A Computational Biology Study on the Structure and Dynamics Determinants of Thermal Stability of the Chitosanase from Aspergillus fumigatus

Chitosan, derived from chitin, has poor solubility at high molecular weights, while chitosan oligosaccharides (COS) exhibit favorable properties (e.g., antibacterial, anticancer) suitable for applications in food, pharmaceuticals, agriculture, and environmental protection. Enzymatic degradation using chitosanases is preferred over chemical/physical methods. Chitosanases occur in GH5, GH7, GH8, GH46, GH75, and GH80 families; GH46, GH75, and GH80 are specific to chitosan. Many microbial chitosanases have optimal temperatures of 50–70°C but often poor stability at elevated temperatures, limiting industrial use. The study focuses on an Aspergillus fumigatus chitosanase (CAZy Q875I9; GH75) previously shown to produce DP2–DP5 COS with optimal pH 6.5 and temperature 65–70°C, yet remaining stable only below 55°C. Catalytic residues (Asp143 and Glu152 after signal peptide removal) have been identified by mutagenesis. The research question is how elevated temperature affects structural determinants of thermal stability and binding capacity. The authors hypothesize that high temperature induces structural changes at the binding site that hinder ligand binding and reduce catalytic stability. They employ AlphaFold2 structure prediction and molecular simulations to interrogate temperature effects on structure, dynamics, and binding energetics, aiming to guide stability-enhancing modifications.

The paper surveys chitosanase families (GH5, GH7, GH8, GH46, GH75, GH80) and their specificities, noting GH46/GH75/GH80 as chitosanase-specific. Prior work identified catalytic residues in GH75 fungal and viral members via mutagenesis (e.g., Asp175/Glu188 in Fusarium solani; Asp148/Glu157 in soil virus AMG), with conservation across GH75. The A. fumigatus enzyme (Q875I9) exhibits optimal pH 6.5 and temperature 65–70°C but loses stability above 55°C. Only one GH75 experimental structure (viral V-Csn) was recently solved, motivating the use of AlphaFold2 for structure prediction. The review emphasizes thermal stability challenges in glycoside hydrolases and the importance of structural determinants like hydrogen bonds, salt bridges, and disulfide bonds from related literature, as well as the role of MD simulations and MMPBSA in probing protein dynamics and binding.

- Structure prediction and validation: AlphaFold2 (ColabFold) predicted the chitosanase (signal peptide removed; 221 residues). Confidence assessed via predicted lDDT. Model quality evaluated using PROCHECK (Ramachandran analysis), VERIFY_3D (3D/1D compatibility), and PROSA (Z-score and per-residue energy). AlphaFold2 accuracy was also benchmarked on chitosanases from GH2 (2x05), GH8 (5xd0), GH46 (4olt), and GH80 (5b4s). - Protonation and binding site: Protonation states at pH 6.5 using H++ (notably ASP143 and GLU152 states). Binding site identified with DeepSite. DP6-chitosan protonation set to 50% (three amino groups protonated) based on its pKa and optimum pH. - Docking: AutoDock Vina 1.2.0 used to dock DP6-chitosan into the predicted binding site. - Molecular dynamics simulations: Systems built with AmberTools22; protein with ff19SB, ligand with GAFF2; OPC water model; neutralizing ions; cubic box with 10 Å buffer; periodic boundary conditions. Energy minimization, gradual heating to target temperature (1 ns) under 1 bar, equilibration monitoring potential energy and temperature, production runs with GROMACS 2022.1. Algorithms: SETTLE for water, LINCS for bonds to H, PME for electrostatics, V-rescale thermostat, Berendsen barostat during NPT equilibration, Parrinello–Rahman during production. Enzyme-only simulations: three replicates of 100 ns at each of 300 K and 350 K. Enzyme–ligand simulations: three replicates of 50 ns at 300 K for binding free energy. - Analyses: RMSD, RMSF (Cα), conversion to B-factors; radius of gyration; dynamic cross-correlation (Correlationplus) including distance-dependent scatter and cross-correlation matrices; PCA and free energy landscapes (first two PCs), clustering via Gromos to identify dominant conformations; hydrogen bond counts and probabilities (geometric criteria D–A ≤ 3.5 Å, H–D–A angle ≤ 30°) separated into intra- and inter-group for highly flexible regions; salt bridges (VMD saltbr plugin; O–N COM ≤ 4 Å) and probabilities; disulfide bond inference and tests by comparing simulations with and without disulfides; MMPBSA binding free energy (gmx_MMPBSA) total and per-residue decomposition to identify key contributors.

- AlphaFold2 structure and validation: Predicted model had high confidence (most residues lDDT >95). PROCHECK showed 89.4% residues in most favored regions; VERIFY_3D average 3D/1D scores all >0.2 (min 0.28); PROSA Z-score within acceptable range; per-residue energies negative. Structural alignment with viral GH75 V-Csn yielded RMSD 1.33 Å and near-identical binding site geometry and catalytic residue positioning (Asp143 and Glu152 ~6.0 Å apart). - Protonation analysis: H++ predicted pKa shifts suggest ASP143 pKa increased to 8.53 (from 3.86), GLU152 to 6.85 (from 4.25), implying roles opposite to prior expectation (ASP143 likely protonated/general acid, GLU152 deprotonated/general base). Strong local interaction network (ASP61, ASP63, TYR103, ASP143) contributed to ASP143 pKa elevation. - Dynamics at different temperatures: Overall RMSD stabilized after ~10 ns at both temperatures; fluctuations larger and more frequent at 350 K. Binding-site RMSD exceeded overall RMSD at both temperatures; difference modest at 300 K (~0.05 nm) but increased significantly at 350 K (~0.1 nm), indicating reduced site stability at high T. Radius of gyration remained ~1.58–1.62 nm; mean increased by ~0.01 nm at 350 K (<1%), indicating no global denaturation. - Flexibility mapping: At 300 K, loop 3 had highest RMSF (>0.5 nm), loop 6 >0.3 nm; loops 7 and 10 <0.2 nm. At 350 K, flexibility increased markedly in parts of loops 3, 6, and 11 (some residues +>0.2 nm); loop 11 rose to the third most flexible region. B-factor differences localized increased flexibility to these loops, defining a highly flexible region near the binding site. - Dynamic cross-correlation: At 350 K, overall correlation magnitudes decreased (narrower “boat” in distance–correlation scatter). At 300 K, loops 10/11 and loop 6/helix 3 showed strong internal positive correlations and negative correlations with loop 3, reflecting coordinated binding-site motions; these correlations weakened or disappeared at 350 K, especially between loops 6 and 11 and between loops 10/11 and loop 3, indicating disrupted coordinated dynamics at high T. - Major conformational states: PCA free energy landscape showed one dominant basin at 300 K (structure A) and multiple basins at 350 K, with the deepest corresponding to a dominant cluster (structure B) and a second cluster (structure C, ~16%). At 300 K, loop 6 opening narrowed slightly (10.3 Å to 9.2 Å) but remained open. At 350 K, loop 6 shifted downward, closing the site: opening narrowed to 5.5 Å, with an effective vertical distance near catalytic residues ~3.8 Å, blocking ligand accommodation. Structure C featured outward shifts of loops 3 and 11, increasing loop 6–11 separation from 4.2 Å to 10.1 Å. - Interaction determinants: In the highly flexible region, intra-group hydrogen bonds decreased after heating (average count from ~10 to ~7); inter-group counts (~12–13) changed less. Among intra-group H-bonds present ≥20% at 300 K (n=15), all dropped by >30% at 350 K, with 73% dropping by >70%, implicating H-bond disruption as the main driver of conformational change. Some new intra- and inter-group H-bonds formed at 350 K, stabilizing new conformations and explaining modest changes in inter-group counts. - Salt bridges and disulfides: Six salt bridges had ≥20% occurrence; most were helical and distant from the binding site. High temperature did not significantly damage salt bridges; several increased in probability (e.g., LYS209–ASP213 had highest overall probability, ~74%). Sequence analysis indicated six conserved cysteines forming three disulfide bonds (Cys18–Cys40, Cys62–Cys72, Cys132–Cys161). Removing disulfides increased RMSD at both temperatures (by >0.1 nm at 350 K) and radius of gyration (>0.03 nm at 350 K), indicating key roles in maintaining global stability. - Binding energetics (MMPBSA at 300 K): Total binding free energy for DP6-chitosan was −24.93 kcal/mol. Molecular mechanics electrostatics contributed more than double van der Waals; polar solvation (PB) opposed binding strongly (+83.44 kcal/mol). Per-residue decomposition showed four strongest contributors in loop 6, with additional contributions from loop 3 and residues at the site opening; ASH59 and GLH152 inside the site favored binding. Notably, 78.7% of favorable binding energy arose from residues in the highly flexible regions; most had RMSF increases ≥3× average, linking increased flexibility to reduced binding stability at high temperature.

The study demonstrates that while the global fold of the A. fumigatus GH75 chitosanase remains stable upon heating to 350 K, the binding site becomes significantly more flexible and undergoes conformational closure primarily due to disruption of intra-regional hydrogen bonds. Coordinated motions that stabilize the open conformation at 300 K are weakened at 350 K, particularly among loops 3, 6, and 11. The dominant high-temperature conformations exhibit a narrowed or closed binding site that sterically hinders DP6-chitosan binding, explaining the experimentally observed decline in activity stability above ~55°C despite an optimal temperature near 65–70°C. Salt bridges and disulfide bonds maintain the integrity of the overall fold, consistent with minimal changes in radius of gyration, but they do not prevent local site closure. Binding free energy analysis further shows that residues most responsible for ligand binding are concentrated in the regions whose flexibility increases most with temperature, amplifying the loss of binding capacity. These insights directly address the hypothesis by identifying specific structural-dynamic determinants (loop motion, H-bond disruption) that compromise thermal stability, thereby providing concrete targets for protein engineering.

AlphaFold2 accurately predicted the GH75 chitosanase structure, including the binding site and catalytic residue arrangement matching the viral V-Csn. Structure-based pKa analysis suggests catalytic roles for ASP143 and GLU152 opposite to prior expectations. Molecular dynamics reveal that elevated temperature increases flexibility and disrupts coordinated motions in loops 3, 6, and 11, driving a transition from an open to a closed binding-site conformation and sharply reducing ligand-binding capacity. H-bond disruption within the flexible region is the primary interaction change underlying this transition, while salt bridges and three conserved disulfide bonds underpin global structural stability. Residues that contribute most to binding reside in the highly flexible regions, linking increased flexibility to diminished activity stability at high temperature. These findings identify specific residues/loops for engineering (e.g., mutations to strengthen local H-bonding networks or introduce stabilizing interactions) to improve thermal stability, and they illustrate how computational approaches can guide studies of proteins lacking experimental structures.

The structure used is computationally predicted (AlphaFold2) and, while validated and consistent with a related GH75 structure, key inferences (e.g., catalytic residue protonation roles and interaction networks) require experimental verification. Simulations were conducted at two temperatures with finite timescales, which may not capture all long-timescale conformational states or thermal denaturation pathways.