Interaction of DPP4 inhibitors in the SARS-CoV-2 Mpro protomer
The interaction of 12 DPP4 inhibitors (Fig. 1) was investigated by conducting molecular docking studies using the crystal structure of SARS-CoV-2 Mpro protomer.8 The viral protease protomer is made up of 306 amino acids and consists of three domains; domains I, II and III. The substrate binding site is present between domains I and II.8 The molecular docking protocol was validated by docking the known SARS-CoV-2 Mpro inhibitor ((tert-butyl 1-((2S)-1-((2S)-4-(benzylamino)- 3,4-dioxo-1-(2-oxopyrrolidin-3-yl)butan-2-ylamino)-3-cyclopropyl-1-oxopropan-2-yl)-2-oxo- 1,2-dihydropyridin-3-ylcarbamate 1, Fig. 1)8 using the CDOCKER alogorithm, after building 1 in 3D from scratch using the computational software Discovery Studio Structure-Based Design (BIOVIA, Dassault Systemes®). This investigation shows that the peptidomimetic inhibitor 1, exhibits a similar binding mode as per the crystallized inhibitor structure reported (all heavy atom RMSD = 1.70 Å, Fig. 2).8 Consequently molecular docking of 12 DPP4 inhibitors was carried out using CDOCKER algorithm. The binding modes were analyzed by ranking the best poses obtained using CDOCKER energies and CDOCKER interaction energies (Table 1). These investigations identified three DPP4 inhibitors – gemigliptin, linagliptin and evogliptin based on their superior CDOCKER energies which is a function of enzyme-ligand complex energy (Table 1). Their ranking was of the order: gemigliptin (CDOCKER energy = –39.55 kcal/mol) > linagliptin (CDOCKER energy = –34.15 kcal/mol) > evogliptin (CDOCKER energy = –33.95 kcal/mol).
Interestingly, these studies show that the dipeptide nitrile containing covalent DPP4 inhibitors vildagliptin, saxagliptin and a nonpeptide inhibitor, omarigliptin formed high energy complex with the SARS-CoV-2 Mpro protomer and were not considered for further investigation. Fig. 3 shows the binding modes of gemigliptin, linagliptin and evogliptin in the substrate binding region of SARS-CoV-2 Mpro protomer. Gemigliptin was oriented in a linear conformation and the bicyclic dihydropyridopyrimidine ring was in the catalytic site and underwent hydrophobic interactions with catalytic site residues His41 and Cys145 (distance < 5.0 Å, Fig. 3A). The C2 trifluoromethyl substituent underwent hydrogen bonding interactions with Gly143 and backbone NH of Cys145 (distance < 2.8 Å). The protonated primary amine, formed salt-bridge with side chain of Glu166 (distance = 1.79 Å). Interestingly, the difluoropiperidinone substituent was in a solvent exposed area and was forming an intramolecular hydrogen bonding interaction with one of the hydrogens of the pronated amine substituent, suggesting its role in reducing the flexibility across the C1-C2 single bond of oxobutyl spacer, linking the piperidine and dihydropyridopyrimidine rings (Fig. 3A). Docking linagliptin in the SARS-CoV-2 protomer shows that the bicyclic, planar, 4- methylquinazoline ring was oriented toward the catalytic site in a perpendicular fashion with respect to the purine-dione ring and was in hydrophobic contact with His41, Met49 and Cys145 (distance < 5.0 Å, Fig. 3B). The purine-dione ring was in contact with Gly143, Ser144 and His163 through polar and nonpolar interactions (distance < 5 Å) and the piperidine-amine substituent was closer to the entrance of the substrate binding region (Glu166, Leu167, Pro168 and Gln170). Similar to gemigliptin, the protonated primary amine was able to undergo salt-bridge with carboxylate side chain of Glu166 (distance = 4.4 Å, Fig. 3B). Modeling DPP4 inhibitor evogliptin, shows that it exhibited a linear conformation such that the 2,4,5-trifluoromethylbenzene ring was oriented in the catalytic region (His41, Cyst145 and His163, Fig. 3C) and the protonated primary amine formed a salt-bridge with carboxylate of Glu166 (distance < 2.90 Å). The methylpiperazinone substituent was closer to Leu167, Pro168 and Gln189. The C2 tert-butoxy substitutent was in van der Waal’s contact with side chain of Met49 (Fig. 3C). These studies demonstrate that the amino acid residues in the S1 (Phe140, Leu141, Glu166 and Leu167) and S4 (Leu167, Pro168 and His172) pockets are flexible and can accommodate DPP4 inhibitors. Docking studies of other DPP4 inhibitors anagliptin, alogliptin, trelagliptin, sitagliptin, teneligliptin and gosogliptin, also shows their ability to interact in the substrate binding region of SARS-CoV-2 Mpro protomer, which shows their potential to act as inhibitors of viral protease (Fig. S1 and S2, Supplementary Information) although they were not as efficient, compared to gemigliptin, linagliptin and evogliptin based on their CDOCKER energies (Table 1).
Interaction of DPP4 inhibitors gemigliptin, linagliptin and evogliptin in the SARS-CoV-2 Mpro dimer
The catalytically active form of SARS-CoV-2 Mpro is a dimer.8 Therefore, we carried out molecular docking studies of DPP4 inhibitors using the crystal structure of SARS-CoV-2 Mpro dimer, to further understand their binding interactions with SARS-CoV-2 Mpro. We selected DPP4 inhibitors gemigliptin, linagliptin and evogliptin in this study since they exhibited superior CDOCKER scores during our earlier docking investigation using the SARS-CoV-2 Mpro protomer (Fig. 2 and Table 1). Molecular docking studies on the whole SARS-CoV-2 Mpro dimer structure were conducted by sequentially docking DPP4 inhibitors at two substrate binding sites. As an example, interaction of two molecules of gemigliptin on the dimer is shown in Fig. 4. In each protomers, the substrate binding site of SARS-CoV-2 Mpro is located at the surface of domains I and II. Interestingly, the shape of S1 pocket in the SARS-CoV-2 Mpro dimer is maintained by the interactions between Glu166 from chain A of one protomer, with Ser1 from chain B of anotherprotomer.8 The details of enzyme-ligand interaction of gemigliptin, linagliptin and evogliptin in the substrate binding site of SARS-CoV-2 Mpro is shown in Fig. 5. As observed with the protomer docking studies, top ranked pose of gemigliptin exhibited superior CDOCKER energy (–32.62 kcal/mol) compared to linagliptin (CDOCKER energy = –28.90 kcal/mol) and evogliptin (CDOCKER energy = –28.50 kcal/mol) in the dimer. The known inhibitor 1, exhibited superior binding (CDOCKER energy = –56.50 kcal/mol; CDOCKER interaction energy = –69.67 kcal/mol), which is expected, as it is a larger molecule that spans the entire substrate binding region of SARS-CoV-2 Mpro.8 Molecular docking of gemigliptin shows that it was exhibiting a U-shaped conformation and was involved in several contacts in the SARS-CoV-2 Mpro substrate binding site (Fig. 5A). The bicyclic dihydropyridopyrimidine ring was in the catalytic site surrounded by amino acids His41, Asn142, Gly143, Cys145 and His163 and the C4 trifluoromethyl substituent was in contact with the catalytic amino acids His41 and Cys145 via hydrogen bonding and hydrophobic interactions (distance < 5.0 Å). The C2 trifluoromethyl substituent underwent several hydrogen bonding interactions with Ser144 and His163 (distance < 3.0 Å) and was in contact with Cys145 (hydrophobic interactions, distance < 5 Å). The protonated primary amine formed a salt-bridge with Glu166 (distance = 1.79 Å) and the piperidine substituent was in a region consisting of Met49, Met165, Glu166, Leu167 and Gln189. The C5 difluoro-substituent of piperidinone was closer to Met165 side chain. These observations demonstrate that gemigliptin exhibits different binding modes in both protomer and dimer models of SARS-CoV-2 Mpro (RMSD = 3.83 Å). Next, we conducted docking studies of linagliptin in the SARS-CoV-2 Mpro dimer. It shows an L-shaped conformation in the catalytic site and the quinazoline substituent was in contact with Met49, Cys145 and Met165 (π-alkyl and π-sulfur interactions, distance < 5.0 Å, Fig. 5B) whereas the purine-dione was closer to Leu141, Asn142, His163 and His172 and underwent both polar and nonpolar contacts (distance < 5.0 Å). As observed with gemigliptin, the protonated primary amine underwent salt-bridge/electrostatic interactions with Glu166 (distance < 5.0 Å). Its binding mode was similar to that observed in the SARS-CoV-2 Mpro protomer (RMSD = 1.94 Å, Fig. 3B) and the only difference was in the orientation of piperidine substituent, which was closer to Leu167, Pro168 and Gln170 in the protomer binding (Fig. 3B). This can be attributed to the flexibility in the S4 pocket region of SARS-CoV-2 Mpro dimer. Then, we investigated the binding interactions of evogliptin in the SARS-CoV-2 Mpro dimer. This DPP4 inhibitor was in a linear conformation with the 2,4,5-trifluoromethylbenzene ring oriented in the catalytic site (His41, Met49 and Cys145). The aromatic ring underwent π-π T-shaped interactions with His41 aromatic ring (distance < 5.0 Å), π-sulfur interaction with Met49 side chain (distance < 5.0 Å) and π-alkyl interaction with Cys145 (distance < 5.0 Å) as shown in Fig. 5C. The protonated primary amine, underwent cation-π and hydrogen bonding interactions with His41 and His164 respectively and the butanone ketone underwent hydrogen bonding interactions with backbone NH of Glu166 (distance = 2.07 Å). The tert-butoxymethylpiperazinone substituent was oriented in a lipophilic region comprised of Met165, Leu167, Pro168 and Gln189 and the lipophilic tert-group was in van der Waal’s contact with Met165, Leu167 and Pro168 (distance < 5.0 Å). Evogliptin exhibits different contacts and binding mode (RMSD = 3.92 Å) when compared to its binding in the SARS- CoV-2 Mpro protomer (Fig. 3C). These studies demonstrate that the substrate binding site in SARS- CoV-2 Mpro dimer is well defined due to the interactions of Ser1 from chain B at the N-terminus with Glu166 and Phe140 in the S1 pocket that helps to maintain the shape and activity of SARS- CoV-2 Mpro dimer.8 This also explains the differences observed in the binding modes of gemigliptin, linagliptin and evogliptin during the docking studies using either the SARS-CoV-2 Mpro protomer or dimer crystal structures. Lack of N-terminal Ser1 from chain B, can make the S1 and S4 regions of SARS-CoV-2 Mpro protomer, more flexible by exposing them to solvent; whereas presence of Ser1 in the dimer makes those regions less flexible and buries the Glu166 in the S1 pocket. Comparing the electrostatic potential energy surface of SARS-CoV-2 Mpro protomer and dimer substrate binding sites, clearly show these differences (Fig. S3, Supplementary Information). Modeling the manually built known peptidomimetic inhibitor 1 (Fig. 1) in the SARS-CoV-2 Mpro dimer, shows that it was exhibiting superior binding interactions (CDOCKER energy = –56.50 kcal/mol, CDOCKER interaction energy = –69.67 kcal/mol) compared to the top three DPP4 inhibitors from our study: gemigliptin (CDOCKER energy = –32.62 kcal/mol, CDOCKER interaction energy = –43.88 kcal/mol), linagliptin (CDOCKER energy = –28.90 kcal/mol, CDOCKER interaction energy = –45.34 kcal/mol) and evogliptin (CDOCKER energy = –28.51 kcal/mol, CDOCKER interaction energy = –37.07 kcal/mol), suggesting that these DPP4 inhibitors would exhibit reduced binding and inhibition of SARS-CoV-2 Mpro dimer compared to the peptidomimetic inhibitor 1.
Interaction of DPP4 inhibitors gemigliptin, linagliptin and evogliptin in the SARS-CoV Mpro dimer
We investigated the binding interactions of DPP4 inhibitors gemigliptin, linagliptin and evogliptin in the active site of another coronavirus viral protease SAR-CoV Mpro. This virus was responsible for the SARS outbreak in 2003 and has no FDA approved treatment till now.21, 22 The SAR-CoV Mpro shares 96% sequence identity with SARS-CoV-2 Mpro and is also a cysteine protease.8, 23 The dimer form of this viral protease is active. Similar to SARS-CoV-2 Mpro dimer, the catalytic site of SARS-CoV Mpro contains His41 and Cys145 and the N-terminal Ser1 from chain B maintains the shape of S1 pocket.23, 24 Molecular docking studies of gemigliptin shows that, it was binding in an extended conformation and was interacting in the entire span of SARS-CoV Mpro dimer binding site (Fig. 6). The dihydropyridopyrimidine ring with trifluromethyl substituents was oriented in the catalytic site closer to His41, Met49, Asn142, Cys145 and Met165 and underwent hydrogen bonding (distance < 2.5 Å) and hydrophobic interactions (distance < 5.0 Å, Fig. 6). The protonated amine was forming a salt-bridge (distance = 2.70 Å) with Glu166 carboxylate. The difluoropiperidinone was oriented toward a flexible region made up of Pro168 and Gln189, and were solvent exposed. Interaction of linagliptin in the SARS-CoV Mpro dimer binding site, shows that it exhibits an L-shaped conformation and the quinazoline ring was in contact with the catalytic site His41 (hydrogen bonding interaction, distance < 2.75 Å) and Cys145 (π-sulfur hydrophobic interactions). The planar purine-dione ring was closer to Met49, Leu167, Pro168 and Gln189 and the linear butynyl substituent was in van der Waal’s contact with Cys145 and aromatic ring of His163 (distance < 5.0 Å, Fig. 6). The protonated amine of piperidine substituent, underwent salt- bridge with Glu166 carboxylate (distance = 4.78 Å). Molecular docking studies of evogliptin in the SARS-CoV Mpro dimer, shows that it adapts an S-shaped conformation with the trifluorobenzene substituent oriented closer to Leu167, Pro168 and Gln189, with one of the fluorines, undergoing polar interactions with Gln189 side chains (distance = 2.39 Å). Evogliptin exhibited weak interactions with catalytic site amino acid residues, except for the hydrophobic interactions of the Boc-substituent with the side chain of Cys145 (Fig. 6). The piperazinone substituent was closer to Met49 and Met165 and the protonated amine formed salt-bridge with Glu166 carboxylate (distance = 3.52 Å) as observed with gemigliptin and linagliptin. The CDOCKER energies show that gemigliptin exhibited superior interactions in the SARS-CoV Mpro dimer compared to linagliptin and evogliptin. Their CDOCKER energies were of the order: gemigliptin (CDOCKER energy = –35.26 kcal/mol) > evogliptin (CDOCKER energy = –31.89kcal/mol) ≈ linagliptin (CDOCKER energy = –31.81 kcal/mol). These results show that DPP4 inhibitors can bind to SARS-CoV viral protease.
Interaction of DPP4 inhibitors gemigliptin, linagliptin and evogliptin in the MERS-CoV CLpro dimer
The MERS-CoV virus is another contagious disease, which exhibited significantly greater mortality rate compared to SARS-CoV outbreak.25 The x-ray crystal structure of MERS-CoV viral protease, MERS-CoV CLpro has been solved. It shows 50% sequence identity with SARS-CoV protease.8, 26–31 The MERS-CoV CLpro is a cysteine protease, and is made up of three domains similar to SARS-CoV Mpro and SARS-CoV-2 Mpro.27 The dimer is the active form, whereas the monomer is inactive. The catalytic site contains Cys148 and His41 residues.27, 31 Docking the top three DPP4 inhibitors gemigliptin, linagliptin and evogliptin, based on our SARS-CoV Mpro and SARS-CoV-2 Mpro dimer modeling studies, shows that these three DPP4 inhibitors undergo efficient interactions in the MERS-CoV CLpro dimer (Fig. S4, Supplementary Information). Gemigliptin exhibited a U-shaped conformation in the MERS-CoV CLpro dimer binding site and the trifluromethyl substituted dihydropyridopyrimidine ring, underwent numerous polar and nonpolar contacts with His41, Ser147, Cys148 and His166 (distance < 5.0 Å). The piperidinone was closer to Met25, Leu27 and Cys145; whereas the protonated amine underwent hydrogen bonding interaction with His41 backbone (distance = 2.53 Å, Fig. S4, Supplementary Information). Linagliptin exhibited an L-shaped conformation, and the purine-dione aromatic ring was closer to the catalytic site and underwent π-π stacked, π-alkyl and alkyl-alkyl interactions with His41 and Leu49 (distance < 5.0 Å). The butynyl substituent, was oriented toward Cys148 and His166 and underwent alkyl-alkyl and π-alkyl interactions, respectively (distance < 5.0 Å). One of the purine- dione ketones, underwent hydrogen bonding interactions with Gln192 (distance = 2.48 Å) and the protonated amine group formed two hydrogen bonding interactions with Met25 and His41 (distance < 3.0 Å). A similar modeling of evogliptin in MERS-CoV CLpro shows that, it exhibits an extended conformation and the trifluorobenzene substituent, was in van der Waal’s contact with Met168 and Gln192 (distance < 5.0 Å, Fig. S4, Supplementary Information). Interestingly, the Boc-substituent was oriented toward the catalytic site and was in van der Waal’s contact with His41 and Cys148 (distance < 5.0 Å). The CDOCKER energies obtained demonstrate that, gemigliptin was forming the most stable complex with MERS viral protease (CDOCKER energy = –38.55 kcal/mol), followed by linagliptin (CDOCKER energy = –31.65 kcal/mol) and evogliptin (CDOCKER energy = –31.60 kcal/mol). This study shows that DPP4 inhibitors have the potential to bind and inhibit MERS-CoV CLpro dimer.
Pharmacophore model of DPP4 inhibitors toward SARS-CoV-2 Mpro dimer
The top ranked poses obtained from the CDOCKER algorithm for gemigliptin, linagliptin and evogliptin were used to obtain pharmacophore model containing the common structural features required for SARS-CoV-2 Mpro dimer binding. This study shows that the minimum structural feature requirements, to bind in the substrate binding region of SARS-CoV-2 Mpro dimer, includes at least two hydrogen bond acceptors (HBA), two hydrophobic aliphatic (HPA) groups and at least one positively charged ionizable (POS) group, as shown in the pharmacophore model (Fig. S6, Supplementary Information). This figure also provides distance parameters separating these chemical structure parameters and imparts further insights on designing novel inhibitors of SARS- CoV-2 Mpro protease. The chemical structure (Fig. 1) and binding mode of the DPP4 inhibitor gemigliptin in the SARS-CoV-2 Mpro dimer binding site (Fig. 5) was analyzed. This shows that, the pyrimidine N2 and ketone substituent of piperidinone, act as HBA, with the C4 trifluoromethyl of the pyrimidine ring and aliphatic methylenes, which are part of the cyclic piperididone substituent, acting as hydrophobic aliphatic (HPA) groups, whereas the protonated amine group was acting as the POS group. This study also shows that the minimum structural requirements identified from the pharmacophore model, can interact in the S1, S2 and S3 pockets of SARS- CoV-2 Mpro dimer substrate binding site.
Physicochemical properties of DPP4 inhibitors
The 2D and 3D physicochemical properties of twelve DPP4 inhibitors and the reported SARS- CoV-2 Mpro inhibitor 1, were determined to understand the significance of these parameters in SARS-CoV-2 Mpro inhibition and drug design. Parameters including molecular weights, the number of hydrogen bond acceptors, hydrogen bond donors, number of aromatic rings, number of rotatable bonds, polar surface area, molecular volume and AlogP values were calculated (Table 2). These studies show that all the DPP4 inhibitors, obey Lipinski’s rule of five (RO5) such as possessing molecular weights below 500 Daltons, having less than 10-hydrogen bond acceptors, less than 5 hydrogen bond donors and log P values below 5 (Table 2).32 The known inhibitor 1, complied with all the rules, except for the molecular weight, which was slightly over (MW: 593.67, Table 2). The DPP4 inhibitors gemigliptin, linagliptin and evogliptin which exhibited superior binding interactions in the SARS-CoV-2 Mpro protomer and dimer structures, exhibited molecular volumes in the range of 316–380 Å3. This shows that these molecules are smaller as compared to the known peptidomimetic SARS-CoV-2 Mpro inhibitor 1 (molecular volume = 474.71 Å3, Table 2). This is along the expected lines as compound 1 is much larger and therefore is able to bind in the entire span of SARS-CoV-2 Mpro protomer and dimer substrate binding sites (Fig. 2 and Fig. 5).8 Remarkably the flexibility of ligands appears to play a significant role in their ability to bind in the SARS-CoV-2 Mpro dimer substrate binding site, as the known peptidomimetic inhibitor 1, has fourteen C–C rotatable bonds. This is reflected by the fact that DPP4 inhibitors with superior binding interactions, contain at least 3 or more C–C rotatable bonds, with evogliptin containing 7- rotatable bonds (Table 2), followed by linagliptin and anagliptin containing 6-each rotatable C–C bonds respectively.