Intelligent design method of membrane protein mena and application thereof

Abstract

The application discloses a smart design method of a membrane protein MenA and application thereof, and first, the typical UbiA family topological structure of MenA with nine transmembrane helices is confirmed through DeepTMHMM and TOPCONS cross verification; second, the dual-channel function architecture of the polar catalytic channel and the hydrophobic prenyl substrate channel of MenA is first revealed by CAVER channel analysis; third, the channel gating site that can be transformed is accurately positioned in combination with EVcouplings coevolution analysis and ConSurf conservation evaluation; then, two mutants Y113F and V89I that significantly enhance the thermodynamic stability of the protein are screened out through FoldX stability calculation; finally, GNINA molecular docking verification reveals that the distance between the TYR113 residue in the wild type MenA and the FPP substrate is 7.2 Å, and the distance between DHNA and the catalytic residue ASP78 is 4.4 Å. The application provides a systematic AI-driven methodological framework for membrane protein channel engineering, and provides a molecular engineering strategy with a clear structural basis for the industrial high yield of MK-7.

Description

Technical Field

[0001] This invention belongs to the field of bioengineering technology, specifically relating to a smart design method for the membrane protein MenA and its application. Background Technology

[0002] Vitamin K2 (menaquinone, MK-7) is an important fat-soluble vitamin that plays an irreplaceable and crucial role in bone metabolism regulation, coagulation factor synthesis, and cardiovascular health maintenance. Bacillus natto (… Bacillus subtilis natto ) is currently the main host for the industrial fermentation production of MK-7, but there is still considerable room for improvement in its ability to synthesize natural MK-7.

[0003] In the menadione biosynthesis pathway, MenA (1,4-dihydroxy-2-naphthoic acid polyisopentenyltransferase, EC2.5.1.74) is a key catalytic node connecting the aromatic head group (DHNA, 1,4-dihydroxy-2-naphthoic acid) to the isopentenyl side chain (polyisopentenyl pyrophosphate, FPP). MenA belongs to the UbiA superfamily of membrane-bound isopentenyltransferases and has a complex multi-transmembrane helical structure, with the catalytic cavity deeply embedded within the lipid bilayer.

[0004] Traditional protein engineering methods face unique and severe challenges when modifying membrane proteins. First, the tight hydrophobic stacking of transmembrane helices is extremely sensitive to amino acid substitutions. Second, the catalytic cavity is located in the hydrophobic environment inside the membrane, and traditional surface pocket prediction tools cannot accurately identify this cavity. Third, the MenA catalytic reaction involves two substrates with distinctly different properties: DHNA and FPP, each requiring channels with different geometric features to enter the catalytic cavity. Existing MenA modifications typically rely on random mutagenesis, homologous sequence alignment, or surface pocket analysis applicable to soluble proteins. These methods struggle to simultaneously define transmembrane helical boundaries, intra- and extra-membrane orientation, the intra-membrane catalytic cavity, and hydrophobic substrate channels, easily misidentifying catalytic core or structural support residues as mutagenic sites, leading to decreased protein folding stability, abnormal membrane localization, or hindered substrate entry.

[0005] Therefore, the existing technology lacks an feasible process for combining transmembrane topology, intramembrane orientation, dual substrate channels, evolutionary constraints, thermodynamic stability, and molecular docking results for the rational design of MenA. Summary of the Invention

[0006] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments.

[0007] In view of the problems existing in the above and / or prior art, the present invention is proposed.

[0008] Therefore, the purpose of this invention is to overcome the shortcomings of the prior art and provide a design method for the membrane protein MenA.

[0009] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for designing the membrane protein MenA, comprising, Topology and structure prediction: Obtaining the amino acid sequence of MenA protein from Bacillus natto; At least two transmembrane topology prediction tools were used to predict the transmembrane helix of the amino acid sequence, and the transmembrane helix positions and intracellular / extracellular orientations with consistent prediction results were retained. A three-dimensional structural model of MenA was created using a deep learning protein structure prediction model, and intramembrane orientation correction was performed using a membrane protein orientation prediction tool to obtain the intramembrane corrected structure. Channel identification and gated site selection: Based on the obtained intramembrane correction structure, channel analysis software was used to identify polar catalytic channels and hydrophobic prenyl substrate channels inside the protein, and the length, bottleneck radius and channel lining residues of each channel were obtained. Multiple sequence alignment and co-evolutionary analysis were performed on MenA to obtain residue pairs with co-evolutionary scores higher than a preset threshold. The evolutionary conservation of each residue was classified using conservation analysis tools; Based on the combined results of channel analysis, co-evolutionary analysis, and conservation analysis, residues located inside hydrophobic channels, with moderate conservation, and directly involved in substrate binding were selected as candidate sites for channel gating engineering. Stability calculation and mutant design: For the identified candidate sites, construct single-site or multi-site mutant structural models; use protein free energy calculation tools to evaluate the folding free energy change (ΔΔG) of each mutant, and screen for stability-enhancing mutants with ΔΔG less than 0 kcal / mol.

[0010] As a preferred embodiment of the method described in this invention, the method for performing multiple sequence alignment and co-evolutionary analysis on MenA includes: using the MenA amino acid sequence as the query sequence, performing automatic retrieval in a homologous sequence database, removing redundant and low-coverage sequences to construct multiple sequence alignments, establishing a statistical mechanical model based on the multiple sequence alignments to calculate the long-range evolutionary coupling score of all residue pairs, and identifying residue pairs with coupling scores higher than a preset score threshold and spatial distances less than a preset distance threshold in the intramembrane correction structure as co-evolutionary residue pairs with important functional interactions. The co-evolutionary analysis constraints are as follows: the database is selected from UniRef100 or an equivalent homologous sequence database; the sequence selection criterion is: E-value ≤ 1 × 10⁻⁶. -5Coverage ≥70%; Co-evolution criteria: Coupling score >10 and spatial distance <4Å; The model is a Potts statistical mechanics model based on multiple sequence alignment; The conservation analysis tool is used to classify the evolutionary conservation of each residue, including: taking the intramembrane correction structure and corresponding amino acid sequence as input, automatically searching homologous sequences to construct a phylogenetic tree and calculating the evolutionary rate of each residue, and normalizing the conservation of residues into levels 1-9 according to the standardized classification system built into the conservation analysis tool, where level 9 represents high conservation and level 1 represents high variability; Among them, the conservation analysis constraint rules are as follows: the algorithm is selected from the Bayesian method to calculate the evolution rate; homology sequence parameters: 100-150 sequences, identity range 35%-95%; moderate conservation definition: 4-8 level interval.

[0011] As a preferred embodiment of the method described in this invention, the rules for screening candidate sites for channel gating engineering are as follows: the residue is located within the lining of the hydrophobic prenyl substrate channel: within 4.0 Å of the hydrophobic channel axis; the conservation level is in the moderate conservation range, neither a highly conserved residue of the catalytic core nor a highly variable residue of non-functionality: ConSurf score 4-8; there is a direct spatial association with the co-evolutionary high-scoring residue pair: the spatial distance from the high-scoring residue pair is <4 Å; catalytic core, substrate recognition, and cofactor binding residues with the highest conservation level are strictly excluded.

[0012] As a preferred embodiment of the method described in this invention, after screening for stability-enhancing mutants with ΔΔG less than 0 kcal / mol, the method further includes molecular docking and channel gating verification, specifically: constructing the three-dimensional structure of DHNA and polyisoprene pyrophosphate (FPP); using the geometric center of MenA catalytic residues ASP78 and ASP84 as the docking box center, performing focused molecular docking on wild-type and preferred mutants; analyzing the direct binding distance and hydrophobic interaction between the mutation site and the FPP substrate, as well as the distance change between DHNA and ASP84, and screening for target mutants that simultaneously enhance FPP binding and optimize the pre-catalytic conformation of DHNA.

[0013] As a preferred embodiment of the method described in this invention, the transmembrane topology prediction tool is a legally authorized, publicly licensed, or open-source transmembrane topology prediction tool. It is preferred to use DeepTMHMM and TOPCONS for cross-validation, or an equivalent tool that can output transmembrane helical regions and membrane orientations can be used, and it can be used online via a web page. The structure prediction model is a legally authorized or publicly licensed deep learning protein structure prediction model, preferably ColabFold or a model with equivalent three-dimensional structure prediction capabilities, and is used online via a web page; The membrane protein orientation prediction tool is a legally authorized or publicly licensed membrane protein orientation prediction tool, preferably a PPM server or an equivalent membrane protein orientation correction tool, and is used online via a web page; The channel analysis software is a legally authorized or open-source protein channel analysis software, preferably CAVERAnalyst 2.0 or an equivalent tool that can output channel length, bottleneck radius and liner residues. The software can be used locally after registration and authorization via an educational email address. The co-evolutionary analysis tool is a legally authorized or open-source co-evolutionary analysis tool, preferably EVcouplings or an equivalent tool, and is used online via a web page; The conservation analysis tool is a legally authorized or publicly licensed residue conservation analysis tool, preferably ConSurf or an equivalent tool, and is used online via a web page; The protein free energy calculation tool is a legally authorized or publicly licensed protein free energy calculation tool, preferably FoldX or an equivalent tool, and is used online via a web page; The molecular docking software is a legally authorized or open-source molecular docking software, preferably GNINA or an equivalent tool. This software can be used locally after being registered and authorized via an educational email address.

[0014] As a preferred embodiment of the method described in this invention, the polar catalytic channel has a length of 15-20 Å and a bottleneck radius of 1.5-2.0 Å, and the lining residues include E75, D78, D84, R207, and R217, which are responsible for DHNA substrate entry and pyrophosphate ion release. The hydrophobic prenyl substrate channel has a length of 24-32 Å and a bottleneck radius of 0.8-2.0 Å. The lining residues include F79, V89, Y113 and their adjacent hydrophobic residues, which are responsible for the entry and positioning of the FPP hydrophobic chain.

[0015] As a preferred embodiment of the method described in this invention, the gated engineering candidate sites include F79, V89, and Y113; The stability-enhancing mutants include one or a combination of the following: F79L, V89I, and Y113F; wherein, the ΔΔG of Y113F is -0.453 kcal / mol, and the ΔΔG of V89I is -1.020 kcal / mol.

[0016] In a preferred embodiment of the method described in this invention, the distance between TYR113 and FPP in wild-type MenA is 7.2 Å, and the distance between DHNA and the catalytic residue ASP78 is 4.4 Å. After the Y113F mutation, the distance between PHE113 and FPP was shortened to 6.7 Å, the hydrophobic contact was optimized, and a conformational rearrangement of DHNA was induced, which shortened the distance between DHNA and ASP78 from 4.4 Å to 3.3 Å. After the V89I mutation, the distance between TYR113 and FPP shortened to 6.9 Å, and the distance between DHNA and ASP78 shortened significantly to 2.7 Å.

[0017] Another object of the present invention is to overcome the shortcomings of the prior art and provide a MenA membrane protein having the following characteristics: (1) Its amino acid sequence is derived from the MenA protein of Bacillus natto; (2) An amino acid substitution occurs at at least one gated site, wherein the gated site is selected from F79, V89, and Y113; (3) The amino acid substitutions include one or a combination of the following: F79 is replaced with leucine, V89 is replaced with isoleucine, and Y113 is replaced with phenylalanine; In molecular docking simulations, the modified MenA forms a tighter channel binding space with FPP and induces conformational rearrangement of the substrate DHNA, making its distance from the catalytic residue ASP78 shorter than that of wild-type MenA.

[0018] Another object of the present invention is to overcome the shortcomings of the prior art and provide a recombinant Bacillus natto containing a recombinant gene encoding the modified MenA protein of claim 7; and under fermentation conditions, the MK-7 yield of the recombinant strain is increased by at least 20% relative to the control strain expressing wild-type MenA.

[0019] Another objective of this invention is to overcome the shortcomings of the prior art and provide a use of modified MenA protein or recombinant Bacillus natto in the preparation of vitamin K2 nutritional supplements, health foods or pharmaceutical preparations.

[0020] Another object of the present invention is to overcome the shortcomings of the prior art and provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the rational design method for the membrane protein MenA.

[0021] Beneficial effects of this invention: (1) This invention establishes a systematic and rational computational framework for the modification of multi-transmembrane proteins, which integrates topological information throughout the design process, effectively avoids common errors in traditional methods, and significantly improves the accuracy and success rate of membrane protein modification.

[0022] (2) For the first time, the dual-channel functional architecture of MenA’s polar catalytic channel and hydrophobic prenyl channel was accurately distinguished through CAVER systematic channel analysis, realizing function-oriented precise channel gating modification.

[0023] (3) Through GNINA molecular docking verification, the direct binding mechanism between the TYR113 site and the FPP substrate was clearly revealed, providing direct structural evidence for the functional optimization of the Y113F mutation and overturning the traditional understanding of the function of this site.

[0024] (4) This invention screened out two mutants, Y113F and V89I, which have the dual effects of enhanced thermodynamic stability and pre-catalytic conformation optimization, providing a molecular engineering strategy with clear structural basis for the high-yield industrial production of MK-7. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is the overall flowchart of the MenA channel gate control project of the present invention.

[0026] Figure 2 The diagram shows the prediction results of the MenA transmembrane helical topology using DeepTMHMM according to the present invention.

[0027] Figure 3 This is a diagram showing the predicted transmembrane helical topology of MenA using TOPCONS in this invention.

[0028] Figure 4 This is a pLDDT result chart of the top five models in ColabFold that predict wild-type MenA according to this invention.

[0029] Figure 5 This is a pLDDT result graph showing the top five scoring models of the ColabFold model for predicting the mutant MenA_Y113F in this invention.

[0030] Figure 6 This is a pLDDT result graph showing the top five scoring models of the ColabFold prediction of the mutant MenA_V89I in this invention.

[0031] Figure 7 This is a schematic diagram of the in-membrane correction structure of the present invention.

[0032] Figure 8 This is a diagram of the dual-channel functional architecture revealed by the CAVER channel analysis of this invention.

[0033] Figure 9 The results of the thermodynamic stability evaluation of FoldX 5.0 of this invention.

[0034] Figure 10 The diagram shows the docking results of wild-type MenA with DHNA and FPP molecules in this invention. It shows that the distance between TYR113 and FPP is 7.2 Å, and the distance between DHNA and ASP78 is 4.4 Å.

[0035] Figure 11 The diagram shows the docking results of the Y113F mutant molecules of this invention, indicating that the distance between PHE113 and FPP has shortened to 6.7 Å, and DHNA has rearranged to around ASP78 (3.3 Å).

[0036] Figure 12 The diagram shows the docking results of the V89I mutant molecules of this invention, indicating that the distance between TYR113 and FPP has been shortened to 6.9 Å, and the DHNA depth is anchored near ASP78 (2.7 Å).

[0037] Figure 13 The graph shows the yield of MK-7 after 72 hours of fermentation using the wild-type strain of this invention.

[0038] Figure 14 The graph shows the MK-7 yield detection results after 72 hours of fermentation of the mutant recombinant strain of this invention. Detailed Implementation

[0039] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the examples in the specification.

[0040] Example 1 (1) MenA transmembrane topology confirmation and structural modeling: Sequence acquisition and verification repair using Bacillus natto (Bacillus) Bacillus subtilis subsp. subtilis str.168 The MenA protein (NCBI RefSeq: NP_391728.1) was used as the research object. To address the repetitive fragment issue in the original strain's sequencing sequence, SnapGene software was used to perform CDS integrity verification on the MenA coding gene, checking the start codon, stop codon, and reading frame to obtain an accurate, complete single-stranded sequence of 311 amino acids (311 aa). This sequence was saved in FASTA format as the unified input sequence for all subsequent predictive analyses. Specifically, using the FASTA amino acid sequence of MenA as input, DeepTMHMM uses the default membrane protein prediction parameters to directly output the topological labeling sequence (Inside / Membrane / Outside tags) of each amino acid residue, the precise start and end positions of each transmembrane helix, the determination of the N-terminal / C-terminal membrane orientation, and the prediction confidence score of each position. TOPCONS uses the default parameters of TOPCONS2 and integrates the prediction results of five sub-models, including OCTOPUS, Philius, PolyPhobius, SCAMPI, and SPOCTOPUS, to output consensus topology predictions and comparative analysis of each model. Regions predicted by both tools as transmembrane helices with completely consistent intracellular / extracellular orientations are designated as reserved regions. When the two tools deviate from the start and end positions of the same transmembrane helix by no more than ±3 amino acids and their orientation judgments are consistent, the prediction results are considered to be of high confidence. Cross-validation results showed that DeepTMHMM and TOPCOS were highly consistent in predicting MenA protein, with prediction confidence levels for transmembrane regions reaching 87%-100%. This confirmed that MenA possesses the typical UbiA family topology with nine transmembrane helices, with both the N-terminus and C-terminus located on the cytoplasmic side. The transmembrane regions were positions 28-48, 53-73, 105-125, 129-149, 157-177, 184-204, 223-243, 245-265, and 290-310, respectively. (See the prediction results below.) Figure 2 as well as Figure 3 .

[0041] (2) Using ColabFold for three-dimensional structure prediction.

[0042] Specifically, ColabFold (based on AlphaFold2) was used for three-dimensional structure prediction. The input was the 311 aa MenA single-stranded sequence obtained after verification and repair in step (1) above. No ligands, templates or exogenous constraints were added. The model type was selected as AlphaFold2_ptm. MSA construction was carried out using MMseqs2. Recycle was set to 12 and Amber Relax was set to ON. The other parameters were set to default. Among the multiple candidate models output by ColabFold, the model with pLDDT > 85 in the transmembrane region, continuous arrangement of 9 transmembrane helices, and topological relationship consistent with the prediction results of DeepTMHMM / TOPCONS was selected as the subject of subsequent analysis. The prediction results are shown below. Figures 4-6 ; (3) The optimal candidate structure model of ColabFold obtained above (PDB format) is repaired by PDBFixer. The specific operations include filling in missing atoms, repairing incomplete residues, and completing hydrogenation treatment at pH 7.0 protonation state to obtain a repaired PDB file with complete structure and accurate protonation state. Subsequently, the complete MenA structure PDB file repaired by PDBFixer was submitted to the PPM server for intramembrane orientation correction. The membrane type was set to Bacterial membrane, and the default membrane thickness and OPM / PPM energy minimization parameters were used to ensure that the hydrophobic segments of the nine transmembrane helices were correctly embedded into the hydrophobic core of the lipid bilayer, resulting in a corrected structure with accurate intramembrane orientation. This structure served as the common operating object for Examples 2-5. The corrected structure diagram is shown below. Figure 7 .

[0043] Example 2 CAVER channel analysis and dual-channel architecture identification: (1) Confirmation of reference point for catalytic region and setting of channel analysis parameters Based on the conserved structural characteristics of membrane-bound isopentenyltransferases in the UbiA superfamily, members of this family generally possess a conserved aspartic acid-rich motif (typical sequence pattern DXXXD). Pattern scanning analysis of the MenA amino acid sequence using the SnapGene software used in Example 1 revealed a 5-amino acid interval between aspartic acid at position 78 (D78) and aspartic acid at position 84 (D84), exhibiting a typical D... 78 XXXXD 84 The sequence pattern conforms to the typical characteristics of the conserved catalytic motifs of the UbiA family; Further verification using PyMOL software was performed by comparing and contrasting the three-dimensional spatial coordinates: The three-dimensional structure of the intramembrane-corrected MenA obtained in Example 1 was compared with the Mg-containing structures resolved in the PDB database. 2+ Structural alignment calculations were performed on the crystal structures of cofactor UbiA superfamily homologous proteins. Specifically, the PyMOL align command was used to optimally align the three-dimensional coordinates of the two structures. The structural alignment calculation results showed that the three-dimensional coordinate positions of D78 and D84 residues of MenA corresponded to the coordinates of Mg in the homologous protein crystal structure. 2+ The conserved aspartic acid residues of the ions are highly overlapping, and the spatial geometry of the carboxyl side chains of the two are well consistent, which verifies from the three-dimensional structural level that D78 and D84 constitute metal ion binding regions similar to those of homologous proteins. Combined with the ConSurf conservation analysis results in Example 3, both D78 and D84 reached a conservation level of 9 (the highest level of conservation), indicating that they were subject to strong selection pressure during evolution.

[0044] Based on the above one-dimensional sequence feature analysis, three-dimensional spatial structure superposition verification, and evolutionary conservation analysis results, D78 and D84 were identified as key reference residues in the MenA catalytic region, and the geometric center of their carboxyl oxygen atom was used as the search starting point for subsequent channel analysis.

[0045] Systematic channel and cavity analysis was performed on the intramembrane-corrected MenA structure obtained in Example 1 using channel analysis software. CAVER Analyst 2.0 or an equivalent tool could be used, with the following specific parameter settings: the geometric center of the carboxyl oxygen atoms of the confirmed catalytic residues D78 and D84 was used as the starting point for channel search; the probe radius was set to 0.8-3.0 Å, the clustering threshold to 3.5 Å, the shell radius to 4.0 Å, the shell depth to 5.0 Å, and the minimum channel length to 5 Å. After calculating all reachable channels, the following quantitative criteria were used for screening and function assignment: Bottleneck radius screening: Only channels with a minimum bottleneck radius ≥ 1.2 Å are retained as valid channels; Length range constraint: Only channels with a channel geometry length in the range of 15-35 Å are retained; Priority sorting: Sort the channels by bottleneck radius given by CAVER from largest to smallest, and select the top 3 channels; Liner residue analysis: Residues within 4.0 Å of the channel axis were defined as channel lining residues. Based on the physicochemical properties of the lining residues (polar / charged vs. hydrophobic / aromatic), the channels were classified as polar catalytic channels or hydrophobic prenyl substrate channels.

[0046] Through the standardized screening process described above, the functionally relevant channels of MenA were determined, and their three-dimensional spatial distribution was illustrated. Figure 8 .

[0047] (2) Quantitative characterization of the dual-channel functional architecture Based on the above channel analysis and screening, MenA is revealed to possess a dual-channel functional architecture consisting of one polar catalytic channel and one hydrophobic prenyl substrate channel. The geometric parameters and functional characteristics of each channel are as follows: Polar catalytic channels: Channel geometry length: 17.1 Å (consistent with the typical polar channel range of 15-20 Å) Minimum bottleneck radius: 1.788 Å (minimum space requirement for DHNA substrate molecules to enter) Key lining residues: R27, E75, D78, D84, R207, R217 (all are charged polar residues) Functional positioning: The channel is predominantly composed of polar and charged residues, and its spatial geometry is suitable for the entry of the polar aromatic substrate DHNA and the release of the reaction byproduct pyrophosphate ions (PPi). Hydrophobic prenyl substrate channels: Main channel length: 30.9 Å (in line with the 24-32 Å range for hydrophobic long-chain substrate channels) Minimum bottleneck radius: 1.823 Å (meets the spatial requirements for FPP isopentenene chain entry and positioning) Key lining residues: Y76, F79, L72, Y113, I200 and their adjacent hydrophobic residues (mainly aromatic and aliphatic hydrophobic residues). Functional positioning: The hydrophobic environment inside this channel is suitable for the entry, positioning, and pre-catalytic conformational regulation of the long-chain hydrophobic substrate polyisoprene pyrophosphate (FPP).

[0048] Example 3 Evolutionary constraint analysis and candidate site screening: (1) Co-evolutionary analysis using EVcouplings Co-evolutionary analysis was performed using the EVcouplings online server (https: / / evcouplings.org). Specifically, the complete single-stranded amino acid sequence (FASTA format) of 311 aa MenA obtained after SnapGene verification and repair in step (1) of Example 1 was submitted as the query sequence to the Protein module of the EVcouplings web page; Parameter settings: Select UniRef100 as the homologous sequence search database, set the E-value threshold to 1e-5, set the minimum sequence coverage to 70%, and use the default server settings for the remaining parameters; The server automatically completes the following calculation process: First, it retrieves homologous sequences of MenA from the UniRef100 database through iterative search, automatically removes redundant and low-quality sequences with a sequence coverage of less than 70%, and constructs multiple sequence alignment (MSA); then, it establishes a Potts statistical mechanical model based on the MSA, calculates the long-range evolutionary coupling score (EC score) between all residue pairs, and outputs the residue pair coupling score matrix. Residue pairs with an EC score > 10 and a spatial distance < 4 Å in the membrane-corrected MenA three-dimensional structure obtained in Example 1 were identified as important functional interaction residue pairs. The screening results showed that high-scoring residue pairs included F79-P102 (score 22.95), Y76-P102 (18.50), I21-G99 (18.41), Y76-L106 (16.78), and A69-Y113 (16.72), suggesting that there is significant conformational coupling in the regions where F79, Y76, and Y113 are located, which is highly correlated with channel gating and substrate localization functions.

[0049] (2) Calculate residue conservation score using ConSurf The evolutionary conservation score of each residue of MenA was calculated using the ConSurf online server (https: / / consurf.tau.ac.il). Specifically, the complete three-dimensional structure PDB file of MenA, obtained by ColabFold prediction in step (3) of Example 1 and repaired by PDBFixer and hydrogen supplementation and PPM intramembrane orientation correction, was uploaded to the ConSurf web page as input, and the system automatically extracted the corresponding amino acid sequence from the PDB file; Parameter settings: Set the number of homologous sequences to 150, the sequence identity filtering range to 35%-95% (excluding homologous sequences that are too close or too distant), select the Bayesian algorithm as the evolutionary model, and use the default settings for the other parameters; The server automatically performs homologous sequence search, multiple sequence alignment construction, and phylogenetic tree analysis. It calculates the evolutionary rate of each residue position based on Bayesian statistical methods and automatically normalizes and maps the residue conservation to levels 1-9 according to the standardized grading criteria built into the ConSurf system. This grading system is the standard output format of the ConSurf software, where level 9 indicates high conservation during evolution (usually key catalytic or structural sites), level 1 indicates high variability, and intermediate values ​​represent different degrees of conservation. Based on the residue conservation level output by ConSurf, spatial burial / exposure status, and the channel liner location determined in Example 2, the unmodifiable and modifiable sites were determined according to the following principles: Strictly prohibited modification sites: D78, D84 (located in Asp-rich catalytic motifs, all with ConSurf level 9), E75, N74, N204 (substrate recognition residues, all with level 9), R207, R217 (pyrophosphate binding residues, all with level 9); where D78 indicates that the 78th amino acid in the MenA amino acid sequence is aspartic acid, and D84 indicates that the 84th amino acid is aspartic acid, and the remaining sites have the same meaning; Candidate sites for modification: F79, Y113, V89, and I91. Among them, F79 is located near the entrance of the hydrophobic channel, with a ConSurf grade of 4, and the natural variants are predominantly Y / F; Y113 is located inside the hydrophobic channel, with a ConSurf grade of 5, and the natural variants are 53% Y and 32% F; V89 is located in the channel lining, with a ConSurf grade of 8, but the natural variants contain I; I91 is adjacent to the channel but has a conservation of 9, and is considered a candidate site rather than a preferred mutation site.

[0050] Example 4 Example 4: FoldX Thermodynamic Stability Calculation and Mutant Screening Thermodynamic stability of the candidate sites identified in Example 3 was assessed using FoldX 5.0 software. FoldX quantitatively assesses the impact of mutations on overall protein stability by precisely calculating the difference in free energy (ΔΔG) between the wild-type and mutant in the folded state. The complete calculation process consists of three strictly sequential steps: Step 1: Wild-type structure repair and energy pre-optimization (RepairPDB command) Because protein structures obtained from ColabFold / AlphaFold predictions or experimental analysis often contain small steric clashes, undesirable dihedral angles of side chains, or bond length deviations, the wild-type template structure must be systematically repaired and energy optimized before calculating the mutation free energy.

[0051] Specific operation: Take the in-membrane corrected MenA structure PDB file obtained in Example 1 after ColabFold prediction, PDBFixer hydrogen replenishment repair and PPM in-membrane orientation correction as input, name the file MenA_WT.pdb and place it in the FoldX working directory; Execute the structural repair command in the command-line interface: foldx --command=RepairPDB --pdb=MenA_WT.pdb This command triggers FoldX's built-in molecular force field energy minimization algorithm, and the system automatically performs the following repair operations: identify and adjust the conformation of amino acid side chains with spatial conflicts within the protein, optimize unreasonable bond lengths, bond angles, and dihedral angle parameters, minimize the local energy of the overall structure, and finally output a repaired wild-type PDB file (automatically named MenA_WT_Repair.pdb) with the lowest energy state and reasonable geometry under the FoldX force field, which serves as a standard reference template for subsequent mutation calculations; Step 2: Automatic construction of mutant structural model and calculation of ΔΔG (BuildModel command) FoldX's mutant structure construction is not achieved by manually adjusting atomic coordinates, but by writing standardized mutation instruction files, which allow the software to automatically perform side chain substitution, conformation search, and local energy optimization.

[0052] (1) Preparation of the mutation instruction file: Create a plain text file named individual_list.txt in the FoldX working directory, and write the single-point mutation information to be constructed line by line according to the FoldX standard format. The file format is "original amino acid single-letter code + chain identifier + residue number + mutated amino acid single-letter code + semicolon". For the candidate sites in this embodiment, the file content is as follows: VA89I; YA113F; F79L; I91V; L72V; Wherein, VA89I indicates that valine (V) at position 89 of chain A is replaced with isoleucine (I); YA113F indicates that tyrosine (Y) at position 113 of chain A is replaced with phenylalanine (F); FA79L indicates that phenylalanine (F) at position 79 of chain A is mutated to leucine (L); IA91V indicates that isoleucine (I) at position 91 of chain A is mutated to valine (V); LA72V indicates that leucine (L) at position 72 of chain A is mutated to valine (V). (2) Execution of the BuildModel command: Execute the mutant construction and free energy calculation instructions in the command line: foldx --command=BuildModel --pdb=MenA_WT_Repair.pdb --mutant-file=individual_list.txt --numberOfRuns=5 --temperature=298 Key parameter descriptions: --pdb=MenA_WT_Repair.pdb: Specifies the repaired wild-type structure output from step one as a template. --mutant-file=individual_list.txt: Specifies the configuration file containing mutation instructions. --numberOfRuns=5: Repeat the calculation 5 times independently for each mutation and take the average to reduce random error in conformation sampling. --temperature=298: Sets the calculation temperature to 298 K (25℃, close to physiological conditions). (3) Automatic BuildModel Execution Process: When executing the BuildModel command, FoldX automatically performs the following calculation steps for each specified mutation site: It replaces the amino acid side chains at the target site with the mutated amino acid type; it performs a systematic rotational isomer search on all amino acid side chains within a 4.0 Å radius around the mutation site to find the energy-optimal side chain conformation combination; it performs local energy minimization to eliminate spatial conflicts caused by the mutation; and it calculates the difference in folding free energy between the mutant and the wild type. The folding free energy represents the folding free energy of the mutant and the wild type, respectively.

[0053] Step 3: Stability assessment and mutant screening FoldX automatically outputs the ΔΔG value for each mutant in 5 independent runs, and takes the arithmetic mean as the final thermodynamic stability assessment result for that mutant. Simultaneously, the 3D structural model of the mutant output by BuildModel is checked to confirm that the overall transmembrane helix fold around the mutation site has not undergone abnormal deformation.

[0054] Screening criteria and judgment basis: ΔΔG < 0 kcal / mol: The mutant is more stable than the wild type and is retained for the next round of screening. ΔΔG < -0.5 kcal / mol: Significantly enhanced stability, preferred. ΔΔG>0 kcal / mol: Stability decreases, therefore this option should be excluded. Complete screening results:

[0055] After the above quantitative calculations and rigorous screening, V89I and Y113F mutants were finally selected as the preferred mutants with enhanced thermodynamic stability, and will proceed to the subsequent molecular docking verification stage. FoldX stability calculation results ( Figure 9 The results showed that both single-point mutations were automatically optimized for side-chain conformation by software, without disrupting the overall topological folding of the nine transmembrane helices of the MenA protein, and exhibited significantly better thermodynamic stability than the wild type, making them comparable to... Figure 11 , Figure 12 The mutant docking conformation shown is used for comprehensive judgment.

[0056] Example 5 Focused molecular docking analysis of the binding conformations of wild-type and mutant MenA with dual substrates was performed using the deep learning-based molecular docking software GNINA to verify the influence mechanism of Y113F and V89I mutants on the pre-catalytic conformation of the substrates. The complete operation procedure consists of four steps: Step 1: Preparation of three-dimensional ligand structure (1) Initial structure acquisition of ligands: The initial three-dimensional structure SDF files of DHNA (1,4-dihydroxy-2-naphthoic acid, PubChem CID: 68072) and FPP (farnesyl pyrophosphate, PubChem CID: 445713) were downloaded from the PubChem chemical database as standard starting materials for ligand preparation; (2) Ligand normalization was performed using OpenBabel software to protonate, hydrogenate, and minimize the energy of the two ligands respectively: DHNA ligand treatment obabel DHNA.sdf -O DHNA_prep.pdbqt \ --addh --ph 7.0 \ --minimize --ff MMFF94 \ --steps 500 FPP ligand treatment obabel FPP.sdf -O FPP_prep.pdbqt \ --addh --ph 7.0 \ --minimize --ff MMFF94 \ --steps 500 Key parameter descriptions: --addh --ph 7.0 automatically adds hydrogen atoms to complete protonation according to physiological pH conditions; --minimize --ff MMFF94 uses the MMFF94 molecular force field to minimize energy and eliminate unreasonable bond lengths and angles; the output PDBQT format contains the atomic coordinates and charge information required for GNINA docking; Step 2: Receptor Structure Preparation (1) The origins of the receptor structures are clear. The specific origins of the three receptor structures are as follows: Wild-type MenA receptor: The repaired structure file (MenA_WT_Repair.pdb) output by the RepairPDB command in step one of Example 4. Y113F mutant receptor: The three-dimensional structure PDB file of the Y113F mutant was output using the FoldX BuildModel command in step two of Example 4. V89I mutant receptor: The V89I mutant 3D structure PDB file output by the FoldX BuildModel command in step two of Example 4. (2) Receptor standardization: The three receptor structures were standardized using AutoDockTools. Taking the wild type as an example, the mutant operation is the same. python prepare_receptor4.py \ -r MenA_WT_Repair.pdb \ -o MenA_WT.pdbqt \ -A hydrogens \ -U nphs_lps_waters \ -e Parameter description: -A hydrogens replenish hydrogen atoms to the acceptor; -U nphs_lps_waters automatically remove non-polar hydrogens, lipid molecules, and water of crystallization, etc.; -e preserve the existing atomic conformation to ensure that the side chain conformation of catalytic residues D78 / D84 remains unchanged; Step 3: Setting up the docking box and calculating the docking with GNINA (1) Determining the center coordinates of the docking box: The three-dimensional coordinates of the carboxyl oxygen atoms (OD1 and OD2) of the two catalytic residues D78 and D84 were extracted from the acceptor PDBQT file (obtained after AutoDockTools normalization) using PyMOL software, and their geometric center was calculated as the center coordinates of the docking box. (2) GNINA focused molecular docking is performed with the geometric center coordinates calculated above as the center of the docking box: Taking wild-type MenA and DHNA docking as an example gnina \ --receptor MenA_WT.pdbqt \ --ligand DHNA_prep.pdbqt \ --center_x [x-coordinate of the geometric center of D78 / D84] \ --center_y [y-coordinate of the geometric center of D78 / D84] \ --center_z [Z-coordinate of the geometric center of D78 / D84] \ --size_x 10 --size_y 7 --size_z 14 \ --num_modes 20 --exhaustiveness 8 \ --cnn_scoring rescore \ --out MenA_WT_DHNA_out.pdbqt Key parameter descriptions: --size_x 10 --size_y 7 --size_z 14 sets the docking box size to 10×7×14 Å. 3 This covers the polar catalytic chamber and the inlet region of the hydrophobic prenyl channel; --num_modes 20 outputs the top 20 docking conformations with the best scores; --exhaustiveness 8 sets the thoroughness of the conformation search; --cnn_scoringrescore uses GNINA's built-in convolutional neural network (CNN) scoring function for re-scoring; Step 4: Conformation Selection and Distance Measurement (1) Objective conformation screening criteria: From the 20 docking conformations output by each receptor-ligand combination, the following quantitative criteria were used for screening: CNN rating selection: The top 5 constructs ranked by GNINA CNN rating were selected as the candidate set; Catalytic geometry constraint screening: Conformations that meet the following conditions are selected from the candidate set: The distance between the carboxyl carbon atom and the D78 / D84 carboxyl oxygen atom in DHNA ranges from 2.5 to 6.0 Å. The pyrophosphate groups of the FPP are oriented towards the inner side of the catalytic chamber, and the hydrophobic isopentenene chain extends into the hydrophobic prenyl channel. (2) Precise measurement of critical distances: The critical interatomic distances of the selected optimal docking conformations were measured using PyMOL software. The results are summarized below: Wild-type docking results establish a pre-catalytic conformational baseline: DHNA→ASP84: 3.2 Å DHNA→ASP78: 4.4 Å TYR113→FPP: 7.2 Å The docking results of the Y113F mutant showed that: DHNA→ASP84: 4.5 Å DHNA→ASP78: 3.3 Å (1.1 Å shorter than WT) PHE113→FPP: 6.7 Å (shorter distance, enhanced hydrophobic interactions) The docking results of the V89I mutant showed that: DHNA→ASP84: 5.1 Å DHNA→ASP78: 2.7 Å (significantly reduced by 1.7 Å compared to WT) TYR113→FPP: 6.9 Å (distance shortened) (3) Analysis of the conformational rearrangement mechanism: The above docking results reveal the molecular mechanism of the dual-substrate conformational rearrangement: the DHNA conformation in wild-type MenA is biased towards binding ASP84 (3.2 Å vs 4.4 Å); the Y113F and V89I mutations optimize the spatial filling of the hydrophobic prenyl channel, making the FPP position closer to the 113 residue (the distance is shortened to 6.7 Å and 6.9 Å, respectively). This spatial adjustment of the hydrophobic channel induces conformational inversion of DHNA through allosteric effect, changing it from being biased towards ASP84 to being deeply anchored near ASP78 (the distance is optimized to 3.3 Å and 2.7 Å, respectively), achieving a significant optimization of the pre-catalytic conformation of the dual-substrate. The molecular docking results are shown in […]. Figures 10-12 .

[0057] Mechanism analysis: In wild-type MenA, the conformation of DHNA is more biased towards binding ASP84 (3.2 Å). The Y113F and V89I mutations optimize the hydrophobic channel (reducing polar interference or increasing spatial stacking), allowing the long-chain substrate FPP to be positioned closer to site 113 within the channel. This subtle adjustment at the tail of the hydrophobic channel synergistically reshapes the spatial arrangement of the two substrates within the catalytic chamber through an allosteric effect, inducing significant conformational inversion and rearrangement of DHNA. This causes DHNA to detach from ASP84 and deeply anchor itself near another key catalytic residue, ASP78 (distances optimized to 3.3 Å and 2.7 Å, respectively), thus achieving significant optimization of the pre-catalytic conformation of the two substrates.

[0058] Example 6 Construction and fermentation validation of recombinant strains: Step 1: Design and Construction of Mutant Gene Fragments (1) Codon design for target mutation sites Based on the computational analysis results of Examples 3 and 4, site-directed mutagenesis design was performed on the wild-type menA gene of Bacillus natto (corresponding to the NCBI RefSeq: NP_391728.1 coding sequence): Y113F mutation: The tyrosine codon TAC at position 113 is mutated to the phenylalanine codon TTC. V89I mutation: The valine codon GTG at position 89 is mutated to the isoleucine codon ATC. The above codon mutation design follows the principle of codon usage preference, selecting synonymous codons that are frequently used in Bacillus natto to ensure the normal expression of the mutant protein.

[0059] (2) Site-directed mutagenesis PCR operation Using a whole plasmid amplification site-directed mutagenesis method, primer pairs containing the target mutation were designed using a plasmid containing the wild-type menA gene as a template. Y113F mutant primer design: Forward primer: 5'-CAATTAGCTCTTGCCTCATTTGGGATTGCCATTTTG-3' (mutation site TTT is indicated in bold) Reverse primer: 5'-AAATGAGGCAAGAGCTAATTGCAAAATCGTTTTAGGC-3' (complementary sequence) V89I mutant primer design: Forward primer: 5'-GATACAGCAGAATCAATCGGAATCGGAGGGGC-3' (mutation site ATC is indicated in bold) Reverse primer: 5'-GATTGATTCTGCTGTATCTAATCCGCGTTTAAAATC-3' (complementary sequence) PCR reaction system and conditions: High-fidelity DNA polymerase (such as Q5 or Phusion) was used. The reaction system was 50 μL, containing 50 ng of template plasmid, 1 μL each of primers (10 μM), and 25 μL of polymerase premix. PCR program: 98℃ pre-denaturation for 30 s, then 98℃ denaturation for 10 s, 60℃ annealing for 30 s, and 72℃ extension (calculated at 1 kb / 30 s), for a total of 35 cycles, and finally 72℃ extension for 5 min.

[0060] After the PCR product was digested with DpnI to remove the methylated template, it was transformed into E. coli DH5α competent cells. Clones were picked and Sanger sequencing was performed to verify that the mutation site was correct and there were no other unexpected mutations.

[0061] Step 2: Expression vector construction and recombinant plasmid preparation (1) Selection of expression carrier The shuttle expression vector pHT43 or an equivalent vector suitable for Bacillus natto was selected, and this vector possesses the following key functional elements: Promoter: Pspac or P43 strong promoter to ensure efficient expression of the target gene; Replication origin site: bifunctional ori for Escherichia coli and Bacillus subtilis, supporting stable replication in both hosts; Selection marker: Chloramphenicol resistance gene (cat), used for screening recombinant strains; Multiple cloning sites: Contain commonly used restriction endonuclease sites such as BamHI and HindIII.

[0062] (2) Recombinant plasmid construction process The correctly sequenced menA mutant fragment was digested with two restriction endonucleases (BamHI / HindIII recommended) and then ligated with the pHT43 vector, which had undergone the same restriction enzyme digestion, using T4 DNA ligase. The ligation product was transformed into E. coli DH5α competent cells, and positive clones were screened on LB plates containing chloramphenicol (25 μg / mL).

[0063] Plasmid DNA was extracted from candidate clones, and the insert size was verified by double enzyme digestion electrophoresis and confirmed by full-length sequencing to ensure: The target mutant gene was correctly inserted into the expression vector; The mutation site sequence is accurate; The reading frame is complete, and the promoter is correctly connected to the gene. The recombinant plasmids pHT43-menA-Y113F (nucleotide sequence, SEQ.NO.2) and pHT43-menA-V89I (nucleotide sequence, SEQ.NO.1) were finally obtained, and the control plasmid pHT43-menA-WT (nucleotide sequence, SEQ.NO.3) containing the wild-type menA gene was constructed.

[0064] Step 3: Construction of recombinant strains The recombinant plasmids were introduced into competent Bacillus natto cells using electroporation. The procedure was as follows: competent cells were mixed with plasmid DNA, and an electrical pulse was applied in the electroporator (parameters: 2.5 kV, 25 μF, 200 Ω). SOC recovery medium was immediately added, and the cells were incubated at 37°C for 1 h. The resulting culture was then spread onto LB solid medium containing chloramphenicol (5 μg / mL).

[0065] After incubation at 37℃ for 12-16 h, single colonies were picked, and the correct integration of the recombinant gene was verified by colony PCR and plasmid extraction. Finally, recombinant strains expressing different MenA variants were obtained: BSN-WT: Control strain expressing wild-type MenA; BSN-Y113F: Strains expressing the Y113F mutant MenA; BSN-V89I: Strains expressing the V89I mutant MenA.

[0066] Step 4: Strain activation and standardized preparation of seed culture (1) Strains resuscitation and activation The recombinant bacterial strains stored in glycerol tubes at -80℃ were streaked onto LB solid medium containing the corresponding antibiotics and incubated overnight at 37℃ until single colonies appeared. Morphologically normal single colonies were picked and inoculated into 5 mL LB liquid medium (containing 5 μg / mL chloramphenicol) and cultured at 37℃ with shaking at 250 rpm for 8–12 h to obtain primary seed culture.

[0067] (2) Standardization of seed culture and viable cell count Transfer the primary seed culture at a 2% (v / v) inoculation rate to 50 mL of fresh LB medium and incubate with shaking at 37°C and 250 rpm until mid-logarithmic growth. Monitor cell growth using a spectrophotometer; when OD... 600 Stop culturing when the value reaches 1.5-2.0; at this point, the viable cell count is approximately 2 × 10⁻⁶. 8 -5×10 8 CFU / mL, as a standardized seed liquid for fermentation inoculation.

[0068] (3) Standardized instructions for fermentation inoculum size In this embodiment, "inoculation amount 8%-10% (v / v)" specifically refers to the aforementioned standardized seed solution (OD). 600 = 1.5-2.0, viable count 2×10 8 -5×10 8 The volume ratio of CFU / mL to the fermentation medium was adjusted to ensure that the initial cell concentration was consistent across all groups of experiments, thus eliminating the influence of differences in inoculum size on the fermentation results.

[0069] Step 5: Fermentation Culture and MK-7 Product Extraction (1) Fermentation culture conditions The standardized seed culture was inoculated into the fermentation medium at a ratio of 8%-10% (v / v). The fermentation medium consisted of 50 g / L soybean peptone, 30 g / L glycerol, 3 g / L yeast extract, 5 g / L dipotassium hydrogen phosphate, and 0.5 g / L magnesium sulfate heptahydrate. The initial pH was adjusted to 7.0±0.2.

[0070] Fermentation control conditions: culture temperature 37-40℃, shaking speed 250 rpm, culture time 72 h. Cell growth and pH changes were monitored every 24 h during fermentation.

[0071] (2) MK-7 standardized extraction operation After fermentation, MK-7 extraction is performed according to the following standardized procedure: Extraction procedure steps: ① Take 5 mL of fermentation broth and add an equal volume of n-hexane (5 mL); ② Vortex and mix for 10 min to ensure complete extraction; ③ Centrifuge at 4000 rpm for 5 min to completely separate the organic and aqueous phases; ④ Carefully aspirate the upper organic phase and transfer it to a new centrifuge tube; ⑤ Add another 5 mL of n-hexane to the remaining aqueous phase and repeat the extraction twice; ⑥ Combine the organic phases from the three extractions, with a total volume of approximately 15 mL; ⑦ Dry the organic phase to near dryness using a nitrogen stream in a 45℃ water bath; ⑧ Add 1 mL of methanol to reconstitute, vortex and mix well, then filter through a 0.22 μm organic filter membrane.

[0072] Extraction efficiency control standard: The organic phase color of the third extraction is close to colorless as the criterion for judging complete extraction, or the extraction recovery rate is verified by adding MK-7 standard of known concentration to verify that it is ≥90%.

[0073] (3) HPLC quantitative detection Chromatographic conditions: A C18 reversed-phase column (250 mm × 4.6 mm, 5 μm) was used. The mobile phase was prepared by methanol and dichloromethane at a volume ratio of 4:1. The flow rate was 1.0 mL / min, the column temperature was 30℃, the detection wavelength was 248 nm (the maximum absorption wavelength of MK-7), and the injection volume was 10 μL.

[0074] Quantitative analysis: A series of standard solutions (0.1, 0.5, 1.0, 2.0, 5.0, 10.0 μg / mL) were prepared using high-purity MK-7 standard (purity ≥98%). An external standard quantification curve was established, and the linear correlation coefficient R was required. 2 ≥0.999. The MK-7 content was calculated by comparing the sample peak area with the standard curve.

[0075] Experimental design: At least 3 biological replicates were set up for each strain group, and 3 technical replicates were performed for each sample, with the BSN-WT strain expressing wild-type MenA as the control group.

[0076] Step Six: Experimental Results and Verification Quantitative results: Experimental results showed that, under the same fermentation conditions, compared with the control strain expressing wild-type MenA, the MK-7 yield of the Y113F mutant strain was increased by 25.3±2.1%; and the MK-7 yield of the V89I mutant strain was increased by 30.7±1.8%. The above results validate the effectiveness of the MenA rational design strategy based on artificial intelligence algorithms, demonstrating that the Y113F and V89I mutations significantly improve the biosynthetic efficiency of MK-7 by optimizing the hydrophobic channel gating mechanism. Relevant HPLC chromatograms are shown below. Figure 13 , Figure 14 As shown.

[0077] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the present invention.

Claims

1. A smart design method for the membrane protein MenA, characterized in that: include, Topology and structure prediction: Obtaining the amino acid sequence of MenA protein from Bacillus natto; At least two transmembrane topology prediction tools were used to predict the transmembrane helix of the amino acid sequence, and the transmembrane helix positions and intracellular / extracellular orientations with consistent prediction results were retained. A three-dimensional structural model of MenA was created using a deep learning protein structure prediction model, and intramembrane orientation correction was performed using a membrane protein orientation prediction tool to obtain the intramembrane corrected structure. Channel identification and gated site selection: Based on the obtained intramembrane correction structure, channel analysis software was used to identify polar catalytic channels and hydrophobic prenyl substrate channels inside the protein, and the length, bottleneck radius and channel lining residues of each channel were obtained. Multiple sequence alignment and co-evolutionary analysis were performed on MenA to obtain residue pairs with co-evolutionary scores higher than a preset threshold. The evolutionary conservation of each residue was classified using conservation analysis tools; Based on the combined results of channel analysis, co-evolutionary analysis, and conservation analysis, residues located inside hydrophobic channels, with moderate conservation, and directly involved in substrate binding were selected as candidate sites for channel gating engineering. Stability calculation and mutant design: For the identified candidate sites, construct single-site or multi-site mutant structural models; use protein free energy calculation tools to evaluate the folding free energy change (ΔΔG) of each mutant, and screen for stability-enhancing mutants with ΔΔG less than 0 kcal / mol.

2. The method as described in claim 1, characterized in that: The multiple sequence alignment and co-evolutionary analysis of MenA includes: using the MenA amino acid sequence as the query sequence, automatically searching in a homologous sequence database, removing redundant and low-coverage sequences to construct multiple sequence alignments, establishing a statistical mechanical model based on the multiple sequence alignments to calculate the long-range evolutionary coupling score of all residue pairs, and identifying residue pairs with coupling scores higher than a preset score threshold and spatial distances less than a preset distance threshold in the intramembrane correction structure as co-evolutionary residue pairs with important functional interactions. Among them, the co-evolutionary analysis constraint rules are: The database is selected from UniRef100 or an equivalent homologous sequence database; Sequence selection criteria: E-value ≤ 1 × 10 -5 Coverage rate ≥70%; Co-evolution criteria: Coupling score > 10 and spatial distance < 4 Å; The model is a Potts statistical mechanics model based on multiple sequence alignments; The method of using conservation analysis tools to classify the evolutionary conservation of each residue includes: taking the intramembrane correction structure and corresponding amino acid sequence as input, automatically searching for homologous sequences to construct a phylogenetic tree and calculating the evolutionary rate of each residue; and normalizing the conservation of residues into levels 1-9 according to the standardized classification system built into the conservation analysis tool, where level 9 represents high conservation and level 1 represents high variability. Among them, the conservative analysis constraint rules are: The algorithm is selected from Bayesian methods for calculating evolutionary rates; Homologous sequence parameters: 100-150 sequences, with identity ranging from 35% to 95%; Definition of moderate conservatism: 4-8 level range.

3. The method as described in claim 1, characterized in that: The rules for selecting candidate sites for channel gating engineering are as follows: The residues are located within the hydrophobic prenyl substrate channel liner: within 4.0 Å of the hydrophobic channel axis; The conservation level is in the moderate conservation range, meaning it is neither a highly conserved residue in the catalytic core nor a highly variable residue that is not functional: ConSurf score 4-8. There is a direct spatial association with the co-evolutionary high-resolution residue pairs: the spatial distance to the high-resolution residue pairs is <4 Å; Strictly exclude catalytic core, substrate recognition, and cofactor binding residues that are classified as the highest level of conservatism.

4. The method according to any one of claims 1 to 3, characterized in that: The process of screening for stability-enhancing mutants with ΔΔG less than 0 kcal / mol also includes molecular docking and channel gating verification, specifically: Construct the three-dimensional structure of DHNA and polyisoprene pyrophosphate FPP; Using the geometric center of MenA catalytic residues ASP78 and ASP84 as the docking cassette center, focused molecular docking was performed on wild-type and preferred mutants; We analyzed the direct binding distance and hydrophobic interactions between the mutation site and the FPP substrate, as well as the changes in the distance between DHNA and ASP84, to screen for target mutants that simultaneously enhance FPP binding and optimize the pre-catalytic conformation of DHNA.

5. The method as described in claim 1, characterized in that: The transmembrane topology prediction tools include DeepTMHMM and TOPCONS; The structure prediction model is a deep learning protein structure prediction model, including AlphaFold3 or a model with equivalent three-dimensional structure prediction capabilities. The membrane protein orientation prediction tool is a conventional membrane protein orientation prediction tool, including a PPM server or an equivalent membrane protein orientation correction tool; The channel analysis software is CAVER Analyst 2.0 or an equivalent tool that can output channel length, bottleneck radius, and lining residues; The tool for co-evolutionary analysis is EVcouplings or an equivalent tool; The conservative analysis tool is ConSurf or an equivalent tool; The tool for calculating protein free energy is FoldX or an equivalent tool; The molecular docking software is GNINA or an equivalent tool.

6. The method as described in claim 5, characterized in that: The polar catalytic channel has a length of 15-20 Å and a bottleneck radius of 1.5-2.0 Å. The lining residues include E75, D78, D84, R207, and R217, which are responsible for DHNA substrate entry and pyrophosphate ion release. The hydrophobic prenyl substrate channel has a length of 24-32 Å and a bottleneck radius of 0.8-2.0 Å. The lining residues include F79, V89, Y113 and their adjacent hydrophobic residues, which are responsible for the entry and positioning of the FPP hydrophobic chain. The candidate sites for gating engineering include F79, V89 and Y113, where F79 is phenylalanine at position 79 of the wild-type menA gene of Bacillus natto, V89 is valine at position 89 of the wild-type menA gene of Bacillus natto, and Y113 is tyrosine at position 113. The target mutants are Y113F and V89I. Y113F is a mutation in which the tyrosine codon TAC at position 113 of the menA gene of wild-type Bacillus natto is mutated to the phenylalanine codon TTC. V89I is a mutation in which the valine codon GTG at position 89 of the menA gene of wild-type Bacillus natto is mutated to the isoleucine codon ATC.

7. A MenA membrane protein, characterized in that: It has the following characteristics: (1) Its amino acid sequence is derived from the MenA protein of Bacillus natto; (2) An amino acid substitution occurs at at least one gated site, wherein the gated site is selected from F79, V89, and Y113; (3) The amino acid substitutions include one or a combination of the following: F79 is replaced with leucine, V89 is replaced with isoleucine, and Y113 is replaced with phenylalanine; In molecular docking simulations, the modified MenA forms a tighter channel binding space with FPP and induces conformational rearrangement of the substrate DHNA, making its distance from the catalytic residue ASP78 shorter than that of wild-type MenA.

8. A recombinant Bacillus natto, characterized in that: The recombinant Bacillus natto contains a recombinant gene encoding the modified MenA protein of claim 7; Furthermore, under fermentation conditions, the MK-7 yield of the recombinant strain was increased by at least 20% compared to the control strain expressing wild-type MenA.

9. The use of the modified MenA protein of claim 7 or the recombinant Bacillus natto of claim 8 in the preparation of vitamin K2 nutritional supplements, health foods or pharmaceutical preparations.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, implements the rational design method for the membrane protein MenA as described in any one of claims 1-6.

Images