. 3.0. Nat Protoc. 2019;14(3):639–702. pmid:30787451
- 55. Orth JD, Thiele I, Palsson BØ. What is flux balance analysis? Nat Biotechnol. 2010;28(3):245–8. pmid:20212490
- 56. Oh Y-K, Palsson BO, Park SM, Schilling CH, Mahadevan R. Genome-scale reconstruction of metabolic network in Bacillus subtilis based on high-throughput phenotyping and gene essentiality data. J Biol Chem. 2007;282(39):28791–9. pmid:17573341
- 57. Armstrong D, Cooney M, Johns M. Growth and amino acid requirements of hyaluronic-acid-producing Streptococcus zooepidemicus. Appl Microbiol Biotechnol. 1997;47(3):309–12.
- 58. Pan NC, Vignoli JA, Baldo C, Pereira HCB, Silva R, Celligoi M. Agroindustrial byproducts for the production of hyaluronic acid by Streptococcus zooepidemicus ATCC 39920. Int J Sci Technol Res. 2015;4(4):114–8.
- 59. Zhang Y, Cai J, Shang X, Wang B, Liu S, Chai X, et al. A new genome-scale metabolic model of Corynebacterium glutamicum and its application. Biotechnol Biofuels. 2017;10(1):169. pmid:28680478
- 60. Narasimha SM, Malpani T, Mohite OS, Nath JS, Raman K. Understanding flux switching in metabolic networks through an analysis of synthetic lethals. npj Syst Biol Appl. 2024;10(1):104. pmid:39289347
- 61. Mekanik M, Motamedian E, Fotovat R, Jafarian V. Reconstruction of a genome-scale metabolic model for Auxenochlorella protothecoides to study hydrogen production under anaerobiosis using multiple optimal solutions. Int J Hydrogen Energy. 2019;44(5):2580–91.
- 62. Motamedian E, Naeimpoor F. LAMOS: A linear algorithm to identify the origin of multiple optimal flux distributions in metabolic networks. Comput Chem Eng. 2018;117:372–7.
- 63. Sahu A, Blätke M-A, Szymański JJ, Töpfer N. Advances in flux balance analysis by integrating machine learning and mechanism-based models. Comput Struct Biotechnol J. 2021;19:4626–40. pmid:34471504
- 64. Gao H-J, Du G-C, Chen J. Analysis of metabolic fluxes for hyaluronic acid (HA) production by Streptococcus zooepidemicus. World J Microbiol Biotechnol. 2006;22(4):399–408.
- 65. Hirose Y, Zielinski DC, Poudel S, Rychel K, Baker JL, Toya Y, et al. A genome-scale metabolic model of a globally disseminated hyperinvasive M1 strain of Streptococcus pyogenes. Msystems. 2024;9(9):e00736–24. pmid:39158303
- 66. Pastink MI, Teusink B, Hols P, Visser S, de Vos WM, Hugenholtz J. Genome-scale model of Streptococcus thermophilus LMG18311 for metabolic comparison of lactic acid bacteria. Appl Environ Microbiol. 2009;75(11):3627–33. pmid:19346354
- 67. Levering J, Fiedler T, Sieg A, van Grinsven KW, Hering S, Veith N, et al. Genome-scale reconstruction of the Streptococcus pyogenes M49 metabolic network reveals growth requirements and indicates potential drug targets. J Biotechnol. 2016;232:25–37. pmid:26970054
- 68. Liu L, Du G, Chen J, Wang M, Sun J. Comparative study on the influence of dissolved oxygen control approaches on the microbial hyaluronic acid production of Streptococcus zooepidemicus. Bioprocess Biosyst Eng. 2009;32(6):755–63. pmid:19205747
- 69. Charbonneau AR, Forman OP, Cain AK, Newland G, Robinson C, Boursnell M, et al. Defining the ABC of gene essentiality in streptococci. Bmc Genomics. 2017;18(1):1–11. pmid:28569133
- 70. Li H, Xia Y, Tian Z, Jin Y, Bai F, Cheng Z, et al. Dihydrolipoamide acetyltransferase AceF Influences the Type III secretion system and resistance to oxidative stresses through RsmY/Z in Pseudomonas aeruginosa. Microorganisms. 2022;10(3):666. pmid:35336241
- 71. Weng Y, Chen F, Liu Y, Zhao Q, Chen R, Pan X, et al. Pseudomonas aeruginosa enolase influences bacterial tolerance to oxidative stresses and virulence. Front Microbiol 2016;7:1999. pmid:28018326
- 72. Ma C, Guo Y, Gu H, Zhang L, Liu H, Feng H, et al. A novel monoclonal antibody against FbaA can inhibit the binding of the complement regulatory protein factor H to group A streptococcus. Clin Vacc Immunol. 2011;18(4):552–8. pmid:21228138
- 73. Ihnen ED, Demain AL. Glucose-6-phosphate dehydrogenase and its deficiency in mutants of Corynebacterium glutamicum. J Bacteriol. 1969;98(3):1151–8. pmid:5788701
- 74. Rathinavelu S, Pavan SS, Sivaprakasam S. Hybrid Model-based Framework for Soft Sensing and Forecasting Key Process Variables in the Production of Hyaluronic Acid by Streptococcus zooepidemicus. Biotechnol Bioprocess Eng. 2023;28(1):203–14.