Network Dynamics as Fingerprints of Thermostability in an In Silico-Engineered DyP-Type Peroxidase

Abstract

Stabilizing industrial enzymes is crucial for advancing environmentally responsible bioprocesses; however, the structural basis of thermostability remains incompletely understood. Here, we engineered thermostable variants of a tetrameric dye-decolorizing peroxidase (DyP) using two independent open-source design algorithms, yielding enzymes with significantly improved thermal performance and prolonged activity at elevated temperatures. Subsequent recombination strategies minimize the mutational burden while maintaining or enhancing stability. Structural and dynamic analyses of the thermostable variants revealed convergent features, including increased compactness, rigidity, and an enriched network of hydrogen bonds and hydrophobic interactions. Despite differing mutation profiles, stabilizing substitutions clustered in similar structural regions. Notably, the integration of dynamic modeling with protein correlation network analysis uncovered a previously unrecognized fingerprint of stabilization: highly connected structural networks characterized by denser and more persistent intra- and intermonomer interactions, greater internal cohesion, and enhanced cooperative dynamics. Tetramers exhibit long-range communication pathways and redundant routes, supporting coordinated motions that can hinder local unfolding and tetramer dissociation. These findings identify dynamic interaction networks as hypothetical new indicators of protein stability and offer a previously unexplored framework for rational enzyme design.

Introduction

Enzyme stability is paramount for biocatalysts, particularly in industrial setups, where proteins must endure high temperatures, extreme pH values, and denaturing organic solvents. The improving predictive power of in-silico algorithms and computational methods has become an attractive tool for engineering protein thermostability. These methods predict enzyme discovery, protein solubility, activity, specificity, and stability. − Several enzymes have been stabilized relying on multiple primary sequence alignments and consensus analysis by replacing a hotspot position with conserved residues in their thermophilic counterparts. − At a different level, structure-based engineering allows the prediction of stabilizing mutations based on classical rational design and energy calculations. For instance, flexible regions in a protein are known to “weaken” the overall structure by increasing its flexibility and propensity for unfolding and denaturation, making them typical targets for engineering. Combining this knowledge with molecular dynamics simulations enabled the prediction of mutations that enhanced rigidity and protein stability. , Additionally, the Gibbs free energy increment (ΔΔGfold) can be calculated using advanced computational tools, such as FoldX and Rosetta, which comprehensively analyze potential stabilizing mutations. , Thus, by using computational methods capable of simulating site-saturation or site-directed mutagenesis in silico, researchers can efficiently screen mutations likely to result in improved stability. This provides invaluable insights for protein engineering and design, significantly accelerating the identification of stabilizing mutations. Furthermore, unlike traditional methods, de novo design allows scientists to build proteins from scratch, guided by theoretical models and computational tools. − Computational analysis provides a complete, multi-approach prediction of stabilizing mutations and is now available from web servers such as FireProt. , PROSS, and FRESCO. These platforms use a multilevel framework that unites consensus analysis, energy calculations, and, in FRESCO, de novo design to analyze possibly stabilizing mutations from sequence-based and energy-based analysis, calculate possible antagonistic effects among predicted mutations, and provide a list of target positions for experimental verification.

Protein stability is a complex, multivariable equation. A single factor does not dominate the relationship between protein structure and thermostability; instead, several factors contribute to this relationship. ,, Standard mechanisms include a better-packed hydrophobic core favoring enthalpically driven van der Waals interactions, increased rigidity, and more hydrogen bonds, salt bridges, aromatic interactions, and disulfide bonds. However, the most frequently observed strategies are enhanced ionic interactions, greater compactness, and the distribution of intramolecular interactions throughout the structure rather than clustering them in specific regions. − These features contribute to improved robustness of the native state (thermodynamic stability) and protect proteins from irreversible degradation reactions at high temperatures (kinetic stability), such as aggregation, deamidation, and oxidation of specific amino acid residues. ,

Dye-decolorizing peroxidases (DyPs) are heme-containing microbial enzymes that catalyze the reduction of hydrogen peroxide to water with concomitant oxidation of various substrates, including anthraquinone dyes, lignin-related phenolic and nonphenolic compounds, and metal ions (e.g., Mn­(II), Fe­(II)), holding considerable promise for biotechnological applications, particularly in biorefineries. We have previously provided structural and mechanistic insights into their catalytic mechanism, − enhanced their properties through directed evolution, − and expanded their functional scope by modulating redox behavior. − ,− Here, we employed free web-based design tools exclusively for academics to stabilize the DyP-type peroxidase from Pseudomonas putida (PpDyP) and combined biochemical, biophysical, and computational analyses to dissect the underlying stabilization mechanisms. We applied residue interaction network analysis derived from molecular dynamics simulations to explore how mutations affect protein dynamics and stability. − In this framework, amino acids are represented as nodes, with edges defined by the strength and number of noncovalent interactions. This approach, previously used to investigate protein folding and allostery, , as well as proteins’ conformational dynamics, particularly within enzyme redesign, , provides system-specific insights into dynamic features that may accompany or support enhanced thermostability in designed enzymes, highlighting properties that are not yet routinely optimized in current design pipelines.

Methods

In Silico Stability Design

The monomeric structure (PDB Code; Chain A) and dimeric structure (PDB Code; Chain B and D) of PpDyP were submitted to PROSS (https://pross.weizmann.ac.il/) and FireProt (https://loschmidt.chemi.muni.cz/FireProtweb/) algorithms. Submissions were performed in standard settings, except that the conserved residues 132, 197, and 214 were fixed in PROSS. The original mutation sets provided were manually reviewed in Tables S1 and S2.

Bacterial Strains, Plasmids, and Media

E. coli Tuner (DE3, Novagen) was used to express the ppDyP gene previously cloned in pET15-b (+) plasmid (Novagen) (the in-silico variants (PpDyP_PR and PpDyP_FP) and all DNA Shuffling variants. In the Tuner strain, genes are under the control of the T7 promoter, and IPTG induces their expression. E. coli strain DH5α (Novagen) was used to amplify plasmid constructs. LB medium was used to grow E. coli strains, supplemented with 100 μg·mL^–1 ampicillin.

DNA-Shuffling

DNA shuffling was performed after gene coding for PpDyP wild-type, PpDyP PROSS and PpDyP FireProt were amplified using primers pET21Fw2 (5′ CTTCCCCATCGGTGATGTCGGCGATATAG 3′) and pET21Rv (5′ CCAAGGGGTTATGCTAGTTATTGCTCAG 3′). Two libraries were constructed: PpDyP^WT shuffled with PpDyP^PR, and PpDyP^WT shuffled with PpDyP^FP. A mixture containing 280 ng of each parental gene was digested for each library with 0.15 U of DNase I in 200 mM Tris-HCl, pH 7, with 0.2 M MnCl₂ for 20 min at 15 °C in a thermocycler. Digestion was stopped by adding 6 μL of 0.5 M EDTA. The PCR reassembly was performed in a 20 μL reaction volume containing 5 μL of DNA fragments, 200 μM of dNTPs, NZYProof polymerase buffer, and 2.5 U of NZYProof polymerase (NZYTech, Lisboa, Portugal). After an initial denaturation period of 3 min at 96 °C, the subsequent steps were repeated for 45 cycles in a thermal cycler: 1 min at 94 °C, 90 s at 59 °C, 90 s at 56 °C, 90 s at 53 °C, 90 s at 50 °C, 90 s at 47 °C, 90 s at 44 °C, 90 s at 41 °C and 1 min +5 s/cycle at 72 °C followed by a final 10 min period at 72 °C. The PCR reassembly products were amplified by PCR using the primers PpDyPFw (5′ GGATTAGCCTCATATGCCGTTCCAGCAAGG 3′) and PpDyPOptRv (5′ GTGTTTCTGTATCTGGATCCTTACGCACGCAGCAGCG 3′). The amplified products were purified using the GFX PCR DNA and Gel Band Purification kit (GE Healthcare, Chicago, IL, USA). Ligation was performed through T4 Ligase, and the constructed libraries were first introduced into E. cloni 10 G electrocompetent cells, the plasmids were extracted and then introduced into E. coli BL21 Star. The transformed cells were spread into an LB solid medium supplemented with 100 μg mL^–1 of ampicillin, and plates were incubated overnight at 37 °C.

Activity and Stability Screenings in 96-Well Plates

Colonies were picked from the original culture plate and transferred to 96-well plates containing 200 μL of LB supplemented with 100 μg mL^–1 of ampicillin. Four wells on each plate were used to inoculate the parental strain of each generation as a control. Plates were incubated for 6 h at 37 °C, 750 rpm in a Titramax 1000-plate shaker (Heidolph, Schwabach, Germany), after which 0.1 mM IPTG and 30 μM Hemin were added. Cultures were cultivated overnight at room temperature at 750 rpm. Cells were harvested by centrifugation (5 min at 3220× g, 4 °C) and disrupted in 100 μL of 30% Bacterial Protein Extraction Reagent (B-PER, Thermo Scientific) in 20 mM Tris-HCl buffer, pH 7.6, and incubated for 20 min at room temperature at 750 rpm. Plates were centrifuged (30 min at 3220 g, 4 °C), and the supernatants (crude cell extracts) were collected and used for enzymatic activity assays. Cell crude extracts (20 μL) were transferred into two replicas of 96-well plates. One plate was assayed for initial activity (Ai) by adding 180 μL of the reaction mixture. The activity was in 100 mM sodium acetate, pH 4.5, 1 mM ABTS, and 0.5 mM H₂O₂. The progress of reactions was monitored at 420 nm (ABTS, ε420 nm = 36,000 M^–1 cm^–1) using a Synergy2 microplate reader (BioTek, Winooski VT, USA). One unit (U) of enzymatic activity was defined as the enzyme required to reduce 1 μmol of substrate per minute. Cell crude extracts in the second plate were incubated at 50 °C for 1 h, cooled on ice for 5 min, incubated at room temperature for 5 min, and assayed for residual activity (Ar) using the same conditions indicated for assaying the initial activity. Thermostability was assessed using the ratio of the residual activity to the initial activity of the variant (v), normalized to the parent type (p)–(Ar/Ai)­v/(Ar/Ai)­p. The variants’ activity relative to the parent was calculated using the ratio of the initial activity of the variant to the parent type (Aiv/Aip).

Enzyme Production and Purification

Recombinant enzymes were produced in 0.5 L of LB medium in 2L-SHOTT Baffled Erlenmeyer Flask (SHOTT) and purified. Cell sediments were suspended in 20 mM Tris–HCl buffer, 20 mM Imidazole, pH 7.6, containing DNase I (10 μgmL^–1 extract), MgCl₂ (5 mM), and a mixture of protease inhibitors, antipain, and leupeptin (2 μgmL^–1 extract). Cells were disrupted in a French press (Thermo EFC). Cell debris was removed by centrifugation (18,000×g, 2 h, 4 °C), and the resulting crude extracts were used for protein purification using an AKTA purifier (GE Healthcare, BioSciences) at room temperature. For purification, the crude extract was loaded onto a HisTrap (Cytiva) column equilibrated with 20 mM Tris–HCl buffer, 20 mM Imidazole, pH 7.6. Elution was carried out with 0.5 M imidazole in the buffer. The active fractions were pooled and concentrated before applying them to a Superdex 200 Increase 10/300 GL column (Cytiva) equilibrated with 20 mM Tris–HCl buffer, pH 7.6, with 0.2 M NaCl. The fractions with the tetrameric form of PpDyP and variants were pooled together. For the crystallization assays, PpDyP^PR and PpDyP^FP were further purified using a 6 mL RESOURCE Q anion exchange chromatography column (Cytiva) equilibrated with 20 mM Tris-HCl buffer, pH 7.6. Elution was carried out with 1 M NaCl in the buffer. The fractions corresponding to PpDyP and variants were pooled and concentrated until 10–15 mg/mL was reached. Purified proteins were stored at −20 °C.

Spectroscopic Analysis

At room temperature, the UV–Visible absorption spectra of purified enzymes were recorded on a Nicolet Evolution 300 spectrophotometer from Thermo Scientific (Madison, USA). The pyridine ferrohemochrome method determined the heme content using an extinction coefficient of εR-O 556 nm (28.32 mM^–1 cm^–1).

Apparent Steady-State Kinetic Analysis

The apparent kinetic parameters (k _cat and K _m were estimated for hydrogen peroxide and reduced substrates 2,2-azino bis (3-ethylbenzthiazoline-6sulfonic acid (ABTS) and 2,6-dimethoxy phenol (DMP). Reactions were monitored at 420 nm (ABTS, ε_420 nm = 36,000 M^–1 cm^–1) and 468 nm (DMP, ε_468 nm = 49,600 M^–1 cm^–1) using the Synergy2 microplate reader (BioTek, Vermont, USA) at 25 °C. Reactions to estimate the kinetic parameters for ABTS (0.1–3 mM) were performed in the presence of 1.5 mM H₂O₂ (wild-type) and 0.5 mM H₂O₂ (PpDyP^PR and PpDyP^FP) in 100 mM sodium acetate buffer at pH 4.5. Reactions to estimate kinetic parameters of H₂O₂ (0.01–6 mM) were performed in 2 mM ABTS in 100 mM sodium acetate buffer at pH 4.5. The reactions for DMP (0.01–2 mM) were performed in 100 mM sodium acetate or phosphate buffer in the presence of 0.5 mM H₂O₂ at the optimal pH of the enzymes. The kinetic data were fitted directly using the Michaelis–Menten equation or the equation for the nonlinear curve that fits enzyme kinetics affected by substrate inhibition (v = Vmax­[S]/(Km + [S]­(1 + [S]/K _i))) (Origin software).

Kinetic and Thermodynamic Stability

Thermal inactivation assays were performed as previously described. In brief, enzyme preparations in 20 mM Tris–HCl buffer, 200 mM NaCl, pH 7.6, were incubated at 45, 50, 55, 60, 65, 70, and 75 °C. At fixed time intervals, sample aliquots were withdrawn and tested for activity following ABTS oxidation at 25 °C. The thermal inactivation appears to obey first-order kinetics, and the half-life values (t1/2) were calculated using t1/2= ln2/k. The activation energy for irreversible unfolding was determined using the Arrhenius equation (k = Aê(−Ea/RT)), where k is the rate constant obtained from thermal denaturation kinetics, A is the pre-exponential factor, Ea is the activation energy, R is the gas constant, and T is the absolute temperature in Kelvin. The thermodynamic stability of purified PpDyP^WT, PpDyP^PR, and PpDyP^FP was assessed by steady-state fluorescence measurements using a Cary Eclipse spectrofluorometer (Agilent Technologies) at excitation wavelengths of 296 nm and emission wavelengths of 350 nm, as previously reported. In contrast, the thermodynamic stability of variants obtained using DNA Shuffling was measured with nano DSF Prometheus NT.48 (Nanotemper) at emission wavelengths of 330 and 350 nm using 150 μg/mL purified protein in 20 mM Tris–HCl, 200 mM NaCl, pH 7.6. For the equilibrium unfolding studies, guanidine hydrochloride (GdnHCl) concentrations in the 0–3.5 M range in 20 mM Tris–HCl buffer, 200 mM NaCl, pH 7.6, were used to induce protein unfolding as previously reported. For the thermal unfolding assays, samples containing the enzymes (20 μM in 20 mM Tris–HCl buffer, 200 mM NaCl, pH 7.6) were heated at a rate of 1 °C/min to 100 °C. Protein aggregation was monitored by measuring static light scattering at 500 nm with excitation and emission wavelengths. The thermodynamic stability of enzymes was analyzed according to a two-state process using the previously described equations.

Crystallization and Cryoprotection

Vapor-diffusion crystallization trials of PpDyP^PR and PpDyP^FP (∼15 mg/mL) in 20 mM Tris–HCl pH 7.6 and 200 mM NaCl were made with the commercial screens PACT premier (MD1–29), Structure Screen 1 + 2 HT-96 (MD1–30), and JCSG-plusTM (MD1–37) from Molecular Dimensions (Sheffield, U.K.). Crystallization drops were set using a robot Mosquito (SPT Labtech) and 96-well sitting drop iQ plates (SPT Labtech), mixing 2:1, 1:1, and 1:2 protein:reservoir solutions against 40 μL of reservoir solution. PACT premier screen generated PpDyP^PR and PpDyP^FP crystals within 2 and 5 days, respectively, at 20 °C. PpDyP^PR crystallized in 0.2 M calcium chloride dihydrate, 0.1 M MES pH 6.0, 20% (w/v) PEG 6000, while PpDyP^FP crystallized in 0.2 M sodium iodide, 0.1 M Bis Tris Propane pH 6.5, 20% (w/v) PEG 3350. The crystallization hits were optimized at microliter scale using hanging drop vapor diffusion trials in XRL 24-well crystallization plates (Molecular Dimensions) and equilibrated against 500 μL of the reservoir solution. The best PpDyP^PR and PpDyP^FP crystals were obtained in 1:1 and 2:1 protein:reservoir drops, respectively. Crystals were harvested and transferred to their reservoir solutions supplemented with 20% (v/v) glycerol before flash-cooling in liquid nitrogen.

Data Collection and Processing

Diffraction data were collected at 100 K in ALBA synchrotron (Barcelona, Spain) using the beamline BL13-XALOC. Diffraction images of PpDyP^PR were obtained with a DECTRIS PILATUS3 X 6 M detector, using 0.97926 Å radiation wavelength, a crystal-to-detector distance of 432.07 mm, and an oscillation width of 0.10° in a total of 180° rotation. The diffraction data of PpDyP^FP were obtained using the same detector and radiation wavelength as PpDyP^PR, crystal-to-detector distance of 508.30 mm, and oscillation width of 0.15° in a total of 360° rotation. Data were indexed and integrated with XDS in a space group determined with POINTLESS, and the data scaled with AIMLESS. These programs were used within the autoPROC data processing pipeline.

Structure Determination and Refinement

Estimating the unit cell contents indicated two and four molecules in the PpDyP^PR and PpDyP^FP asymmetric units, respectively. , The structures were determined by molecular replacement (MR) using the wild-type structure (PDB 7QYQ) as a search model. The MR solutions for variants were obtained using PHASER. within the PHENIX suite, and their final translation-function Z-scores TFZ of 58.7 and 79.0, respectively, indicated a successful solution. The two crystal structures were refined using PHENIX.REFINE, − followed by inspection of σ_A-weighted 2|Fo|−|Fc| and |Fo|−|Fc| electron density maps for manual model improvement and completion using COOT. Cycles of iterative structure refinements included atomic positions and atomic displacement parameters (a.d.p.s) optimization, automatic water solvent completion, as well as polypeptide chain regions for translation, libration, and screw refinement of anisotropic a.d.p.s defined by the TLSMD server (http://skuld.bmsc.washington.edu/~tlsmd). Although refinement included standard stereochemistry libraries, interatomic distances involving iron sites were refined without target restraints. The stereochemistry of the refined structures was analyzed with MolProbity. All figures were prepared with PYMOL. The electrostatic potential of PpDyP and variants was calculated with the Adaptive Poisson–Boltzmann Solver (APBS) plugin in PYMOL and PDB 2PQR The web server Protein Interactions Calculator was used to identify protein interactions, and DogSiteScorer to detect protein activities and their dimensions. The structure factors and atomic coordinates of PpDyP^PR and PpDyP^FP were submitted to the Protein Data Bank (PDB) with accession codes 9F1O and 9F1Q, respectively.

Loop Modeling

The nonvisible loop 121–128 in the PpDyP^PR structure was built using the Rosetta loop modeling. Loop regions were generated with the cyclic coordinate descent (CCD) loop closure algorithm using fragments from proteins with known structure; a full-atom refinement step was performed with the next-generation kinematic (NGK) closure robotics-inspired conformational sampling protocol. The crystal structure of PpDyP^PR was kept intact except for the generated loop region, with the side chains repackaged within 10 Å of the remodeled region. 500 loops were generated, and each model was evaluated and ranked based on its Rosetta energy score to identify the optimal loop candidate.

Molecular Dynamics Simulations

The crystal structures of the tetrameric forms from wild-type and PpDyP^FP were used as initial coordinates for the corresponding simulation setup. The following protocol was followed for each variant. Hydrogen atoms were added, and the protonation state of titratable residues was determined at pH 7.6 with the H++ server (http://biophysics.cs.vt.edu/H++). The tetramer was then solvated in an orthorhombic box of water molecules, ensuring a minimum distance of 10 Å from any protein atom to the edge of the box. Sodium atoms were added to neutralize the system. The system was prepared in leap and included in the AMBERTools22 suite of Amber22, University of California, San Francisco, USA (2022). The Amber force field 14 Stony Brook ff14SB was used for the protein, the TIP3P model for water molecules, and GAFF-derived parameters for the heme cofactor. The solvated protein was fully energy-minimized before starting the molecular dynamics (MD) simulations, which were performed using the Amber22 package. 100 ps of heating with restraints on the protein’s heavy atoms was performed, followed by a 100 ps equilibration in the NVT ensemble (with restraints applied to the protein’s backbone and the cofactor heavy atoms), and 5 ns of free equilibration at constant temperature and pressure. A time step of 2 fs, combined with the SHAKE algorithm, was used to constrain bonds involving hydrogen atoms in the simulations. A cutoff value of 14.0 Å was used for nonbonding interactions, and the Particle Mesh Ewald method for long-range electrostatics. A production run of 1 μs was then run in triplicate. A lower temperature (300 K) and a higher temperature (331 K) were simulated for each variant. Molecular dynamics simulations of the monomeric forms of wild-type, PpDyP^FP, and PpDyP^PR were performed at the same temperatures, and additionally for PpDyP^PR at 343 K (i.e., 26.9, 57.9, and 69.9 °C), as this variant shows significantly higher apparent melting temperatures compared to the other enzymes. The initial coordinates were taken from selecting one monomer from the corresponding crystal structures. As explained above, the missing loop in the PROSS structure was modeled with Rosetta. The simulations were analyzed with an in-house script that uses pytraj (https://github.com/amber-md/pytraj), a Python package binding to the cpptraj program and, for the contacts analysis, the Yasara program. Protein network analysis on the monomeric and tetrameric molecular dynamics trajectories at 300 K of all variants was conducted following the Shortest Path Map method as implemented in SPMweb (https://spmosuna.com/). For each system, the server was provided with the three-dimensional structure of the protein, corresponding to the representative structure of the most populated cluster obtained after clusterization by cpptraj of the 3 μs simulation, as well as the correlation and distance matrices computed from the MD trajectory using the cpptraj program. The default distance and SPM significance thresholds of 6.0 Å and 0.3 were used. In the case of the tetramers, a significance threshold of 0.1 was also tested using SPM, as explained in the Results.

Other Methods

The concentration of purified proteins was estimated using the molar absorption coefficient of PpDyP (ε280 = 34,850 M^–1·cm^–1), calculated from the protein sequence using the ExPASy Bioinformatics Resource Portal (http://web.expasy.org). The oligomerization state of PpDyP and its variants was typically assessed by injecting 100 μL of purified enzyme preparations into a gel filtration Superose 12 10/300 GL (GE Healthcare Biosciences, NJ, USA) column, which was equilibrated with 20 mM Tris-HCl buffer, pH 7.6, and 0.2 M NaCl. The calibration curve was generated using the retention times of the Protein Standards preparation against their molecular masses (Bio-Rad Laboratories, CA, USA).

Results and Discussion

One-Step Design of PpDyP Variants for Thermophilic-like Properties

Monomeric and dimeric structures of wild-type were submitted to the PROSS and FireProt web servers, which combine atomistic energy-based and evolution-based approaches to predict multiple additive mutations that increase protein stability. The proposed mutation list was curated based on two key criteria: first, residues near the catalytic site (heme cofactor) were excluded to prevent any disruption of enzymatic activity; second, priority was given to mutations identified in the dimer interface over those suggested for monomeric structures (Table S1). Notably, across the two sets of proposed mutations, one comprising 29 mutations derived from the PROSS output and the other containing 21 mutations from the FireProt outputeight mutations were mutually identified by both web servers: E91D, L110F, L120R, D139G, D174E, V217M, Q222G, and E237D (these will be highlighted in bold for clarity) (Table S1). Two synthetic genes were designed and synthesized, encoding PpDyP^PR and PpDyP^FP variants that contain the 29 PROSS and 21 FireProt mutations, respectively. The genes encoding these variants were overexpressed in Escherichia coli, leading to successful protein production, purification, and characterization. Optimization of growth conditions significantly increased protein yields from 14 to 150 mg L^–1. However, this improvement came at the cost of reduced heme incorporation, yielding 0.5–1 mol of heme per mole of protein, instead of the usual 1:1 ratio (Table S2). The UV–visible spectra of the enzymes exhibited the characteristic Soret band at 404 nm, Q bands at 520 nm, and a charge transfer band at ∼640 nm (Figure S1). Size-exclusion chromatography analysis revealed that, under these conditions, the tetrameric form of PpDyP was the dominant oligomerization state in solution, with only trace amounts of monomeric and dimeric forms detected (Figure S2), in stark contrast to previous observations.

Increased Stability and Activity of the Designed Variants

Interestingly, the maximal activity was achieved at 65 and 70 °C for PpDyP^FP and PpDyP^PR, a significant shift from the optimal temperature in the wild-type (between 20 and 30 °C), reflecting the higher thermostability of enzymes (Figure a). The thermodynamic stability of the proteins is assessed by determining their melting temperature (T _m). T _m is the temperature at which 50% of the protein is unfolded. It reflects the equilibrium between the native, functional protein and the unfolded state, providing a measure of thermodynamic stability. Please note that in the simplest case, protein unfolding data can be described by the two-state equilibrium model, involving only the native (N) and unfolded forms (U): N ↔ U. The T _m of the PpDyP enzymes are presented as apparent values (T _m(app)), − since we observed temperature-dependent aggregation precedes enzymes’ unfolding by following static light scattering with emission and excitation at 500 nm in the spectrofluorometer: wild-type starts aggregating at 54 °C, and PpDyP^FP and PpDyP^PR variants at 60 and 72 °C (Figure S3). Both variants exhibit apparent T _m(app) values of 72 °C (PpDyP^FP) and 82 °C (PpDyP^PR), which are 10 and 20 °C higher than the wild type (62 °C), respectively (Figure b and Table S3). The improvements obtained are among the best reported in the literature for computationally guided engineered enzymes (Table S4). Furthermore, the thermophilic variants exhibit higher stability against chemically induced denaturation, displaying a significant increase in guanidine hydrochloride (GdnHCl) midpoint values compared to the wild-type (Table S5). Both variants demonstrate a remarkable 150-fold increase in half-life at 60 °C compared to the wild type (Figure c), which loses activity after 1 h at this temperature. Kinetic or operational stability refers to the free-energy barrier that “separates” the native state from non-functional forms. The unfolding process can become irreversible, particularly if temperature-induced unfolding occurs. The Lumry–Eyring model schematizes this scenario, where D is the irreversibly denatured state. Kinetic stability assays performed at increasing temperatures (45, 50, 55, and 60 °C for the wild type, and 60, 65, 70, and 75 °C for PpDyP^FP and PpDyP^PR) allowed us to calculate the free-energy barriers for irreversible denaturation. The thermostable variants displayed activation energies of 51 and 53 kcal/mol, respectively, while the wild type showed an activation energy of 37 kcal/mol (Table S6 and Figure S4). These higher values in variants indicate that they undergo irreversible denaturation at a significantly lower rate, allowing them to remain functional for extended periods. Higher energy input (i.e., elevated temperatures) is also required to accelerate denaturation in the variants. Variants achieve increased thermostability by combining mutations that, e.g., promote interactions between residues, increase compactness, increase rigidity and strengthen the connectivity of interaction networks; as most frequently observed in stability engineering, no single factor dominates.

1.

Biochemical characterization of thermostable and hit variants from DNA shuffling experiments. (a) Temperature profile of wild-type (black), PpDyP^PR (blue), and PpDyP^FP (red) variants. (b) Thermodynamic stability, where apparent melting temperatures were estimated 62, 82, and 72 °C for wild-type and the PpDyP^PR and PpDyP^FP variants, (c) Kinetic stability, measured at 60 °C and halve-lives of 0.14, 23, and 20 h were estimated for wild-type, PpDyP^PR and PpDyP^FP. (d) Phenotype of hit variants from DNA shuffling experiments using genes coding for wild-type and PpDyP^PR (10G9 and 5F8) and PpDyP^FP (11B8); the number of mutations, melting temperature, half-life at 60 °C, and k _cat for ABTS. (e) Genotype of hit variants from DNA shuffling experiments ordered by the highest T _m(app). In bold are the eight mutations predicted simultaneously by both PROSS and FireProt web servers. (f) Difference in apparent melting temperature relative to wild-type (T _m(app) of 62 °C) of enzyme variants from the DNA-shuffling of wild-type × PpDyP^PR (in blue) and wild-type × PpDyP^FP (in red). (g) Fold change of the half-life time at 60 °C relative to wild-type (t _1/2 = 0.14 h) of enzyme variants that resulted from the DNA-shuffling of wild-type × PpDyP^PR (in blue) and wild-type × PpDyP^FP (in red). (h) Fold change of k _cat for ABTS of wild-type (55 s^–1) to variants from the DNA-shuffling of wild-type × PpDyP^PR (in light blue) and wild-type × PpDyP^FP (in light red).

An exciting aspect is that neither variant enzyme exhibits the typical trade-off between stability and enzyme activity; indeed, both variants show a 10-fold increase in activity (k _cat) for ABTS and H₂O₂ at room temperature, with comparable Km values, resulting in up to 1 order of magnitude higher catalytic efficiency and higher resistance to H₂O₂ inhibition (Table and Figure S5a,b). Furthermore, steady-state activity for the oxidation of the lignin-related phenolic 2,6-DMP (Table S7), kinetics at room temperature, demonstrate that whereas PpDyP^FP shows comparable kinetic parameters to the wild-type, PpDyP^PR variant exhibited 10-fold higher activity, comparable Km and thus, 1 order of magnitude higher catalytic efficiency (k _cat/K _m) than the wild-type, and a 4-unit shift in the optimum pH to 8.0, features previously observed in variant 6E10 from directed evolution. This higher activity of lignin-related phenolics at alkaline pH is an attractive property from a biotechnological perspective, as it enhances the solubility of technical lignin raw materials (e.g., kraft, organosolv, and soda lignins) and accelerates reaction rates. ,

1. Apparent Steady-State Kinetic Parameters for the Reduction of H₂O₂ and the Oxidation of ABTS .

	H2O2				ABTS
	k cat (s–1)	K m (mM)	k cat/K m (M–1 s–1)	K i (mM)	K m (mM)	k cat/K m (M–1 s–1)
wild-type	55 ± 2	0.14 ± 0.03	(4.2 ± 1.1) × 105	1.91 ± 0.25	0.18 ± 0.01	(3.2 ± 0.1) × 105
PpDyPPR	503 ± 6	0.15 ± 0.01	(3.5 ± 0.3) × 106	1.16 ± 0.03	0.38 ± 0.03	(1.3 ± 0.1) × 106
PpDyPFP	515 ± 28	0.11 ± 0.01	(5.1 ± 0.6) × 106	0.92 ± 0.05	0.20 ± 0.01	(2.6 ± 0.2) × 106

^aKinetics were measured in 1.5 mM H₂O₂ for the wild-type and 0.5 mM H₂O₂ for the PpDyP^PR and PpDyP^FP variants at 25 °C in sodium acetate buffer (pH 4.5). The data are represented as the mean ± SD of the independent experiments (n = 3). K _i is the inhibition constant for hydrogen peroxide.

DNA-Shuffling to Help Distinguish Beneficial Mutations in Variants

Despite the intention of automated computational design protocols to predict only stabilizing mutations, unexpected local interactions among mutations can result in neutral or even globally destabilizing, adverse structural effects. This occurs because the functional impact of a mutation depends on the genetic background, i.e., the interactions between mutations direct novel protein functions due to epistasis. Therefore, the variants’ genes were recombined with the wild-type gene using DNA shuffling; the selection of chimeric variants that maintain comparable or higher relative activity/stability than the initial variants PpDyP^PR and PpDyP^FP (Figure S6) allowed the identification of the smallest subset of beneficial mutations that have a functional impact (Tables S8 and S9). The frequency of mutations varies among the different variants, suggesting that no mutation was consistently retained or eliminated. We have purified and characterized the top ten variants from each shuffling experiment (Figure f–h; Tables S8 and S9). Interestingly, most variants have fully complemented heme content compared to the initial thermostable templates (Table S10).

Notably, three-hit variants from the shuffling experiments (10G9, 5F8, and 11B8) show similar or higher stability properties than the template variants (Figure d and Tables S11 and S12). Coincidently or not, they contain a high percentage of the eight mutations predicted by both web servers: six in 10G9 and 5F8 and seven in 11B8 (highlighted in bold in Figure e and yellow spheres in Figure S7). Variants 10G9 (with 12 mutations less than PpDyP^PR) and 5F8 (minus eight mutations) exhibit comparable apparent T _m(agg) values (82–83 °C) to PpDyP^PR (Figure d,f). Notably, variant 10G9 is the only one that does not aggregate at temperatures up to 100 °C (Figure S8). On the other hand, variant 5F8 exhibits the highest measured half-life at 60 °C, approximately twice that of the PpDyP^PR, and nearly 250 times longer than the wild-type (Figure d,g). However, both variants exhibit a trade-off between stability and activity, with lower kcat values for ABTS in 10G9 and 5F8 compared to PpDyP^PR (Figure h and Table S11). In the case of 11B8, an increase in the T _m(app) of 8 °C is observed with comparable kinetic stability to PpDyP^FP (Figure f,g). At the same time, the k _cat for ABTS was reduced by 2-fold (Figures h and S7e,f, Table S12). The adverse effects of the four mutations (A155 V, S169A, D174E, V279L) were alleviated when they were removed, which may be hypothetically attributed to a phenomenon known as negative epistasis, highlighting the nonlinear and complex nature of protein structure and function.

Apart from these three hit variants, we observed that removing mutations has, in general, a negative impact on the stability of the other chimeric enzymes, affecting both the thermodynamic and kinetic stability (Figures f,g, S9, and S10), as well as their activity (Figure h). Nevertheless, please note that most chimeric variants exhibit enhanced stability and increased activity when compared to the wild-type enzyme (Figure f–h). A trade-off between stability and activity is only observed in a few cases in point; for example, chimera variants from PpDyP^PR, 10G8, and 7B9 show T _m values of 73 and 76 °C, and 22B4, from PpDyP^FP, reveals a T _m of 74 °C (Figure e); however, they all exhibited reduced activity (and also very low kinetic stability) (Figure h).

Investigation of Structural Features behind Improved Thermostability

The crystal structures of the in silico designed variants were solved using X-ray crystallography (Figure a–c and Table S13). PpDyP^FP, similarly to wild-type, is a tetramer with four subunits in the asymmetric unit. It reveals the canonical dimeric interfacesstructurally conserved across all dimeric DyPsand the noncanonical interfaces unique to PpDyP. The crystal structure of PpDyP^PR is a dimer with the noncanonical interface (Figure a,b). In both variants, most mutations are situated more than 40 Å away from the heme pocket, with only a few (three to five) positioned approximately at 10–15 Å. In DyPs, the heme pocket is surrounded by three loop flexible regions (Figure S11). The access to the heme pocket is made by one cavity and two tunnels (Figures S11, S12, and S13, Table S14). The cavity facilitates access to reduced substrates like ABTS to the heme; interestingly, positively charged residue R181 in the cavity of both thermostable variants is ∼5-fold more solvent-exposed than in the wild-type (Figure d–f, Tables S15 and S16). We hypothesize that the increased exposure of this positively charged residue likely favors accommodation of the anionic substrate ABTS and contributes to the 10-fold enhancement in catalytic activity toward this substrate (Table ). In a previous piece of work, we noticed that the replacement of E188 for a positively charged lysine (E188 K), situated at main heme access channel, in the 9F6 variant, selected during directed evolution and after screening for ABTS oxidation, resulted in a 1 order of magnitude higher catalytic efficiency for ABTS when compared to the wild-type.

2.

Structural and biochemical characterization of PpDyP variants. (a) Alignment based on 3D structural superpositions of wild-type and thermostable variants. The residues marked with an asterisk represent segments 121–128, not visible in the PpDyP^PR structure modeled with Rosetta. The mutations are marked in blue and red boxes for PpDyP^PR and PpDyP^FP, respectively. The common mutations are marked in a yellow box. (b,c) Cartoon representation of monomeric structures of PpDyP^PR and PpDyP^FP, colored in gray, with mutations colored as in (a). The solvent-accessible surface of the cavity provides access to the heme in wild-type (d), PpDyP^PR (e), and PpDyP^FP (f). The residues surrounding the cavity belonging to loop 1 (120–139) and loop 2 (179–205) are colored pink and green, respectively; R181 is colored red. The heme cofactor is depicted as sticks with carbon, oxygen, and nitrogen atoms colored yellow, red, and blue, respectively. The iron atom is represented as an orange sphere.

3.

Representation of protein interaction networks. Cartoon representation of the PpDyP^PR dimer (a) and PpDyP^FP tetramer (b) with mutations located in canonical and noncanonical interfaces, shown as spheres in blue and red. The larger spheres represent mutations closer to the interfaces, and the smaller ones are farther away. In PpDyP^PR (c) the hypothesized most stabilizing mutations in the X-ray structure are near the canonical interface of the tetramer, such as E91D, L96F, L97H, T99A and E103I (α3-helix), and the noncanonical interface, such as L110F (β4-strand), as well as R17 V (β1-strand) and A35R (α1-helix) generating 12 and 13 new H-bonds and hydrophobic interactions with the loss of only 2 H-bonds (d) (Table ). In the case of PpDyP^FP (e) mutations close to canonical E91D, D94E (both in α3-helix), L120R and H125R, and to the noncanonical interfaces, A54 V (α2-helix), L110F (β4-strand), G118A (β4-strand), V217M and S218P, triggered nine new H-bonds and hydrophobic interactions at the expense of losing 3 H-bonds and one hydrophobic interaction (f) as analyzed in X-ray structure (Table ). Notably, mutation S169A, absent in the hit variant 11B8, contributed to the loss of H-bonds in PpDyP^FP. Lines in green represent new H-bonds, and in red, new hydrophobic contacts (dashed lines represent lost interactions in the variant as compared to wild-type); in gray are those previously present in wild-type. In bold, there are common mutations to PpDyP^PR and PpDyP^FP.

Structural analysis revealed that the variants display ∼30% reductions in protein cavity volume and surface area relative to the wild-type (Table ). Additionally, the cavities in the variants contain bulkier apolar residues, likely contributing to increased enzyme packing density. − Notably, 16 out of 29 mutations (55%) in PpDyP^PR and 12 out of 21 mutations (57%) in PpDyP^FP are in the enzymes’ cavities (Table S17). Furthermore, the introduced mutations led to an expanded hydrogen bond network (Table ), primarily driven by 17 and 13 mutations, accounting for approximately 59 and 62% of the total mutations in the PpDyP^PR and PpDyP^FP variants, respectively. New hydrophobic interactions, which are also important for protein stability, emerged due to the incorporation of 11 and 10 mutations in PpDyP^PR and PpDyP^FP, respectively, accounting for 38 and 48% of the total mutations­(Table ). Critical regions near protein interfaces can be identified by analyzing H-bond and hydrophobic networks (Figure a,b). In PpDyP^PR and PpDyP^FP, important mutations near canonical and noncanonical interfaces introduced multiple new hydrogen bonds and hydrophobic interactions, with minimal losses, enhancing interfacial stability (Figure c–f). These structural changes may play a crucial role in improving the protein’s stability. −

2. Total Number of Interactions, Including New and Lost Interactions, and Cavity Number and Dimensions Resulting from Introducing PROSS or FireProt Mutations .

enzyme	cavities number	volume (Å3)	area (Å2)	involved mutations
wild-type	6	1680	2427
PpDyPPR	4	1165	1637	L7I, S115E, L120R, Q144E, V217M, K234H, E237D
PpDyPFP	4	1232	1619	L110F, L120R, H125R, D139G, A155 V, V217M, S218P, L232F, E237D, V279L

enzyme	total no. H-bonds	no. new interactions	no. lost interactions	involved mutations
wild-type	409
PpDyPPR	423 (+34)	88	54	L7I, R17 V, A28D, A35R, E91D, L96F, L97H, T99A, E103I, L110F, S115E, T138Q, Q144E, Q165D, D174E, E237D, P283A
PpDyPFP	425 (+18)	75	57	A54 V, E91D, D94E, Q100R, L110F, G118A, L120R, H125R, S169A D174E, S218P, E237D, V279L

enzyme	total no. hydroph.	no. new interactions	no. lost interactions	involved mutations
wild-type	254
PpDyPPR	268 (+18)	35	17	L7I, R17 V, A28D, A35R, L96F, T99A, E103I, L110F, L120R, A194E, V217M
PpDyPFP	273 (+18)	35	17	A9L, S25F, A54 V, L110F, G118A, L120R, V217M, S218P, L232F, V279L

^aThe specific mutations contributing to some of these interactions and cavities are also highlighted, and the ones in bold are common in the two variants.

^bTo compare wild-type with PpDyP^PR, interactions involving the 121–128 loop, not visible in the PpDyP^PR crystal structure, were excluded (total 409 residues).

The number of salt bridges and aromatic–aromatic interactions appears comparable in the wild-type and thermostable variants (Table S18). ,− In contrast, both variants have a slightly higher number of solvent-exposed charged residues (Figure S14), which is expected to contribute to a higher protein’s solvation energy. , These can lead to stronger interactions among monomeric interfaces. PpDyP^PR has an increase of 5 negative charges compared to the wild-type, whereas PpDyP^FP is more positive than wild-type with a charge of 4 positive charges (Figure S14). The exceptional stability of PpDyP^PR can be partially attributed to its increased negatively charged surface, which enhances protein-sodium interactions with the solvent.

Considering the above analysis and the generally accepted molecular factors influencing enzyme stability, the number of hypothetically relevant mutations can be reduced to 13 and 12 in the PpDyp^PR and PpDyp^FP variants, respectively (see Table S19). Chimeric variants 10G9 (with 17 mutations) contain the hypothetical 13/13 essential mutations, and variant 5F8 (21 mutations) includes 11/13 of those crucial mutations. Furthermore, both web servers predicted five mutations: E91D, L110F, L120R, V217M, and E237D, which are present in establishing new H-bonds and hydrophobic networks and altering the number and size of cavities in most of the thermophilic enzymes investigated. Therefore, these mutations appear to be essential in the molecular mechanisms of PpDyP thermostabilization.

Investigation of Conformational Features behind Improved Thermostability

The effect of mutations on protein’s flexibility was analyzed on the molecular dynamics simulations (MDs) of monomeric and tetrameric forms of wild-type and variants at different temperatures (300, 331, and 343 K (this higher temperature was tested only for PpDyP^PR), which corresponds roughly to 27, 58, and 70 °C). The structural variability and fluctuations along the MD trajectories were assessed using the C-alpha root-mean-square deviations (RMSD) and root-mean-square fluctuations (RMSF).

Comparison among monomers reveals that PpDyP^PR at 343 K exhibits comparable or lower variability and fluctuations compared to the wild-type and PpDyP^FP at 331 K, consistent with its experimentally determined higher thermostability (Figures S15 and S16a,b, and Table S20). PpDyP^PR maintains the integrity of its hydrophobic interactions across all temperatures, unlike wild-type and PpDyP^FP, where these interactions are progressively lost as the temperature increases (Figure S18). Five loop regions in the monomers exhibit differential fluctuations when compared to the wild-type, with four located near key interfaces, regions 1 and 2 near the noncanonical interface, and regions 3 and 5 near the canonical interfaces (Figure S16e). Regions 3 and 4 are more rigid in both thermostable enzymes, whereas regions 1, 2, and 5 are more or less rigid than the wild-type depending on the variant. Notably, the α3-helix (residues 91–103) also appears more rigid and thus further stabilized upon mutation (Figures S16a,b and S17) by the insertion of mutations L120R and H125R, H121Y, and G123D (Figure S16d–f).

The tetrameric forms of PpDyP^FP and the wild-type enzyme exhibited RMSD and RMSF values lower than their monomeric counterparts, indicating that a stable quaternary structure enhances adaptability to high temperatures (Figure S19 and Table S20). Furthermore, the PpDyP^FP tetramer was more rigid than the wild-type, particularly in regions 3 and 4, and demonstrated higher structural resilience to temperature fluctuations, suggesting improved stability under varying thermal conditions (Figure a,b). Tetrameric PpDyP^FP exhibited slightly more hydrophobic and aromatic–aromatic interactions than the wild-type (data not shown).

4.

Structural fluctuations and effect of mutations on intermonomeric interactions in canonical and noncanonical interfaces of wild-type and PpDyP^FP tetramers analyzed using molecular dynamic simulations. C-alpha RMSFs at 300 K for the wild-type (dark green) and PpDyP^FP (dark red) (a), 331 K for the wild-type (light green) and PpDyP^FP (light red); the gray areas represent the regions with higher fluctuations (b). Cartoon representation of the PpDyP^FP tetramer showing the regions belonging to the noncanonical (1 and 2) and canonical (3, 5, and 6) interfaces and more solvent-exposed (4 and 7) (c). In the PpDyP^FP tetramer, the mutations S25F, E91D, D94E, and H125R are represented as spheres, where the carbon atoms are colored according to the corresponding subunit. The oxygen and nitrogen atoms are colored red and blue, respectively. The mutations involved in intramonomeric interactions (A54 V, L110F, and L120R) are shown as sticks (d). Mutations E91D, D94E, and H125R (all spheres) and residues (as sticks) are involved in interactions in the canonical interface (e). Mutation S25F (as spheres) and residues (as sticks) are involved in noncanonical interface interactions (f). Changes in the interactions between residues at the interfaces, in wild type versus FireProt, color-coded according to the number of interactions from light blue (low) to dark red (high) (g).

The tetrameric forms of PpDyP^FP exhibit a significantly lower flexibility compared to the wild-type, particularly at the canonical interface (Figure a–c). The insertion of mutations E91D, D94E, and H125R results in the formation of new intermonomer H-bond interactions (Figure e,g). Notice that at each interface, these interactions appear twice, e.g., E91D (chain A)-H121 (chain D) and E91D (chain D)-H121 (chain A), thereby multiplying their contribution to protein stability. At the noncanonical interface, the rigidification is small, even though mutation S25F promotes new intermonomeric interactions (e.g., S25F–Y166 and E27-R259) (Figure d–g). However, the effect could be slightly higher as the temperature increases, in region 6 that contains Y166 (Figure b) to the E27-R259 salt-bridge (Figure d,f,g). Interestingly, the most essential mutations (E91D, D94E, H125R, and S25F) identified to contribute to protein–protein interface stabilization are retained in the chimeric thermostable 11B8 variant that resulted from shuffling of PpDyp^FP and wild-type (Figure d). This variant shows only four fewer mutations than PpDyp^FP. MD simulations revealed that the absence of two of them (S169A and V279L) resulted in increased intermonomer and solvent-exposed hydrogen bonds (data not shown), which can most likely explain its enhanced stability. In summary, mutations previously identified in the intramonomer are also involved in intermonomer interactions, as seen in the improved stability of the PpDyP^FP tetramer (Figures and ).

Shortest Path Method (SPM) Analysis Reveals Reinforced Protein Networks in Thermostable Variants

We evaluate the effectiveness of the Shortest Path Map (SPM) tool in determining whether mutations and the resulting changes in protein stability are reflected in the dynamics of residue interaction networks and in identifying potential structural signatures of enhanced stability. , In physical terms, a residue interaction network represents the protein as a dynamic, interconnected system where each amino acid residue acts as a node, and the edges between nodes correspond to physical interactions or correlated atomic motions over time. When derived from molecular dynamics (MD) simulations, this network captures both static proximity and time-averaged dynamic correlations that reflect how residues move in concert or influence one another during thermal fluctuations. Notably, the analysis of SPMs of monomers of thermostable variants displays (i) larger spheres, indicating stronger residue correlations and reflecting dense intrasubunit contacts, as well as (ii) thicker edges, which represent more persistent and energetically favorable interactions over time than those of the wild-type. This may indicate tighter internal cohesion (Figures a,b, and S20). SPMs of thermostable variants include a high proportion of mutated residues, including 11 of the 14 mutations most critical for PpDyP^PR stability and all 12 essential mutations for PpDyP^FP (Table S21). Notably, six and seven of the eight mutations predicted by both web servers appear directly or indirectly in the SPM analyses of the variants. Importantly, the mutations are consistently engaged in SPM networks located in similar structural regions of the thermostable monomers, for example, S218 (which presents larger spheres in PpDyP^PR and PpDyP^FP) is connected to α3 (E91D) through β4 (G118A or F119), and β1 (H15) in both thermostable variants but was unconnected in the wild-type (Figures a,b, and S20).

5.

Shortest Path Map (SPM) of monomeric and tetrameric forms of the wild-type and thermostable variant. In the SPM of wild-type (a), the red spheres and sticks (10) represent the residues that will be replaced in the variant and that are directly and indirectly involved in the SPM. In SPMs of PpDyP^FP (b) monomers, colored in light pink, the red spheres represent mutations in the SPM, and the red sticks represent mutations that interact with residues in the SPM. This SPM shows that 18 out of 21 mutations, where 12 are directly (red spheres) or 6 are indirectly (red sticks), involved in the map. The black and blue circles represent the mutations that participate in the interaction at the tetramer interfaces, with blue indicating the non-canonical and black indicating the canonical interfaces, as shown in (e). For simplicity, in the SPMs of the tetramers, only the mutations from chains B and C are shown for the wild-type (c) and PpDyP^FP (e) structures. SPM of PpDyP^FP tetramer shows that 10 (out of 21) mutations are directly (3, spheres with the same color as the corresponding chain but in a darker tone) or indirectly (7, sticks with the same color as the corresponding chain but in a darker tone) involved on the map. Simplified schemes of the detailed SPMs of the tetramers have been drawn for the wild-type (d) and PpDyP^FP (f) proteins, with the sphere’s size and the line’s thickness proportional to the degree of correlation in the map. The residues highlighted by circles represent the ones involved in intermonomeric dynamic linkage, which are also described in (c) and (e).

Tetrameric PpDyP^FP displays a more complex and solid network than wild-type, evident from the larger spheres and thicker connections (Figure c–f). Notably, of the five common mutations identified as more relevant for the thermal stability of the enzyme (E91D, L110F, L120R, V217M, and E237D; Table S19), four are present in the SPM network, with L120R and L110F appearing in central nodes. The identical four residues are involved in the wild-type SPM directly or indirectly, which suggests that the SPM approach, even when applied independently, is likely to highlight these residues as relevant. Interestingly, new distinct architectural structural aspects are now visible in the networks of tetramers: the maps reveal a (iii) reinforced dynamic connectivity between monomers through their interfaces, forming interwoven, ring-like communication chains linking the four monomers, highlighting structural interplay and expanded interaction networks within the tetrameric assembly (Figure c,e). For clarity, the standard threshold of 0.3 was used to depict SPMs in half of the tetramer. However, lowering the threshold to 0.1 revealed the presence of networks across the entire assembly, reflecting longer-range communication pathways that support coordinated motion across monomers (Figure d,f, and Table S21). In the wild-type, the G93-H121 and P77-T79 connections maintain monomer association between interfaces (Figure c,d). The PpDyP^FP thermostable variant features not one, but three distinct subnetworks per monomer: G93-Q24, H121-R259, and (to a lesser extent) H121-Q24, resulting in two dynamic connections at each interface (Figure d,f). Therefore, the thermophilic variant introduces new (and even redundant) stronger links in regions where the wild-type enzyme forms only minimal or weak contacts, and interactions occur at conventional contact points, thereby enhancing overall cohesion. These alterations result in a more interconnected and resilient architecture that reflects a shift from the wild-type’s sparse and less-correlated motions to a tightly knit, cooperative network in thermostable variants.

It is interesting to note that, even though some of the regions involved in the SPM correspond to less flexible regions (e.g., region 3), not all rigidified regions participate in the SPM (e.g., region 4). Therefore, we hypothesize that correlated motion is an emergent feature of the introduced mutations, which may contribute to the more efficient propagation of structural perturbations, making the native state more stable. Still, this is not necessarily due to enhanced rigidity, or at least, not all rigidified regions become more correlated. Thus, correlation analysis is adding an extra layer of information that goes beyond rigidity, with stronger networks emerging as a new potential fingerprint of thermostability. Further case studies will be necessary to determine whether this is a general characteristic of all thermostable proteins.

Conclusions

The structural and dynamic analyses of two thermostable DyP-type peroxidase variants, each designed using independent computational platforms, revealed convergent stabilizing features despite distinct mutation profiles. Although the specific mutations differed, they clustered in comparable structural regions and collectively enhanced protein compactness, intermonomeric interactions, and backbone rigidity. Residue interaction networks derived from molecular dynamics simulations illuminate how these mutations reshape protein dynamics. Thermostable variants display modular monomers that exhibit greater internal cohesion and cooperative dynamics, and their quaternary structures feature long-range communication pathways, facilitating coordinated motions across monomers. The data suggest that, in this enzyme, the mutations that enhance stability also increase both rigidity and dynamic coupling between regions. These findings highlight dynamic connectivity as a potential signature of thermostability in this systemone that remains hidden in conventional static structural analysesand offer insight beyond the traditional rigidity-focused view. We hypothesize that such enhanced dynamic coupling may contribute to a more efficient redistribution of local perturbations throughout the structure, thereby reducing the probability of localized unfolding events and tetramer dissociation. This interpretation aligns with the broader framework of dynamic allostery and network communication in proteins, where strongly connected or correlated networks have been proposed to help maintain global structural integrity in the face of local disruptions.

Although we cannot conclusively establish a mechanistic link, our data support the notion that specific dynamic features may emerge as properties of stabilizing mutations and may be a fingerprint of thermal resistance. We therefore propose this as a working hypothesis, grounded in our observations and consistent with existing theoretical models, while acknowledging that further experimental and computational studies are needed to test its generality.

All computer analysis scripts used for molecular dynamics are available on GitHub (https://github.com/insilichem/utils_PpDyP)

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acscatal.5c03333.

• UV–vis spectra; SEC of enzymes; temperature induced aggregation; Arrhenius plots; kinetic analysis; activity and stability screening; cartoons of structures; temperature-induced aggregation; kinetic analysis shuffling variants; kinetic analysis shuffling variants; crystal packing; Heme access; T2 obstruction; surface electrostatic potential; C-alpha RMSDs for the monomers; structural fluctuations; structural flexibility; surface electrostatic potential; C-alpha RMSDs for the tetramers; SPM in monomers; mutations suggested by Web servers; UV Vis spectra; thermodynamic stability; thermodynamic stability; list of works; kinetic stability; kinetic parameter MM; top 10 variants PROSS; top 10 variants FIREPROT; Heme content; thermostability and kinetics of PROSS; variants PROSS; top 10 variants FIREPROT; Heme content; thermostability and kinetics of PROSS; X-ray data collection, processing and refinement statistics; molecular dimensions; molecular dimension; accessible surface area; internal cavities; salt bridges; role of PROSS and FP; mean RMSD and RMSF; and SPM in tetramers (PDF)

§.

C.F.R. and D.S. contributed equally to this work.

L.O.M., D.S., and L.M. conceived this study; D.S. and C.F.R. performed the mutagenesis, kinetic, and biochemical characterization of the enzymes; C.L. and P.B. solved the crystal structures and performed the structural analysis; P.B. and L.M. performed molecular dynamics (MD) simulations and computational analysis. All authors contributed to the experimental design, results interpretation, and paper writing.

The authors declare no competing financial interest.