Probing the role of cation-π interaction in the thermotolerance and catalytic performance of endo-polygalacturonases

Understanding the dynamics of the key pectinase, polygalacturonase, and improving its thermotolerance and catalytic efficiency are of importance for the cost-competitive bioconversion of pectic materials. By combining structure analysis and molecular dynamics (MD) simulations, eight mutagenesis sites having the potential to form cation-π interactions were identified in the widely used fungal endo-polygalacturonase PG63. In comparison to the wild-type, three single mutants H58Y, T71Y and T304Y showed improved thermostability (the apparent Tms increased by 0.6−3.9 °C) and catalytic efficiency (by up to 32-fold). Chromatogram analysis of the hydrolysis products indicated that a larger amount of shorter sugars were released from the polygalacturonic acid by these three mutants than by the wild-type. MD analysis of the enzyme-substrate complexes illustrated that the mutants with introduced cation-π interaction have modified conformations of catalytic crevice, which provide an enviable environment for the catalytic process. Moreover, the lower plasticity of T3 loop 2 at the edge of the subsite tunnel appears to recruit the reducing ends of oligogalacturonide into the active site tunnel and initiates new hydrolysis reactions. This study demonstrates the importance of cation-π interaction in protein conformation and provides a realistic strategy to enhance the thermotolerance and catalytic performance of endo-polygalacturonases.


Results
Identification of mutagenesis sites in PG63 to form the cation-π interaction. Based on the primary sequence of PG63 (342 residues, Fig. 1A), amino acids that are likely involved in the formation of the cation-π interaction were identified, including 15 Lys, 7 Phe, 6 Tyr, 5 Arg, and 5 Trp residues (38 in total). Because Lys and Tyr lead to favorable cation-π interactions in thermophilic proteins 24 , residues in close proximity to the 38 targets of modeled PG63 were substituted to Lys or Tyr in silico. As a result, 43 mutants were obtained in silico (Additional file 1). On this basis, the brute-force approach combined with 10-ns molecular dynamics (MD) simulations was used to examine the occupancy rate during the trajectory of each potential cation-π interaction. The possibility of each mutant forming a cation-π interaction was assessed according to two rules: 1, the distance cutoff between the mass center of the positively charged group and aromatic residue containing the π -system should be less than or equal to 6 Å, and 2, the absolute value of the dihedral angle defined between the side chain of two residues should be less than or equal to 60°. In combination with the MD simulation result, only eight mutants have an occupancy rate of > 50% (Additional file 1). These potential mutagenesis sites, i.e. G41Y, H58Y, T71Y, A74Y, Q129Y, D220Y, N294Y, and T304Y, are all located on the protein surface and far from the catalytic center except for Gln129 (as shown in Fig. 1B). Substitution of these residues is presumed to have no direct effect on catalytic efficiency, but probably makes an indirect impact through non-covalent interactions.

Experimental characterization of selected mutants.
To test the effect of cation-π interaction on enzymatic properties, the eight mutants described above were successfully constructed by site-directed mutagenesis. The wild-type PG63 and its mutants were then produced in P. pastoris GS115 and secreted into the culture media. The aim of this study is to find out whether and how cation-π interaction can affect the catalytic performance of the enzyme, and the specific activity of each enzyme towards PGA (pH 4.0, 50 °C and 10 min) were first compared. Among the eight mutants, mutants H58Y, T304Y and T71Y exhibited significantly higher specific activities than wild-type PG63 (14,900 ± 838, 2,600 ± 67, and 1,900 ± 59 U/mg vs. 1,600 ± 51 U/mg, respectively). Therefore, the triple mutant H58Y/T71Y/T304Y was produced as described above and had a specific activity of 5,100 ± 204 U/mg under the same conditions, indicating that cation-π interaction at the specific site contributes to the improvement of catalytic efficiency; however, more cation-π interactions have no additive effects. The reason might be that more cation-π interactions mitigate the adjacent interaction network by competition (e.g., as with a salt bridge) 25 .

Isothermal-titration calorimetry (ITC) assay.
To study the substrate binding mechanism, inactive wild-type PG63 and its three single mutants were each constructed by replacing the catalytic Asp182 with Asn 16 . Using GalpA3 as the substrate, the thermodynamics of enzyme-substrate were determined using the ITC. The raw binding curves (Additional file 3) all fit well to the independent binding models, with a calculated value (stoichiometry) of n = 1 ± 0.08. As shown in Table 2, all mutants had stronger affinity to GalpA3 than wild-type PG63, with an increased K a value (4.20− 5.19 × 10 2 M −1 vs. 3.34 × 10 2 M −1 ). The substrate binding of all enzymes is enthalpy-driven, as the enthalpy values (Δ H) are all negative. The highly favorable Δ H values offset the negative entropy values (Δ S) and result in negative Δ G 0 values. Compared to the wild-type PG63 (− 3.51 kcal/mol), the three single mutants showed increased Δ G 0 values of − 2.97, − 0.16, and − 1.52 kcal/mol, respectively. These Δ G 0 increases might be ascribed to the interaction strength between substrate and enzyme and the changes in conformational entropy 26 . These results are consistent with the decreased K m values.
Hydrolysis product analysis. Wild-type PG63 hydrolyzed PGA into small oligogalacturonides in an endo-action mode 23 . The hydrolysis profiles of PGA by wild-type PG63 and its three single mutants showed a similar mode of action but varied in hydrolysis efficiencies (Fig. 3). After incubation with PGA for 60 min, the main hydrolysis products by wild-type PG63 were GalpA2 and GalpA3. However, a larger amount of shorter sugars (i.e. GalpA, GalpA2 and GalpA3) were detected in the three single mutants (H58Y, T71Y and T304Y). Among them, mutant H58Y had the highest hydrolytic activity than mutants T71Y and T304Y.

Molecular docking.
To elucidate the underlying mechanisms of cation-π interaction in endo-PG thermotolerance and catalysis improvement, the GalpA6 was docked into the substrate-binding pocket of wild-type PG63, followed by minimization. Based on the active-site architecture of galacturonate units co-crystallized with Stereum purpureum endo-PG I (1KCD) (Additional file 4), which authorizes efficient contacts of the GalpA6 with the catalytic residues, the wild-type PG63 binds to GalpA6 along the − 4 to + 2 subsites, yielding one GalpA2 and one GalpA4 as the hydrolysis products (Fig. 4A). In addition to the conserved side chains of Asn97, residues Arg237, Lys239, and Tyr272 are all involved in binding GalpA at + 1 subsite. The atom NE2 of His150 of S. purpureum endo-PG I (1KCD, corresponding to His158 in PG63) forms a hydrogen bond with the O2 hydroxy group of GalpA at + 1 subsite 16 . However, the side chain of His158 is disordered and does not interact with GalpA6. 158HNTD161 (PG63 numbering) is one of the four strictly conserved sequence segments in fungal PGs 9 ; therefore, the residue His158 probably has an evolutionary relationship with the functions of the enzyme.
Scientific RepoRts | 6:38413 | DOI: 10.1038/srep38413 MD simulation. In order to gain insights into the substrate binding in the catalytic crevice of the enzyme, the wild type enzyme-substrate complex was submitted to in silico mutagenesis with PyMOL. During the course of these simulations, the root mean square deviation (RMSD) plots of the α -carbon atoms of four systems (wild-type PG63 and its three single mutants) showed stable plateaus after 30 ns (Additional file 5), indicating that each system maintains equilibrium within the last 20 ns. Meanwhile, the plasticity of wild-type PG63 and its three single mutants was evaluated by root mean square fluctuations (RMSF) analysis through two independent runs , where (k cat /K m ) mut and (k cat /K m ) wt are the k cat /K m ratios of the mutant and wild type enzymes, respectively, R is the ideal gas constant, and T is the temperature in Kelvin.
( Fig. 5A). A few regions, especially the loops between β -sheets, displayed significant differences in mobility. For instance, the T3 loops 1 (residue 90− 106) and 2 (residue 147− 162) of mutants H58Y and T71Y exhibit lower mobility relative to that of wild-type PG63, and the T3 loop 2 (residue 147− 162) and T1 loop (residue 269− 289)    in wild-type PG63 are more flexible than those in mutant T304Y. These local effects caused subtle differences and could have altered the interactions between some specific residues, as depicted in Fig. 5B-D. In mutant H58Y (Fig. 5B), one of the positively charged groups in the side chain of Arg89, i.e. the NH2 atom, participates in the formation of a cation-π interaction with Trp90 (4.0 Å), and another NH1 atom forms a cation-π interaction with Tyr58 (5.2 Å). This double dentate mode appears to play an important role in stabilizing the protein.
Furthermore, the Phe56 located on the T3 loop interacts with Arg89 via two hydrogen bonds in mutant H58Y (one more than that in wild-type PG63), restricting the mobility of the T3 loop in the N-terminus. In mutant T71Y (Fig. 5C), residues Tyr71 and Arg113 form a cation-π interaction (5.7 Å) that is located on the two adjacent loops. Meanwhile, four hydrogen bonds are identified as in wild-type PG63, which stabilizes the three loop regions. Similarly, residue Tyr304 interacts with Lys330 through a cation-π interaction (4.8 Å), and two more hydrogen bonds (S301@O/K330@NZ and S302@O/K330@NZ) are identified in mutant T304Y than in wild-type PG63 (Fig. 5D). Located on the reducing end of the active site, where a long loop freely swings, these bonds contribute to stabilizing the last turn of the loop. The MD trajectories also provide valuable information on the interactions between substrate and enzyme. It is worth noting that the T3 loop 2 region (residue 147− 162) has much lower plasticity in all single mutants than in wild-type PG63 (Fig. 5A). The rigidity of the T3 loop 2 region follows the order of mutant H58Y > T304Y > T71Y > wild-type PG63, which is consistent with the catalytic activity of the enzymes. Hence, the catalytic crevices of the enzymes were compared with each other. The resolved endo-PG's crystal structure shows that His158 and Tyr272 are located at the two sides of the entrance of the catalytic cleft and function primarily in substrate recognition 16,17 . The computed GalpA6 trajectories were inspected regularly, as shown in Fig. 6. The GalpA6 is positioned corresponding to the catalytic triad of wild-type PG63 along the − 4/+ 2 subsites, and remains in a state of equilibrium to enable catalysis during the simulation. In contrast, the three single mutant systems show that GalpA6 starts to slide from the binding cleft at about 18 ns, although it remains connected to the active-site residues. During the final simulation stage of mutant H58Y, the conformation of GalpA6 in the crevice was displaced by two binding sites with respect to its starting position, to evolve from − 4/+ 2 to − 2/+ 4 subsites. A similar phenomenon occurred in mutants T71Y and T304Y systems with one binding site to evolve from − 4/+ 2 to − 3/+ 3 subsites.

Discussion
The noncovalent cation-π interaction relevant to structural biology and other diverse roles has been gaining increasing attentions. For example, cation-π interactions are well known to have a stabilizing role in the secondary, tertiary, and quaternary structure of proteins through electrostatic interactions 24,27 . The Arg residues scattered over the surface of the fibronectin III-like (FnIII) domain of a β -glucosidase play an important role in the interaction with lignin by means of cation-π stacking with the aromatic rings 28 . Additionally, the aromatic side chain of Tyr74 in the amino acid antiporter AdiC is a critical pH sensor that is mediated by cation-π interaction 29 . However, the contribution is currently unknown with respect to cation-π interaction's role in the catalytic performance of enzymes. Our goal herein is thus to probe the role of cation-π interaction in catalysis by introducing cation-π interaction into the widely used GH28 endo-PG, PG63, from Penicillium sp. CGMCC 1669.
The intramolecular cation-π interaction is a well-known factor responsible for protein thermostability. For example, our study is in agreement with other studies that thermostability can be significantly enhanced by the cation-π interaction between Lys202 and Trp204 of an alkaline α -amylase from Alkalimonas amylolytica and Lys7 and Trp99 of a tyrosinase from Streptomyces kathirae SC-1 30,31 . The double dentate mode of cation-π interactions (Arg89-Trp90 and Arg89-Tyr58) in mutant H58Y, the cation-π interaction Arg113-Tyr71 located on the two adjacent loops in mutant T71Y, and the cation-π interaction Lys330-Tyr304 stabilizing the C-terminal long loop freely swings in mutant T304Y showed a significant increase in apparent T m value of ~3.9 °C, 0.6 °C, and 3.4 °C, respectively. In summary, these conformational shifts within the active loop may account for the increased thermostability of three mutants. On the other hand, the thermal performance of the mutants follows the order of wild-type PG63 < T71Y < T304Y < H58Y and corresponds to their occupancy rates (54.8%, 81.9%, and 83.0%; Additional file 1). This indicates that the computational method we used is reliable and can provide decisive guidance in designing proteins for greater thermotolerance.
Theoretically, the amounts of reducing ends generated from PGA hydrolysis after treatment with the same amount of enzymes should be the same. However, significant differences were observed between wild-type PG63 and its mutants (Fig. 3). A larger amount of small oligogalacturonides were released from PGA by the three single mutants than by the wild-type. That is, wild-type PG63 failed to degrade relatively longer oligogalacturonides, whereas the three mutants had this capacity. These results are also supported by the profound changes of k cat /K m of the three single mutants on the PGA (Table 1). However, this action mode is far from processivity due to the similar product ratios of GalpA2 and GalpA3. Beyond that, processivity is dependent on an arginine residue at a position commensurate with subsite − 5 in PG 32 . PG63 has a serine residue at this position and thus functions in a non-processive mode. The mutants' capacity to degrade relatively longer oligogalacturonides is thought to be critical to improving the catalytic efficiency of substrate hydrolysis 33 .
For the MD trajectories, GalpA6 can move along the catalytic crevice without completely leaving the cleft, which helps catalytic residues immediately initiate continuous hydrolysis. Based on the hydrolysis profiles of the products, the mutants hydrolyzed more relatively long oligogalacturonides to GalpA, GalpA2, and GalpA3. These results suggest that the introduction of a cation-π interaction may modify the conformation of the catalytic crevice and provide an enviable environment for catalytic processes. Given the findings described above, the authors assume that the lower plasticity of T3 loop 2 at the edge of the subsite tunnel recruits the reducing ends of oligogalacturonide into the active site tunnel and initiates new hydrolysis reactions. It is widely acknowledged that some critical residues can stack with the ring faces of sugar units, forming carbohydrate-π interactions, and play a vital role in enzyme-substrate recognition. For instance, three residues (Tyr7, Trp23, and Tyr112) constitute a harboring wall to contact the substrate through hydrophobic interactions as they occurred in the endoglucanase 3 from Trichoderma harzianum did 34 . In the wild-type PG63-GalpA6 complex, the side chain of strictly conserved His158 located on the T3 loop 2 region is disordered and observed to have different conformations; however, in the mutant complexes it converges into a single conformation. The plasticity of residue His158 is much lower in mutant enzymes by introducing a cation-π interaction than in wild-type PG63. The residue His158, located on the T3 loop 2 region, is more flexible in wild-type PG63 than those mutants, and is unable to act as a director fixed on the entrance of the catalytic cleft. From this perspective, it is similar to processivity in which the substrate binds in a tunnel, with multiple attacks occurring on the single carbohydrate chain 32 , although the three mutant enzymes are typically nonprocessive enzymes as the amounts of GalpA, GalpA2, and GalpA3 increase simultaneously. In other words, the importance of the conserved His158 located at T3 loop 2 might be comparable to that of Trp40 of cellobiohydrolase TrCel7A, the key residue to load chain ends and to initiate continuous hydrolysis 35 .

Conclusion
The thermotolerance and catalytic efficiency of the widely used endo-PG are remarkably improved through introduction of a cation-π interaction. This bond might restrict the mobility of the active loop and modify the conformation of the catalytic crevice, thus providing an enviable environment for catalytic processes. As a result, this engineering approach enhanced the efficiency of substrate degradation and the production of oligogalacturonides of lower polymerization degrees. This study reveals the significance of cation-π interaction in protein conformation and provides a realistic strategy to enhance the catalytic performance of proteins.

Methods
Plasmids, strains, culture media and materials. The recombinant plasmid used in this work was pPIC9-pg63, which contained the gene pg63 (HQ446162) from the strain Penicillium sp. CGMCC 1669 23 . Sitedirected mutagenesis was carried out using the specific primers (Additional file 6) and two-step polymerase chain reactions (PCRs). All mutants were verified by double-stranded plasmid sequencing. Escherichia coli Trans I-T1 (TransGen, Beijing, China) and P. pastoris GS115 (Invitrogen, Carlsbad, CA) were used for plasmid amplification and heterologous expression, respectively. The pEASY-T3 vector (TransGen) and pPIC9 (Invitrogen) were used for plasmid construction, respectively. Culture media for His + transformants selection and P. pastoris growth Scientific RepoRts | 6:38413 | DOI: 10.1038/srep38413 and induction were prepared according to the manual of the Pichia Expression kit (Invitrogen). Mono-, di-, and trigalacturonic acid (GalpA, GalpA2 and GalpA3) and the substrate PGA were purchased from Sigma-Aldrich (St. Louis, MO). All other standard chemicals were of analytical grade.
Identification of potential mutagenesis sites. Theoretically, when the cationic side chain of lysine or arginine is in close proximity to the side chain of an aromatic residue (tryptophan, tyrosine, or phenylalanine), the geometry is biased toward the formation of a favorable cation-π interaction 36 . Moreover, the side chains of electron-rich tryptophan, tyrosine, and phenylalanine can engage in a variety of supramolecular interactions 20 . Hence, the potential targets (Lys, Arg, Phe, Trp, and Tyr) involved in the formation of a cation-π interaction were identified in the deduced protein sequence of PG63. The CaPTURE program 36 and a brute-force approach were used to assess the possibility of each residue substitution to form a cation-π interaction with one of the candidate targets identified above. Then, the 10-ns MD simulations were carried out with an AMBER 14 simulation package for further screening, and the occupancy rate during the trajectory of each putative cation-π interaction was calculated. The mutants that formed a cation-π interaction with an occupancy rate of > 50% were selected for site-directed mutagenesis.
Enzymatic activity assay and biochemical characterization. The endo-PG activity was determined by measuring the release of reducing sugars from PGA using the 3,5-dinitrosalicylic acid (DNS) method with d-(+ )-GalpA as the standard, as described previously 37,38 . One unit of endo-PG activity was defined as the amount of enzyme that released reducing sugars equivalent to 1 μ mol of d-(+ )-GalpA per minute under assay conditions (pH 4.0, 50 °C and 10 min). All reactions were repeated three times. The protein concentrations were determined using the Bio-Rad protein assay kit (Boston, MA).
The pH-activity profiles of wild-type PG63 and its mutants were determined at 50 °C for 10 min in 100-mM McIlvaine buffer (pH 3.0-8.0). The optimal temperature was determined by performing the same process at temperatures ranging from 0 to 80 °C in 100-mM McIlvaine buffer (pH 4.0) for 10 min. The enzyme half-lives (t 1/2 ) were determined by measuring the residual activities under standard conditions after various durations of incubation at 55 °C without substrate. For the determination of T 50 (a temperature at which 50% of the maximal activity of an enzyme is retained), the enzymes (50 μ g/mL) were incubated at a temperature range of 30 to 80 °C for 30 min without substrate, and the residual activities were then measured at standard conditions. The Michaelis constants (K m ) and kinetic constants (k cat ) of wild-type PG63 and its mutants were determined under standard conditions (pH 4.0, 50 °C, and 5 min) with PGA of 0.5 to 10 mg/mL as the substrate. The kinetic values were calculated by fitting the Lineweaver-Burk plot. All reactions were performed in triplicate.
Differential-scanning calorimetry (DSC) assay. The thermodynamic stabilities (T m , melting temperature) of wild-type PG63 and its mutants were determined on a MicroCal ™ VP-Capillary DSC (GE Healthcare) as described previously 13 . Aliquots of proteins (0.35 mg) were dissolved in 1 mL of 20-mM McIlvaine buffer (pH 4.0) and loaded onto the capillary automatically. The melting temperature, T m , corresponded to the maximum of the transition peak over 20-90 °C within a heating and scanning rate of 120 °C/h. The test was repeated at least twice. The same McIlvaine buffer was used as the blank control. ITC assay. ITC was employed to measure the binding affinities between wild-type PG63 or the three single mutants with GalpA3 at 25 °C using an ITC 200 microcalorimeter (GE Healthcare, Waukesha, WI) 39 . To avoid a catalysis reaction in the titration experiment, site-directed mutagenesis was conducted to obtain inactive wild-type PG63 and its three single mutants by replacing the catalytic residue Asp182 with Asn by overlap PCR with specific primers (Additional file 6). Protein expression and purification were conducted as described above. Hydrolysis product analysis. The capacities of wild-type PG63 and its mutants to hydrolyze PGA were evaluated using high-performance anion-exchange chromatography (HPAEC; Thermo Fisher Scientific, Sunnyvale, CA) as described previously 40 . Purified recombinant proteins (0.1 U) were incubated with 3 mg of PGA in McIlvaine buffer (pH 4.0) at 50 °C for 60 min. Samples (200 μ L) were extracted regularly, heated in a bath of boiling water for 5 min to inactivate the enzymes, and centrifuged at 12,000 × g at 4 °C for 10 min. Each sample (25 μ L) was injected into a high-performance anion exchange chromatography (HPAEC) system model 2500 equipped with a CarboPac PA200 column (3 × 250 mm). The oligosaccharides were eluted by NaOH (0.2-1 M) at a flow rate of 0.45 mL/min. GalpA, GalpA2, and GalpA3 were used as the standards.
Homology modeling, molecular docking and molecular dynamic simulation. The homology modeling of wild-type PG63 was built based on the X-ray structure of endo-PG from Colletotrichum lupini (2IQ7_A, 61% identity) by using the Swiss Model server. The structure was optimized (especially loops) for energy minimization by YASARA software (www.yasara.org) to remove steric clashes. PyMOL (Delano Scientific, Portland, OR) was applied to visualize and analyze the generated models.
AutoDock Vina was selected as the docking tool to determine enzyme-and substrate-binding conformation according to the user's guide 18,41 . Wild-type PG63 was used as the receptor, and the substrate hexagalacturonic acid (GalpA6) coordinates of the X-ray crystal structure of a pectin methylesterase in complex with GalpA6 (2NTB) was used as ligand. The C α atom of Lys239, which is an important contributor for recognizing the carboxyl group of the substrate, was selected as the center of the box. A docking grid with a size of 60 Å × 60 Å × 60 Å was used to encompass the ligand partial-protein complex. The set number generated for the docking poses was 20. The best binding conformation was selected by the criteria of interacting energy combined with the geometrical matching quality and the binding affinity (Δ G, kcal/mol) of the AutoDock Vina software. After that, energy minimization was used to relax atomic clashes for the complex using YASARA software 3,13 . This process ultimately obtained a reliable model for the PG63-GalpA6 complex.
The simulations were carried out with the AMBER 14 simulation package and repeated twice for each system. The standard AMBER force field ff99SB and GLYCAM_06j-1 were used to generate the topologies and parameters of the enzyme and substrate, respectively 42,43 . Each system was immersed in a dodecahedral periodic box of TIP3P water molecules that extended 10 Å from the protein atoms 44 . Six sodium ions were added to maintain the system's neutrality. Before the MD simulations, the water molecules/ions were resolved to minimize energy via 1,000 steps, followed by 20,000 steps for the side chains of the protein, and then 4,000 steps for the whole system to remove potentially poor contacts between the solute and solvent. After energy minimization, the systems were gradually heated from 0 to 300 K over 100 ps, followed by 50-ns production of MD simulations with a time step of 2 fs at a temperature of 300 K and pressure of 1.0 atm that were controlled by the Langevin algorithm 45 . Long-range electrostatic interactions were treated using the particle-mesh Ewald (PME) method 46 . Bonds involving hydrogen atoms were constrained by the SHAKE algorithm 47 .