Marijan Logarušić ¹ Karla Šubar ¹ Maja Nikolić ¹ Ana Tušek ¹ Anja Damjanović ¹ Mia Radović ¹ Ivana Radojčić Redovniković ¹ Polona Žnidaršič-Plazl ¹ Wolfgang Kroutil ² Marina Cvjetko Bubalo ^1*

• ¹ University of Zagreb, Zagreb, Croatia
• ² University of Graz, Graz, Styria, Austria

Deep eutectic solvents (DESs) have emerged as green solvents with versatile applications, demonstrating significant potential in biocatalysis. They often increase the solubility of poorly watersoluble substrates, serve as smart co-substrates, modulate enzyme stereoselectivity, and potentially improve enzyme activity and stability. Despite these advantages, screening for an optimal DES and determining the appropriate water content for a given biocatalytic reaction remains a complex and time-consuming process, posing a significant challenge. This paper discusses the rational design of DES tailored to a given biocatalytic system through a combination of experimental screening and computational tools, guided by performance targets defined by solvent properties and process constraints. The efficacy of this approach is demonstrated by the reduction of CO2 to formate catalyzed by NADH-dependent formate dehydrogenase (FDH). By systematically analyzing FDH activity and stability, NADH stability (both long-term and short-term stability after solvent saturation with CO2), and CO2 solubility in initially selected glycerol-based DESs, we were able to skillfully guide the DES screening process. Considering trade-offs between experimentally determined performance metrics of DESs, 20% solution of choline chloride:glycerol in phosphate buffer (ChCl:Gly80%B) was identified as the most promising solvent system for a given reaction. Using ChCl:Gly as a co-solvent resulted in an almost 15-fold increase in FDH half-life compared to the reference buffer and stabilized the coenzyme after the addition of CO2. Moreover, the 20% addition of ChCl:Gly to the buffer improved the volumetric productivity of FDH-catalyzed CO2 reduction in a batch system compared to the reference buffer. The exceptional stability of the enzyme in this cosolvent system shows great potential for application in continuous operation, which can significantly improve process productivity. Additionally, based on easily measurable physicochemical solvent properties and molecular descriptors derived from COSMO-RS, QSAR models were developed, which successfully predicted enzyme activity and stability, as well as coenzyme stability in selected solvent systems with DESs.