In order to find docking poses that were compatible with the pharmacophore, the resulting ligand poses were filtered again with Phase through the structure-based common pharmacophore using the same filtering conditions as in the first Phase run but without reorienting the poses (i.e., the score in place option was used). From these two pharmacophore screens, we obtained 4,952 compounds (see Figure 4) with at least one pose that was both compatible with the DPP-IV active site and had functional groups that match the 3D location of the two compulsory sites and at least one of the optional sites of the structure-based common pharmacophore. Finally, the poses for the 4,952 compounds from the second pharmacophore screen were submitted to a shape and electrostatic-potential comparison with the experimental pose of the DPP-IV inhibitor at the PDB file 3C45 (that has the smallest IC50 for all the non-peptide reversible inhibitors found in DPP-IV-inhibitor complexes at the PDB [14]; see Figure 1). The shape andelectrostatic-potential comparison identified 446 hit molecules with potential DPP-IV inhibitory activity (see Figure 4).
Finding New Scaffolds of Natural Origin for DPP-IV Inhibitors
One of the most important challenges of any VS workflow is the ability to find molecules with the required activity but without trivial similarity (in terms of chemical structure) to known active compounds. To determine which of the 446 potential DPP-IV inhibitors predicted by our VS workflow could be considered as new lead molecules, we merged the 446 potential DPP-IV inhibitors with 2,342 known DPP-IV inhibitors that were obtained from the BindingDB database [24]. After calculating the 2D fingerprints of these inhibitors, the resulting set was classified into 50 clusters by means of a hierarchical cluster analysis (data not shown). Notably, 12 out of the 50 clusters obtained consisted exclusively of NPs that were previously unidentified as DPP-IV inhibitors. The 219 molecules that belong to these 12 clusters are scaffold-hopping candidates for DPP-IV inhibition (see Table S1).
Figure 2. The relative location of the experimental poses of the ligands in Figure 1 after DPP-IV superposition. The experimental pose for the most potent inhibitor (i.e., the one at 3C45) is shown in black for reference. For each ligand, the energetically relevant pharmacophore sites are shown. Light red and light blue spheres represent the acceptor and donor features, respectively. The green spheres and orange torus display the hydrophobic regions and aromatic rings, respectively. Blue spheres represent positively charged regions. Furthermore, Figure 6 shows that from all the tested molecules, C5 is the most potent inhibitor with an IC50 of 61.55 mM (see Figure 7). With the exception of C1, which significantly inhibited DPP-IV only at 1 mM, the rest of the molecules significantly inhibit DPP-IV at 0.25 mM (see Figure 6) showing a doseresponse effect. Moreover, a SciFinder search (Chemical Abstracts Service, Columbus, Ohio, USA; http://www.cas.org/products/ sfacad) of the literature revealed that none of these 7 moleculesTable 2. Site contribution to the energy-optimized pharmacophores obtained from PDB complexes in bold from Table 1.Required and optional sites at the structure-based common pharmacophore are shown in bold and italics, respectively. The other sites are not part of the structurebased common pharmacophore. Data at the same raw for different PDB complexes indicate that the pharmacophore site is shared by these complexes.
Figure 3. The structure-based common pharmacophore derived from the alignment of the poses in Figure 2 and shown in the context of the 3C45 active site. The pharmacophore is formed by two hydrogen-bond acceptors (i.e., A1 and A2), one positive/hydrogen-bond donor feature (i.e., P/D) and 4 hydrophobic/aromatic ring sites (i.e., H/R1, H/R2, H/R3 and H/R4). The associated tolerances (i.e., radii) of the ???pharmacophore are 1.8A for P/D, A1 and A2, 2.0A for H/R1, H/R3 and H/R4 and 3.3A for H/R2. Two out of these seven sites (i.e., P/D and H/R1) are required during pharmacophore-based searches whereas the remaining five are optional. The P/D site interacts with the Glu205/Glu206 dyad whereas the H/R1 site potentially fills the S1 pocket. The residues are colored according to the type of intermolecular interactions involved. For example, blue residues interact with donor sites, pink residues interact with acceptor sites and green residues are involved in hydrophobic contacts. Light green residues are a part of the S1 pocket.have been reported as antidiabetic drugs. In fact, no bioactivity has been described for these 7 molecules.
With the exception of C7 in which the positive charge of the tertiary amine forms a salt bridge with Glu205/Glu206 (see Figure 8D), all compounds use primary or secondary amines to form hydrogen bond interactions with either Glu206 or with the Glu205/Glu206 dyad side chains (see Figures 8 and 9A). Additionally, all molecules filled the S1 pocket (partially in the case of C1 and C8, which could explain why these two molecules have lower activities as DPP-IV inhibitors; see Figure 6) establishing one intermolecular interaction that corresponds to the compulsory H/R1 site of our common structurebased pharmacophore (see Figure 3). Moreover, it is worthwhile to mention that some molecules could potentially form additional hydrogen bonds with DPP-IV. For example, the hydroxyl and the methoxy groups of C1 could hydrogen bond with the side chains of Glu206 and Ser630, respectively (see Figure 8A). C8 forms two additional hydrogen bonds with the side chains of Arg358 and Tyr666 (see Figure 8E). Finally, C9 could form three additional hydrogen bonds with the side chains of Tyr547, Ser630 and Tyr662 (see Figure 8F). Figure 9A shows the best docking pose of C5 in the DPP-IV binding pocket where its tertiary amine hydrogen bonds with
Glu206. The carbonyl oxygen of the 7-hydroxy-2H-chromen-2one moiety could also hydrogen bond with the Tyr666 side chain. The S1 pocket is occupied by the C5 butyl chain that could form hydrophobic interactions with Tyr662, Tyr666 and Val711. Finally, the chromene ring of the 7-hydroxy-2H-chromen-2-one moiety forms p-p interactions with Phe357. Interestingly, this interaction with Phe357 has been shown to be directly related to the increased potency of synthetic DPP-IV inhibitors relative to those that lack this interaction [13,15,25?7]. Therefore, the fact that this interaction is only present at C5 (see Figures 8 and 9A) would explain why this molecule shows higher bioactivity than the other compounds assayed (see Figure 6). Moreover, an electrostatic and shape comparison of the 7 poses in Figures 8 and 9A revealed that the molecule with the highest similarity to the 3C45 ligand (with the lowest IC50; see Figure 1) is C5 (results not shown). The ET_combo score for this comparison is 1.050, which corresponds to a shape and electrostatic contribution of 0.628 and 0.422, respectively. Remarkably, the same analysis with C2 (which shows a significant bioactivity as DPP-IV inhibitor; see Figure 6), also has a significant ET_combo score of 1.038.
