Levan production by B. subtilis LY7.16 under different initial sucrose concentrations

Levan production is primarily influenced by substrate concentration and the metabolic potential of the producing microorganism. Drawing data from levan synthesis under varying sucrose concentrations (Fig A in S1 Text), we examined the interplay between these factors to identify strategies for surpassing current production limits. The primary analysis of levan production by B. subtilis LY7.16 showed that the maximal levan production positively correlated with the increase of initial sucrose concentration. The association showed steeply increased levan production up to 200 g · L^-1 of initial sucrose concentration, then steadily when concentration greater than 250 g · L^-1 (Fig 1A and 1B). The changes in rate of levan production upon the given sucrose concentration are corresponding to the decline of B. subtilis LY7.16 growth rate (Fig 1B and 1D). The results showed that cell biomass production rate steadily high at sucrose concentrations 50 g · L^-1 to 200 g · L^-1 and started decreasing at 250 g · L^-1 (Fig 1D). At the high sucrose concentrations (250 g · L^-1 to 300 g · L^-1), B. subtilis LY7.16 exhibited slow biomass production, which were presumed to result in lower levansucrase levels. However, under this condition, the highest levan yield by substrate use was apparently observed (Fig 1E and Table 1). The contrasting effects of sucrose concentration on maximal levan production and cell biomass formation create a complex challenge in achieving the optimal balance of these influential factors for levan production. This phenomenon is hypothesized to result from differences in the equilibrium states of sucrose utilization for B. subtilis LY7.16 growth and its availability for levan synthesis via extracellular levansucrase at varying sucrose concentrations.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Analysis of levan production by B. subtilis LY7.16 under varying initial sucrose concentrations.

https://doi.org/10.1371/journal.pcbi.1014273.t001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Levan production by B. subtilis LY7.16 under varying initial sucrose concentrations, (A) maximal levan concentration, (B) levan production rate, (C) maximal B. subtilis LY7.16 biomass concentration, (D) maximal specific growth rate, and (E) levan yield by sucrose use.

https://doi.org/10.1371/journal.pcbi.1014273.g001

Metabolite profiling during levan production (Fig A in S1 Text) suggested that the equilibrium shift is likely to occur when initial sucrose concentrations were greater than 200 g · L^-1. At low range of sucrose concentrations (50 g · L^-1 and 100 g · L^-1), sucrose was rapidly converted and showed a sharp increase of cell biomass production and levan within a short time. This rapid conversion occurred only momentarily due to lacking sucrose to sustain biomass and levan production (Fig A (A, B) in S1 Text). Following sucrose depletion, an increase in fructose levels was observed, accompanied by a decline in cell biomass and levan. This phenomenon may be attributed either to insufficient levansucrase production by B. subtilis LY7.16 or to levansucrase-mediated levan degradation as a compensatory response, wherein the enzyme utilizes accumulated levan as an alternative fructose source under sucrose-limited conditions. In this regard, sucrose availability plays a crucial role in levan production, as it not only promotes levan synthesis but also prevents its degradation. At high range of sucrose concentrations (250 g·L^-1 and 300 g·L^-1), sucrose gradually decreased during the initial phase before undergoing a steep decline, while fructose showed significantly decreased then slowly rebounded after sucrose depletion. These changes in substrate profiles were associated with a slower rate of cell biomass and levan production and decay compared to the observation at lower range of sucrose concentrations (Fig A (C-E) in S1 Text). The high sucrose availability likely hindered B. subtilis LY7.16 growth during the early period, as observed deceleration in growth rate compared to lower concentrations, but the initial futile level of sucrose appeared to mitigate the sharp degradation of levan after sucrose depleted. The results suggested distinct mechanisms of substrate conversion and possibly differences in intracellular metabolic activity between low and high initial sucrose concentrations, ultimately influencing total yield and substrate utilization efficiency. These findings highlight the strong dependence of levan production on the balance between sucrose utilization for cell growth and its conversion into levan, which was greatly influenced by the initial sucrose availability.

Dynamic metabolic modeling of levan production by B. subtilis LY7.16 under low and high sucrose culture conditions

To maximize levan yield and ensure efficient substrate utilization by B. subtilis LY7.16, a process-based model of levan production was developed, as illustrated in Fig 2. The hypothetical cascade of levan production by B. subtilis LY7.16 described the conversion of sucrose as a substrate for microbial levansucrase productions and extracellular levan synthesis via the secreted levansucrase enzyme (Fig 2A). These processes are proposed as key determinants of final levan yield and substrate utilization efficiency. The presence and level of sucrose concentrations triggered the expression of sacB gene that encoded levansucrase enzyme, which was secreted outside the cell. Levansucrase converts sucrose via two biochemical reactions: hydrolysis – breaking down sucrose into fructose and glucose, and transfructosylation – converting sucrose to levan and glucose. The apparent antagonistic relationship of levan synthesis by levansucrase and the sucrose-dependent enzyme production in B. subtilis LY7.16 introduces complexity in optimizing the levan production process to achieve both efficiency and effective substrate utilization.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Scheme of a dynamic metabolic modeling framework of levan production by B. subtilis LY7.16 (ly716-Bs-dMM).

(A) Hypothetical cascade illustrates a process-based model of levan production by B. subtilis LY7.16. All rates of metabolite conversion (v) are presented in mmol·(L·h)^-1. Intracellular metabolic fluxes and cell growth rate are presented in mmol·(gDW·h)^-1 and h^-1, respectively. (B) The integrated dynamic metabolic modeling framework demonstrates the coupling of the dynamic (BsODE) and genome-scale metabolic (BsCBM) models in representing the association of extracellular substrate conversion and microbial metabolic-dependent levansucrase production in levan synthesis.

https://doi.org/10.1371/journal.pcbi.1014273.g002

The integrated dynamic metabolic modeling framework, denoted as ly716-Bs-dMM (Fig 2B), combines dynamic (BsODE) and genome-scale metabolic models (BsCBM) to simultaneously analyze environmental and cellular metabolism under specific conditions. The BsODE model integrates extracellular reactions (Fig 2A) to simulate microbial biomass growth, sucrose consumption, and the synthesis of glucose, fructose, and levan under constrained parameters. This framework enables simulation of dynamic profiling of metabolite concentrations over time and facilitates the prediction of intracellular metabolic responses to varying initial sucrose concentrations.

The ly716-Bs-dMM was first employed to simulate sucrose conversion into levan synthesis at an initial sucrose concentration of 100 g·L^-1. The simulation results in Fig 3A demonstrated that the model effectively captured sucrose consumption in relation to microbial cell biomass and levan production, though it slightly over-estimated the maximal levan production in this framework. The model also successfully simulated changes in all relevant biochemical concentrations during levan production. The simulation revealed that sucrose was rapidly hydrolyzed into glucose and fructose, and was also efficiently utilized by B. subtilis LY7.16. Its rapid consumption led to the swift accumulation of fructose substrate and levansucrase enzyme, facilitating levan synthesis. As sucrose levels declined, fructose remained steady while glucose continued to increase, indicating heightened transfructosylation activity in converting fructose to levan after six hours. Unlike hydrolysis, transfructosylation produces both levan and glucose and is significantly influenced by sucrose and levansucrase availability. Upon sucrose depletion, cell growth declined, and cell-secreted levansucrase enzyme was predicted to level off. During this phase, levan production appeared to decrease, in contrast to a rise in fructose levels. This shift suggested a reversal in the levansucrase-mediated reaction, wherein the degradation of existing levan into fructose helped sustain equilibrium between cell growth and levan synthesis. The ability of levansucrase to degrade levan into its fructose monomer has been well established [10,34]. Recent studies indicate that levan degradation progresses only up to a certain level and does not achieve complete conversion, likely due to the chemical structure of levan [34]. As a result, levan degradation and the subsequent synthesis of fructose reached a steady state once this condition was met.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Simulation of ly716-Bs-dMM under low range of initial sucrose concentrations.

(A) Model simulation at initial sucrose concentration of 100 g·L^-1. (B) Model prediction for validation the simulatability of ly716-Bs-dMM under initial sucrose concentration of 50 g·L^-1 condition.

https://doi.org/10.1371/journal.pcbi.1014273.g003

The predictive performance of ly716-Bs-dMM at low range of substrate concentration was evaluated using an independent experiment of levan production that was initialized at sucrose concentration of 50 g·L^-1 (Fig 3B). Similar phenomena were observed, with a less rapid sucrose consumption under this condition. During biomass growth and levansucrase secretion, hydrolysis exhibited a sharp increase at the sixth hour, after which transfructosylation became the dominant reaction, rapidly producing levan and glucose. Additionally, once levan reached its maximum concentration, it underwent degradation, accompanied by an increase in fructose levels. The ly716-Bs-dMM model successfully captured these key phenomena, ensuring accurate representation at low sucrose concentrations. It is worth noting that while overall metabolite conversion was well simulated, the model appeared to underestimate microbial cell growth and subsequent levan-associated products, including glucose and fructose. This difference may suggest that levansucrase activity in the model may lower than in reality under these conditions. The average normalized root mean squared error (NRMSE), which quantifies the deviation of model simulations from experimental values, was comparable between the 50 g·L^-1 (16.46%) and 100 g·L^-1 (13.24%) conditions. These results demonstrate that the ly716-Bs-dMM model with the optimal parameter values (α, = 18.19 mg_levansucrase·gDW_biomass^-1 and v_{max,trans,levansucrase} = 6.42 mmol·mg_levansucrase^-1 h^-1, µ_max = 0.2454 h^-1, Table A in S2 Text) can robustly simulate sucrose conversion into levan across a low range of initial sucrose concentrations.

The ly716-Bs-dMM was further applied to simulate levan production under high initial sucrose concentration, primarily at 250 g·L^-1. The simulation presented in S1 Text (Fig B) deviated from the experimentally measured metabolite profiles, indicting limitations of the current ly716-Bs-dMM framework in accurately representing sucrose-to-levan conversion dynamics under high initial sucrose concentrations. The deviations were particularly evident in the profiles of levan, sucrose, and especially fructose, for which the simulated patterns diverged noticeably from the experimental measurements. The model was able to mimic the initial sharp increase of fructose levels (0–6 h), but failed to reproduce the subsequent rapid decline and the later rebound in accumulation following sucrose depletion (6–36 h). The distinct fructose conversion dynamics under low and high initial sucrose concentrations indicate divergent equilibrium states governing sucrose utilization toward levan synthesis and microbial-mediated levansucrase production. This behavior may be attributed to altered transfructosylation activity and changes in microbial metabolism under the sucrose abundance gradient. To test the hypothesis, the ly716-Bs-dMM model was reparametrized to reflect sucrose-dependent transfructosylation activity in levan biosynthesis (α, = 45 mg_levansucrase·gDW_biomass^-1 and v_{max,trans,levansucrase} = 6.42 mmol·mg_levansucrase^-1 h^-1, Table A in S2 Text), along with microbial metabolic adaptation in response to elevated sucrose availability (µ_max = 0.1 h^-1, Table A in S2 Text). Fig 4A showed that the re-optimized ly716-Bs-dMM model successfully captured key dynamic behaviors under high initial sucrose concentration, including sucrose consumption, biomass production, levan production, and changes in fructose and glucose concentrations. The NRMSE of model simulation against measured profiles ranged from 6.8% to 24%, with average NRMSE in high concentration is 12.54% for 250 g·L ⁻ ¹ and 11.54% for 300 g·L ⁻ ¹. Moreover, the reliability of the ly716-Bs-dMM model was further reinforced by its validated simulation of levan production at an initial sucrose concentration of 300 g·L ⁻ ¹ (Fig 4B). These validation results demonstrated that the re-optimized model can robustly simulate levan production dynamics, covering high range of initial sucrose concentrations with NRMSE values from 5.5% to 20.7%.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Simulation of ly716-Bs-dMM under high range of initial sucrose concentrations.

(A) Model simulation at initial sucrose concentration of 250 g·L^-1. (B) Model prediction for validation the simulatability of ly716-Bs-dMM under initial sucrose concentration of 300 g·L^-1 condition.

https://doi.org/10.1371/journal.pcbi.1014273.g004

Overall, the ly716-Bs-dMM model effectively simulated sucrose-to-levan production across a broad range of initial sucrose concentrations. The optimal parameter sets tailored to initial sucrose concentrations revealed a shift in the levansucrase-mediated biochemical conversion pathway, underpinning the distinct conversion dynamics observed in levan synthesis initiated under low versus high sucrose conditions. In addition, the optimal parameters which were obtained from fitting shown the consistency with previously reported enzyme kinetics (S2 Text), indicating the levansucrase mechanism is also consistent with prior study [22]. These results underscore the critical importance of fine-tuning initial sucrose levels to maximize yield within a practical timeframe and emphasize the necessity of balancing the unique mechanistic advantages offered by both low and high sucrose availability.

Simulation of B. subtilis LY7.16-based levan production under low-to-high transitional sucrose concentrations

The optimal initial substrate concentration has been shown to determine maximal yield, substrate use efficiency, and production rate. As shown in Table 1, a shift in sucrose conversion efficiency and levan production was observed between the low (50–100 g·L ⁻ ¹) and high (250–300 g·L ⁻ ¹) initial sucrose concentration ranges. This shift resulted in a 5- to 12-fold increases in maximal levan production, albeit with approximately twice the production time due to limited levansucrase abundance from the ~ 2.5-time reduction in growth rate of B. subtilis LY7.16 under the high sucrose concentration. The appearing trade-off phenomena over the range of initial sucrose concentrations pinpointed an optimal concentration near the transition between low and high levels, potentially around 200 g·L ⁻ ¹.

At an initial sucrose concentration of 200 g·L ⁻ ¹, the time-course profiles of all biochemical conversions exhibited intermediate behavior relative to those observed across the substrate concentration gradient (Fig A in S1 Text), consistent with its corresponding yield performance (Table 1). To assess whether the conversion process at 200 g·L ⁻ ¹ sucrose reflects a mechanistic blend of low- and high-range dynamics, the ly716-Bs-dMM model was employed to simulate levan production using parameter sets specific to each condition. The simulated result in Fig 5 showed that the measured metabolite conversion at 200 g·L ⁻ ¹ (Fig 5 (dot)) was only partially mimicked by either low-range (Fig 5 (dash line)) or high-range (Fig 5 (solid line)) modeling scenarios, and exhibited slightly closer resemblance to the high-range condition. Re-parameterization of the ly716-Bs-dMM model to improve its representation of sucrose-to-levan conversion at an initial sucrose concentration of 200 g·L ⁻ ¹ yielded optimized parameters indicative of a transitional state between low and high sucrose regimes (Table A in S2 Text). This is reflected in a combined profile of key conversion parameters: a biomass yield on substrate (Y_X/S)) of approximately 0.015 gDW_biomass·mmol sucrose ⁻ ¹, positioned between optimal values for low and high concentrations; an α (45 mg_levansucrase·gDW_biomass ⁻ ¹) value resembling that of the high sucrose condition; and a μ_Monod (0.2454 h^-1) value closely aligned with the low concentration regime. These parameter values suggest that at this transitional concentration, B. subtilis LY7.16 may begin shifting its metabolic priorities from biomass production toward levansucrase synthesis to supply the higher abundance of sucrose. While entering to the high range of initial sucrose conditions, demand for levansucrase is greater upon the initial sucrose availability. It led to a rebalancing metabolic equilibrium of B. subtilis LY7.16 to reduce biomass growth for levansucrase production. This hypothetical scenario suggests that the microbial metabolic potential and optimization upon the initial substrate concentration is a critical factor for leaping forward the overall performance of levan production.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Simulation of ly716-Bs-dMM at initial sucrose concentration of 200 g·L^-1 based upon the condition-specific parameters, dash line – low initial sucrose concentration and solid line – high initial sucrose concentration.

The results demonstrate the marginal range of low to high initial sucrose substrate concentration in effect to sucrose-to-levan conversion equilibrium.

https://doi.org/10.1371/journal.pcbi.1014273.g005

Prediction of metabolic alteration in B. subtilis LY7.16 upon initial sucrose concentration

Simulations of the ly716-Bs-dMM model demonstrated that initial sucrose concentrations influenced levan synthesis not only through substrate availability, but potentially also via modulating microbial levansucrase production. Optimized values of key parameters—μ_Monod, Y_X/S, and α—across varying initial sucrose concentrations reveal metabolic adaptations in B. subtilis LY7.16 that promote levansucrase synthesis under elevated sucrose availability. To investigate potential trade-offs between microbial cell biomass accumulation and levansucrase production across sucrose gradients, carbon flux distributions in B. subtilis LY7.16 were analyzed to elucidate metabolic responses to changing substrate levels.

The metabolic model of B. subtilis LY7.16 was simulated under initial sucrose concentrations of 100 g·L ⁻ ¹ (low-range), 200 g·L ⁻ ¹ (transitional), and 250 g·L ⁻ ¹ (high-range). In each scenario, sucrose utilization was modulated based on the optimized biomass yield on substrate (Y_X/S), ensuring alignment with the corresponding extracellular environment. Fig 6 presents the predicted metabolic flux distributions, illustrating metabolite conversion in B. subtilis LY7.16 across gradients of the initial sucrose concentrations. Fluxes are expressed as flux-sums (see Methods), facilitating comparative analysis across the different simulation conditions. The results indicate that microbial metabolism in B. subtilis LY7.16 prioritizes energy expenditure for levansucrase synthesis, while allocating fewer resources toward cellular biomass biosynthesis under higher sucrose concentrations. Simulations revealed a decline in extracellular oxygen levels (Fig 6, Number 1) with increasing initial sucrose concentrations. As oxygen plays a critical role in supporting aerobic bacterial growth, this reduction suggests a deceleration of metabolic activity relevant to biomass production under high-sucrose conditions. The simulated results correspond well with the observed decrease in maximum specific growth rate (μ_max) under elevated sucrose levels.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Simulation of metabolic flux conversion in B. subtilis LY7.16 under low (100 g·L^-1), low-to-high transitional (200 g·L^-1), and high (250 g·L^-1) sucrose concentrations.

The number values represent normalized fluxes expressed as flux-sums, enabling comparative analysis of metabolite fluxes across varying initial sucrose concentrations: low (grey), transitional low-to-high (blue), and high (black).

https://doi.org/10.1371/journal.pcbi.1014273.g006

Under high-sucrose conditions, B. subtilis LY7.16 exhibited a metabolic shift, diverting carbon flux away from biomass formation and toward levansucrase production to support enhanced levan biosynthesis (Fig 6, Number 2). Nevertheless, the simulation result of levansucrase secretion seemed higher in transition condition. This phenomenon may happen due to the rapid growth of biomass in transition condition compared to high sucrose concentration in the snapshot of FBA simulation (time = 9 h). Therefore, at that time, the secretion is higher on transition condition, but it does not represent the overall production of levan production, since the levan synthesis takes longer time in higher sucrose concentration. The altered succinate conversion from more energy intensive via succinyl CoA synthase (SUCOAS, red arrow, Number 3) under low sucrose concentration condition to energy-free reaction via 3-oxoacid CoA-transferase (OCOAT1, green arrow, Number 4) under high sucrose concentration condition supports the metabolic redirection corresponding to the sucrose-induced sacB gene expression.

In silico analyses for enhancing levan production by B. subtilis LY7.16

Analysis of levan conversion across varying initial sucrose concentrations suggests that the optimal condition lies within the transitional range, specifically around 200 g·L ⁻ ¹. Simulations using the ly716-Bs-dMM model further revealed that levan production in B. subtilis LY7.16 can be enhanced under this condition by mitigating metabolic trade-offs, thereby promoting more efficient substrate utilization and maximizing product yield at a sustainable production rate. In this study, in silico analyses involving amino acid pathway interventions and systems-level metabolic engineering were conducted to provide predictive hypotheses that may help suggest strategic approaches for improving levan biosynthesis.

The first scenario was aimed to maximize levansucrase synthesis to evaluate the effect of amino acid supplementation on its production. Threonine was selected as a representative amino acid due to its distinct and differential production under low and transitional-to-high sucrose concentrations (Fig 6, Number 5). Through maximizing levansucrase synthesis, simulation in Fig 7A demonstrates a 1.342-fold increase in both levansucrase and biomass levels under this modification. These results suggest that targeted metabolic interventions, exemplified by threonine supplementation, may offer a promising strategy to redirect carbon flux toward a near-optimal state for microbial levansucrase production and levan synthesis.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. The levan enhancement scenario by sacB perturbation.

(A) Simulation of levansucrase and cell biomass synthesis under threonine supplementation scenarios. (B) Simulation of the extracellular metabolite profiles, and (C) the corresponding conversion rate under altered levansucrase protein yield (α = 22.5 mg_levansucrase gDW_biomass^-1 – Pink, α = 45 mg_levansucrase gDW_biomass^-1 – Black, α = 184.1 mg_levansucrase gDW_biomass^-1 – Blue.

https://doi.org/10.1371/journal.pcbi.1014273.g007

An alternative strategy was explored based on the targeted metabolic engineering rationales. In silico perturbation of sacB was investigated through modulating levansucrase protein yield (α values) within the ly716-Bs-dMM framework and assess its impact on metabolic dynamics. Fig 7B shows simulations of levan production under reduced (α = 22.5 mg_levansucrase·gDW_biomass ⁻ ¹), optimized (α = 45 mg_levansucrase·gDW_biomass ⁻ ¹), and enhanced (α = 184.1 mg_levansucrase·gDW_biomass ⁻ ¹) α values. A reduced α value led to increased biomass concentration but resulted in diminished levansucrase synthesis and lower levan output. Conversely, the enhanced α value stimulated levansucrase expression and levan production while suppressing biomass accumulation. Given that levan biosynthesis is constrained by substrate availability, i.e., sucrose and fructose, these findings suggest that fed-batch cultivation may help maintain substrate supply, thereby enabling sustained levan production.

The simulation results presented in Fig 7B, along with the proposed strategy, are supported by the observed changes in sucrose consumption flux toward biomass production (v₁), which decreases with increasing α values. This trend suggests a metabolic reallocation of carbon towards levan synthesis in B. subtilis at elevated α levels (Fig 7C). As a result, biomass formation is diminished under maximal α conditions, while levansucrase secretion and levan production are markedly enhanced, as shown in Fig 7B. Furthermore, the declining hydrolysis rate (v₃) is associated with a reduced hydrolysis-to-transfructosylation (H/T) ratio, indicating a metabolic shift favoring transfructosylation activity in response to elevated levansucrase levels. Collectively, these results suggest that sacB overexpression, when combined with sustained substrate availability, may serve as an effective strategy to enhance and stabilize levan production.

Furthermore, the ly716-Bs-dMM framework demonstrated additional potential, as illustrated in Fig 8. Overexpression of sacB resulted in widespread enhancement of metabolic fluxes—indicated by the predominance of green pathways—while maintaining a growth rate comparable to the fitted value. However, Fig 8 highlights a flux reduction at the branching point between the oxidative pentose phosphate pathway (OPP) and glycolysis, implicating glucose-6-phosphate dehydrogenase (G6PDH) and phosphoglucoisomerase (PGI) as potential metabolic targets for further levansucrase enhancement. A simulation was performed in which PGI flux was downregulated to 0.006 mmol gDW_biomass^-1 h^-1, successfully redirecting flux toward G6PDH and increasing NADPH generation, an energy exchange molecule. These findings suggested that a combined strategy involving sacB and G6PDH overexpression, PGI downregulation, and fed-batch cultivation may significantly enhance levan production.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 8. The proposed levan enhancement scenario using ly716-Bs-dMM framework by downregulating phosphoglucoisomerase (PGI) expression.

Blue number indicating the original flux prior to PGI downregulation, while green number indicates an increasing shift towards PGI downregulation, and red number indicates the decreasing flux towards PGI knockdown.

https://doi.org/10.1371/journal.pcbi.1014273.g008

Materials

Fructose, glucose, sucrose, yeast extract, beef extract, and peptone were purchased from HiMedia Chemicals India Pvt. Ltd. (Mumbai, India), and all other reagents were obtained from Sigma-Aldrich (St. Louis, MO, USA). All chemicals were of analytical grade. Unless otherwise stated, all other reagents were obtained from Sigma-Aldrich (St. Louis, MO, USA).

Microorganisms, media and culture conditions

Bacillus subtilis strain LY7.16 was isolated from Thai traditional fermented soybean and identified as previously described [18]. The 16S rRNA gene sequence was deposited in the GenBank database under accession number PV436786.1. The strain was deposited at the Thailand Bioresource Research Center (TBRC), Thailand, under accession number TBRC11053.

B. subtilis LY7.16 was maintained on nutrient agar (NA) containing peptone (5 g·L^-1), NaCl (5 g·L^-1), yeast extract (1.5 g·L^-1), beef extract (1.5 g·L^-1), and agar (20 g·L^-1), and incubated aerobically at 37°C. For inoculum preparation, a loopful of culture was transferred into sterile nutrient broth (NB) with the same composition as NA without agar and incubated at 37°C with agitation at 200 rpm for 24 h. The culture was harvested when the optical density at 600 nm (OD₆₀₀) reached approximately 0.6, corresponding to a cell concentration of ~10⁸ colony-forming units (CFU)·mL^-1, and used as the inoculum for subsequent fermentation experiments.

Production of levan from sucrose by B. subtilis LY7.16

A 2% (v/v) inoculum of B. subtilis LY7.16 was transferred into nutrient broth (NB) medium supplemented with sucrose at concentrations of 50, 100, 200, 250, and 300 g·L^-1. The medium was formulated to provide cells with sufficient nutrients from the complex culture environment, while ensuring that sucrose availability exerted a stronger influence on their growth and metabolism [2,5,13,14]. The cultures were incubated at 37°C with agitation at 200 rpm for 120 h. Samples were collected at 6, 12, 18, 24, 30, 36, 48, 60, 72, 96, and 120 h and centrifuged at 10,000 × g for 10 min to separate the cells from the culture supernatant. The concentrations of sucrose, glucose, and fructose in the culture supernatants were determined using high-performance liquid chromatography (HPLC) equipped with a refractive index detector (RID-10A; Shimadzu, Japan). Separation was performed on an Aminex HPX-42C column (Bio-Rad, Hercules, CA, USA) using HPLC-grade water (Fisher Scientific, UK) as the mobile phase at 85 °C and a flow rate of 0.6 mL·min^-1.

For levan recovery, the cell-free supernatant was mixed with chilled 70% (v/v) ethanol at a ratio of 3:1 (ethanol to supernatant) and incubated at 4°C overnight. The precipitated levan was collected by centrifugation at 10,000 × g for 10 min. The resulting levan precipitate and the corresponding fresh cell pellets were dried to constant weight in an oven at 50°C to determine levan yield and dry cell biomass, respectively.

Construction of ly716-Bs-dMM model

A dynamic metabolic model of B. subtilis LY7.16 (ly716-Bs-dMM) was developed by integrating an ordinary differential equation (ODE)-based kinetic model (denoted as BsODE) with a constraint-based genome-scale metabolic model (denoted as BsCBM). This hybrid approach was employed to simulate microbial growth and dynamic changes of metabolites over time under different initial sucrose concentrations. The model combined kinetics of extracellular biochemical reactions with genome-scale metabolic flux distributions inside microbial cells to investigate conversion of sucrose into levan under different initial substrate concentrations. In the model, sucrose consumption was assumed to follow Michaelis–Menten, and growth of microbial cell was assumed to exhibit Monod kinetics [34,36,43].

The BsODE-based kinetic model was developed to describe the time-dependent dynamics of biomass growth over time, which is represented as , sucrose consumption, which is represented as v₁, and metabolite production in batch culture. It was constructed based on prior knowledge of levansucrase synthesis and relevant literature. Experimental data demonstrated that initial sucrose concentration markedly affected levan yield and the synthesis of glucose and fructose (Fig 1). These trends were related to levansucrase enzymatic activities — hydrolysis, transfructosylation, and levan degradation — represented as v₃-v₅ in Fig 2A. The hydrolysis/transfructosylation (H/T) ratio has been reported to inversely correlate with sucrose concentration [22,26].

Accordingly, BsODE was designed to simulate biomass growth, sucrose uptake, and product formation (levan, fructose, and glucose) based on the dynamic mass balance (Fig 2B). Levansucrase, which is an extracellular enzyme, catalyzed levan synthesis; thus, levan formation was assumed to occur extracellularly in the presence of sucrose and levansucrase, consistent with prior findings [2,6,44]. The system dynamics are described by the ODE model in Equations (1–7), with the related kinetic functions detailed in Equations (8–11).

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

µis specific growth rate (h^-1), µ_max is maximum specific growth rate (h^-1), and k_d is death constant (h^-1). The levansucrase-associated reaction, v₃-v₅ also includes the enzymatic parameters v_{max,hyd,levansucrase} (mmol·mg_levansucrase^-1·h^-1), representing maximum hydrolysis rate; v_{max,trans,levansucrase} (mmol·mg_levansucrase^-1·h^-1) representing maximum transfructosylation rate; and v_{max,levan degradation} (mmol·mg_levansucrase^-1·h^-1), representing maximum levan degradation rate. Each reaction is further characterized by its Michaelis–Menten constant (K_{M, hyd, levansucrase}, K_{M, trans, levansucrase}, K_{M, levan degradation}) expressed in mM.

Auxiliary parameters were incorporated to describe the metabolic shifts under different sucrose levels. These included the biomass yield on sucrose (Y_x/s) and α, representing the levansucrase production yield (mg) on basis of a gram dry weight of biomass (gDW_biomass). A Heaviside function and reparameterization of kinetic parameters were employed to introduce sucrose-dependent activity of levansucrase across low-to-high sucrose concentrations regime [45], particularly simulating degradation of levan in an absence of sucrose. The onset of levan degradation (t_{levan degradation}) was used as a parameter constraining levan synthesis activity. The parameter α was embedded in the biomass growth function to estimate levansucrase secretion, as levansucrase was considered a growth-associated metabolite [ 46].

The BsCBM model was reconstructed from the genome-scale metabolic model iYO844 of B. subtilis 168, as shown evolutionary closed to B. subtilis LY7.16 based on high 16S rRNA gene sequence similarity (S3 Text). Levansucrase-related reactions absent in iYO844 were incorporated to reflect B. subtilis LY7.16’s capacity for levan polymerization [8,47,48]. The BsCBM model was further refined to ensure mass balance and network completeness, with manual gap-filling based on KEGG and MetaCyc and literature evidence [47,49]. The model curation and analysis were performed using COBRA Toolbox v3.0 [50] implemented in MATLAB (R2023b, The MathWorks). The BsCBM framework was calibrated to simulate the growth of Bacillus subtilis LY7.16 under varying sucrose concentrations in levan production experiments, ensuring that the model captured the metabolic behavior of this strain. This strategy enabled simulation of metabolic conversion in metabolism of B. subtilis LY7.16 despite the absence of genome sequence data, though potential strain-specific responses may not be fully represented. Each simulation run using the ly716-Bs-dMM framework took approximately 10 minutes on a laptop computer equipped with an AMD Ryzen 4700U processor and 8 GB of memory.

The two submodels, BsODE and BsCBM, were coupled via the constraining linker parameters from BsODE (μMonod, Y_X/S, and α) to BsCBM, resulting in the integrated ly716-Bs-dMM model. The details of the submodel integration are illustrated in Fig 2B. The integrated framework, ly716-Bs-dMM, along with the condition-specific data and the script for running ly716-Bs-dMM, is available in the GitHub repository KMUTT-CASB/ ly716-Bs-dMM (https://github.com/KMUTT-CASB/ly716-Bs-dMM).

Model simulation

To investigate the dynamic behavior of B. subtilis LY7.16, simulations were performed at initial sucrose concentrations of 100, 200, and 250 g·L^-1, representing low, transition, and high levels, respectively. Parameter estimation within BsODE was achieved by minimizing the sum of squared errors between model predictions and experimental data using the Levenberg–Marquardt algorithm [51]. All kinetic parameters were independently estimated for the low and high sucrose conditions, except for levansucrase activity parameters, which were held constant. The transition condition (200 g·L^-1) was simulated using interpolated parameters between the two conditions to assess intermediate systems dynamics. Estimated parameters and their fitted values are summarized in the S2 Text.

Intracellular flux distributions were simulated using flux balance analysis (FBA), a constraint-based optimization technique estimating steady-state metabolic fluxes from stoichiometric and physiological constraints. In the ly716-Bs-dMM model, BsCBM was constrained using auxiliary parameters derived from BsODE, including μMonod (h^-1), sucrose uptake rate (mmol gDW_biomass^-1 · h^-1), and specific levansucrase yield (α, mg_levansucrase⋅gDW_biomass^-1). Model was optimized towards minimal total flux [ 52,53] to examine metabolic flux distributions across sucrose variations, ensuring the representation of suboptimal growth states where nutrient utilization was partitioned between biomass formation and levansucrase secretion [30,32,33]. Simulation focused on mid-exponential growth phase (9 h). The complete mathematical formulation is given in Equations 12–16.

(12)

(13)

(14)

(15)

(16)

The stoichiometric matrix S_irrev,i,j contains the stoichiometric coefficient of metabolite i in reaction j. Subscription “m” represents total metabolite, while “i” represents a specific metabolite. The variable flux_irrev,j, represents intracellular fluxes within the metabolic model, ranging from 0 to a pre-defined upper bound (flux_j,max). The subscript “irrev” indicates that all reactions were converted to irreversible form to minimize the total absolute flux [54].

Model validation

The model was validated against measured data from independent experiments. The structure and associated parameters of the ly716-Bs-dMM model were evaluated using experiments with initial sucrose concentrations of 50 g·L⁻¹ and 300 g·L⁻¹, representing the low and high modeling conditions, respectively. Normalized root mean squared error (NRMSE) was used to quantify the deviation between predicted and observed metabolite profiles, thereby assessing the predictive reliability of the model.

Model analysis

Flux distribution analyses were performed to assess metabolic adaptations under different sucrose levels. Flux-sum analysis was used to normalize each metabolite’s flux by the sucrose uptake rate [55]. Reactions with flux values less than 1 × 10⁻⁶ mmol gDW_biomass⁻¹ h⁻¹ were considered inactive [56]. Fluxes were categorized as (i) active across the whole range of sucrose concentrations, (ii) active only at low range of sucrose concentrations, (iii) active at transition and high range of sucrose concentrations, or (iv) inactive under all conditions. Differences in active fluxes between conditions were analyzed to understand mechanisms underlying variable levan synthesis performance.

In silico analysis of levan production enhancement

The ly716-Bs-dMM framework was used to explore strategies for improving levan yield at a transitional sucrose concentration (200 g·L^-1). Three simulation scenarios were conducted: (i) maximizing levansucrase synthesis by modified the objective function to assess amino acid concentrations to levansucrase production; (ii) examining the effect of sacB downregulation and overexpression on extracellular metabolite patterns by varying α to; and (iii) investigating the NADPH-rich metabolic pathways to evaluate intracellular cofactor balance impacted on levan biosynthesis. These analyses provided a comprehensive system-level scenario of environmental and intracellular factors influencing levan production.