Selected residues of Est924 were mutated by site mutagenesis using primers in Supplementary Table 1 with wild-type Est924 plasmid as a template. The introduction of single site mutations into Est924 was constructed using the QuickChange kit (Stratagene, CA, United States) by following the manufacturer’s instructions. The PCR product was incubated with DpnI (New England Biolabs, MA, United States) at 37°C for 1 h to degrade the parental template. Variants were transformed into Escherichia Coli BL21 (DE3) competent cells. Since naproxen is sold as the (S)-enantiomer, the recombinants were picked up and inoculated on lysogeny broth (LB) solid mediums containing 10 mM racemic naproxen ethyl ester and then cultured at 37°C for 12 h. We picked recombinants with translucent zones and those with transparent zones smaller than Est924. The standard expression and purification conditions were followed to obtain mutant enzymes. The racemic naproxen ethyl ester was subsequently used as substrate and the enantiomeric excess (ee) of products was analyzed by HPLC. The extracted product was purified by flash column chromatography with petroleum ether and ethyl acetate to separate acid and ester. The acid was esterified by the previously described method (Jin et al., 2003) and the ee of the product was analyzed using suitable column with the chiral stationary phase. (Supplementary Table 2) mounted on HPLC (Agilent, CA, United States) and UV detector at 214 nm (Godinho et al., 2012). The enantioselectivity of the esterase was expressed as the enantiomeric ratio (E) and calculation of E-value was performed using the method of Kim et al. (2017). Finally, the genes of the enzymes with increased enantioselectivity were verified by DNA sequencing (Sangon Biotech, Shanghai, China).

A single colony was picked up and grown in LB supplemented with kanamycin (50 μg/mL). The overnight culture was used to inoculate main cultures (800 mL LB medium containing 50 μg/mL kanamycin) to an initial OD600 of 0.05. Cells were grown at 37°C to an OD600 of 0.6, and IPTG was added to a final concentration of 0.1 mM. The cultures were incubated at 30°C for an additional 10 h, and cells were harvested by centrifugation at 10,800 g at 4°C for 20 min. Cell pellets were washed with a NaCl solution (0.9%, w/v) and frozen at −20°C for later purification.

The cell pellets were thawed on ice, resuspended in 10 ml lysis buffer (50 mM phosphate buffer containing 300 mM NaCl and 10 mM imidazole, pH 8.0) per gram of cell pellet and disrupted by sonication (Sonicator QSonica Q500 Ultra Sonicator) at 0°C. Cell debris was removed by centrifugation at 18,000 g for 30 min at 4°C. The supernatant was filtered through a 0.25 μm PVDF filter and was loaded onto a Ni-NTA column (Thermo scientific) equilibrated with lysis buffer. The column was then washed with six column volumes of wash buffer (50 mM phosphate buffer containing 300 mM NaCl and 20 mM imidazole, pH 8.0) after loading the filtered lysate. Finally, the protein was eluted with elution buffer (50 mM phosphate buffer containing 300 mM NaCl and 250 mM imidazole), and fractions containing target enzyme (determined by SDS-PAGE) were pooled and dialyzed at 4°C against Tris–HCl buffer (20 mM, pH 7.5). Protein expression and purity were assessed by SDS-PAGE (Supplementary Figure 1).

Esterase activity was measured using the standard assay: 50 μg purified esterase was reacted with 10 mM racemic ethyl 2-arylpropionates in 200 mM Tris–HCl buffer (pH 8.0) with 5% DMSO for 30 min at 30°C. The reaction was terminated by adjusting the pH to 2 using 6 M HCl and extracted with ethyl acetate. The organic layers were dried over anhydrous Na₂SO₄ and evaporated under vacuum. Conversion ratio was measured using NMR. The ee of products was analyzed by HPLC as described above.

Homology modeling of Est924, and M3 (I202F/A203W/G208F) were performed using the SWISS-MODEL¹ online system using esterase EstD11 (PDB id: 7AT3) as a template (sequence similarity: 49%) (Waterhouse et al., 2018; Miguel-Ruano et al., 2021). The structure of SsoEst1 was simulated in the same manner using 5LK6 as the template (sequence similarity 91.78%). The final models were validated by PROCHECK² and Verify 3D (Laskowski et al., 1996). Pymol was employed to visualize and analyze the 3D structure of Est924 and its mutants (DeLano, 2002).

Est924 and M3 were docked with (R) and (S)-naproxen using the Surflex-Dock module of Sybyl-X 2.0 program, respectively. All docking calculations were set to equal parameters (20 poses each), only using the pose that was highest ranked by Surflex-Dock. The conservative degree of residues was determined by the WebLogo 3.0 tool (Crooks et al., 2004).

After induction, cells were collected by centrifugation at 8,000 × g for 10 min, washed twice with 0.9% brine, and resuspended in the reaction mixture containing 50 mM potassium phosphate buffer (pH 8.0), 10 mM ketoprofen ethyl ester, and 5% DMSO to form cell suspension (OD600 nm = 3) with a total volume of 10 mL (in a 50 mL shake flask). The bioconversion reactions were performed at 30°C and 200 rpm. At different time points, 100 μL of sample was taken and mixed with 4 μL 6 M HCl and extracted with EtOAc. The resulting supernatant was centrifuged at 12,000 × g for 5 min and analyzed by HPLC as described above. For the recycling of the catalyst, the whole cell and supernatant are separated by centrifugation at the end of the batch reaction. Add a new reaction solution to the cell and restart the catalysis process.

It is widely known that bHSLs are promiscuous by nature (Pascale and Arnaud, 2012). For instance, esterase EH1 and EstD11 belong to the bHSL family and demonstrate promiscuous substrate specificity (Martínez-Martínez et al., 2018; Miguel-Ruano et al., 2021). Organic solvents and the alcohol moiety of esters often affect the enantioselectivity of esterases. Est25, EstD11, and Est924 have relatively low (R)-enantioselectivity to ethyl 2-arylpropionates, for this reason are the suitable template for the engineering of an (S)-enantioselective biocatalyst. Methanol, DMSO, and Triton-X100 are good cosolvents for Est924. However, the addition of methanol produces transesterified products that are difficult to be separated from the reaction mixture (data not shown). The foam formed by Triton-X100 is not conducive to industrial applications. Therefore, ethyl 2-arylpropionates and DMSO were selected as the substrates and cosolvent, respectively. The substrate specificity of Est924 was determined at 30°C in pH 8.0 buffer by using racemic ethyl (R, S)-2-arylpropionates as substrates. As shown in Figure 1 and Table 1, Est924 could hydrolyze all tested substrates, indicating that it had a wide substrate scope. In the pharmaceutical industry, there is an increasing demand for esterase-mediated chiral resolution of racemic 2-arylpropionate esters due to the high activity and low side effects of the (S)-enantiomers. The enantioselectivity of Est924 was subsequently determined using (R) and (S)-naproxen ethyl ester as substrates. Similar to Est25 (Kim et al., 2006) and SsoEST1 (Sehgal and Kelly, 2003), wild-type Est924 had activity toward (R)-naproxen ethyl ester and (S)-naproxen ethyl ester, with a slight preference for (R) -enantiomer (10%ee_R).

Enzymes with broad-spectrum substrate applicability generally owe this to their larger active pocket, similar to bHSL family esterases (Martínez-Martínez et al., 2018). According to the classification of the Esther database (Lenfant et al., 2013), 111 esterase structures of the bHSL family were resolved. We superimposed the crystal structures of Est25 (PDB: 4J7A) and 41 wild-type enzymes of bHSL family (Supplementary Figure 2). Three-quarters of bHSL family esterases had highly similar crystal structures, although the lowest sequence similarity observed between 5IQ3 and 3D7R was only 11.7% (Supplementary Table 3). The residue L255 of Est25 attracted our attention, because the variant Est25_L255W reversed the enantioselectivity of the racemic ketoprofen ethyl ester, resulting in (S)-ketoprofen with 80%ee_p (Yoon et al., 2014). Residue L255 was located at the loop between helix 10 and 11 with an isobutyl side chain extension to catalytic serine (Figure 2A). The enantiomer discrimination by enzymes is a very accurate mechanism, which often involves a few amino acids located at the active site. Based on structural information, about three-quarters of the 41 analyzed structures contain the same loop, indicating that it is conserved in the bHSL family (Supplementary Table 3). Alternatively, carboxylesterase SsoEST1 has the same residue (L198) as Est25 (L255), but selectively produces (S)-naproxen when racemic naproxen methyl ester was used as a substrate (Sehgal and Kelly, 2003). This indicates that other residues in the active pocket probably control the stereoselectivity of the α carbon. By docking (R)-naproxen ethyl ester into Est25 (Figure 2B), several residues around the Cα were selected to test whether they played a role in the stereoselectivity of the enzymes, due to the opposite stereoselectivity of Est25 and SsoEST1. As shown in Table 2, the residues at position A2 in all crystal structures were hydrophobic amino acids, and the residues at position B1 exhibited lower conservation (Figure 2C). Since residues at position A2 and position B1 directly extend their side chains toward the α carbon of substrates, their low conservation suggested that these residues might be responsible for the different stereoselectivity of enzymes.

Est924 can hydrolyze a variety of ethyl 2-arylpropionates, demonstrating the same enantioselectivity as Est25. We simulated the structure of Est924 using the crystal structure of EstD11 as a template (Figure 3A). The key residues of Est924 that might control its stereopreferences were analyzed. The selectivity of Est25 mutants with L255W to (S)-ketoprofen ethyl ester was significantly improved, because tryptophan has the largest side chain, which increases the binding steric hindrance of the (R)-enantiomer (Kim et al., 2017). Superposition of the Est924 and Est25_F72G/L255W crystal structure (5IQ3) revealed that the A203 with a small side chain in Est924 might be responsible for the low enantioselectivity. The Est25_F72G/L255W residue was transferred to Est924 by A203W substitution to investigate whether the variant Est924_A203W could make (S)-naproxen as the main product (Figure 3B, yellow residues). Similarly, the structure of SsoEst1 was also simulated and superimposed with the Est25_F72G/L255W. The imidazole group of H202 in SsoEst1 occupies the indole ring of W255, which might be the second key residue responsible for the enantioselectivity of the enzymes (Figure 3C, yellow residues). Multiple sequence alignment of reported bHSL family esterases with (R)-enantioselective resolution of 2-arylpropionates also showed that the B1 position was occupied by residues with smaller side chains (Figure 3D). Therefore, G208 at position B1 was also selected as a mutation site to prove our hypothesis. In addition, because the steric hindrance and hydrophobicity of I202 were lower than the phenylalanine in Est25 (F254), SsoEst1 (F197), and Est3 (F203), I202 was also selected for mutation. Moreover, F206 at position B2 of SsoEst1 may also be beneficial to its (S)-enantioselectivity, and the residue I212 of Est924 was replaced by phenylalanine.

The enzyme activity and enantioselectivity of esterases were measured using (R, S)-naproxen ethyl ester as substrate by HPLC. Firstly, A203 was mutated to tryptophan according to the structural analysis and results obtained by Kim et al. (2017). As predicted, variation of residue A203 significantly impacts enantioselectivity, and the variant M1 (A203W) turned to (S)-selective, producing (S)-naproxen with 65.6%ee_p. The variant Est924_G208F also reversed the stereoselectivity and hydrolyzed (S)-naproxen ethyl ester with moderate selectivity, yielding (S)-naproxen with 56.3%ee_p. However, the variant Est924_I212F with a larger hydrophobic side chain did not affect the enantioselectivity. These results indicated that both positions (204 and 208) were effective in improving the selectivity of Est924. Variant M2 with A203W and G208F was highly (S)-selective, producing (S)-naproxen with 95%ee_s. Finally, the variant with a mutation at I202 still retained the (S)-enantioselectivity as high as M2, and variant M3 (I202F/A203W/G208F) with 95.7%ee_s was obtained.

To clarify why the enantioselectivity of the variants was reversed and increased, we investigated the specific activity (mM/mg/min) for biocatalytic reactions using wild-type Est924, M1, G208F, M2, and M3 with (R)-naproxen ethyl ester and (S)-naproxen ethyl ester as substrates. Est924 showed the lowest enantioselectivity among esterases (Table 3, list 1), which mainly caused similar hydrolysis activity as compared to the enantiomers. M1 retained 90% activity of Est924 on (S)-naproxen ethyl ester, and 15% activity against (R)-enantiomer (Table 3, list 2). Therefore, the enantioselectivity of M1 was reversed and (S)-naproxen was formed as the main product. M3 showed the highest enantioselectivity among the esterases examined. It was almost inactivated in its ability to process (R)-naproxen ethyl ester, but still retained 80% of the activity of Est924 when (S)-naproxen ethyl ester was used as a substrate (Table 3, list 5). Subsequently, the structure of the optimal variant M3 was simulated and docked with (R) and (S)-naproxen ethyl ester, respectively. As shown in Figure 4A, (S)-naproxen ethyl ester fitted well with the active pocket but failed to obtain an effective docking of (R)-enantiomer with M3. Next, we docked (R) and (S)-naproxen ethyl ester into the structure of Est924, as shown in Figure 4B. Both the (R) and (S)-enantiomer pointed their naphthalene rings to the acyl binding site region. The distance between the protonated Nε atom of the active histidine residue and the ester oxygen atom (d_NE–O) was usually taken as a geometrical probe to indicate enantioselectivity (Henke et al., 2003). The d_NE–O of Est924 with (R) and (S)-naproxen ethyl ester were measured in the most favorable binding mode. The Δd_NE–O of (R)-naproxen ethyl ester (3.3Å) was 0.3Å shorter than (S)-naproxen ethyl ester (3.6Å), which corresponded to the slight (R)-stereopreference of Est924. We superimposed the modeled structure of M3 and Est924, and found that the α-methyl of (R)-naproxen ethyl ester was blocked by the steric hindrance of residue W203, which prevented it from being effectively bound. In addition, the molecular docking results also suggested that (S)-naproxen ethyl ester was not affected by the newly generated steric hindrance from large side chain amino acids in M3. However, mutations reduced the activity pockets of enzymes which were not conducive to the entry of substrates, so their activities on (S)-naproxen ethyl ester were reduced. This suggested that the enantioselectivity reversal of the enzyme was due to the steric hindrance effect that made it difficult for the enzyme to bind (R) -enantiomer, which was consistent with the results of specific activity tests.

In addition to (S)-naproxen, (S)-ibuprofen and (S)-ketoprofen are also widely used. A variety of esterases have been used to produce them through the resolution of corresponding racemic esters (Kim et al., 2002; Choi et al., 2003; Sathishkumar et al., 2010; López et al., 2014; Yoon et al., 2014; Memarpoor-Yazdi et al., 2018). Est924 could effectively hydrolyze both racemic ibuprofen ethyl ester and ketoprofen ethyl ester (Table 1). The optimal variant (M3) was tested for enantioselective resolution of these racemic esters. The results showed that M3 preferred to hydrolyze their (S)-conformation esters, generating (S)-ibuprofen (74.5%ee_s) and (S)-ketoprofen (95%ee_s), respectively. These results showed that we successfully generated Est924 variants with inverted and highly increased enantioselectivity toward ethyl (S)-2-arylpropionates and provided Est924 mutants with potential application in industrial biocatalysts. On the other hand, the low enantioselectivity of M3 to ibuprofen ethyl ester also suggested that the stereoselectivity of the enzyme to ethyl 2-arylpropionates was also regulated by other residues in the active pocket, especially the acyl binding site.

Compared with free enzyme and immobilized enzyme, whole-cell catalysts avoided the separation and purification of enzymes, which might be the better method for reducing the cost of the process. Ketoprofen ethyl ester and naproxen ethyl ester were selected as substrates due to the high enantioselectivity of M3. Control cells harboring the empty vector (in the absence of M3) were verified to catalyze the conversion of ketoprofen ethyl ester and naproxen ethyl ester was less than 1% within 24 h, which indicated that the whole-cell catalyst did not have undesired reactions. Besides, although the cell envelope of the whole-cell catalysts helps stabilize enzymes, it reduces the mass transfer efficiency of substrates and products. Considering the activity and recovery of the whole-cell catalyst, freeze-thaw treatment of E. coli cells was a common method (Schwaiger et al., 2021). Using cell lysate as a reference, the cells achieved about 81% of the relative activity and still maintained high (S)-enantioselectivity toward ketoprofen ethyl ester (95%ee_S) and naproxen ethyl ester (96%ee_S) after a single freeze-thaw cycle. The recyclability of biocatalyst was an important factor in the industrial application of the whole-cell biocatalytic process. As shown in Figure 5, the cells could be reused for at least five cycles without sacrificing the activity and enantioselectivity.