3.2 Investigation of the key reaction parameters
The compatibility of all involved catalysts and their reaction conditions, especially the solvent, is a critical factor that requires a compatibility window that is often difficult to meet in one pot (Gómez Baraibar et al., 2016). To construct the tandem reaction, each sub-step was pre-tested. The kinetics of Knoevenagel condensation in H2O was investigated. The highest concentration of1b was set as 50 mM due to the tolerance of the enzymes. As shown in Figure 2 and Table S2, the condensation efficiency markedly accelerated with increasing the concentration of 1b in first 1 hours, and significantly slowed down and reached equilibrium in 2 hours, due to a decrease in substrate concentration and reversibility of the Knoevenagel condensation. In addition, another possible reason may be because 1,3-diketones tend to form stable six membered cyclic enols resulting in lower reactivity in water (Hu, Guan, Deng, & He, 2012). To solve this problem, we considered the solvent in this system. Traditionally, Knoevenagel condensation is always conducted in DMF and DMSO because aprotic solvents may lead to the deprotonation of the active hydrogen substrate, which is why water is seemed as negative factor (Siebenhaar, Casagrande, Studer, & Blaser, 2001). It was understood that all enzymes need essential bound water, and that enzymatic activity in organic solvents depends on water content (C.-H. Wang, Guan, & He, 2011). Thus, we selected an organic solvent optimized by mixing with water to promote Knoevenagel condensation. As expected, the species of the solvent affected the yield obviously (Figure S1 and S2). 30% DMSO promoted the yield up to 35%, in contrast to the yield of only 20% in water. This shows that deprotonation is indeed vital in a reaction that is more facile in non-protonic solvents, as expected according to the mechanism of Knoevenagel condensation. Meanwhile, compared to classical Knoevenagel condensation in DMF combined with a molecular sieve, this sub-step presented a very low yield due to aqueous environment (C.-J. Li & Chen, 2006).
Although the condensation process is significantly different in presence and absence of DMSO, the gap may be narrowed when the condensation and reduction are synergistic. Firstly, the high cost of NADH requiresin situ regeneration, as a common cofactor cycling system, GDH was selected. 1b and EAA were chosen as substrates to conduct tandem reactions. As shown in Figure 3A, the addition of GDH greatly increases the yield. But 3b was not synthesized in GDH alone, while GDH promotes the Knoevenagel condensation. Meanwhile, all available NADH were used and 8% yield was obtained in NerA alone, indicating that NerA may accomplish the condensation of 1b and EAA and reduction of 2bindependently. To verify this conclusion, the reaction with NerA or GDH on their own were conducted, shown in Table S3. It was found that NerA on its own was able to facilitate the reaction, achieving a yield of 84%.
Then, the yield was determined in both water and 30% DMSO. As shown in Figure 3B, the yield in 30% DMSO achieved 53% in 1 hour, whereas only reached 36% in aqueous solvent. However, we noticed that the product3b grows linearly in both systems (Figure 3B), prolonging the reaction time and improving the loading of enzymes may compensate for the loss of yield due to the removal of organic solvent. To test the idea, the same amount of NerA (2.4 mg mL-1) was added in the aqueous phase, along with 30% DMSO and the reaction time was extended to reach equilibrium. As shown in Figure 3C, the yield in water reached 89% in 4 hours, still lower than that after 3 hours in 30% DMSO (90%). There was almost no residue of 2b in water, and the yield was the same as in 30% DMSO at 5 hours, so that condensation is a limiting step in water. In fact, some ERs, such as YersER, tend to reduce the E -isomer and the isomerization is required to reduce the isomeric mixture (Litman, Wang, Zhao, & Hartwig, 2018). Both theE and Z isomers formed by condensation of aldehydes with 1, 3-dicarbonyl compounds in this reaction can be effectively reduced by NerA. Molecular docking experiments showed that carbonyl oxygen atom in either of the acetyl or carboxylic acid group of 2b can form hydrogen bonds with H178 and N181 in catalytic pocket, and the side chain extends into the hydrophobic cavity formed by Y65 and Y356 (Figure 4). Therefore, the unsaturated double bond is correctly fixed and efficiently reduced. Next, 3 mg mL-1 NerA was added with 1b and EAA in aqueous solvent, as shown in Figure 3D, equilibrium was achieved in 3 hours in both systems. The yield in water was comparable with that of 30% DMSO (90% vs. 91%, respectively) while 30% DMSO encumbered the separation and purification of the products. In addition, using water as the medium has received considerable attention in organic synthesis due to its economic, environmental, safety, and other advantages (Patel, Sharma, & Jasra, 2008). Based on these results, a neat aqueous solvent was chosen.
Next, the dosage of NerA and the molar ratio of substrates were tested to maximize the final yield. Shown in Table 1, the yield was promoted by the addition of NerA. Furthermore, Figures 3C and 3D also showed that when the amount of NerA was increased to 3 mg mL-1, the formation of 3b was significantly accelerated compared with 2.4 mg mL-1, and more 2b accumulated in the first hour. Indicating that the increase of enzymes obviously promotes the condensation and reduction. However, 3 mg mL-1 of NerA is sufficient, since the yield of 3b is similar to that of 3.6 mg mL-1. To further optimize the reaction, we observed that the yield was maximized when the substrate molar ratio was 1:2. Finally, 3 mg mL-1 NerA and a substrate molar ratio of 1:2 were used.