3 ADP expulsion increases PcAPSK activity
To increase Pc APSK activity, the enzyme’s catalytic mechanism and structural characteristics were first studied. The catalytic process ofPc APSK can be divided into three steps (Figure 3A ): (i) ATP binds to the enzyme, causing the lid structure to close; (ii) APS reaches the binding site and reacts with ATP; and (iii) the lid structure is opened to first release PAPS and then ADP. If ADP is not released in time, a complex (Enzyme-ADP-APS) will be formed and hinders the catalysis progress. To study the interaction between Pc APSK and ADP, ADP was docked into the Pc APSK based on its crystal structure (PDB ID: 1M7H) (Figure 3B ), and a bottleneck of 6.4 Å in the release channel was identified using CAVER (Figure 3C ). Then, molecular dynamics (MD) simulation was performed to calculate the root mean square fluctuation (RMSF) values and conformation ofPc APSK. As shown in Figure 3D , two regions (A and B) displayed higher RMSF values, suggesting that these two motifs could undergo noticeable movement and influence protein conformation. In addition, MD simulations revealed a conformational change of the lid, which regulated the release of ADP by controlling the ”closed-open” movement (Figure 3E ). ADP was immobilized by residues, whose binding free energy reached -55.81 kcal·mol-1 in the closed conformation. These results proved that the release of ADP was unfavorable; hence, promoting it might speed up the catalytic reaction.
Based on the structure of ADP-APSK and MD simulations, an “ADP expulsion” strategy was designed to promote the release of ADP by (i) reducing the binding affinity of target residues, (ii) changing the conformation of the lid to widen the release tunnel. Thirty-four candidate residues in regions A and B were selected for alanine scanning (Figure 4A ). Six residues whose activity was ≥30% higher were identified: binding residues S36A (35.2%), K38A (74.7%), and T40A135.17 (57.2%) in region A, and lid constituent residues K151A (107.2%), D139A (43.6%), and G167A (67.3%) in region B. Then, two smart mutation libraries (A and B) were constructed and screened for NNK‐based site‐saturation mutagenesis (Figure 4B ). In library A, saturation mutation of K38 increased APSK activity by 5.31-fold in variant L1K38G, whereas iterative saturation mutation of K38 and T40 produced mutant L2K38G/T40S, which presented 12.12-fold higher activity. Notably, when mutation S36 was introduced into L2, resulting in variant L3S36G/K38G/T40S, a significant drop in activity, amounting to only 4.99-fold of the wild-type (L0), was observed (Figure 4B ). In library B, saturation mutation of K151 increased the activity by 4.50-fold in variant L4K151V, and saturation mutation of G167 increased it by 7.80-fold in variant L5K151V/G167I. Iterative saturation mutation of K151, G167, and D139 produced mutant L6D139V/K151V/G167I, which presented 19.65-fold higher APSK activity. Finally, the best mutants from the two libraries were combined to generate variant L7K38G/T40S/D139V/K151V/G167I. As shown in Table 2 , variant L7 exhibited a 31.89-fold increase in kcat , from 0.61 to 19.45 s-1·mM-1, and a 45.74-fold higher specific activity (48.94 U·mg-1) compared to L0.
To explain the improved Pc APSK activity from a mechanism point of view, binding affinity and release channel width were compared between variants L0 and L7. In the latter, binding affinity for ADP was reduced due to an increase in binding energy from -55.82 to -21.57 kcal·mol-1, as calculated by MD simulations. As shown in Figure 4C , the average number of hydrogen bonds decreased from 31.1 to 27.3, which might have led to a weaker ADP–enzyme interaction. As a result, the root mean square distance increased from 0.55 to 0.63 Å (Figure 4D ), suggesting a reduced stability ofPc APSK. Analysis of the release channels revealed that the bottleneck in L7 increased by nearly 0.73 Å to 7.13 Å (Figure 4E ) and was more conducive to the release of ADP. In addition, the lid of the L7 variant changed to a more open state following engineering of the lid and hinge residues (Figure 4F ). These, caused a kink in the loop (from I162 to A170) that shifted two α-helix moieties to a more open state (0.52 and 0.55 Å, respectively). In particular, compared to the L0 variant, L7 showed an increase in RMSF of nearly 0.18 and 0.25 Å around regions A and B, respectively (Figure 4G ). These results indicate that the change in hinge residues led to a more flexible channel, which favored ADP release. In summary, the mechanism promoting the release of ADP might have resulted from a lower binding affinity due to weaker inner interactions, as well as increased flexibility of hinge residues, leading to a more open conformation of the lid and a wider release channel.