Abstract
Plant natural products are vital sources of medicines and health supplements, but plant extraction results in high costs and limited accessibility. Microbial production offers a sustainable alternative, yet poor stability and low activity of plant enzymes in heterologous hosts remain major bottlenecks. We developed REvoDesign(rational evolutionary-involved enzyme redesign), a rational evolutionary-guided workflow that integrates structural modeling with co-evolutionary data to optimize minimal key residues, improving both stability and activity while minimizing experimental workload and mitigating negative epistasis through cross-region mutation design. A user-friendly interface with customizable modules makes the tool easy to use. REvoDesign’s effectiveness was validated by improving bifunctional phytoene synthase/lycopene cyclase(CarRP) and taxadiene-5α-hydroxylase(T5αH), leading to collaborations with multiple research teams and companies, including a signed strategic partnership. This data-driven, structure-guided approach provides a powerful solution for engineering enzymes to enable efficient microbial production of plant natural products.
Introducation
According to the World Health Organization, both communicable diseases (malaria, hepatitis) and non-communicable diseases (cancer, cardiovascular disease) pose a serious threat to human health. Regarding communicable diseases, malaria will account for 263 million people worldwide (by 2023), and hepatitis will account for approximately 254 million (by 2022)（1,2）. Regarding non-communicable diseases, cardiovascular disease will account for over 330 million people in China alone, and cancer will be diagnosed with 20 million new cases worldwide in 2022, resulting in 9.7 million deaths (3-6). These diseases significantly increase mortality rates in impoverished areas due to a lack of medical resources and limited affordability.
Fig.1 Patients with major infectious and non-communicable diseases worldwide.
The United Nations' Sustainable Development Goal to ensure healthy lives and promote well-being for all at ages mentions the development and universal access to affordable vaccines and medicines (6).
For ages, traditional medical systems have relied on natural plant products to treat diseases (7).To date, numerous secondary metabolites with diverse structures and pharmacological activities have been identified in plant species (e.g., artemisinin, paclitaxel, notoginseng saponins, and vinblastine) (8,9). These components are not only the foundation of traditional therapies but also a vital source of modern innovative medicines (10).
Plant extraction is currently the primary method for producing natural plant products. This traditional production method is plagued by resource scarcity, reliance on arable land, and long growth cycles. To overcome these limitations and challenges, synthetic biology has become a green and efficient new resource production model, effectively controlling raw material supply while protecting natural resources and the environment (11-13).
Fig. 2 Basic route for the artificial evolution of natural enzymes into industrial enzymes based on rational design.
With the increasing application scenarios, decreasing costs, and technological advancements, the synthetic biology industry market is expected to expand rapidly, with broad market prospects. According to a report, the global synthetic biology market is expected to reach US$50 billion in 2028, with the healthcare and food and pharmaceutical sectors accounting for approximately 46.4%.
Fig. 3 Forecast of the global market size of synthetic biology.
The synthesis of natural products is often hampered by the poor stability and low activity of plant enzymes expressed in microbial hosts. This is because plant enzymes are highly evolved to adapt and function in the microenvironment of plant cells, which often results in poor physical properties when expressed in heterologous systems. This poses a significant obstacle to meeting the requirements for economically viable industrial-scale production of desired metabolites (14-15).Currently, enzyme modification strategies are mainly divided into two categories: directed evolution and rational design. Directed evolution relies on random introduction of mutations, high-throughput experimental screening, and iteration; rational design relies on spatial structure analysis and prior knowledge to guide the design of site-directed mutagenesis (16).
Fig. 4 Protein engineering strategies: directed evolution and rational design.
The tremendous progress in deep learning-based protein modeling and de novo protein design has accelerated our understanding of the complex interrelationships among enzyme sequence, structure, and function, which is expected to enhance the catalytic ability of enzymes (17). In addition, the rapid growth of protein databases has also promoted the study of enzyme evolutionary trajectories, which may provide a holistic perspective to guide protein design (18-21). Therefore, hybrid protein engineering methods have been developed that combine enzyme evolutionary insights with computational power. For example, PROSS, FuncLib, FireProt and HotSpot (22,23).
| PROSS | FuncLib | ProteinMPNN | |
|---|---|---|---|
| Crystal structure | YES | YES | No |
| Design site | Surface Backbone | Active Center |
Surface Active center Backbone |
| Library size | +++++ | +++++ | +++++ |
| Experimental burden | +++++ | +++++ | +++ |
| Dependance on computing resources | +++++ | +++++ | +++ |
| Time consumption | +++++ | +++++ | ++ |
| Computer knowledge | Few | Few | Many* |
| Major application field | Industrial enzyme for themostability | Industrial enzyme for themostability | Industrial enzyme for themostability |
Table 3 Comparison of protein design platform characteristics.
*Building your own service requires both strong computer and Rosetta knowledge. + represents the level, the more + represents the higher the level
Although these computational tools for protein optimization are promising, several challenges remain for enzyme engineering, especially for plant enzymes.Shortcomings of Current Protein Modification Tools for Plant Enzymes
Our YNNU-China 2025 team shares a common vision: to develop a suite of natural enzyme design tools with a graphical interface, high integration, modularity, and customizability, starting from the perspective of protein science beginners. This approach connects various published protein design tools and the data they generate. While achieving the goal of teaching introductory students about rational enzyme design, we also integrate or adapt these high-performance protein design tools as working components, lowering the barriers to secondary development, deployment, and use, and improving design efficiency. We also aim to decouple the ideal enzyme design tool from its service architecture, offloading computationally intensive tasks to the server and enabling interaction through a unified data or communication interface.
Our Solution：REvoDesign
This study developed an evolution-based enzyme redesign (REvoDesign) workflow to guide the rapid evolution of enzymes and improve the efficiency of plant enzyme engineering for natural product synthesis in microbial hosts. REvoDesign combines computational design tools with conservation and coevolutionary data to screen a limited number of residues from the enzyme surface and active center and construct minimal variants through amino acid clustering and stacking, significantly reducing experimental testing and synergistically improving enzyme stability and activity.
Fig.5 Overview of REvoDesign pipeline.
a Protein structure model. b Conservation information. c Coevolutionary information. d, e The constructed structure is used to identify hotspots for active center mutations and protein surface mutations, respectively, to improve enzyme activity and stability. f Cross-model filtering is used. g Sequence clustering. h Minimization experimental testing. i Iterative optimization of mutant library size. j Combinatorial mutations across the active center and surface regions are used to achieve a balance between enzyme activity and stability without triggering negative epistatic effects.
Furthermore, REvoDesign is built as a user-friendly Python package with a customizable framework, an intuitive interface, and a well-organized configuration system, making the tool easy to use. REvoDesign acts as a comprehensive assistance tool, requiring no extensive programming experience.
Application testing and Evaluation
To verify the applicability and versatility of the REvoDesign strategy, we designed lycopene cyclase in the lycopene biosynthesis pathway and taxadiene-5α-hydroxylase in paclitaxel biosynthesis as models for study. In the CarRP modification experiment, REvoDesign quickly identified key residues (F81A and Y145F) that influence lycopene cyclization and, through rational selection, preserved P domain function. The modified strain significantly reduced or even completely lost cyclization activity, resulting in efficient lycopene accumulation. The Y145F mutant achieved the highest yield, becoming the optimal substrate for subsequent fermentation.
In the T5αH optimization experiment, REvoDesign helped define active and stability sites, guiding mutant design and, combined with energy evaluation, selecting 23 candidate mutants. Furthermore, a combination of non-overlapping double mutations significantly increased T5αH yield, demonstrating the effectiveness of the multidimensional optimization strategy and validating the practicality and predictive reliability of REvoDesign for synergistic modification of activity and stability.
These results demonstrate that REvoDesign can rapidly identify key sites, design high-performance mutations, and achieve significant increases in target metabolites through combinatorial optimization, providing an efficient and scalable technical pathway for the heterologous synthesis of plant natural products.
Fig. 6 The complete process of REvoDesign protein design (design, construction, testing, and learning).
Expert evaluation: The robustness and effectiveness of this strategy give it broad application potential in engineering plant enzymes required for microbial production of target compounds.
Company evaluation: It is believed that REvoDesign has clear innovations and strong explainability, and can effectively solve industry pain points, especially in improving the adaptability of key enzymes and reducing the cost of high-throughput screening.
Summary and Outlook
Fig. 7 Flow chart illustrating the user-friendly graphical interface for REvoDesign workflow.
REvoDesign interface features a modular system in which the core functionalities and auxiliary tools are integrated, and is structured in line with enzyme engineering workflow.
In this study, we developed a computational enzyme redesign strategy, REvoDesign, that integrates structure-guided design and co-evolutionary information to improve the stability and activity of enzymes of varying structural complexity. REvoDesign primarily targets the protein surface and active center regions, engineering them to improve stability and activity, overcoming the trade-off between stability and activity and mitigating epistatic effects caused by spatially independent mutations. REvoDesign's performance was subsequently validated using two enzymes with distinct structures and catalytic mechanisms. The optimal single-point mutant established a novel lycopene biosynthetic pathway in yeast, providing an alternative route for industrial-scale production. Using heterologous lycopene production as an example, we integrated our research expertise to connect with interested companies and signed a strategic collaboration agreement with Meiqi Biotech. Furthermore, we are collaborating with multiple research teams, utilizing this tool to assist them in heterologous enzyme engineering, improving mutant design efficiency and significantly reducing screening workload.
In summary, we demonstrated that REvoDesign, combined with pathway engineering, provides a cost-effective strategy to enhance the metabolic flux of high-value plant natural products in heterologous hosts. From an industrial application perspective, enzyme engineering remains one of the most direct approaches to address scale-up bottlenecks, such as optimizing key enzymes for catalytic efficiency or environmental adaptability. We foresee that advances in new technologies and evolutionary insights gained from big-omics data will further accelerate the design of plant proteins, which holds great potential for future synthetic biology research.Our REvoDesign tool, a rational evolutionary enzyme redesign tool, provides foundational-advance in optimizing heterologous expression of plant enzymes.