(673g) Computational Protein-Ligand Co-Design for Enhancing Extracellular Electron Transfer in Lactiplantibacillus Plantarum

While directed evolution (DE) of proteins has been tremendously successful in improving protein selectivity, specificity, expression, solubility, and stability, goal-oriented protein design remains a challenging problem. As the number of possible protein mutants is large, even when limiting potential mutation to a small number of residues, a deliberate design strategy is required. Rational design [1, 2] and DE are two useful experimental options that are now complemented by machine learning (ML) techniques that elucidate the protein-function landscape by computational means [3]. An equally important and challenging problem is engineering of ligands with specific structural motifs or chemical properties. Similarly to protein engineering, the space of small molecules is large, likely in excess of 10⁶⁰ structures. Generative deep learning models, either conditioned on properties of interest, or guided by optimization algorithms have been proposed as a promising strategy for identifying molecules with desired characteristics [4].

The prior approaches subscribe to the individual design paradigm, where proteins and small molecules are engineered individually. This paradigm is sufficient in many endeavors where the goal is to engineer an entity that will function within an existing system (e.g., a drug targeting a particular receptor, or an enzyme to increase the yield for a target molecule). However, emerging applications in bioelectronic sensing [5] and synthetic biosynthesis [6] are not necessarily bound by the constraints of existing systems and thus may benefit from the mutual co-design of ligands and proteins, where both are engineered together. Our hypothesis is that co-design offers greater engineering flexibility, and thus a higher likelihood of discovering highly functional protein-ligand combinations, compared to individual design alternatives. Unfortunately, there are currently no methods that allow effective exploration of this joint protein-ligand design space.

We explore this co-design paradigm by considering type II NADH:quinone oxidoreductase (Ndh2) and its interaction with quinones to power extracellular electron transfer (EET). EET has been previously harnessed to sense various chemicals, thereby enabling creation of inexpensive whole-cell bioelectronic sensors [7]. Ndh2 is a ubiquitous protein that exists in many organisms as a key component of the respiratory chain, but was also found to facilitate EET in L. plantarum when in the presence of exogenous quinones [8]. As the EET response is dependent on the type of quinone and other factors, Ndh2 presents an exciting opportunity for designing sensors for pharmacologically relevant quinones [7]. The complex dependency of Ndh2 variants and quinones underscores the need to consider the Ndh2 and the quinone in tandem if the objective is to design for maximum EET rate.

In this work, we develop a framework for exploring the co-design landscape of proteins and ligands and present its application to enhancing EET transfer when considering Ndh2-quinone interactions. Our method for co-design is composed of three major steps. First, we design a process for creating separate libraries of protein mutants and quinone molecule derivatives. The Ndh2 variants were constructed by enumerating all single-point mutations around the quinone binding site, while the ligands were made by iteratively adding functional groups, e.g., phenyl, hydroxy, and other groups, in up to 6 locations on 1,4-dihydroxy-2-naphthoic acid and menadione, two quinones with a significant effect on EET. Second, we implement two search strategies for exploring protein-ligand combinations. In individual design, either the protein or the ligand is allowed to vary, so we construct those protein-ligand pairs by selecting a protein or ligand from our libraries and combining it with a wildtype Ndh2 or one of the base quinones. In co-design, both the protein and ligand can vary, so we create those pairs by combining a protein variant and a derivative quinone sampled from the libraries. Finally, we develop a protocol for evaluating the goodness of fit between a quinone and Ndh2 using molecular docking with AutoDock Vina and predicted free energy of binding calculations with MM/GBSA. Our proposed protocol is designed to approximate the favorability of the Ndh2-quinone interaction, which we believe is a major contributor to EET activity, and we use it to identify protein-ligand pairs with greatest interaction favorability.

We report several findings. First, we demonstrate that our proposed co-design approach can consider a significantly larger number of protein-ligand pairs than protein- or ligand-centric individual design methods alone. Second, we show that the co-design approach identifies pairs with the highest interaction favorabilities, making it the superior alternative to traditional individual design. Finally, we present a number of promising Ndh2-quinone combinations and highlight their key features that we believe would help improve EET. By simultaneously considering both the ligand and the protein, we are able to design novel biochemical systems in ways that were not possible with prior methods.

References

1. Hellinga, H.W. Rational protein design: Combining theory and experiment. Proceedings of the National Academy of Sciences. 1997, 94, 19. doi: 10.1073/pnas.94.19.10015

2. Korendovych, I.V. Rational and Semirational Protein Design. Protein Engineering, Methods in Molecular Biology. 2018, 1685, 15-23. doi: 10.1007/978-1-4939-7366-8_2

3. Wang, Y., Xue, P., Cao, M., Yu, T., Lane, S.T., Zhao, H. Directed Evolution: Methodologies and Applications. Chem Rev. 2021;121(20):12384-12444. doi: 10.1021/acs.chemrev.1c00260

4. Gao, W., Coley, C.W. The Synthesizability of Molecules Proposed by Generative Models. J. Chem. Inf. Model. 2020, 60, 12, 5714-5723. doi: 10.1021/acs.jcim.0c00174

5. Shanbhag, M.M., Manasa, G., Mascarenhas, R.J., Mondal, K., Shetti, N.P. Fundamentals of bio-electrochemical sensing. Chemical Engineering Journal Advances. 2023, 16. doi: 10.1016/j.ceja.2023.100516

6. Bowie, J.U., Sherkhanov, S., Korman, T.P., Valliere, M.A., Opgenorth, P.H., Liu, H. Synthetic Biochemistry: The Bio-inspired Cell-Free Approach to Commodity Chemical Production. Trends in Biotechnology. 2020, 38, 7, 766-778. doi: 10.1016/j.tibtech.2019.12.024

7. Li, S., De Groote Tavares, C., Tolar, J.G., Ajo-Franklin, C.M. Selective bioelectronic sensing of pharmacologically relevant quinones using extracellular electron transfer in Lactiplantibacillus plantarum. Biosens Bioelectron. 2024 Jan 1:243:115762. doi: 10.1016/j.bios.2023.115762

8. Stevens, E.T., Van Beeck, W., Blackburn, B., Tejedor-Sanz, S., Rasmussen, A.R.M., Mevers, E., Ajo-Franklin, C.M., Marco, M.L. Lactiplantibacillus plantarum uses ecologically relevant, exogenous quinones for extracellular electron transfer. bioRxiv 2023.03.13.532228. doi: 10.1101/2023.03.13.532228