Predicting ZINC02132035 Derivatives with Higher Binding Affinity on DPP-IV
Although none of the seven in vitro assayed VS hits showed activity in the nanomolar range, these hits incorporate scaffolds with no previously described effects on DPP-IV activity and, consequently, are of interest from a medicinal chemistry point of view as lead compounds for more potent DPP-IV inhibitors.Figure 4. Schematic overview of the VS workflow and the procedure used for selecting the VS hits that were tested for DPP-IV inhibitory activity. For the VS, the number of compounds that passed each step and the programs used are showed. For the selection of VS hits for bioactivity testing, the numbers show either how many VS hits are scaffold-hopping candidates for DPP-IV inhibition (Fingerprint similarity analysis step) or how many molecules were experimentally tested for bioactivity (Biological test step).C5; see Figure 6) by (1) using a fragment-based docking approach to identify which intermolecular interactions with the DPP-IV binding site could improve the binding affinity of C5 derivatives relative to C5; (2) using this information to identify where changes in C5 should be made; and (3) accordingly building C5 derivatives and predicting their relative binding affinities. The comparison of the XP descriptors from C5 and from the docked poses of the fragments showed that while some of the terms of the scoring function are 0.00 Kcal/mol for C5, their corresponding value for 13 out of 50 fragments is in the [-2.48, 0.83] Kcal/mol range (see Table S2). Interestingly, 12 out of 13 of these fragments bind at the locations of three of the sites of our structure-based common pharmacophore (i.e., H/R1, H/R2 and H/R4), whereas the remaining fragment is close to the H/R1 site (see Table S2). These findings demonstrate that our pharmacophore is able to capture all of the essential features for an inhibitor interaction with the DPP-IV binding-site, which would explain why all of the assayed molecules show activity as DPP-IV inhibitors (see Figure 6). Moreover, the analysis in Table S2 shows that C5 activity can be improved if (1) its butyl group matching the H/R1 site is replaced by a group that is able to interact with the lipophilic atoms of the S1 pocket either by producing the so-called hydrophobic enclosure reward (that would consist of enclosing the two sides of the substituent t a 180u angle?on the hydrophobic environment of the S1 pocket) or by making p-cation interactions with the aromatic side chains in this pocket and (2) groups that match the H/R2 site (optional in our pharmacophore but present in half of the ligands used to derive the pharmacophore; see Table 2) and that are able to make hydrophobically packed correlated H-bonds are added to C5.
The substituents that were attached to the C5 core to obtain the top five derivatives with the highest predicted binding affinity for the DPP-IV binding site are shown in Table S3. None of the five molecules are currently registered in ChemSpider (http://www. chemspider.com), whereas their analysis with FAF-Drugs2 shows that all of these derivatives exhibit the proper ADMET properties. Therefore, these derivatives are undescribed drug-like molecules that, according to their XP GScores (see Table S3), would show a significant increase in their binding affinity relative to C5 (i.e., 4.2 Kcal/mol). Figure 9 shows the docked poses for C5 derivatives compared with C5 and can be used to explain the structural basis of the expected increase in binding affinity. Remarkably, the XP GScores for these poses are in the -9.5 to -11.8 Kcal/mol range (see Table S3), whereas the GScores for the experimental poses of the DPP-IV inhibitors shown in Figure 1 are in the -5.8 to 11.0 Kcal/mol range (results not shown). Therefore, the C5 derivatives reported in Table S3 are likely to exhibit nanomolar activity as DPP-IV inhibitors. As shown in Figure 9, the C5 derivatives usually maintained the most important protein-ligand interactions found for the C5 core. Moreover, Table S3 also shows that all of the substituents thatFigure 5. Chemical structures and ZINC codes for the 9 molecules selected for experimentally testing whether these compounds exhibited DPP-IV inhibitory activity. The insolubility of C4 and C6 prevented these compounds from being assayed for DPP-IV inhibitory activity. Positions in the C5 structure that will be replaced by substituents to identify derivatives with higher binding affinity on the DPP-IV binding site are (a) indicated with a grey background and (b) annotated with the label of the corresponding site in the common structure-based pharmacophore (see Figure 3).
have replaced the original C5 butyl group (i.e., at the H/R1 site) have a common positive formal charge that, according to results shown in Figure 9, allows them to form p-cation interactions with two of the aromatic residues in the S1 pocket (i.e., Tyr662 and Tyr666). Additionally, some of the substituents at this location (i.e., 97 in C5-97-786, 100 in C5-100-563 and 274 in C5-274-536; see Table S3) also aid in increasing the protein-ligand binding affinity by enclosing the two sides of the corresponding ring in the lipophilic protein environment in the S1 pocket (results not shown). Furthermore, all substituents at the H/R2 site (except the one in C5-309-787) are able to make hydrogen bonds either with the S2 pocket residue Arg358 (i.e., 786 in C5-97-786, 784 in C5137-784 and 563 in C5-100-563; see Figures 9B, 9C and 9D) or with Arg669 (i.e., 536 in C5-274-536; see Figures 9F). The 786 substituent in C5-97-786 is also able to make a hydrogen bond with the Ser209 side chain (see Figure 9B). Remarkably, there are SAR studies with a structurally distinct series of DPP-IV inhibitors that show (1) a 4-fold loss of potency when substituents that interact with the side chains of Ser209 and Arg358 are removed[28], (2) a 2-fold increase in inhibition when a carboxylic acid that interacts with Arg358 is introduced [27], and (3) a 6-fold increase in inhibition when a 3-pyridyl group that interacts with Ser209 is introduced [29]. Therefore, the substituents selected for the H/R2 site by the combinatorial screen are able to form the intermolecular interactions with the S2 pocket that previous SAR studies with anti-diabetic drugs have shown to increase the affinity for the DPP-IV binding site.
Conclusions
The challenge of any VS protocol consists of using in silico tools to predict which molecules in a database have the required activity against a specific target. The results of the present study demonstrate that our VS protocol is highly successful in the non-trivial identification of DPP-IV inhibitors with no chemicalstructure similarities to known activities. Therefore, scaffold hopping on this target can be achieved.
Figure 6. Dose-response results for the in vitro inhibition of DPP-IV by C1, C2, C3, C5, C7, C8 and C9. The relative DPP-IV inhibitory activity with or without the selected NPs (vehicle, 1% DMSO) is shown where each column represents the average 6 SEM (n = 3 or 4).
