Engineered saccharomyces cerevisiae strains for production of 2,3-butanediol and methods of making 2,3-butanediol

Abstract

The application provides an engineered Saccharomyces cerevisiae strain for producing 2,3-butanediol and a method for preparing 2,3-butanediol. The engineered Saccharomyces cerevisiae strain contains a dual light control system for regulating an ethanol synthesis pathway and a 2,3-butanediol synthesis pathway, respectively, and heterologously expresses an NADH oxidase. The method for preparing 2,3-butanediol comprises fermentation using the engineered Saccharomyces cerevisiae strain. In the fermentation process, including a growth phase and a product synthesis phase, dark culture is performed in the growth phase, and light stimulation is performed using dynamic light in the product synthesis phase. A light parameter sequence of the dynamic light is generated by conditional optimization of a differential evolution algorithm. In the conditional optimization process of the differential evolution algorithm, a trained convolutional neural network model is used as a fitness function.

Description

Technical Field

[0001] This invention relates to the field of biotechnology, specifically to an engineered brewing yeast for the production of 2,3-butanediol. Saccharomyces cerevisiae ) strain and method for preparing 2,3-butanediol. Background Technology

[0002] 2,3-Butanediol (2,3-BDO) is an important platform compound widely used in fuels, chemicals, food, and pharmaceuticals. Saccharomyces cerevisiae, as an important industrial microbial chassis, possesses advantages such as robustness, acid resistance, and tolerance to high osmotic pressure, making it an ideal host for 2,3-butanediol production. However, Saccharomyces cerevisiae naturally tends to produce ethanol through fermentation, and its strong ethanol fermentation capacity makes it difficult to effectively transfer carbon metabolic flux to the exogenously introduced 2,3-butanediol synthesis pathway. Traditional gene knockout techniques to completely block the ethanol pathway lead to severe growth defects in Saccharomyces cerevisiae, preventing the accumulation of sufficient biomass.

[0003] In recent years, optogenetic-based dynamic metabolic regulation technology has successfully achieved spatiotemporal separation of the growth and production stages, solving the problem of inhibited cell growth. However, existing light control technologies still face three main problems in practical applications: First, redox imbalance. In the process of forcibly redirecting carbon metabolic flux using light control, mismatched coenzyme consumption can easily lead to intracellular redox imbalance, resulting in the accumulation of byproducts such as glycerol and severe carbon flux loss. Second, a lack of analytical and transferable process models of the nonlinear interaction mechanism between light and metabolism. Existing dynamic regulation relies heavily on manual experience, making it difficult to achieve precise matching of complex temporal light parameters. Third, uneven light distribution during fermentation scale-up. In high-density fermentation scale-up in large-volume reactors, high-density cell populations can produce a severe "self-shading effect," leading to a decrease in light transmittance within the system. Currently, there is a lack of effective engineering solutions.

[0004] Therefore, it is urgent to propose a dynamic metabolic regulation framework that integrates optogenetics and machine learning to achieve global optimization of illumination strategies. At the same time, it is necessary to establish a light-controlled amplification strategy based on consistent light intensity per unit cell to ensure stable transmission of light signals and effectiveness of metabolic regulation in photobioreactors of 5L and above. Summary of the Invention

[0005] This invention provides an engineered brewing yeast for the production of 2,3-butanediol. Saccharomyces cerevisiaeThis study aims to address the problems of intracellular redox imbalance and byproduct accumulation caused by mismatched coenzyme consumption during the forced redirection of carbon metabolic flow in Saccharomyces cerevisiae using optogenetics. It also aims to solve the existing limitations in effectively analyzing and generating optimal time-series light control strategies based on the nonlinear relationship between complex dynamic lighting (such as pulsed and gradient lighting) and metabolism, and to address the "self-shading effect" that occurs during high-density fermentation in the culture scale-up process, which leads to a decrease in light transmittance within the system.

[0006] To achieve the above objectives, one embodiment of the present invention provides an engineered brewing yeast for the production of 2,3-butanediol. Saccharomyces cerevisiae The engineered Saccharomyces cerevisiae strain is characterized by containing a dual photocontrol system that regulates the ethanol synthesis pathway and the 2,3-butanediol synthesis pathway respectively, and heterologously expressing NADH oxidase.

[0007] In one embodiment of the present invention, the dual-light control system includes a photosensitive degradation module and a photoinduced stabilization module. The photosensitive degradation module degrades pyruvate decarboxylase under light irradiation, and the photoinduced stabilization module stabilizes the protein of 2,3-butanediol dehydrogenase in the 2,3-butanediol synthesis pathway under light irradiation.

[0008] In one embodiment of the present invention, the engineered Saccharomyces cerevisiae strain contains an endogenous pyruvate decarboxylase isoenzyme gene. pdc5 and pdc6 Knockout endogenous pyruvate decarboxylase isoenzyme gene pdc1 The photosensitive degradation module is integrated; the 2,3-butanediol dehydrogenase gene is present in the engineered Saccharomyces cerevisiae strain. bdh1 The light-induced stabilization module is integrated.

[0009] In one embodiment of the present invention, the NADH oxidase is derived from Lactobacillus casei, Lactobacillus paracasei, Bacillus subtilis, Lactococcus lactis, or Lactobacillus reuteri.

[0010] In one embodiment of the present invention, the NADH oxidase is LPnox2 derived from Lactobacillus casei or Lactobacillus paracasei.

[0011] In one embodiment of the present invention, the engineered Saccharomyces cerevisiae strain also heterologously expresses the α-acetolactate synthase gene derived from Bacillus subtilis. alsS and α-acetolactate decarboxylase gene alsD .

[0012] To achieve the above objectives, another method for preparing 2,3-butanediol is provided by the present invention, the method comprising the following steps: Fermentation was carried out using the engineered Saccharomyces cerevisiae strain described above for the production of 2,3-butanediol; The fermentation process includes a growth stage and a product synthesis stage. The growth stage is carried out in darkness, and the product synthesis stage is stimulated by dynamic lighting. The light parameter sequence of the dynamic lighting is generated by condition optimization through differential evolution algorithm. During the conditional optimization process of the differential evolution algorithm, a trained convolutional neural network model is used to predict the yield of 2,3-butanediol corresponding to the illumination parameter sequence.

[0013] In one embodiment of the present invention, the trained convolutional neural network model uses the light intensity and duration sequences at different stages as input features and the predicted yield of 2,3-butanediol as output features, and maximizes the predicted yield during the conditional optimization process of the differential evolution algorithm.

[0014] In one embodiment of the present invention, the training data of the trained convolutional neural network model is obtained through the following steps: during the fermentation process, different photo-induced initiation times, periodic pulsed light with different duty cycles, and gradient pulsed light are tested respectively to collect the light parameter sequence and the yield data of 2,3-butanediol to construct a training set.

[0015] In one embodiment of the present invention, the method further includes the following steps: Yeast extract and citric acid were added to the fermentation reactor for fed-batch fermentation; and The light intensity is dynamically adjusted in real time based on the cell density in the fermentation broth to maintain a constant amount of light energy received by each strain.

[0016] Compared with the prior art, the present invention has achieved the following beneficial effects: By heterologously expressing NADH oxidase, a non-productive NADH consumption pathway was constructed. During the metabolic transition from the light-controlled inhibition of the ethanol pathway (high NADH consumption) to the 2,3-butanediol pathway (low NADH consumption), the excess NADH accumulated intracellularly was effectively consumed, and NAD+ was regenerated. This alleviated the intracellular redox imbalance, maintained the efficient operation of glycolysis, and significantly blocked the glycerol synthesis pathway (glycerol is usually generated as a "pressure relief valve" for NADH). The carbon source that originally flowed to glycerol was pulled back to the 2,3-butanediol synthesis pathway. Ultimately, the yield and production of 2,3-butanediol were further improved.

[0017] The method for preparing 2,3-butanediol provided by this invention combines photo-induction timing, intermittent lighting, and multi-stage pulsed lighting strategies for empirical optimization of light intensity. This achieves dynamic control of the growth and fermentation production of 2,3-butanediol by *Saccharomyces cerevisiae*, resulting in a 2.52-fold increase in 2,3-butanediol yield. Furthermore, a prediction model for "light-controlled time series – yield" is established based on a convolutional neural network (validation set R² = 0.9827, MSE = 0.0053), calculated as follows:

[0018]

[0019] Meanwhile, by combining differential evolution algorithm to optimize the time-series illumination strategy, the yield was further increased by 15.60% based on the empirically optimal solution. Finally, based on consistent light intensity per strain, this light control strategy was applied to a 5L photobioreactor, achieving a yield of 3.65 g / L, which can be further increased to 23.95 g / L through a feeding strategy. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. It should be understood that the drawings described below are only some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0021] Figure 1 and Figure 2 A schematic diagram of the metabolic network and dual-light-controlled switch working mechanism of the engineered Saccharomyces cerevisiae strain provided in an embodiment of the present invention; Figure 3 This is a comparative graph showing the effect of heterologous expression of NADH oxidases from different sources on the yield of 2,3-butanediol in embodiments of the present invention. Figure 4 This is a comparison chart of 2,3-butanediol yield under different gradient light pulse strategies in the embodiments of the present invention; Figure 5 This is a flowchart of the dynamic optimization process for lighting strategies based on convolutional neural networks and differential evolution algorithms, provided in an embodiment of the present invention. Figure 6 This is a schematic diagram of dynamic illumination control for a 5L photobioreactor provided in an embodiment of the present invention; Figure 7 The kinetic curve of high-density fed-batch fermentation in a 5L photobioreactor provided in this embodiment of the invention; Figure 8 The graph showing the effect of different concentrations of yeast extract on the growth of the strain under dark culture conditions is provided for embodiments of the present invention. Figure 9 The graph shows the effect of different concentrations of yeast extract on the growth of the strain under blue light induction conditions, as provided in the embodiments of the present invention. Figure 10 The graph showing the effect of different concentrations of citric acid on the growth of the strain under dark culture conditions is provided for embodiments of the present invention. Figure 11 The graph shows the effect of different concentrations of citric acid on the growth of the strain under blue light induction conditions, as provided in the embodiments of the present invention. Detailed Implementation

[0022] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. It should be understood that the specific embodiments described herein are only for illustration and explanation of the present invention and are not intended to limit the present invention.

[0023] Any specific numerical value (including the endpoints of the numerical range) disclosed in this invention is not limited to the exact value, but should be understood to also cover values ​​close to the exact value, such as all possible values ​​within ±5% of the exact value. Furthermore, for the disclosed numerical range, one or more new numerical ranges can be obtained by arbitrarily combining the endpoint values ​​of the range, the endpoint values ​​with specific point values ​​within the range, and the specific point values. These new numerical ranges should also be considered as specifically disclosed in this invention.

[0024] The terminology used in this invention is for the purpose of describing specific exemplary embodiments only and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “the” as used herein are intended to include the plural forms as well. The terms “comprising,” “including,” “containing,” and “having” are inclusive and thus describe the presence of said features, elements, compositions, steps, integers, operations, and / or components, but do not exclude the presence or inclusion of one or more other features, integers, steps, operations, elements, components, and / or sets thereof. Although the open-ended term “comprising” should be understood as a non-limiting term used to describe and claim the various embodiments described in this invention, in some aspects it may instead be understood as a more restrictive and limiting term, such as “consisting of” or “essentially composed of.” Thus, for any given embodiment describing compositions, materials, components, elements, features, integers, operations, and / or method steps, the invention also particularly includes embodiments consisting of or substantially consisting of such compositions, materials, components, elements, features, integers, operations, and / or method steps. In the case of “consisting of…”, alternative embodiments exclude any additional compositions, materials, components, elements, features, integers, operating and / or method steps. In the case of “essentially composed of…”, any additional compositions, materials, components, elements, features, integers, operating and / or method steps that substantially affect the essential and novel characteristics are excluded from such embodiments. However, any compositions, materials, components, elements, features, integers, operating and / or method steps that do not substantially affect the essential and novel characteristics may be included in the embodiments.

[0025] Any method steps, procedures, and operations described in this invention should not be construed as necessarily requiring them to be performed in the specific order discussed or shown, unless explicitly specified. It should also be understood that, unless otherwise stated, additional or alternative steps may be used.

[0026] In this invention, except where expressly stated, any matters or issues not mentioned are directly applicable to those known in the art without any modification. Furthermore, any embodiment described in this invention can be freely combined with one or more other embodiments described in this invention, and the resulting technical solutions or concepts are considered part of the original disclosure or original record of this invention, and should not be regarded as new content not disclosed or anticipated by this invention, unless those skilled in the art consider the combination to be clearly unreasonable.

[0027] Unless otherwise stated, the terminology used in this invention has the same meaning as commonly understood by those skilled in the art. If a term is defined in this invention and its definition differs from the common understanding in the art, the definition of this invention shall prevail.

[0028] It is worth noting that the Latin scientific name of *Lactobacillus casei* mentioned in this invention specification is... Lactobacillus paracasei The Latin scientific name of Lactobacillus paracasei is Lactobacillus paracasei or Lacticaseibacillus paracasei, The Latin scientific name of Bacillus subtilis is Bacillus subtilis The Latin scientific name of Lactococcus lactis is Lactococcus lactis The Latin scientific name of Lactobacillus reuteri is Limosilactobacillus reuteri .

[0029] This invention provides an engineered *Saccharomyces cerevisiae* strain that produces 2,3-butanediol (2,3-BDO). This engineered *Saccharomyces cerevisiae* strain contains a dual photocontrol system that separately regulates the ethanol and 2,3-butanediol synthesis pathways, and heterologously expresses NADH oxidase. By constructing a dual photocontrol system and heterologously expressing NADH oxidase, the excessive intracellular NADH consumption and NAD+ regeneration are achieved when the ethanol pathway is inhibited and the 2,3-butanediol synthesis pathway is enhanced under photocontrolled conditions. This alleviates the problem of excessive accumulation of glycerol byproducts and carbon flux loss in *Saccharomyces cerevisiae* due to redox imbalance during carbon metabolic flux redirection.

[0030] In some embodiments of the present invention, the NADH oxidase is derived from *Lactobacillus casei*, *Lactobacillus paracasei*, *Bacillus subtilis*, *Lactococcus lactis*, or *Lactobacillus reuteri*. By introducing NADH oxidase derived from specific lactic acid bacteria or Bacillus, stable expression and effective catalysis of this exogenous coenzyme regulatory element in the chassis of *Saccharomyces cerevisiae* are achieved, thereby further solving the problems of poor adaptability and uncertain catalytic efficiency of exogenous redox regulatory elements in heterologous hosts.

[0031] In some embodiments of the present invention, the NADH oxidase is LPnox2 derived from *Lactobacillus casei* or *Lactobacillus paracasei*. By specifically introducing LPnox2 derived from *Lactobacillus casei* or *Lactobacillus paracasei* as the NADH oxidase, more efficient regulation of coenzyme oxidation rate and metabolic balance is achieved, thereby further solving the problem of difficulty in timely and effective response to rapidly accumulating NADH in dynamic light-controlled fermentation to minimize byproduct generation.

[0032] In some embodiments of the present invention, the dual-photocontrol system includes a photosensitive degradation module and a photoinduced stabilization module, which are used to regulate the protein stability of pyruvate decarboxylase in the ethanol synthesis pathway and 2,3-butanediol dehydrogenase in the 2,3-butanediol synthesis pathway, respectively. By deploying the photosensitive degradation module and the photoinduced stabilization module separately, rapid reverse regulation of key enzymes in ethanol and 2,3-butanediol synthesis at the post-translational modification level is achieved, thereby further solving the problems of response lag in traditional gene transcriptional regulation and strain growth defects caused by static gene knockout.

[0033] In some embodiments of the present invention, the engineered Saccharomyces cerevisiae strain contains an endogenous pyruvate decarboxylase isoenzyme gene. pdc5 and pdc6 Knockout endogenous pyruvate decarboxylase isoenzyme gene pdc1 The photosensitive degradation module (PSD) is retained and fused to the engineered Saccharomyces cerevisiae strain; the 2,3-butanediol dehydrogenase gene bdh1 is fused with the photoinduced stabilization module (SULI). By knocking out redundant isoenzyme genes and fusing the PSD and SULI modules to the retained core enzyme gene, rapid degradation of pyruvate decarboxylase and simultaneous stable accumulation of 2,3-butanediol dehydrogenase under light irradiation are achieved, thereby further solving the problems of incomplete metabolic flux cutoff in the context of multiple isoenzymes and the inability of single regulation to achieve efficient switching of metabolic nodes.

[0034] In some embodiments of the present invention, the engineered Saccharomyces cerevisiae strain also heterologously expresses the α-acetolactate synthase gene derived from Bacillus subtilis. alsS and α-acetolactate decarboxylase gene alsD By introducing exogenous high-efficiency catalytic elements alsS and alsD The metabolic flux of the upstream node (pyruvate to acetoin) in the 2,3-butanediol synthesis pathway was enhanced, thereby further solving the problem of insufficient 2,3-butanediol precursor supply caused by the lack of efficient precursor conversion pathways in the natural chassis of Saccharomyces cerevisiae.

[0035] This invention provides a method for preparing 2,3-butanediol, comprising a fermentation step using a *Saccharomyces cerevisiae* strain, wherein the *Saccharomyces cerevisiae* strain is the engineered *Saccharomyces cerevisiae* strain for producing 2,3-butanediol described in the first aspect. The fermentation process includes a growth stage and a product synthesis stage. During the growth stage, the organism is cultured in darkness. During the product synthesis stage, a dynamic illumination strategy is employed for blue light stimulation. The illumination sequence of the dynamic illumination strategy is generated through differential evolution algorithm optimization. In the optimization process of the differential evolution algorithm, a trained convolutional neural network model is used as the fitness function. The trained convolutional neural network model uses the light intensity and duration sequences of different stages as input features and the predicted yield of 2,3-butanediol as the output feature. The optimization objective is to maximize the predicted yield. Dividing the fermentation process into a dark culture growth stage and a blue light-stimulated product synthesis stage achieves spatiotemporal separation of cell biomass accumulation and target product synthesis, reducing the metabolic burden during the growth period. The dynamic illumination strategy generated by combining a convolutional neural network model with a differential evolution algorithm can automatically calculate and match complex temporal illumination parameters based on the algorithm. This data-driven control mechanism replaces traditional manual experience-based exploration, enabling precise quantitative allocation of light intensity and duration at different fermentation stages, thereby maximizing the synthesis efficiency of 2,3-butanediol.

[0036] In some embodiments of the present invention, the training data for the trained convolutional neural network model is obtained as follows: During the fermentation process, different blue light induction initiation times, periodic pulsed light with different duty cycles, and gradient pulsed light are tested respectively, and the corresponding light parameter sequences and 2,3-butanediol yield data are collected to construct a training set. By collecting actual fermentation data containing different induction initiation times, periodic pulsed duty cycles, and gradient pulses, a training set with multi-dimensional features is constructed. These diverse test data cover the complex mapping relationship between changes in light conditions and cell metabolic output, providing sufficient learning samples for the convolutional neural network model. This improves the accuracy of the model's prediction of 2,3-butanediol yield and ensures the reliability of the output results as a fitness function during the algorithm optimization process.

[0037] In some embodiments of the present invention, yeast extract and citric acid are jointly added to the culture medium used for fermentation. By specifically adding yeast extract and citric acid in combination to the fermentation culture medium, the problem of decreased cell viability is solved.

[0038] In some embodiments of the present invention, the method further includes a fermentation scale-up step: during fed-batch fermentation in a fermenter, the incident light intensity is dynamically increased in real time according to the cell density in the fermentation broth to maintain a constant light energy received per unit cell. By feeding in batches and dynamically increasing the incident light intensity according to the biomass, the consistency and stability of the light stimulation received by each cell in the photobioreactor are achieved, thereby further solving the problem of uneven light distribution and severe attenuation of deep light signals caused by the "self-shading effect" of high-density cell populations during fermentation scale-up culture. In scale-up culture in fermenters and fed-batch high-density fermentation, the increase in cell concentration leads to a decrease in the light transmittance of the fermentation broth and mutual shading between cells. Dynamically increasing the incident light intensity in real time according to the cell density to maintain a constant light energy received per unit cell can effectively alleviate the aforementioned self-shading effect. This mechanism ensures that the optogenetic system within the engineered strain receives a consistent light stimulation signal in fermentation systems of different scales and cell densities, maintaining the stability of metabolic pathway control and ensuring the overall conversion efficiency during scale-up production.

[0039] In some embodiments of the present invention, the fermentation tank has a capacity of 5L.

[0040] This invention establishes a dynamic lighting strategy (including optimized light-induced timing, periodic light pulses, and gradient light pulses) to alleviate the metabolic stress caused by continuous light by using appropriate dark intervals. It also finely coordinates the competition for carbon flux between cell growth and product synthesis in different stages of fermentation, thereby further improving the yield of 2,3-butanediol compared to constant light conditions.

[0041] This invention overcomes the limitations of manual empirical optimization and the complex nonlinear coupling problem of time-series illumination parameters by constructing a machine learning framework based on convolutional neural networks (CNN) and differential evolution (DE) algorithms. This algorithm framework can output a stage-specific optimal sequence of dynamic light intensity oscillations, further increasing fermentation yield by 15.60% on top of the empirically optimal strategy.

[0042] This invention establishes a dynamic light-based amplification process based on cell density. During fermentation amplification, the incident light intensity is adjusted in real time according to the biomass, maintaining a constant light energy received per cell and effectively overcoming the "self-shading effect" of high-density cell populations. Combined with specific nutrient supplementation (yeast extract and citric acid), high-density fed-batch fermentation was achieved in a 5L photobioreactor, yielding a production of 23.95 g / L, providing a technical foundation for the industrial-scale amplification of light-controlled biomanufacturing processes.

[0043] Unless otherwise described, the raw materials mentioned in this specification can be obtained through general commercial channels.

[0044] All raw materials used in this invention are shown in Tables 1 to 3 below. There are no particular restrictions on their sources; they can be purchased from the market or prepared using conventional methods known to those skilled in the art.

[0045] Table 1: Main Reagents

[0046] Table 2: Enzymes

[0047] Table 3: Reagent Kits

[0048] In the specific implementation of this invention, the relevant experimental materials, operating parameters, and construction details are as follows: The starting strain of the Saccharomyces cerevisiae, BY4742, has the accession number ATCC 201389. The introduced 2,3-butanediol synthesis gene can be derived from Bacillus subtilis (such as Bacillus subtilis 168, accession number ATCC 23857), Bacillus amyloliquefaciens, or Lactococcus lactis.

[0049] The photostimulation or photoinduced treatment involved in this invention has a specific wavelength range of 400 to 500 nm, preferably 460 nm, which falls within the blue light range.

[0050] In the gene construction process of engineered Saccharomyces cerevisiae strains, the gene amplification template and construction vector involved plasmids pYES2 (SEQ ID NO: 7) and pYES2L_SD (SEQ ID NO: 8). Specific selection markers involved in the gene editing and recombination process included: the use of... his3 Gene knockout pdc5 (Using histidine-deficient culture medium for screening), using KanMX gene knockout pdc6 (Use with a culture medium containing G418 for screening) ura3 Gene introduction into the PSD module (for screening in uracil-deficient medium), and the use of leu2 Genes were introduced into the SULI module (in conjunction with leucine-deficient culture medium for screening).

[0051] The relevant gene sequences involved in the embodiments of this invention can all be obtained by searching the following public accession numbers or gene IDs in public bioinformatics databases (such as NCBI or GenBank): α-acetyllactate synthase gene derived from Bacillus subtilis alsSThe reference GenBank accession number is Z93767 (or NC_000964.3, Gene ID: 939943).

[0052] α-acetyllactate decarboxylase gene derived from Bacillus subtilis alsD The reference GenBank login number is: AY780804.1.

[0053] Endogenous pyruvate decarboxylase isoenzyme gene derived from Saccharomyces cerevisiae pdc1 The reference GenBank accession number is: NM_001181931.1 (Gene ID: 850733).

[0054] Endogenous 2,3-butanediol dehydrogenase gene derived from Saccharomyces cerevisiae bdh1 The reference GenBank accession number is: NM_001178691.1 (Gene ID: 851680).

[0055] NADH oxidase gene derived from Lactobacillus casei LPnox2 Its homologous sequence reference GenBank accession number is: GCA_000026485.1.

[0056] The sources of the non-natural light-controlled module sequences involved in this invention are as follows: Photoinduced stable tag sequence, see the published literature Nature Communications (2023, Vol. 14, No. 1, p. 13).

[0057] The photosensitive degradation tag sequence is referenced from the published literature *Yeast* (2013, Vol. 30, p. 90). Table 4: Primers related to the introduction of the NADH oxidase gene.

[0058] Example 1: Construction of heterologous NADH oxidase (LPnox2) To achieve stable expression of heterologous NADH oxidase (LPnox2) in Saccharomyces cerevisiae, a complete gene expression module was constructed in this embodiment. This expression module drives and terminates the transcription of the target gene by cloning the endogenous FBA1 promoter (PFBA1) and FBA1 terminator (TFBA1) of Saccharomyces cerevisiae. In the specific amplification operation, in addition to using the forward and reverse primers Nox2-F (SEQ ID NO: 1) and Nox2-R (SEQ ID NO: 2) in Table 4 above to amplify the LPnox2 gene fragment, FBA1p-F (SEQ ID NO: 3) and FBA1p-R (SEQ ID NO: 4) were used to amplify the PFBA1 promoter fragment, and FBA1t-F (SEQ ID NO: 5) and FBA1t-R (SEQ ID NO: 6) were used to amplify the TFBA1 terminator fragment.

[0059] The promoter, target gene, and terminator fragments obtained from the above amplification were purified and integrated into a pre-linearized plasmid vector using conventional assembly techniques such as seamless cloning, ultimately yielding a successfully constructed recombinant plasmid (e.g., pYES2L_SDnox2, SEQ ID NO: 9). In this recombinant plasmid, LPnox2 Gene expression is strictly controlled by the upstream FBA1 promoter, forming a complete expression cassette of PFBA1-nox2-TFBA1, to ensure that the intracellular redox balance can be effectively remodeled after subsequent transformation into engineered strains.

[0060] Example 2: Construction of a dual-light-controlled engineered Saccharomyces cerevisiae strain BYSD_BSPPnox2 This embodiment describes the specific process of constructing an engineered Saccharomyces cerevisiae strain that includes a dual light control system and NADH oxidase.

[0061] First, using Saccharomyces cerevisiae BY4742 as the starting strain, an α-acetolactate synthase gene heterologously derived from Bacillus subtilis was introduced. alsS and α-acetolactate decarboxylase gene alsD This opens up the synthetic pathway from pyruvate to acetoin. Based on this, homologous recombination technology is used to sequentially knock out endogenous pyruvate decarboxylase isoenzyme genes. pdc5 and pdc6 To eliminate interference from background enzyme activity, a photosensitive degradation module PSD (SEQ ID NO: 10) was constructed, and the photosensitive degradation tag sequence was fused into the endogenous enzymes retained on the chromosome. pdc1 An intermediate strain capable of blue light-induced degradation of Pdc1 protein was constructed by targeting the C-terminus of the gene; simultaneously, a light-induced stabilization module SULI (SEQ ID NO: 11) was constructed to store the 2,3-butanediol dehydrogenase gene. bdh1By fusing expression with a light-induced stable tag, the Bdh1 protein is made stable only under blue light.

[0062] Figure 1 and Figure 2 This illustration shows a schematic diagram of the metabolic network and working mechanism of the engineered Saccharomyces cerevisiae strain provided in this embodiment of the invention. Figure 1 and Figure 2 It was found that by fusing a photosensitive degradation tag to endogenous pyruvate decarboxylase (Pdc1) and knocking out the secondary isoenzyme gene, dynamic downregulation of the ethanol pathway was achieved; simultaneously, by fusing a photoinducible stabilization module (SULI) with 2,3-butanediol dehydrogenase (Bdh1), dynamic upregulation of the 2,3-butanediol synthesis pathway was achieved. In darkness (growth mode), Pdc1-psd remained stable to support cell growth, while Bdh1-SULI was degraded; under blue light (production mode), Pdc1-psd was targeted for degradation to limit ethanol flux, while Bdh1-SULI gained stability to maximize 2,3-butanediol synthesis.

[0063] Example 3: Screening of NADH oxidases and their effect on metabolic flux This example verifies the effect of NADH oxidases from different sources on the 2,3-butanediol production performance of engineered strains.

[0064] To screen for the optimal NADH oxidase suitable for the brewer's yeast chassis, five NADH oxidase genes derived from Lactobacillus casei (Lp), Lactobacillus paracasei (Lpa), Bacillus subtilis (Bs), Lactococcus lactis (Ll), and Lactobacillus reuteri (Lr) were selected for heterologous expression testing.

[0065] Figure 3 This paper presents a comparative graph showing the effects of heterologous expression of NADH oxidases from different sources on 2,3-butanediol production based on a dual-light-controlled strain in this invention. These NADH oxidase elements include those from *Lactobacillus casei* (LPnox1, LPnox2), *Bacillus subtilis* (BsnoxC), *Lactococcus lactis* (LLnoxE), and *Lactobacillus reuteri* (LRnox). The results show that expression of LPnox2 from *Lactobacillus casei* (i.e., the LPnox2 bar chart) has the most significant effect on increasing 2,3-butanediol production, demonstrating its ability to effectively consume excess intracellular NADH and alleviate the redox imbalance caused by photoinduced metabolic reorientation.

[0066] Specifically, under blue light-induced fermentation conditions, the intracellular NADH / NAD+ ratio of the strain expressing LPnox2 was significantly reduced compared to the control strain that did not express the enzyme, effectively alleviating reducing stress. At the same time, the production of glycerol, a byproduct of NADH "pressure relief", was greatly reduced, and the carbon metabolic flux was redirected back to the 2,3-butanediol synthesis pathway.

[0067] Final tests showed that the introduction of LPnox2 increased the yield of 2,3-butanediol by approximately 23.16% compared to the original dual-light-controlled system, demonstrating the key role of this specific enzyme element in solving the problem of light-controlled metabolic imbalance.

[0068] Example 4: Empirical Optimization of Dynamic Illumination Strategy This embodiment describes the optimization of light timing and pulse parameters during fermentation to balance cell growth and 2,3-butanediol synthesis.

[0069] (1) Timing of light induction: Blue light was turned on for the engineered strain BYSD_BSPPnox2 at different growth stages. The results showed that initiating blue light induction in the middle of exponential growth (about 8 hours after fermentation) could maximize the product synthesis, with a yield of 2.47 g / L, which was significantly better than lag induction or culture in complete darkness.

[0070] (2) Periodic light pulses: After 8 hours of dark culture, different light duty cycles (50%-100%) with a period of 3 hours were applied. Using 70% light pulses (i.e., 2.1 hours of light / 0.9 hours of darkness) increased the yield to 2.83 g / L, indicating that moderate dark intervals help alleviate the metabolic homeostasis stress caused by continuous light.

[0071] (3) Gradient light pulse: In order to coordinate the allocation of resources throughout the process, low-frequency pulses of ≤30% are used to maintain growth in the early stage of fermentation (early growth stage), and then the duty cycle is gradually increased, reaching 70% light pulses in the middle of exponential growth (corresponding to Figure 4 Gradient strategy G). This strategy ultimately increased the yield of 2,3-butanediol to 3.27 g / L, an improvement of 32.39% compared to constant light conditions.

[0072] Figure 4 The bar chart shows the comparison of 2,3-butanediol yield obtained by fermentation under different gradient light pulse strategies in the embodiments of the present invention.

[0073] Example 5: Dynamic Prediction and Global Optimization of Light Control Strategy Based on CNN-DE Algorithm Framework This embodiment describes the process of using a machine learning model to perform high-dimensional space optimization of temporal illumination parameters.

[0074] Based on the aforementioned fermentation database, including constant, pulsed, gradient pulsed, and different light intensities, a convolutional neural network (CNN) regression model was constructed. This model comprises two temporal convolutional layers (each containing 32 3×1 filters) and a fully connected layer, used to capture the nonlinear relationship between temporal illumination characteristics and 2,3-butanediol yield. The performance of the CNN regression model was calculated and verified using the following formula. Online verification showed that the model has extremely high prediction accuracy (R² = 0.9827, MSE = 0.0053).

[0075]

[0076]

[0077] Among them, R 2 The coefficient of determination is denoted as , and MSE is the mean squared error.

[0078] Subsequently, using the CNN prediction model as the fitness function, the Differential Evolution (DE) algorithm was employed for optimization (setting 481 decision variables and introducing adaptive intermittent probabilities). The DE algorithm outputs a set of globally optimal illumination sequences that are dynamic, stage-specific, and exhibit light intensity oscillations.

[0079] Figure 5 This illustration shows a schematic diagram of the dynamic optimization process of lighting strategies based on convolutional neural networks (CNN) and differential evolution (DE) algorithms provided in an embodiment of the present invention. Figure 5 As shown, the process first collects fermentation experimental data under different lighting conditions to train a CNN regression model. Then, the trained CNN model with high prediction accuracy is used as the fitness function and optimized using a differential evolution (DE) algorithm. By performing crossover, mutation, and selection operations on the light intensity sequence, the globally optimal time-series light intensity sequence that maximizes the predicted yield of 2,3-butanediol is finally output. Figure 4 This study compared several dynamic lighting schemes, including strategies A through J, which employed low-frequency pulses in the early stages of fermentation and gradually increased the duty cycle in the later stages. The results clearly show that the optimized gradient pulse strategy (e.g., ...) is superior. Figure 4 Strategy G in this study effectively balances the energy and metabolic flux allocation needs of cells at different fermentation stages, thereby significantly increasing the final fermentation yield of 2,3-butanediol. When this strategy was implemented in actual fermentation using a programmable intelligent light controller, the final yield of 2,3-butanediol reached 3.78 g / L, a further increase of 15.60% compared to the empirically optimal strategy G in Example 3.

[0080] Specifically, the method for constructing a CNN prediction model can be carried out in the following way.

[0081] First, time-series data of light intensity recorded in multiple batches of fermentation experiments were collected, and the final 2,3-butanediol concentration obtained in the corresponding experiments was used as the model prediction target. To enhance the robustness of the model and expand the amount of training data, Gaussian noise with an intensity of 5% of the standard deviation of the original data was added to the original light sequence to generate new training samples. Two samples were generated for each experimental batch: the original light sequence sample and the light sequence sample with added noise, both corresponding to the same product concentration label. Subsequently, all samples were merged to form a training dataset, and the one-dimensional light time-series data was reshaped into a four-dimensional tensor structure suitable for input to a convolutional neural network.

[0082] In terms of network structure, the CNN model consists of an input layer, a feature extraction layer, and a regression output layer. The input layer receives the time series of illumination intensity and performs Z-score normalization. The feature extraction part consists of two concatenated convolutional layers, each using 32 one-dimensional convolutional kernels. A "Same" padding mode is used to ensure that the temporal dimension remains consistent during the extraction process. Each convolutional layer is followed by a ReLU activation function to enhance the model's non-linear feature mapping capability. In the output stage, the high-dimensional features extracted by the convolutions are flattened and then enter a fully connected layer for dimensionality reduction and compression. Finally, the predicted 2,3-BDO concentration is output through a single-neuron regression layer. The model training uses the Adam optimization algorithm, with 120 training epochs and a mini-batch size of 16. Cross-validation, mean squared error, and coefficient of determination are used to systematically evaluate the model's generalization performance and prediction accuracy.

[0083] The trained CNN model was used as the objective function, and the illumination strategy was globally optimized using the Differential Evolutionary Algorithm (DE). First, the entire fermentation process was set to 0 to 72 hours and discretized with a time step of 0.05 hours to form a complete illumination time series. To reduce the dimensionality of the optimization variables, three consecutive time points were grouped together, with consistent illumination intensity within each group. This significantly reduced the number of optimization variables from the original time series length to the number of groups.

[0084] In the differential evolution algorithm, a population of 1000 individuals is first initialized, containing multiple candidate lighting strategies. Some individuals are generated from lighting strategies that have performed well in historical experiments, while others are randomly generated while satisfying upper and lower bound constraints on light intensity. The upper limit of light intensity is set in stages according to the fermentation stage; for example, darkness is maintained in the early stages of fermentation, while the upper limit of light intensity is gradually increased in subsequent stages to meet the actual needs of microbial growth and metabolic regulation. Simultaneously, a staged intermittent lighting probability control mechanism is introduced into the algorithm, allowing certain time periods to be set to darkness with a certain probability, thereby simulating the intermittent lighting patterns commonly encountered in actual fermentation processes.

[0085] In each generation of evolution, the Differential Evolutionary Algorithm (DEA) generates new candidate lighting strategies through mutation, crossover, and selection operations. For each candidate strategy, the grouped lighting sequences are first expanded into a complete time series, and then input into a pre-trained CNN prediction model to obtain the corresponding 2,3-BDO predicted output. Since the DEA typically aims to minimize the fitness function, the negative value of the predicted output is used as the fitness function. Through multiple generations of iterative evolution, the algorithm continuously retains lighting strategies with higher predicted outputs and gradually approaches the optimal solution.

[0086] Once the algorithm reaches the set maximum number of iterations, the illumination strategy with the highest predicted yield is selected as the optimal illumination control scheme. Simultaneously, this strategy is compared with the best-performing illumination strategies from historical experiments, and the one with the higher predicted yield is chosen as the final output. The resulting illumination sequence typically exhibits a dynamic light intensity control pattern that varies with fermentation stages, incorporating appropriate intermittent light characteristics, thereby achieving fine-tuning of metabolic flux while ensuring cell growth.

[0087] This method combines machine learning prediction models with evolutionary optimization algorithms to achieve efficient searching of the complex fermentation illumination strategy space. It can obtain near-optimal dynamic illumination control conditions with a significant reduction in the number of experiments, providing a general technical route for intelligent optimization of photogenetic regulation of fermentation processes.

[0088] Example 6: High-density scale-up culture and feeding process in a 5L photobioreactor This embodiment describes the process of scale-up verification and high-density fed-batch fermentation in a 5L stirred photobioreactor.

[0089] (1) Dynamic light intensity control: To overcome the severe cell self-shading effect and light attenuation in high-density fermentation, the principle of "consistent light intensity per unit cell" was implemented. The automated light controller is based on the cell density (OD) of the fermentation broth. 600The system dynamically adjusts the incident light intensity of the internal LED light columns based on real-time changes, ensuring stable transmission of deep-layer light signals.

[0090] Figure 6 This diagram illustrates the dynamic light control during the scale-up cultivation process of the 5L photobioreactor provided in this embodiment of the invention. Figure 6 As shown, to overcome the cell self-shading effect caused by high-density culture, the reactor is equipped with an automated light controller. This controller can dynamically adjust the incident light intensity of the internal LED light columns in real time to ensure that the light energy received by each cell remains constant.

[0091] Figure 7 This paper presents kinetic curves of the engineered strain provided in this invention undergoing high-density fed-batch fermentation in a 5L photobioreactor. The figures detail the glucose consumption and cell density (OD) during fermentation under optimized fermentation conditions (combined addition of yeast extract and citric acid) and dynamic light control. 600 The curve shows the growth of [unspecified substance] and the dynamic changes in the concentrations of metabolites such as 2,3-butanediol, ethanol, glycerol, and acetoin. This curve demonstrates that the optogenetic regulation strategy of this invention possesses excellent robustness and high-yield potential in the scale-up system.

[0092] (2) Nutritional fortification and supplemental feeding in batches: See Figures 8 to 10 In the scale-up system, experiments confirmed that the combined addition of 1 g / L yeast extract (YE) and 1 g / L citric acid (CA) to the initial fermentation medium effectively alleviated photoinduced energy imbalance and growth inhibition. Based on this optimized medium, fed-batch fermentation with glucose (dynamically fed high-concentration glucose, KH2PO4, MgSO4, and amino acid mixture) was adopted, and the final cell density in the reactor (OD) was [not specified]. 600 The yield of 2,3-butanediol climbed to 22.24 g / L, and the final yield of 2,3-butanediol reached 23.95 g / L, successfully realizing the efficient operation of optogenetic regulation under the amplification system.

[0093] Figure 8 This invention presents kinetic curves illustrating the effect of adding different concentrations (0.5 g / L to 5 g / L) of yeast extract (YE) to the culture medium on the growth of the strain under dark culture (i.e., growth mode) conditions, as provided in this embodiment. The horizontal axis represents fermentation time, and the vertical axis represents cell optical density (OD). 600 ).

[0094] Figure 9This invention presents kinetic curves illustrating the effect of adding different concentrations of yeast extract (YE) to the culture medium on the growth of the strain under blue light-induced (i.e., production mode) conditions, as provided in this embodiment. The figures show that adding yeast extract effectively alleviates growth inhibition caused by light stress and metabolic redirection, and significantly promotes biomass accumulation in the early stages of fermentation.

[0095] Figure 10 This invention presents kinetic curves illustrating the effects of adding different concentrations of citric acid to the culture medium on the growth of the bacterial strain under dark culture conditions, as provided in this embodiment. These curves allow observation of the influence of citric acid, as an intermediate in the tricarboxylic acid cycle, on basal metabolism and cell proliferation during the dark phase.

[0096] Figure 11 The figures illustrate the kinetic curves showing the effect of adding different concentrations (1 g / L to 3 g / L) of citric acid to the culture medium on the growth of the bacterial strain under blue light induction conditions, as provided in the embodiments of the present invention. The curves in the figures show that supplementing with an appropriate amount of citric acid can effectively alleviate the energy imbalance problem under blue light induction, thereby synergistically enhancing the cell growth capacity under light-controlled fermentation conditions.

[0097] The above measurements of cell biomass and metabolite concentrations were all performed using the following methods.

[0098] 1. Cell biomass determination: The optical density (OD) of the fermentation broth was measured at a wavelength of 600 nm using a spectrophotometer. 600 ).

[0099] 2. Sample pretreatment: Collect the supernatant of the fermentation broth and filter it using a 0.22μm filter.

[0100] 3. Chromatographic analysis conditions: Glucose, glycerol, ethanol, acetoin, and 2,3-butanediol were quantitatively determined using a high-performance liquid chromatograph (1260 Infinity, Agilent, CO, USA) equipped with a differential refractive index (RI) detector; a BioRad Aminex HPX-87H column (300×7.8 mm) was used, and the column temperature was maintained at 50℃; the mobile phase was 5 mM H2SO4 solution, and the flow rate was set to 0.6 mL / min.

[0101] The calculations and simulations of intracellular metabolic fluxes were performed in the following manner.

[0102] 1. Model Construction and Constraints: The ecYeast 8 model was used to simulate the metabolic network of Saccharomyces cerevisiae, and metabolic constraints were set based on the carbon source, oxygen supply and growth rate under experimental conditions.

[0103] 2. Flux Balance Analysis: With the goal of maximizing 2,3-butanediol synthesis, the flux balance analysis (FBA) method was used to calculate the relative flux in a specific pathway and to evaluate the changes in metabolic flux distribution under different light conditions.

[0104] It should be understood that the dosages, reaction conditions, etc., in the various embodiments of this specification are approximate unless otherwise specified, and can be slightly modified according to actual circumstances to obtain similar results. Unless specifically defined, all technical and scientific terms used herein have the same meaning as understood by those skilled in the art. All documents mentioned herein are incorporated herein by reference. The embodiments described in this specification are preferred embodiments for illustrative purposes. Those skilled in the art can implement the invention using similar methods and materials to obtain the same or similar results. Various modifications or alterations to the invention still fall within the scope defined by the appended claims.

Claims

1. An engineered brewing yeast for the production of 2,3-butanediol (Saccharomyces cerevisiae) Saccharomyces cerevisiae The strain is characterized by: The engineered Saccharomyces cerevisiae strain contains a dual photocontrol system that regulates the ethanol synthesis pathway and the 2,3-butanediol synthesis pathway, and heterologously expresses NADH oxidase.

2. The engineered Saccharomyces cerevisiae strain for producing 2,3-butanediol as described in claim 1, characterized in that: The dual-light control system includes a photosensitive degradation module and a photoinduced stabilization module. The photosensitive degradation module degrades pyruvate decarboxylase under light irradiation, and the photoinduced stabilization module stabilizes 2,3-butanediol dehydrogenase in the 2,3-butanediol synthesis pathway under light irradiation.

3. The engineered Saccharomyces cerevisiae strain for producing 2,3-butanediol as described in claim 2, characterized in that: The engineered Saccharomyces cerevisiae strain contains an endogenous pyruvate decarboxylase isoenzyme gene. pdc5 and pdc6 Knockout endogenous pyruvate decarboxylase isoenzyme gene pdc1 The photosensitive degradation module is integrated; the 2,3-butanediol dehydrogenase gene is present in the engineered Saccharomyces cerevisiae strain. bdh1 The light-induced stabilization module is integrated.

4. The engineered Saccharomyces cerevisiae strain for producing 2,3-butanediol as described in claim 1, characterized in that: The NADH oxidase is derived from Lactobacillus casei, Lactobacillus paracasei, Bacillus subtilis, Lactococcus lactis, or Lactobacillus reuteri.

5. The engineered Saccharomyces cerevisiae strain for producing 2,3-butanediol as described in claim 1, characterized in that: The NADH oxidase is LPnox2 derived from Lactobacillus casei or Lactobacillus paracasei.

6. The engineered Saccharomyces cerevisiae strain for producing 2,3-butanediol as described in any one of claims 1 to 5, characterized in that: The engineered Saccharomyces cerevisiae strain also heterologously expresses the α-acetyllactate synthase gene derived from Bacillus subtilis. alsS and α-acetolactate decarboxylase gene alsD .

7. A method for preparing 2,3-butanediol, characterized in that: The method includes the following steps: Fermentation was carried out using the engineered Saccharomyces cerevisiae strain for the production of 2,3-butanediol as described in any one of claims 1 to 6; The fermentation process includes a growth stage and a product synthesis stage. The growth stage is carried out in darkness, and the product synthesis stage is stimulated by dynamic lighting. The light parameter sequence of the dynamic lighting is generated by condition optimization through differential evolution algorithm. During the conditional optimization process of the differential evolution algorithm, a trained convolutional neural network model is used to predict the yield of 2,3-butanediol corresponding to the illumination parameter sequence.

8. The method as described in claim 7, characterized in that: The trained convolutional neural network model uses the light intensity and duration sequences at different stages as input features and the predicted yield of 2,3-butanediol as output features. In the conditional optimization process of the differential evolution algorithm, the predicted yield is maximized.

9. The method as described in claim 7, characterized in that: The training data for the trained convolutional neural network model is obtained through the following steps: During the fermentation process, different photoinduction initiation times, periodic pulsed light with different duty cycles, and gradient pulsed light were tested to collect light parameter sequences and 2,3-butanediol yield data to construct a training set.

10. The method for preparing 2,3-butanediol according to claim 7, characterized in that: The method also includes the following steps: Yeast extract and citric acid were added to the fermentation reactor for fed-batch fermentation; and The light intensity is dynamically adjusted according to the cell density in the fermentation broth to maintain a constant amount of light energy received by each strain.

Images