An In Silico Target Fishing Approach to Identify Novel Ochratoxin A Hydrolyzing Enzyme.

Ochratoxin A (OTA), a mycotoxin that is of utmost concern in food and feed safety, is produced by fungal species that mainly belong to the Aspergillus and Penicillium genera. The development of mitigation strategies to reduce OTA content along the supply chains is key to ensuring safer production of food and feed. Enzyme-based strategies are among the most promising methods due to their specificity, efficacy, and multi-situ applicability. In particular, some enzymes are already known for hydrolyzing OTA into ochratoxin alpha (OTα) and phenylalanine (Phe), eventually resulting in detoxification action. Therefore, the discovery of novel OTA hydrolyzing enzymes, along with the advancement of an innovative approach for their identification, could provide a broader basis to develop more effective mitigating strategies in the future. In the present study, a hybrid in silico/in vitro workflow coupling virtual screening with enzymatic assays was applied in order to identify novel OTA hydrolyzing enzymes. Among the various hits, porcine carboxypeptidase B was identified for the first time as an effective OTA hydrolyzing enzyme. The successful experimental endorsement of findings of the workflow confirms that the presented strategy is suitable for identifying novel OTA hydrolyzing enzymes, and it might be relevant for the discovery of other mycotoxin- mitigating enzymes.

. Docking results of carboxypeptidases (Carboxypeptidase T (CPT) and carboxypeptidase A (CPA)). Proteins are represented in cartoon, ligands in sticks and Zn2+ ions in spheres. (A) Comparison between the calculated (yellow) and crystallographic (white) pose of CXA within CPA (PDB ID: 1IY7). (B) Comparison between the calculated (yellow) and crystallographic (white) pose of CXA within CPT (PDB ID: 4DJL). (C) Comparison between the calculated pose of OTA within CPA (white; model derived from Protein Databank (PDB) structure having code 1IY7) and CPT (yellow). .
Subsequently, the complex underwent MD simulations to check its geometrical stability. Both OTA and CPT showed a marked geometrical stability throughout the simulation, as shown in Figure S2A and S2B. In addition, the analysis of OTA trajectory revealed its persistence within the catalytic site stably adopting an orientation likely suitable to undergo the reaction of OTA hydrolysis into OTα and Phe ( Figure  S2C, Figure 2).

Carboxypeptidase B (CPB)
CPB is a porcine homologous of the bovine carboxypeptidase A (CPA). Similarly to CPT, this enzyme is recorded in PDB in complex with CXA (phenylalanine-N-sulphonate). Therefore, CXA was docked and its pose was compared to its crystallographic architecture of binding in order to assess the model reliability. CXA scored 84.7 GoldScore units. The calculated pose of CXA was in strong agreement with its crystallographic architecture supporting, also in this case, the model reliability, as shown in Figure S3A. Concerning OTA, it scored 95.3 GOLDScore units, suggesting its capability to well satisfy the pocket, as shown within CPA and CPT. In addition, the computed pose of OTA within CPB was compared with the one calculated within CPA. OTA showed the same mode of binding in both enzymes ( Figure S3B), further supporting its substrate-like interaction within this enzyme. Figure S3. Docking results of CPB. Proteins are represented in cartoon, ligands in sticks and Zn2+ ions in spheres. (A) Comparison between the calculated (yellow) and crystallographic (white) pose of CXA within CPB (PDB ID: 5J1Q). (B) Comparison between the calculated pose of OTA within CPA (white) and CPB (yellow). . Subsequently, the OTA-CPB complex underwent MD simulations to check the geometrical stability. The protein was found to be stable throughout the simulation, while OTA showed an early change of geometry, which was kept stable up to the end of simulation ( Figure S4A and S4B). Notably, the analysis of OTA trajectory excluded the existence of outward trajectories. In addition, the thorough analysis of OTA movements pointed out that the coumarin-like moiety was mainly reorganized during the simulation, while the amidic bond and Phe-part were both kept stable and theoretically well organized with respect to Zn ion to undergo the reaction of OTA hydrolysis into OTα and Phe ( Figure S4C, Figure 2).

Serine Carboxypeptidase II (Ser-CP II)
Ser-CP II is a serine-carboxypeptidase from wheat related to the already known OTA-hydrolyzing enzyme Carboxypeptidase Y (CPY) from bakery yeast. The Ser-CP II is recorded in PDB in the bound state with FC0 (N-carboxy-L-phenylalanine), one of the OTA-like compounds identified in the ligand-based virtual screening. FC0 was docked and its pose was compared to its crystallographic architecture of binding to assess the reliability of docking procedure on this model. The score recorded for FC0 was 57.5. The calculated and crystallographic poses were in strong agreement, suggesting the model reliability, as shown in Figure S5A. Subsequently, the interaction between OTA and Ser-CP II was calculated and compared with the calculated pose of OTA within CPY. The scores recorded for OTA within the Ser-CP II and CPY were 54.7 and 73.4, respectively, pointing to the capability of OTA to satisfy the requirements of both pockets well. However, the interaction with the latter could not be considered more favorite, in spite of the higher score recorded, due to the lack of complete training and validation procedures. Concerning the architectures of binding, OTA showed a slight different organization due to the Thr60Tyr substitution in the binding pocket of Ser-CP II that caused a different organization of the Phe moiety of OTA ( Figure 5B). Subsequently, the OTA-serine carboxypeptidase II complex underwent MD simulations to check its geometrical stability. The protein was found stable throughout the simulation, while OTA was found to be more mobile ( Figure S6A and S6B). In addition, the analysis of OTA trajectory revealed a marked trajectory outward the inner part of catalytic site ( Figure S6C).

Neprilysin
Neprilysin is a Zn-dependent endopeptidase that is able to hydrolyze peptides up to 50 residues. This enzyme is recorded in PDB in the bound state with TIO [thiorphan: (2-mercaptomethyl-3-phenylpropionyl)-glycine], one of the OTA-like compounds identified in the ligand-based virtual screening. Therefore, TIO was docked and its pose was compared to its crystallographic architecture of binding in order to assess the reliability of docking procedure on this model. The score recorded for TIO was 64.5. The calculated and crystallographic poses were in strong agreement, suggesting the model reliability, as shown in Figure S7A. Subsequently, the interaction of OTA was calculated, recording a score of 87.9 units. The Phe portion of OTA superimposed the benzene moiety of thiorphan orienting the amide bond to the Zn ion in a reasonably proper position to undergo hydrolysis, as shown in Figure S7B. Subsequently, the OTA-neprylisin complex underwent molecular dynamic simulations to check its geometrical stability. The protein was found stable throughout the simulation while OTA was found to be more mobile ( Figure S8A and S8B). In particular, the analysis of trajectories revealed that OTA had an early increase of RMSD, mainly due to the reorganization of the coumarin-like moiety, but the overall organization of the amidic bound in respect to the Zn ion was kept stable throughout the simulation ( Figure  S8C).

Urokinase
Urokinase is a serine protease that is related to the already characterized OTA hydrolyzing enzyme bovine alpha chymotrypsin. This enzyme is recorded in PDB in the bound state with 9UP [methyl(7carbamimidoylnaphthalen-1-yl)carbamate], one of the OTA-like compounds identified in the ligand-based virtual screening. Therefore, the computed pose of 9UP was compared to its crystallographic architecture of binding to assess the reliability of docking procedure on this model. The score that was recorded for 9UP was 56.0. The calculated and crystallographic poses were in strong agreement suggesting the model reliability, as shown in Figure S9A. Subsequently, the interaction of OTA was calculated, recording a score of 64.3 units. The coumarin-like moiety of OTA superimposed the aromatic portion of 9UP, while the amide bond was in a reasonably well-oriented position in respect to the catalytic serine to undergo hydrolysis, as shown in Figure S9B. Subsequently, the OTA-urokinase complex underwent molecular dynamic simulations to check its geometrical stability. The protein was found to be stable throughout the simulation, while OTA was found more mobile ( Figure S10A and S10B). In particular, the analysis of trajectories revealed that the Phe moiety was the most mobile, while the coumarin-like moiety was found to be quite stable along the all simulation ( Figure S10C). Nevertheless, OTA was found to be stably buried within the catalytic cleft during the all simulation. Notably, the mobility of Phe moiety started to increase early (nearly after 3 nsec of simulation) and tended to get stabilized again similarly to the starting pose toward the end of simulation (in the last 5 nsec), as shown in Figure S10B.

Endothiapepsin
Endothiapepsin is a pepsin-like aspartic peptidase. This enzyme is recorded PDB in the bound state with 46L [6-(dimethylamino)pyridine-3-carboxylic acid], one of the OTA-like compounds identified in the ligand-based virtual screening. Therefore, the computed pose of 46L was compared to its crystallographic architecture of binding to assess the reliability of docking procedure on this model. The score recorded for 46L was 35.4. The calculated and crystallographic poses were in strong agreement suggesting the model reliability, as shown in Figure S11A. Subsequently, the interaction of OTA was calculated, recording a score of 62.4 units. OTA adopted an organization that was markedly different from the one of 46L, as shown in Figure S11A. Nevertheless, the amidic bond was placed in correspondence to the catalytic core of the enzyme adopting a position theoretically suitable to undergoing hydrolysis. Subsequently, the OTA-endothiapepsin complex underwent molecular dynamic simulations to check its geometrical stability. The protein and OTA both showed an overall geometrical instability, as marked RMSD variations were observed, as shown in Figure S12A and S12B. In addition, the analysis of OTA trajectories revealed a marked outward movement from the catalytic site.

Cathepsin A
Cathepsin A is a broad-spectrum serine protease that is homologous to the already characterized OTAhydrolyzing enzyme carboxypeptidase Y (CPY). This enzyme is recorded in PDB the bound state with S61: (S)-3-{[1-(2-fluro-phenyl)-5-hydroxy-1H-pyrazole-3-carbonyl]-amino}-3-o-tolyl-propionic acid, one of the OTA-like compounds identified in the ligand-based virtual screening. Therefore, the computed pose of S61 was compared to its crystallographic architecture of binding to assess the reliability of docking procedure on this model. The score recorded for S61 was 64.7. The calculated and crystallographic poses were in good agreement suggesting the model reliability, as shown in Figure S13A. Subsequently, the interaction of OTA was calculated, recording a score of 59.6 units. OTA adopted a binding pose resembling the one observed within the wheat serine carboxypeptidase II (Ser-CP II; see above) and arranging the amidic bond in a possibly suitable orientation to undergo hydrolysis, as shown in Figure S13B. Subsequently, the OTA-cathepsin A complex underwent MD simulations to check its geometrical stability. Protein and OTA showed an increasing mobility trend along the considered timeframe, though of slight intensity, as shown in Figure S14. In addition, the analysis of OTA trajectory did not show a geometrically stable interaction with the site. This might be due to the incompleteness of the model used (seven residues close to the OTA binding site were missing).

Matrix Metalloproteinase 12 (MMP-12)
MMP-12 is a Zn-and Ca-dependent protease that cleaves a broad range of substrates. This enzyme is recorded in PDB in the bound state with DSV: N-(dibenzo [b,d]thiophen-3-ylsulfonyl)-L-valine, one of the OTA-like compounds identified in the ligand-based virtual screening. Therefore, the computed pose of DSV was compared to its crystallographic architecture of binding to assess the reliability of docking procedure on this model. The score recorded for DSV was 83.4. The calculated and crystallographic poses were in strong agreement, suggesting the model reliability, as shown in Figure S15A. Subsequently, the interaction of OTA was calculated, recording a score of 59.7 units. The superimposition of the calculated pose of OTA with the crystallographic pose of DSV revealed that the carboxylate of OTA arranged similarly to the sulphonyl group of DSV, rather than mimicking its carboxylate group, as shown in figure S15B. Subsequently, the OTA-MMP12 complex underwent molecular dynamic simulations to check its geometrical stability. The protein and OTA were found overall stable along the simulation, as shown in Figure S16. In addition, the analysis of OTA trajectory revealed that the toxin stably persisted close to the catalytic site arranging the amidic bond in the proper orientation to undergo hydrolysis.

Beta-Secretase 1 (BACE1)
The enzyme is an endopeptidase with a relatively non-stringent specificity that is homologous to the aspartic proteases of the pepsin family. The protein is recorded in PDB in the bound state with ZY4, one of the OTA-like compounds identified in the ligand-based virtual screening. Therefore, the computed pose of ZY4 was compared to its crystallographic architecture of binding to assess the reliability of docking procedure on this model. The score recorded for 1HN was 99.9. The calculated and crystallographic poses were in strong agreement suggesting the model reliability, as shown in Figure S17A. Subsequently, the interaction of OTA was calculated, recording a score of 58.2 units. Figure S17B illustrates OTA oriented the amidic bond towards the catalytic aspartates and the superimposition of its pose with the crystallographic pose of ZY4. Subsequently, the OTA-BACE1 complex underwent MD simulations to check its geometrical stability. Protein and OTA got stabilized toward the end of the simulation, as shown in Figure S18. In addition, the analysis of OTA trajectory revealed that the toxin described an outward movement followed by an inward migration toward the binding site in the last part of the simulation.

HPLC-MS Chromatograms
(E) Representation of OTα peak (blue) measured in CPB reaction sample after 180 minutes of incubation (see above Figure C Table S1. First instance list of PDB identification codes of ligands and proteins identified in the virtual screening before the manual inspection.