A method for designing polyketide synthases based on computational experiment closed loop

Abstract

This application relates to the technical fields of synthetic biology and enzyme engineering, and in particular to a computational-experimental closed-loop rational design method for polyketide synthases. The method includes: constructing a full-length FCHS model using AlphaFold3; identifying binding hotspots and verifying dynamic stability through molecular docking and kinetic simulations; and constructing a low-redundancy mutant library containing A, L, and F substitutions. This application solves the problems of difficult protein structure analysis and limited experimental throughput through a closed-loop process of computational pre-screening, dynamic verification, and evolutionary filtering, providing an efficient paradigm for the improvement of complex multifunctional enzymes.

Description

Technical Field

[0001] This invention belongs to the field of synthetic biology and enzyme engineering technology, specifically a method for designing and applying polyketide synthases based on computational closed-loop experiments. Background Technology

[0002] Chalcone is a high-value core precursor of flavonoids, and its heterologous synthesis by microorganisms is a key strategy to overcome traditional production bottlenecks. FCHS is a multifunctional polyketide synthase (approximately 290 kDa) derived from filamentous fungi, possessing both coumaroyl-CoA ligase and chalcone synthase activities. Theoretically, it can achieve "one enzyme, multiple steps" catalytic synthesis of chalcone, with the potential to simplify the pathway and improve efficiency.

[0003] However, the practical application of FCHS faces multiple technical bottlenecks. First, there is a lack of structural information. FCHS has a large molecular weight and complex structure, and lacks homologous templates, resulting in an unknown three-dimensional structure. This leads to a lack of key information such as catalytic mechanisms and substrate binding modes, posing challenges to targeted functional modification. Second, rational design is hindered. Due to the lack of structural guidance, traditional directed evolution methods require the construction of ultra-large-capacity mutant libraries, which results in low throughput and is time-consuming and labor-intensive for proteins like FCHS. Conventional rational design methods based on sequence conservation also suffer from insufficient predictive reliability due to the limited number of homologous sequences for FCHS. Finally, enzymatic performance still needs optimization. The catalytic efficiency (kcat / Km) of wild-type FCHS still has significant room for improvement, requiring modification through protein engineering. However, currently, efficient and precise modification strategies are lacking.

[0004] US Patent 7680601B1 discloses a method for designing polyketide synthase genes. This method uses alphanumeric symbols to represent the structures of polyketide and PKS genes and utilizes recombination of known PKS gene modules or parts thereof to design new genes. This strategy based on module splicing and known sequence retrieval has some effect in constructing novel PKS genes. However, for multifunctional synthases like FCHS with unknown structures and a lack of homologous templates, relying solely on module recombination has limitations in optimizing substrate binding modes and improving specific catalytic efficiencies. Furthermore, it lacks an effective closed-loop feedback iteration for computational prediction and experimental verification.

[0005] Patent CN119790150A discloses a method for engineering non-ribosomal peptides and polyketide synthases, which uses conserved sequence motifs within T-domain genes as fusion sites to generate artificial enzymes. This method utilizes sequence conservation for functional element combination; however, when dealing with complex enzymes like FCHS with limited homologous sequences and unreliable conservation predictions, relying solely on conserved motif design strategies presents challenges in precisely regulating enzymatic performance. Furthermore, its effectiveness is limited for modifying proteins with atypical structures and a lack of template guidance.

[0006] Therefore, there is an urgent need in this field for a rational design method that can overcome the above-mentioned bottlenecks and is specifically designed for large and complex enzymes such as FCHS with unknown structures and lack of homologous templates, so as to achieve rapid and accurate optimization of their functions. Summary of the Invention

[0007] This application provides a computational and experimental closed-loop rational design method and application for polyketide synthases. By integrating high-precision structure prediction, multi-scale molecular simulation, evolutionary conservation analysis, and customized low-redundancy mutant library screening, site-specific modification of polyketide synthases FCHS with molecular weight in the range of 280 to 300 kDa is carried out to obtain mutants with enhanced catalytic activity.

[0008] In a first aspect, this application provides a rational design method for polyketide synthases based on computational and experimental closed-loop methods, comprising the following steps: S1: Computational modeling of protein 3D structure: Obtain the full-length amino acid sequence of FCHS protein, construct a 3D structural model of FCHS using the AlphaFold3 algorithm, and select structural conformations with a template modeling quality score pTM value greater than 0.5 and a local confidence score pLDDT value in the range of 70 to 100 as the initial model; S2: Substrate binding hotspot calculation and identification: Using the initial model in S1 as the acceptor and p-coumaric acid as the ligand, the search space is set as a 50 Å x 50 Å x 50 Å cube region containing all potential active sites. Flexible docking simulation is performed using a molecular docking program. The top 10 conformations are extracted in order of binding energy score from low to high. Amino acid residues within 4 to 6 Å from the center of the substrate molecule are identified as candidate mutation sites. S3: Calculation and verification of the dynamic stability of the complex: Molecular dynamics simulations were performed on the FCHS-substrate complex identified in S2. The temperature of the simulation system was set to 298 to 310 K, the pressure to 1 standard atmosphere, and the simulation duration to 100 to 200 nanoseconds. The root mean square deviation (RMSD), gyroscope radius (Rg), and free energy landscape of the protein backbone atoms in the simulation trajectory were calculated. Conformations with RMSD fluctuations greater than 3 Å were eliminated, and key amino acid residues at the lowest energy point with structural fluctuation differences in the range of 0.5 to 1.5 Å were identified. According to this application, full-length modeling of a protein of approximately 290 kDa was performed using AlphaFold3, solving the problem that conventional homology modeling cannot construct complex multifunctional enzyme structures when templates are lacking. Molecular dynamics simulations were used to observe the dynamic interaction between the substrate and enzyme at the nanosecond scale, eliminating spurious sites that may exist in static docking. Combined with the PSSM matrix, evolutionarily irreplaceable core functional residues were eliminated, precisely narrowing the mutation range from 2500 to 2700 residues in the full sequence to 3 to 8 core residues.

[0009] In some implementations, in step S1, the calculation process of AlphaFold3 includes inputting the FCHS sequence into the multiple sequence alignment (MSA) module, setting the number of iterations to 3 to 5, predicting the relative positions and orientations between residues through the structure module, and outputting a structure coordinate file in PDB format.

[0010] In some embodiments, in step S2, molecular docking is performed using AutoDock Vina or Glide software, with the exhaustiveness parameter set to 8 to 32, and the docking center coordinates set to the center of the domain in the FCHS protein that has chalcone synthase function.

[0011] In some embodiments, in step S3, the molecular dynamics simulation uses GROMACS or Amber software, selects AMBER99SB or CHARMM36 all-atom force field, adds a sodium chloride solution with a mass fraction of 0.85% to 0.95% to the simulation system to balance the system charge, and adopts periodic boundary conditions.

[0012] In some implementations, during step S4, the calculation of the position-specific scoring matrix PSSM is performed with the homologous sequence coverage set to greater than 70% and the sequence consistency set to 20% to 90%. The amino acid substitution frequency at each position is calculated through 3 to 5 rounds of iteration.

[0013] In some embodiments, in step S5, the selected target mutation site includes amino acid 412 in the FCHS protein sequence, which is mutated from leucine L to alanine A.

[0014] In a second aspect, this application provides a mutant that enhances the catalytic activity of polyketide synthase FCHS, said mutant being prepared according to the method described in any embodiment of the first aspect.

[0015] The technical solution of this application solves the structural analysis problem caused by the large molecular weight of FCHS proteins through the tight coupling of specific computational logic and experimental steps. It replaces blind random evolution with a targeted "ALF" small library strategy, significantly reducing the screening workload. Through computational screening from S1 to S4, the experimental scale is controlled within the throughput range of 96-well or 384-well plates, enabling targeted improvement of the performance of complex multifunctional enzymes.

[0016] The technical details of this application are further explained below.

[0017] In step S1, for the ultra-large protein FCHS (approximately 290 kDa), AlphaFold3 employs a diffusion model to directly predict atomic coordinates, unlike previous versions which predicted the distance distribution between residues. During execution, the system first performs three rounds of homologous sequence searches on the FCHS sequence, constructing a deep MSA (Mean Sequence Analysis). Then, it extracts co-evolutionary information between residues using a Transformer architecture. The generated model undergoes relaxation processing to eliminate atomic overlap and bond length anomalies. The selected model must meet the global model quality assessment metric pTM value between 0.55 and 0.85, representing the accuracy of the predicted relative positions between domains, providing a foundation for subsequent identification of cross-domain active pockets.

[0018] In step S2, since FCHS possesses dual activities as both a ligase and a synthase, the focus of molecular docking is on identifying the substrate-binding pocket within the chalcone synthase (CHS) domain. The p-coumaric acid ligand is placed within the predicted cavity by performing charge balance and energy minimization processing. The docking algorithm employs a Lamarckian genetic algorithm or Monte Carlo simulation, performing 1 million to 5 million conformational searches within the search space. The binding free energy is derived by calculating van der Waals forces, electrostatic attraction, and hydrogen bond contributions. Special attention is paid to side-chain residues within 3.5 to 5.5 Å ranges from the substrate carboxyl and phenolic hydroxyl groups; these residues anchor the substrate through hydrophobic interactions or hydrogen bond networks and are key geometric constraint points affecting the catalytic rate.

[0019] In step S3, the static docking model is placed in an explicit water model solvent chamber. The edge of the solvent chamber is at least 12 Å from the protein surface. The ionic strength of the system is adjusted to 0.1 to 0.15 mol / L by adding sodium or chloride ions. The system is first subjected to energy minimization, using the steepest descent method and the conjugate gradient method alternately for 1000 to 5000 steps. Subsequently, equilibrium simulations are performed for 500 to 1000 picoseconds each under isothermal-volume (NVT) and isothermal-bariostatic (NPT) ensembles. A conformation frame is extracted every 10 to 20 picoseconds from the formal production kinetics simulation trajectory. By analyzing the distance evolution between the substrate and active site residues, if a residue maintains a distance of less than 5 Å from the substrate for more than 90% of the simulation time, and the protein secondary structure at that location remains stable (verified by a change in the radius of gyration Rg of less than 5%), then that residue is identified as a high-reliability mutation hotspot.

[0020] The method described in this application is not only applicable to FCHS, but also to other enzymes with molecular weights in the range of 200 to 500 kDa and multi-domain cooperative relationships (such as NRPS / PKS hybrid enzymes). Through this closed-loop process of "computational pre-screening - dynamic validation - evolutionary filtering - simplified experiment", the protein engineering cycle that originally required months or even years can be shortened to 4 to 6 weeks, and high-precision fine-tuning of the catalytic center of complex enzymes can be achieved even in the absence of crystal structures.

[0021] In summary, the technical solution provided in this application overcomes the core obstacles in the rational design of polyketide synthases, such as structural deficiencies, insufficient evolutionary information, and experimental throughput limitations, through specific operational steps and parameter constraints. It provides high-performance enzyme elements and efficient modification paradigms for the biomanufacturing of chalcones and other high-value flavonoids.

[0022] In some implementations, the free energy landscape (FEL) calculated in step S3 is derived by projecting the simulated trajectory onto the first two principal components (PC1 and PC2) using the Boltzmann distribution function. Conformations located in energy deep pits with energy values ​​2 to 5 kcal / mol lower than the surrounding barriers are selected to ensure that the identified binding modes are thermodynamically stable.

[0023] In some embodiments, in step S5.2, the expression host is *Escherichia coli* BL21(DE3), the inducer is IPTG at a concentration of 0.1 to 1.0 mmol / L, and the induction temperature is 16 to 25°C. A lower induction temperature helps prevent the 290 kDa protein from forming inclusion bodies, promoting its folding into an active, soluble state.

[0024] This application establishes a universal engineering modification process for multifunctional enzymes with unknown structures and lacking homologous templates through the aforementioned meticulously designed computational and experimental procedures. This process deeply integrates the predictive capabilities of structural biology, the simulation capabilities of computational chemistry, and the experimental screening capabilities of synthetic biology, significantly improving the certainty and efficiency of biocatalyst development.

[0025] In some embodiments, the FCHS structure model predicted in step S1 contains a triplet of key cysteine, histidine, and asparagine residues in the active center of the CHS domain. The spatial arrangement of these residues in the model is highly similar to the known CHS crystal structure from plant sources, further verifying the accuracy of the modeling.

[0026] In some embodiments, the candidate mutation hotspots identified in step S2 also include residues located at key inflection points of the substrate channel. The side chain orientation of these residues determines the conformation of the substrate pair coumaroyl-CoA within the active site. By changing the polarity of these residues, the residence time of the substrate within the pocket can be adjusted, thereby affecting the catalytic frequency.

[0027] In some implementations, the gyration radius Rg analysis in step S3 shows that during the simulation, each structural domain of the FCHS maintains a compact overall conformation without obvious structural dissociation, proving that the full-length model generated by AlphaFold3 has reasonable integrity in terms of dynamics.

[0028] The method described in this application can effectively transform FCHS, a natural enzyme with great application potential, into a highly efficient industrial catalyst, laying a solid technical foundation for the green and efficient biosynthesis of chalcone.

[0029] In some embodiments, the mutant L412A exhibits excellent stability in the pH range of 6.0 to 8.0 and retains more than 80% of its initial activity after being incubated at 30 to 40°C for 24 hours. This enables it to adapt to complex microbial intracellular metabolic environments and reduce metabolic loss due to enzyme inactivation.

[0030] In some embodiments, the rational design method described in this application is not limited to the substitution of a single amino acid, but also includes increasing or decreasing the length of the loop region surrounding the active site. By calculating and simulating to predict the effect of loop region flexibility on substrate entry and exit, mutants with the deletion or addition of 1 to 3 amino acid residues are designed and synthesized, and their impact on catalytic efficiency is verified using the screening system described in S5.

[0031] In some embodiments, in step S5.2, to increase the soluble expression level of the protein, a molecular chaperone protein, such as GroEL, GroES, or the DnaK, DnaJ, GrpE system, is fused to the expression vector. The co-expression level of the molecular chaperone is regulated by promoters of varying strengths to match the synthesis rate of the FCHS mutant and reduce misfolding.

[0032] The technical solution of this application constructs a complete, closed-loop feedback protein engineering system through meticulous design and parameter control of each technical step. From the initial sequence input to the final mutant verification, each step is supported by specific technical means, eliminating the blindness of human selection and ensuring the reproducibility and scientific rigor of the modification results.

[0033] The rational design method based on computational and experimental closed loop provided in this application opens the door to enzyme structure and function research through high-precision modeling with AlphaFold3, achieves precise capture of catalytic processes through multi-scale computational simulation, maximizes experimental efficiency through evolutionary analysis and simplified library strategies, and finally successfully obtains FCHS mutants with significantly improved performance, providing key core components for the microbial heterologous synthesis of chalcone.

[0034] In some embodiments, the method further includes modifying the thermostability of the mutant. By calculating and predicting the flexible regions on the protein surface that are susceptible to heat denaturation, disulfide bonds or salt bridges are introduced, and the contribution of improved thermostability to chalcone production is verified using the closed-loop process described in this application.

[0035] The technical means and parameter ranges described in this application are all based on the results of extensive experimental exploration and computational optimization, and have clear feasibility and industrial application prospects. Through the implementation of this application, those skilled in the art can achieve rapid and precise evolution of complex, multifunctional polyketide synthases without possessing a crystal structure. Attached Figure Description

[0036] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0037] Figure 1 This is a schematic diagram of the overall process of a computational and experimental closed-loop rational design method for polyketide synthases provided in an embodiment of the present invention. Detailed Implementation

[0038] The various embodiments or implementation schemes in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments.

[0039] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0040] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0041] As described in the background section above, polyketide synthases (PKS), such as FCHS, have always been a challenge in the field of biocatalysis due to their large molecular weight (typically between 280 and 300 kDa) and the presence of multiple functional domains, making the determination and rational design of their three-dimensional structures difficult. Conventional experimental methods are insufficient to cover the mutational space of thousands of amino acid residues.

[0042] Based on this, this application provides a computational and experimental closed-loop rational design method and application for polyketide synthases, which integrates high-precision structure prediction, multi-scale molecular simulation, evolutionary conservation analysis and customized low-redundancy mutation library screening to achieve targeted modification of FCHS.

[0043] In a first aspect, this application provides a rational design method for polyketide synthase FCHS, comprising the following steps: S1: Computational Modeling of Protein 3D Structure: The full-length amino acid sequence of the FCHS protein was obtained, and a 3D structural model of the FCHS was constructed using the AlphaFold3 algorithm. Conformations with a template modeling quality score (pTM) greater than 0.5 and a local confidence score (pLDDT) between 70 and 100 were selected as initial models. During the computation, the FCHS sequence was input into the Multiple Sequence Alignment (MSA) module, with 3 to 5 iterations. The structure module predicted the relative positions and orientations of residues and output a PDB format structural coordinate file. AlphaFold3 directly predicted atomic coordinates using a diffusion model. The system first performed three rounds of homology sequence searching on the FCHS sequence to construct a deep MSA, and then extracted co-evolutionary information between residues using a Transformer architecture.

[0044] S2: Substrate Binding Hotspot Calculation and Identification: Using the initial model from S1 as the acceptor and p-coumaric acid as the ligand, the search space was set as a 50 Å x 50 Å x 50 Å cube containing all potential active sites. Flexible docking simulation was performed using a molecular docking program. The docking center coordinates were set to the center of the chalcone synthase domain in the FCHS protein. The top 10 conformations were extracted in ascending order of binding energy score, and amino acid residues within a range of 4 to 6 Å from the substrate molecule center were identified as candidate mutation sites. Molecular docking was performed using AutoDock Vina or Glide software, with exhaustive parameters set to 8 to 32.

[0045] S3: Complex Dynamic Stability Calculation and Validation: Molecular dynamics simulations were performed on the FCHS-substrate complexes identified in S2. The simulation system temperature was set to 298–310 K, the pressure to 1 atm, and the simulation duration to 100–200 nanoseconds. The root mean square deviation (RMSD), cyclotron radius (Rg), and free energy landscape of the protein backbone atoms in the simulated trajectory were calculated. Conformations with RMSD fluctuations greater than 3 Å were eliminated, and key amino acid residues at their lowest energy points with structural fluctuation differences within the range of 0.5–1.5 Å were identified. The simulations were performed using GROMACS or Amber software, with AMBER99SB or CHARMM36 all-atom force fields selected. A sodium chloride solution with a mass fraction of 0.85%–0.95% was added to the simulation system to balance the system charge, and periodic boundary conditions were used. The solvent chamber edge was at least 12 Å away from the protein surface.

[0046] In some embodiments, the target mutation site selected in step S5 includes amino acid 412 in the FCHS protein sequence, which is mutated from leucine L to alanine A.

[0047] The following describes embodiments of this application.

[0048] Example 1 This embodiment provides a rational design process for FCHS based on a computation-experiment closed loop and the construction of mutants: Structural modeling: The FCHS amino acid sequence was input into AlphaFold3, and MSA was iterated 5 times. The model pTM was 0.78 and the pLDDT mean was 85.4.

[0049] Molecular docking: Using p-coumaric acid as a substrate, AutoDock Vina docking was performed in the CHS domain of FCHS (center coordinates X: 45.2, Y: -12.3, Z: 18.5), and sites such as L412, F415, and V420 were identified as being less than 5 angstroms away from the substrate.

[0050] Kinetic simulations: A 200-nanosecond simulation was performed in GROMACS at a temperature of 300 K. Analysis revealed that L412 maintained a distance of 4.2 Å from the substrate over a period of 180 nanoseconds, with an RMSD fluctuation of 1.2 Å and an Rg stability of 3.8 nm.

[0051] Comparative Example 1 This comparative example uses the conventional homology modeling method to construct the FCHS structure: Using the Swiss-Model with known plant CHS crystal structures as templates (25% consistency), only partial structures could be constructed due to the lack of a full-length template. Subsequent docking identification sites exhibited drastic fluctuations in RMSD greater than 5 Å in kinetic simulations, making it impossible to pinpoint effective mutation sites.

[0052] In step S3, the free energy landscape (FEL) is calculated using the Boltzmann distribution function by projecting the simulated trajectory onto the first two principal components (PC1 and PC2). A conformation located in an energy deep well is selected, with an energy value 2 to 5 kcal / mol lower than the surrounding barrier.

[0053] In step S5.2, to increase the soluble expression level of the protein, molecular chaperone proteins, such as the GroEL or GroES systems, can be fused to the expression vector. The co-expression level of the molecular chaperones is regulated by promoters of varying strengths to match the synthesis rate of the FCHS mutant.

[0054] In some implementations, machine learning models are used to train the experimental data generated in S5.3 and S5.4. The physicochemical properties of the residues are used as input features, and the activity enhancement factor is used as the output label to build a support vector machine (SVM) or random forest (RF) regression model.

[0055] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A rational design method for polyketide synthase based on computational and experimental closed-loop approaches, characterized in that, Includes the following steps: S1: Computational modeling of protein 3D structure: Obtain the full-length amino acid sequence of polyketide synthase FCHS, construct a 3D structural model of FCHS using the AlphaFold3 algorithm, and select structural conformations with a template modeling quality score pTM value greater than 0.5 and a local confidence score pLDDT value in the range of 70 to 100 as the initial model; S2: Substrate binding hotspot calculation and identification: Using the initial model in S1 as the acceptor and p-coumaric acid as the ligand, the search space is set as a 50 Å x 50 Å x 50 Å cube region containing the active site. Flexible docking simulation is performed using a molecular docking program. The top 10 conformations are extracted in order of binding energy score from low to high. Amino acid residues within 4 to 6 Å from the center of the substrate molecule are identified as candidate mutation sites. S3: Calculation and verification of the dynamic stability of the complex: Molecular dynamics simulations were performed on the FCHS-substrate complex identified in S2. The temperature of the simulation system was set to 298 to 310 K, the pressure to 1 standard atmosphere, and the simulation duration to 100 to 200 nanoseconds. The root mean square deviation (RMSD), gyroscope radius (Rg), and free energy landscape of the protein backbone atoms in the simulation trajectory were calculated. Conformations with RMSD fluctuations greater than 3 Å were eliminated, and key amino acid residues at the lowest energy point with structural fluctuation differences in the range of 0.5 to 1.5 Å were identified.

2. The method according to claim 1, characterized in that, In step S1, the calculation process of AlphaFold3 includes inputting the FCHS sequence into the multiple sequence alignment (MSA) module, setting the number of iterations to 3 to 5, predicting the relative positions and orientations between residues through the structure module, and outputting a structure coordinate file in PDB format.

3. The method according to claim 1, characterized in that, In step S2, molecular docking is performed using AutoDockVina or Glide software, with the exhaustiveness parameter set to 8 to 32, and the docking center coordinates set to the center of the domain in the FCHS protein that has chalcone synthase function.

4. The method according to claim 1, characterized in that, In step S3, the molecular dynamics simulation uses GROMACS or Amber software, selects AMBER99SB or CHARMM36 all-atom force field, adds sodium chloride solution with a mass fraction of 0.85% to 0.95% to balance the system charge, and adopts periodic boundary conditions.

Images