(99f) Strategies for Renewable Muconic Acid Production from Lignin-Based Aromatics through Rational Metabolic Engineering of Pseudomonas Putida KT2440

The field of metabolic engineering is rapidly expanding, with a growing focus on using metabolic pathways for the production of biofuels and biochemicals. With the rising demand for sustainable energy sources and a push to reduce dependence on fossil fuels, it has become crucial to optimize these pathways. Lignocellulosic biomass is a promising source of renewable carbon for the production of biofuels and high-value chemicals due to their unique chemical properties, such as their high stability and resistance to degradation. Muconic acid is an important chemical intermediate that can be produced from lignin-based aromatics using microorganisms. It is suitable for polymerization into biobased polyesters since it can be converted through a single-step hydrogenation process to adipic acid, a common monomer in the high-volume product nylon-6,6. Bio-PET's primary building block, terephthalic acid, can be generated by transforming cis,cis-muconic acid into trans,trans-muconic acid via isomerization, followed by subsequent reactions. Pseudomonas putida is a highly promising bacterium for the industrial production of biofuels and biochemicals, due to its strong ability to tolerate toxic compounds, as well as its capacity to grow on a wide range of substrates. In order to enhance its cellular performance, curated genome-scale models of P. putida are developed along with large-scale stoichiometric and kinetic models (Tokic et al., 2020).

In this work we devise rational metabolic engineering strategies for the production of muconic acid from lignin-based aromatics using the bacterium Pseudomonas putida, with a focus on increasing the efficiency and yield of these processes. We develop large-scale kinetic models for biobased muconic acid production through lignin-based aromatics using the ORACLE (Optimization and Risk Analysis of Complex Living Entities) methodology and exploit machine learning methods to optimize parameterization of the kinetic models (Miskovic & Hatzimanikatis, 2010). The developed large-scale kinetic models are used to derive engineering strategies to manipulate the genetic makeup of Pseudomonas putida KT2440 and are then tested in the bioreactor simulations, thus mimicking real-world conditions.

In order to achieve the objectives of this project, a multi-disciplinary approach is followed, combining systems biology, metabolic engineering, and bioprocess optimization. The proposed workflow outline is based on the Optimization and Risk Analysis of Complex Living Entities (ORACLE) methodology, which is to be applied, with the use of the SKiMpy package, to the genome scale model of P. putida KT2440 to develop large-scale kinetic models for biobased muconic acid production through lignin-based aromatics (Weilandt et al., 2022). In the first module of the systematic workflow, we integrate available omics and cultivation data acquired on the host organism, followed by constructing a population of large-scale kinetic models of P. putida. We then simulate the muconic acid yield different lignocellulosic feedstocks and postulate rational metabolic engineering strategies that will increase muconic acid production while increasing microorganism’s robustness. In the last module, we employ clustering and advanced analytics to visualize and identify key enzymes that increase robustness to various feedstocks and muconic acid productivity.

In a supplementary aspect of this project, the utilization of machine learning (ML) methods in the final phase of parameterizing the kinetic models is explored. In particular, the study uses ML techniques such as RENAISSANCE and REKINDLE to improve the parameterization of the kinetic models. RENAISSANCE (REconstruction of dyNAmIc models through Stratified Sampling using Artificial Neural networks and Concepts of Evolution strategies) is a generative machine learning framework for efficiently parameterizing large-scale kinetic models with dynamic properties matching experimental observations (Choudhury et al., 2023). Similarly, REKINDLE (Reconstruction of Kinetic Models using Deep Learning) is a deep-learning-based framework that can efficiently generate kinetic models with dynamic properties matching the ones observed in cells (Choudhury et al., 2022).

In summary, this project contributes to the advancement of sustainable biobased chemicals and assists in designing future biorefineries. Moreover, the knowledge and expertise acquired through this project are valuable for future biobased chemical research and development. The study's results may also facilitate the exploration of alternative feedstocks, microbial strains, and production pathways, thereby propelling the field of biobased chemical production forward.

References

Choudhury, S., Moret, M., Salvy, P., Weilandt, D. R., Hatzimanikatis, V., & Miskovic, L. (2022). Reconstructing Kinetic Models for Dynamical Studies of Metabolism using Generative Adversarial Networks. Nature Machine Intelligence, 4(8), 710–719.

Choudhury, S., Narayanan, B., Moret, M., Hatzimanikatis, V., & Miskovic, L. (2023). Generative machine learning produces kinetic models that accurately characterize intracellular metabolic states. BioRxiv (Cold Spring Harbor Laboratory).

Miskovic, L., & Hatzimanikatis, V. (2010). Production of biofuels and biochemicals: in need of an ORACLE. Trends in Biotechnology, 28(8), 391–397.

Tokic, M., Hatzimanikatis, V., & Miskovic, L. (2020). Large-scale kinetic metabolic models of Pseudomonas putida KT2440 for consistent design of metabolic engineering strategies. Biotechnology for Biofuels, 13(1).

Weilandt, D. R., Salvy, P., Masid, M., Fengos, G., Denhardt-Erikson, R., Hosseini, Z., & Hatzimanikatis, V. (2022). Symbolic kinetic models in python (SKiMpy): intuitive modeling of large-scale biological kinetic models. Bioinformatics, 39(1).