The molecular dynamics simulation of G. stearothermophilus α-amylase GsamyA (GenBank accession number AF032864.1) was conducted under the Amber99SB-ILDN force field. Initially, a reaction box was set up containing 38,651 TIP3P water molecules and 11 sodium ions. Subsequently, temperature equilibration (NVT) was conducted for 5 ns at 50°C, followed by pressure equilibration (NPT) for 1 ns. Finally, a 200 ns molecular dynamics simulation was performed using the NRI mathematical model, allowing for the analysis of residue dynamics. In the simulations mentioned above, the temperature coupling method is V-rescale, the pressure coupling method is Berendsen.

The mutagenesis based on the aforementioned amino acids along the motion pathway was conducted, respectively. The protein structure of the enzyme was predicted using C-I-TASSER (Contact-guided Iterative Threading ASSEmbly Refinement), an enhanced version of I-TASSER that offers high-accuracy predictions of protein structure and function (Zheng et al., 2021).

Proceeding further, we employed a state-of-the-art, deep learning tool based on the three-dimensional structure of the protein predicted above. This tool, MutCompute, is freely available for academic use at www.mutcompute.com. This tool, a 3D convolutional neural network trained to establish connections between amino acids and their neighboring chemical microenvironments, guided the identification of innovative gain-of-function mutations in the docking model. Briefly, the tool rebuilt the neural network architecture and then, to further improve the accuracy, incorporated hydrogen atoms to explicitly recapitulate local hydrogen bonding networks, and added biophysical annotations such as partial charge and solvent accessibility to each atom (Shroff et al., 2020).

The amino acids in the conformational motion pathway of enzyme were predicted by NRI model as described above. The top-10 possible pathways were selected, and imputed into the MutCompute tool to get the candidate mutations. Through this approach, we identified amino acid mutations with the highest potential to enhance both the activity and stability of the amylase enzyme.

The wild-type G. stearothermophilus α-amylase gene GsamyA was artificially synthesized based on the available sequence (GenBank accession No. AF032864.1) and subsequently cloned into the expression vector pET28a to generate the recombinant expression plasmid, pET28-amy. To introduce targeted mutations in the GsamyA amylase gene, two pairs of homology arm primers were designed utilizing the known coding sequence as a reference. The mutated sites of the amylase and the specific sequences for the inverse PCR primers were listed in Table 1. The point mutation was introduced into the gene by using inverse polymerase chain reaction (inverse PCR), with the recombinant plasmid pET28-amy as a template. The inverse PCR was conducted according to the description by Edelheit et al. (2009). Subsequently, a one-step cloning technique (One Step Cloning Kit, Vazyme Biotech, China) was conducted according to the description by the manufacture to form the recombinant plasmids carrying the mutant genes.

The plasmid was transformed into Escherichia coli BL21(DE3) by chemical methods, and the colonies containing amylase gene were selected from LB medium plate (containing 50 μg/mL kanamycin). For inducible expression of amylase, a single colony was picked into 5 mL LB medium, incubated at 37°C, 180 rpm for 6–8 h. Then the liquid culture was inoculated in 500 mL LB liquid medium (containing 50 μg/mL kanamycin) at a ratio of 1:100. The bacteria were incubated at 37°C until the A₆₀₀ value was 0.6. Then, 0.05 mmol/L IPTG was added to induce the expression of amylase gene, and the culture was induced at 16°C for 15 h. The bacteria were collected by centrifugation and resuspended in 20 mL lysis buffer (Tris-HCl, pH 7.5, 0.5 mmol/L MgSO₄, 0.1 mmol/L MnCl₂); After ultrasonic disruption of the bacterial cells, the supernatant containing the target amylase was collected by centrifugation and subsequently purified using a Ni-agarose column. The purified amylase was then analyzed by SDS-PAGE electrophoresis and its protein content was determined using the Bradford assay. The determination of amylase activity employed the 3,5-dinitrosalicylic acid (DNS) method to measure reducing sugar content. One unit (U) of amylase activity is defined as the amount of enzyme required to hydrolyze starch and produce 1 μmoL glucose per minute under specific assay conditions.

The hydrolytic activities of the wild-type and mutants towards different substrates, including soluble starch, amylopectin, pullulan, and potato starch, were measured. The reaction mixture containing 50 μL properly diluted enzyme and 500 μL substrates with 1.0% (w/v) concentration was incubated at 60°C for 5 min and then the amount of the reducing sugar liberated from the substrates was determined by the DNS method. The hydrolysis products of starch were analyzed by thin layer chromatography (TLC). In this process, the components were separated on silica gel plates with the developing solvent N-butyl alcohol, isopropyl alcohol, acetic acid and water with the ratio of Vn:Vi:Va:Vw = 7:5:2:4, and then visualized by phosphoric acid chromogenic agent. Enzyme kinetics measurements were performed using various concentrations (0.1–5.0 mg/mL) of soluble starch as substrate and the amylase activity after a 5-min reaction under optimal temperature. The kinetic parameters K_M, V_max, k_cat, and k_cat/K_M were obtained through data analysis using Lineweaver-Burk double reciprocal plots.

In order to ascertain the optimal temperature for enzymatic activity, the reaction mixtures described as above experiments were incubated at various temperatures ranging from 30°C to 90°C, and the enzyme activity was measured, respectively. To evaluate thermal stability, the enzyme solution was appropriately diluted and subjected to a 1-h treatment within the temperature range of 30°C to 90°C, and then the remaining enzyme activity was measured under the general conditions with 60°C temperature. To evaluate the half-life of mutants, the enzymes were incubated under 60°C, and the remaining activities were checked at intervals. Comparing to the highest activity, which was designated as 100% activity and illustrated as relative activity, the data were normalized. All of the values presented in graphs and tables are the means of three replicates.

Based on the available structural information, the protein channel of both the wild-type and mutants were examined. The analysis utilizes the CAVER software (Chovancova et al., 2012) for tunnel and channel analysis and visualization in protein structures. With the assistance of the PLIP website, the protein-ligand interaction profiler was detected and visualized (Adasme et al., 2021). The force maps of the interactions between adjacent sugar substrate and the corresponding residues surrounding the enzymatic cleavage site were showcased by using PyMOL.¹

The three-dimensional structure of G. stearothermophilus α-amylase revealed that the loop segments are primarily located between Domain A and B, as well as between Domain A and C (Table 2). A total of 23 loop segments were extracted from this amylase, with 17 segments in Domain A and 6 segments in Domain B. The catalytic residues D232, E262, and D329 involved in starch hydrolysis are all located within these loop regions and spatially positioned within an active pocket formed by Domain A and B. Within the vicinity of these catalytic residues, a total of 37 amino acids located within the loop structures contribute to the formation of the active pocket responsible for starch hydrolysis.

Based on the NRI model, this study analyzed the motion pathways of the amylase and extracted the top 10 most probable routes (>80% possibility) for further analysis (Figure 1; Supplementary Table S1). Among these 10 highly probable pathways, the first five amino acid positions remain constant and are as follows: P207- > L202- > V103- > P44- > Y364. Variability in amino acid positions occurs at the 6th, 7th, and 8th positions along these trajectories. At position 6th, variations primarily occur between C361, F363, and D366. At position 7th, variations occur among D341, K345, P346, I352, Y359, and P360. At position 8th, variations take place between L347, F351, T354, P355, Q356, and E357 residues. It indicates that amino acids involved in conformational changes within the enzyme molecule are mainly located at positions P207- > L202- > V103- > P44- > Y364, and the P44 is occupies a “hinge” position during conformational changes (Figure 1).

In order to improve the activity and stability of the enzyme molecule, the amino acids within the top 10 most probable motion pathways were analyzed by MutCompute,² and the mutants were designed with maximum possibility to acquire improved activity and stability (Table 2). As shown by Table 3, the first five amino acids along the motion pathway remain relatively conserved, with mutations primarily occur between the 2nd (L202I) and 4th (P44E) position. However, there is greater variability observed in positions 6th, 7th, and 8th. Combined the candidate mutants in the 10 pathways, this study focused on ten specific mutants: P44E, L202I, K345I, T354C, Q356P, Y359T, C361M, F363H, Y364W, and D366L.

The point mutation was introduced into the amylase gene by inverse PCR. After IPTG inducible expression of these mutant genes, were conBoth the wild-type and mutants achieved active expression (Figure 2A). The TLC analysis of starch hydrolysis products by both the wild-type and mutants revealed the presence of monosaccharides, disaccharides, and oligosaccharides with a degree of polymerization (DP) ranging from 1 to 5. Furthermore, no significant changes observed in the composition of products among different mutants (Figure 2B). The differences in enzymatic activities between the wild-type and mutants were evidenced using soluble starch, amylopectin, pullulan, and potato starch as substrates (Figure 2C; Table 4). Overall, the highest enzymatic activity was observed when amylopectin was utilized as the substrate, whereas the lowest activity was observed with pullulan. When potato starch was used as the substrate, mutants P44E, K345I, Y359T, and C361M exhibited significantly increased hydrolytic capacity compared to the wild-type by 95.0, 68.0, 74.0, and 33.0%, respectively. In contrast, mutants T354C, and F363H showed decreased hydrolytic capacity by 72.0% and 55.0%, respectively. These observations not only suggest possible structural heterogeneity among amylopectin, potato starch, soluble starch, and pullulan but also indicate distinct spatial differentiation among mutants particularly within the catalytic pocket.

The optimal temperature for the wild-type was found to be 60°C. Within the range of 50°C to 90°C, both the wild-type and mutants maintained over 80% of their maximum enzymatic activity, indicating that they are mid-temperature amylases. Mutants Q356P, C361M, and D366L exhibited a decrease in their optimal temperature to 50°C (Figure 3A). However, mutants L202I, K345I, and Y364W exhibited similar optimal temperature to the wild-type (Figure 3B). Conversely, mutants F354C and F363H demonstrated an increase in their optimal temperature to 70°C (Figure 3C). Mutants P44E and Y359T displayed a further elevation in their optimum temperature to 90°C (Figure 3D).

The thermal stability of enzyme molecules was investigated at various temperatures, revealing distinct patterns between the wild-type and mutants (Figure 4). At lower temperatures, both the wild-type and mutants displayed exceptional stability. After a 1-h treatment at 60°C, their residual enzymatic activity remained above 80%. However, when the temperature exceeded 60°C, significant variations in enzyme stability were observed. Following a 1-h treatment at 80°C, mutants C361M and D366L exhibited a decrease in activity by 36%. Subsequently, after treatment at 90°C for 1-h, the remaining enzymatic activity for mutants K354I, Q356P, C361M, and D366L dropped below 40% (Figure 4A), and the half-life for them were 7.5-, 8.0-, 4.0-, and 5.0-h, respectively, when incubated under 60°C which were significantly lower than the 18.2-h of the wild-type (Figure 4C). However, mutants P44EY, Y359T, and T354C demonstrated improved temperature stability. After 1-h of treatment at 90°C, mutant P44E maintained more than 90% of the enzyme activity, and the half-life was 25.0-h, which was significantly better than the wild-type (Figures 4B,D).

The kinetic parameters of the wild-type and mutants are shown in Table 5. Mutants P44E, Y359T, and D366L exhibited significantly increased substrates affinity. The K_M values decreased by 39.1, 54.3, and 33.7% compared to the wild-type, respectively. Mutants L202I, K345I, and F363H exhibited a decrease in substrate affinity, as indicated by an increase in K_M values of 14.1, 17.4, and 26.1%, respectively. Furthermore, mutants P44E, K345I, and Y359T exhibited an enhancement in catalytic efficiency (k_cat/K_M), with increases of 93.8%, 5.0%, and 27.2% respectively, compared to the wild-type.

Protein tunnels and channels play a crucial role in the flexibility, stability, and enzymatic activity of proteins. A comprehensive analysis of the protein channels in both the wild-type and mutants was conducted (Table 6; Figure 5; and Supplementary Figure S1). As illustrated in the figures, significant differences were observed between the wild-type and mutants with respect to channel quantity, radius, and length. The wild-type enzyme contains only one channel, while most of mutants displayed an increase in the number of channel (Figure 5). For instance, mutants P44E, K345I, and Y359T had 8, 5, and 7 channels, respectively. The highest number of channels was found in mutant C361M with up to ten distinct channels. However, mutant T354C appeared to lack any observable channels (Table 5). The lengths and radii of the channels were crucial factors influencing substrate and ion transport. Distinct differences in the length and the radii were evident both among different enzyme molecules and within individual enzymes across various channel regions. The wild-type has a channel length of 16.34 Å, while the longest channel was observed in mutant Y364W, measuring 45.02 Å. In contrast, mutant Q356P displayed shorter channels with a length of only 14.50 Å. Mutant P44E possessed the largest bottleneck radius at 1.00 Å, whereas mutant D366L had a significantly smaller channel bottleneck radius of only 0.71 Å (Table 5; Figure 5).

The interactions between substrates and amino acids within the active pockets of both the wild-type and mutants were analyzed, and the specific interactions occurring between the amino acids located before and after the cleavage site on the substrate molecule were illustrated (Figure 6; Supplementary Figure S2). In the wild-type enzyme, H236 forms an essential hydrogen bond with the −3 subsite, while E262, H328, and H106 each establish distinct hydrogen bonds with the −2 subsite. Y57, an aromatic amino acid, robustly interacted with the sugar ring at the −1 subsite to further enhance the interaction between substrate and enzymes. Mutants T354C, Q356P, and F363H introduce cysteine (C), proline (P), or histidine (H) as replacements for existing amino acids. It leads to a much more compacted enzyme structure overall due to the introduction of hydrophilic, basic amino acids or subamino groups (P) along the pathway, and formed a more complex hydrogen bonding network between amino acids in the catalytic center and substrate molecules (Figure 6). Thus might decrease the enzyme-substrate conversion rate, results and reduced enzymatic activity. Mutants P44E, K345I, and Y359T exhibit significant changes in the hydrogen bonding network within the catalytic center due to the substitution of more hydrophobic amino acids. These substitutions introduce both hydrogen bonds and hydrophobic interactions. Amino acid Y57 was becoming distant to the sugar ring (4 Å), and eliminated the aromatic stacking interactions with the sugar rings (Figure 6; Supplementary Figure S2).