Identification of a Novel Scaffold for Inhibition of Dipeptidyl Peptidase-4

Abstract

Dipeptidyl peptidase-4 (DPP-4) is considered a major drug target for type 2 diabetes mellitus (T2DM). In addition to T2DM, a regulatory role of DPP-4 was also found in cardiovascular diseases. Existing DPP-4 inhibitors have been reported to have several adverse effects. In this study, a computer-aided drug design approach and its use to detect a novel class of inhibitor for DPP-4 are reported. Through structure and pharmacophore-based screening, we identified 13 hit compounds from an ∼4-million-compound library. Physical interactions of these hits with DPP-4 were studied using docking and explicit solvent molecular dynamics (MD) simulations. Later, MMPBSA binding energy was calculated for the ligand/protein simulation trajectories to determine the stability of compounds in the binding cavity. These compounds have a novel scaffold and exhibited a stable binding mode. “Best-in-screen” compounds (or their closest available analogs) were resourced and their inhibition of DPP-4 activity was experimentally validated using an in vitro enzyme activity assay in the presence of 100 and 10 μM compounds. These assays identified a compound with a spirochromanone center with 53% inhibition activity at a 100 μM concentration. A further five spirochromanone compounds were synthesized and examined in silico and in vitro; again, one compound showed 53% inhibitory activity action at 100 μM. Overall, this study identified two novel “spirochromanone” compounds that lowered DPP-4 activity by more than ∼50% at 100 μM. This study also showed the impact of fast in silico drug design techniques utilizing virtual screening and MD to identify novel scaffolds to bind and inhibit DPP-4. Spirochromanone motif identified here may be used to design molecules to achieve drug-like inhibitory action against DPP-4.

1. Introduction

Diabetes mellitus (DM) is likely one of the oldest known diseases; it was first reported in an Egyptian manuscript nearly 3000 years ago (Tattersall, 2010). In the early twentieth century, two forms of this chronic disorder were distinguished and designated as type 1 and type 2 DM (Himsworth, 1936), with the latter also known as metabolic syndrome (Patlak, 2002), hyperglycemia, insulin resistance, and relative insulin deficiency. Whereas type 1 DM is caused by a deficiency in insulin production, type 2 DM (T2DM) sufferers do not respond to insulin and, in later stages of the disease, may also suffer from decreased insulin production. T2DM is the most common form of diabetes and many studies have confirmed the contribution of a variety of genetic, environmental, and behavioral factors to its onset (Chen et al., 2012). According to the World Health Organization, the global disease burden of diabetes in adults has risen from 4.7% in 1980 to 8.5% in 2014 (Mathers and Loncar, 2006).

Insulin is a small peptide hormone released by pancreatic beta cells in response to elevated glucose levels in the blood. Primarily, insulin triggers cellular glucose uptake, but it is also involved in regulating carbohydrate and lipid metabolism by promoting glucose storage in the form of glycogen, lipids, and protein. Furthermore, insulin also stimulates the cellular uptake of amino acids, inhibits protein degradation, and promotes protein synthesis (Joost, 2011). The ubiquitously expressed cell surface protease dipeptidyl peptidase-4 (DPP-4) suppresses the insulinotropic effects of the insulin regulator incretin hormone glucagon-like peptide (GLP-1) through proteolytic degradation of GLP-1 (Gerich, 1998). In line with this molecular mechanism, DPP-4-deficient mice showed improved glucose tolerance and higher plasma concentrations of the incretin glucose-dependent insulinotropic peptide, GLP-1, and insulin (Proud, 2006). Owing to its hyperglycemic activity, GLP-1 has been used extensively for the treatment of T2DM (Barnett, 2006) and, following validation of DPP-4 as a target for control of T2DM, DPP-4 inhibitors have been in clinical use as antidiabetic therapeutics since their action prolongs the activity of endogenously released GLP-1 (Proud, 2006; Shirakawa et al., 2016).

In addition to the therapeutic value of DPP-4 for T2DM, inhibition of this enzyme has also been shown to elicit beneficial effects in cardiovascular diseases, including atherosclerosis and ischemic heart disease, the leading cause of premature adult mortality worldwide (Lusis, 2000). Whereas the detailed biological mechanisms behind atherosclerosis remain unknown, various studies suggest that some DPP-4 inhibitors also possess cardioprotective effects (Chinda et al., 2013; Hausenloy et al., 2013; Ferrannini and DeFronzo, 2015; Gilbert and Krum, 2015; Zhong et al., 2015; Kubota et al., 2016), including beneficial outcomes for atherosclerosis sufferers (Remm et al., 2015; Silva Júnior et al., 2015; Singh et al., 2015). Accordingly, DPP-4 is at present also considered a potential drug target for chronic arterial diseases.

The intrinsic membrane glycoprotein DPP-4 is a serine protease with 766 amino acids and its structural topology comprises (1) a cytoplasmic tail, (2) a transmembrane domain, and (3) an extracellular domain. The extracellular domain is the most relevant for therapeutic intervention because it is composed of the catalytic site and an eight-bladed β-propeller domain that is connected to the catalytic domain (Thoma et al., 2003). Mechanistically, DPP-4 is an exopeptidase that cleaves proline-containing dipeptides from the N-terminus (Röhrborn et al., 2015); proteolysis occurs in the catalytic domain, using a catalytic triad of Ser630, Asp708, and His740. Among the first inhibitors of DPP-4 were saxagliptin (Augeri et al., 2005) and vildagliptin (Mathieu and Degrande, 2008), which target the substrate binding subsites (S1 and S2) of DPP-4. Accordingly, residues of interest for drug discovery include Val656, Trp662, Tyr666, and Val711 (S1 pocket); Arg25, Phe357, Arg358, Tyr547, Pro550, and Asn710 (S2 pocket); and residues in the N-terminal recognition region formed by Glu205, Glu206, and Tyr662 (Arulmozhi and Portha, 2006). Subsequently, optimized inhibitors with improved selectivity and lesser side effects have been developed, including alogliptin (Feng et al., 2007), anagliptin (Kato et al., 2011), evogliptin (Kim et al., 2012), gemigliptin, linagliptin (Eckhardt et al., 2015), omarigliptin (Biftu et al., 2014), sitagliptin (Thornberry and Weber, 2007), and teneligliptin (Yoshida et al., 2012a).

Whereas DPP-4 inhibitors in general are tolerated well by patients, they can provoke hyperglycemia (Salvo et al., 2016), heart failure (Dicker, 2011), and acute pancreatitis (DeVries and Rosenstock, 2017) when used in combination with sulfonylurea derivatives and insulin (Salvo et al., 2016). There are also reports of nasopharyngitis, headache, upper respiratory tract infection, and urinary tract infection (Ligueros-Saylan et al., 2010; Ali and Fonseca, 2013), and thus, while existing inhibitors are potent against DPP-4 (IC₅₀ values in the nanomolar inhibition range) (Berger et al., 2018), they have potentially severe side effects in what is already a vulnerable patient group. Therefore, work is ongoing to identify a new chemical scaffold for inhibiting DPP-4 with less or no adverse effects for patients.

The availability of experimental three-dimensional (3D) structures of the ligand-bound DPP-4 extracellular domains (Feng et al., 2007; Edmondson et al., 2009; Yoshida et al., 2012a, b; Nabeno et al., 2013) allows the application of computational methodologies to discover and design new inhibitors for this therapeutic target, particularly work that aims to discover new hit compounds that can be translated into chemotherapeutics with minimal side effects. In this study, we report an in silico drug design approach to identify new hit compounds to bind and potentially inhibit DPP-4. Following screening of a large compound library (N ≈ 4 million), 50 hits were selected based on the favorability of their predicted interactions and docking energies. The selected molecules were docked and subjected to molecular dynamics (MD) simulations as potential DPP-4:ligand complexes to calculate the binding energies with Poisson-Boltzmann surface area continuum solvation (MM/PBSA). The dynamics behavior of each compound in the substrate binding cavity of DPP-4 was also examined. A total of 13 compounds were selected and sourced commercially and a small library of five readily available synthetic compounds with similar topology was generated. The resulting set of 17 compounds was subjected to an in vitro enzyme inhibition assay.

2. Methods

Earlier studies divided the binding pocket of DPP-4 into two spatial regions labeled the S1 and S2 (Arulmozhi and Portha, 2006) sites. However, an additional binding region was also identified that covers the N-terminal recognition site. Cocrystal structures of known inhibitors with DPP-4 were used here as reference molecules to locate these binding regions: alogliptin, anagliptin, linagliptin, saxagliptin, sitagliptin, and vildagliptin (PDB 3GOB, 3WQH, 2RGU, 3BJM, 4FFW, and 3W2T, respectively) (Eckhardt et al., 2007; Metzler et al., 2008; Robert et al., 2009; Tang et al., 2012; Nabeno et al., 2013; Watanabe et al., 2015). Overall, this study utilized three rounds of virtual screening to identify potential hit compounds. Runs 1 and 2 utilized a protein structure-based approach, while Run 3 was a ligand-based approach using pharmacophores of known inhibitors to search the ZINC database. Compounds identified from all three rounds were further examined by docking into the DPP-4 active site, 20 ns MD simulations and MMPBSA binding energy calculations. These highest scoring compounds were then tested for inhibition of DPP-4 in vitro. Compounds were resourced from MolPort (Latvia) where they are further supplied by ChemDiv, Inc., ENAMINE Ltd., Pharmeks, Ltd., Specs and TargetMol, and Vitas-M Laboratory, Ltd. via MolPort.

2.1. Screening and identification of hit compounds

Virtual screening Runs 1 and 2 used the HitGen module of commercial drug design tool INVENTUS™ (www.novoinformatics.com/inventus.php) for structure-based screening, which uses physicochemical features to screen compounds from an inbuilt compound library consisting of ∼4 million compounds (Mukherjee and Jayaram, 2013). The physicochemical filters for screening were estimated from the computed values of known inhibitors (alogliptin, anagliptin, linagliptin, saxagliptin, sitagliptin, and vildagliptin) selected previously as reference compounds. These physicochemical filters were as follows: (1) a number of hydrogen bond donors (HBD), (2) a number of hydrogen bond acceptors (HBA), (3) logP, (4) a number of aromatic atoms, (5) a number of atoms, (6) carboxyl group count, (7) hydroxyl group count, (8) phosphate/sulfate group count, (9) amide group count, (10) ketone/aldehyde group count, and (11) ester/ether group count. Run 1 did not use any organic functional group filter criteria to screen compounds, while Run 2 explicitly used a count for these functional groups, estimated from their corresponding values in known inhibitors.

Table 1 shows the list of physicochemical filters used in screening and their respective ranges for Runs 1 and 2. The top 50 hit compounds from each run were identified using the scoring function of the screening protocol implemented in INVENTUS (Mukherjee and Jayaram, 2013). These hit compounds were docked into the binding site of DPP-4 (identified using the vildagliptin-DPP-4 structure; PDB 3W2T) using the INVENTUS docking module (NovoDocker), which applies a rigid docking grid search technique (Gupta et al., 2007). The binding energy of each docked pose generated was calculated using an all-atom energy-based computational protocol incorporated in the BioAff module of INVENTUS (Jain and Jayaram, 2005). The docking program generates the best eight poses for each protein/ligand complex and those compounds that had negative binding energy across all docked poses were selected for further analysis. Screening, docking, and binding energy analysis resulted in the selection of six compounds from Run 1 and five compounds from Run 2 from the top 50 compounds in each round.

Table 1.
Parameters Used for Screening in Run 1 and Run 2 Under INVENTUS Package with Their Respective Filtering Range to Generate Top 50 Hit Compounds

Screening features Run 1 Run 2

Minimum Maximum Minimum Maximum

HBD 1 3 1 3

HBA 5 10 5 10

logP 2 4 2 4

Aromatic atoms 5 15 5 15

No. of atoms 40 50 40 50

Carboxyl NF NF 0 10

Hydroxyl NF NF 0 1

Phosphate/sulfate NF NF 0 0

Amide NF NF 0 10

Ketones/aldehyde NF NF 1 4

Ester/ether NF NF 0 10

Binding cutoff −5 kcal mol⁻¹

Screening features	Run 1	Run 2
HBD	1	3	1	3
HBA	5	10	5	10
logP	2	4	2	4
Aromatic atoms	5	15	5	15
No. of atoms	40	50	40	50
Carboxyl	NF	NF	0	10
Hydroxyl	NF	NF	0	1
Phosphate/sulfate	NF	NF	0	0
Amide	NF	NF	0	10
Ketones/aldehyde	NF	NF	1	4
Ester/ether	NF	NF	0	10
Binding cutoff	−5 kcal mol⁻¹

HBA, hydrogen bond acceptors; HBD, hydrogen bond donors; NF, no filters applied on given parameter.

In screening Run 3, the PharmaGist (Schneidman-Duhovny et al., 2008) server was used to generate a three-point pharmacophore from known inhibitors of DPP-4 (alogliptin, anagliptin, linagliptin, saxagliptin, sitagliptin, and vildagliptin). These known drugs share similar geometrical arrangements of one HBD, one HBA, and one aromatic ring, and the resulting pharmacophore was applied using the ZincPharmer (Koes and Camacho, 2012) server to screen the ZINC compound database. This produced 2000 compounds, which were further ranked and screened against the DPP-4 protein using the custom library option in the HitGen module of INVENTUS. In this step, the pharmacophore 2000 was used as a search library for structure-based screening in INVENTUS as in Runs 1 and 2. Again, the top 50 hits were docked using NovoDocker and compounds with negative binding energy across all docked poses selected for further analysis. This process selected two compounds from Run 3. In total, the 13 compounds selected from the three screening runs were redocked using AutoDock version 1.5.6 (Morris et al., 2009) to cross-validate their predicted binding affinities for DPP-4. The highest ranked docked pose for each molecule was analyzed by the stand-alone LIGPLOT tool to sketch an interaction plot (Wallace et al., 1995). Last, the interacting residues shown in LIGPLOT for selected protein/ligand complexes were compared with the residues of existing binding sites: S1, S2, and the N-terminal recognition region. A detailed, stepwise explanation of the INVENTUS and the AutoDock parameters used is provided in Supplementary Material.

2.2. Molecular dynamics simulations

GROMACS 4.6.4 (Berendsen et al., 1995) was used to simulate the highest scoring docked pose of DPP-4 protein with each compound (both screening hits and purchased analogs thereof). These complexes were simulated for 20 ns in an 11.9 × 11.9 × 11.9 Å³ water box under explicit solvent conditions. Initial complexes were energetically minimized with solvent using 5000 steps of a steepest descent algorithm. Postminimization, the complete system was equilibrated for 100 ps and 1 ns under NVT (canonical ensemble) and NPT (isothermal isobaric ensemble) ensembles, respectively. The NVT equilibration ensemble used velocity rescale as temperature coupling, while NPT equilibration had velocity rescale as temperature coupling and Berendsen pressure coupling. During each 20 ns simulation phase, pressure coupling was changed to Parrinello-Rahman. Root mean square deviation (RMSD) and hydrogen bond counts were computed for each 20 ns trajectory using g_rms and g_hbond. The MMPBSA binding energies of DPP-4 with each compound were computed using g_mmpbsa (Kumari et al., 2014), which calculates the nonpolar solvation energy of each protein/ligand complex using the solvent accessible surface area model. This tool allows prediction of molecular interactions comprising both van der Waals forces and electrostatic interactions with polar and nonpolar solvation terms for each simulation frame. During the calculation of polar solvation energy, the solvent dielectric constant was fixed at 80; the protein dielectric constant was set at 1. The linear Poisson/Boltzmann equation was used to solve the continuum solvent model (Fogolari et al., 2002).

2.3. DPP-4 inhibition assays

DPP-4 inhibition was assessed using a DPP-IV Drug Discovery Kit (BML-AK499; Enzo Life Sciences, Farmingdale, NY). All assays were performed at 37°C, in triplicate, according to the kit protocol. Compounds were made up in dimethyl sulfoxide (DMSO) and added such that the final concentration of DMSO was 1%. Controls included DPP-4 enzyme with either buffer (uninhibited) or 10 μM P32/98 (control inhibitor) added. To test compounds for DPP-4 inhibition, 2.5 nM DPP-4 was first incubated with 10 μM AMI compounds 1–11 for 10 minutes. The reactions were then initiated by the addition of 5 μM H-Gly-Pro-pNA substrate and continuously monitored through the absorbance of the pNA product at 405 nm for 35 minutes. The data were processed for a time window of 0–31 minutes: initial rates were obtained from linear fits in R, scaled relative to an uninhibited enzyme control, and plotted with error bars showing the standard error. Significance was assessed using the Student's t-test. Those compounds that showed some inhibition, reducing the relative DPP-4 activity by at least 15%–20%, were retested at 100 μM, as per the lowest concentration used by Guasch et al.'s (2012) work.

3. Results and Discussion

Virtual high-throughput screening (vHTS) is an established technique for rapid screening of large sets of compounds against a given protein target to identify potential binding molecules (Subramaniam et al., 2008). The core principle of this methodology is based on the physicochemical complementarity between a ligand and the defined binding pocket of a protein (Subramaniam et al., 2008). In this study, vHTS was used to screen the ZINC (Irwin et al., 2012) and NCI (Ihlenfeldt et al., 2002) compound libraries (4 million molecules) to discover novel potential inhibitors. This work used the crystal structure of DPP-4 complexed with vildagliptin (PDB 3W2T) (Mathieu and Degrande, 2008) as a reference. Based on previously characterized interactions between DPP-4 inhibitors and various residues in the substrate binding site, a set of filters was generated to guide selection of hit compounds. The filters were designed to consider ligand interactions with the S1 and S2 subsites, as well as the N-terminal recognition region of DPP-4, and were implemented as a quantitative structure/activity relationship equation in the molecular modeling procedure (Mukherjee and Jayaram, 2013). Two independent vHTS campaigns were conducted, each applying a different set of filter criteria (see details in the Methods section). In addition, an independent vHTS was performed using a pharmacophore search method. From the three vHTS campaigns, 50 compounds were selected based on virtual screening scores and estimated protein/ligand binding energies (Mukherjee and Jayaram, 2013).

Recently, several empirical binding energy equations have been proposed to calculate improved estimates of the ligand affinity of a particular protein (Kollman et al., 2000; Schulz-Gasch and Stahl, 2004). Thus, we used the NovoDocker (Gupta et al., 2007) and BioAff (Jain and Jayaram, 2005) modules of the INVENTUS package to further appraise interactions between the 50 selected compounds and DPP-4. Eight binding poses were generated for each of the 50 compounds and binding energy scores were calculated using the INVENTUS scoring function (Jain and Jayaram, 2005). For compounds that yielded negative binding energies for all eight poses (n = 13), protein/ligand complexes were generated using the complex of vildagliptin with DPP-4 as a guide (Mathieu and Degrande, 2008) and subjected to MD simulations extending over 20 ns, with subsequent calculation of free binding energies using MM/PBSA (Kollman et al., 2000). Of the selected 13 compounds, 12 could be sourced commercially (although some of the final compounds purchased were analogs of the original hits with a minimum similarity of 70% [Tanimoto coefficient]).

The analogs were tested in silico against the same docking and MD simulations as their parent compounds. The purchased compounds were tested for inhibition of DPP-4 in an in vitro enzyme assay at concentrations of 10 and 100 μM. A small library of five readily available compounds exploring the recurring spirochromanone-type scaffold was synthesized in-house and included in the enzyme inhibition assays. These five compounds were also subjected to molecular docking and simulations to predict their behavior in the DPP-4 binding pocket.

3.1. Virtual screening using DPP-4 inhibitor-derived filters

In two independent screening runs (Run 1 and Run 2), different sets of filters were applied in vHTS of the large compound library using the screening module of INVENTUS. From each run, the top 50 compounds, as ranked by vHTS protein/ligand complementarity scores (Mukherjee and Jayaram, 2013), were docked into the binding site, which was predicted using the cocrystal structure of the reference ligand vildagliptin and DPP-4 (PDB 3W2T). Eight binding poses were generated for each protein/ligand pair (Supplementary Table S1). Compounds that yielded a negative docking score for all eight poses were selected to be carried forward. This criterion resulted in the selection of six (R1.1–R1.6) and five (R2.1–R2.5) compounds from Runs 1 and 2, respectively. An appraisal of the drug-likeness (according to Lipinski) of these 11 compounds showed that they comply with Lipinski's rule (Owens, 2003) for the number of HBD and HBA (i.e., one to three HBD and four to eight HBA).

These compounds have logP values in the range of 2.14–4.06 and molecular masses of 387–483 Da. Notably, all 11 hit compounds possess two or more aromatic rings, which is a recurring pharmacophore in many known DPP-4 inhibitors (alogliptin, anagliptin, linagliptin, saxagliptin, sitagliptin, and vildagliptin) (Agrawal et al., 2012). To assess the structural diversity of the identified hits, the Tanimoto coefficient (Bajusz et al., 2015) was used as a standard measure of similarity between pairs of compounds. A score below 0.5 is considered an acceptable cutoff for low similarity (Baldi and Nasr, 2010). Among all 56 pairings, a maximum score of 0.6 was obtained for compounds R1.3 and R1.4; all remaining pairings had similarity scores of less than 0.5 (Supplementary Table S2) and can thus be considered sufficiently different in their structures. A comparison of all 11 hits with the six DPP-4 inhibitors mentioned above yielded similarity scores of less than 0.2, confirming their structural novelty.

3.1.1. Pharmacophore screening

In addition to screening using the custom-generated filters, pharmacophore-based screening was also performed (Run 3). Conceptually, 3D pharmacophores are generated based on the structures of known ligands (Moro et al., 2005; Gao et al., 2010; Amaravadhi et al., 2014). In this study, a three-point pharmacophore was generated using the structures of six known DPP-4 inhibitors (alogliptin, anagliptin, linagliptin, saxagliptin, sitagliptin, and vildagliptin) and used to search the ZINC database, which resulted in the identification of 2000 compounds (Supplementary Table S3). Subsequently, this customized library of 2000 compounds was used in vHTS with the INVENTUS screening module (Mukherjee and Jayaram, 2013). As in Runs 1 and 2, the top 50 compounds were selected based on protein/ligand complementarity scores and subjected to docking calculations with the same DPP-4 binding site as described above. Eight poses for each protein/ligand pair were generated and assessed for their binding energies. Two molecules (R3.1 and R3.2) showed negative binding energies for all eight poses and were thus carried forward to the next step. Both compounds comply with Lipinski's rules for drug-like HBA (4 and 5), HBD (3 and 2), logP (0.23 and 1.1), and molecular mass (381 and 370) and possess the recurring aromatic ring pharmacophore of known DPP-4 inhibitors. The two compounds are sufficiently dissimilar with respect to each other (Tanimoto coefficient 0.3) and known DPP-4 inhibitors (maximum Tanimoto coefficient 0.2; Supplementary Table S2).

3.1.2. Virtual screening summary

Using the virtual screening protocol described above, a set of 13 diverse compounds with high predicted binding scores was identified. The diversity of the compounds within the set and when compared with six established DPP-4 inhibitors is illustrated in the hierarchical clustering similarity matrix in Figure 1a, where 0 represents no similarity and 1 indicates 100% similarity.

FIG. 1.

Similarity index and consistency of predicted binding scores for compounds identified by virtual screening. (a) Cross similarity matrix of 13 compounds identified in virtual screening Runs 1–3 and six established inhibitors of DPP-4. The heatmap color scheme renders high similarity (Tanimoto coefficient - >1) in lighter blue and low similarity (Tanimoto coefficient - >0) in dark blue. Figure prepared with “R” package. (b) The predicted binding scores for the identified 13 compounds obtained with INVENTUS™ and AutoDock show a Pearson correlation coefficient of 0.84, indicating consistency of results obtained with both software packages. DPP-4, dipeptidyl peptidase-4.

3.2. Redocking

Correctly distinguishing between binding and nonbinding compounds is a known challenge in virtual screening studies (Deng et al., 2014), owing to the potential shortcomings of the scoring function implemented within a particular software package. Therefore, we sought to corroborate the binding scores derived from INVENTUS by docking the 13 identified compounds using AutoDock (Morris et al., 2009). The calculated binding energies obtained with INVENTUS ranged between −12.37 and −6.91 kcal mol⁻¹; for AutoDock, the corresponding binding energy range was from −11.77 to −3.45 kcal mol⁻¹. Notably, the compounds identified by pharmacophore search (Run 3) exhibited lower binding scores than compounds identified with DPP-4 inhibitor-derived filters (Runs 1 and 2). The best binding scores for each compound from each program were analyzed for correlation (Fig. 1b) and yielded a Pearson correlation coefficient of 0.84, indicating consistency in the predicted binding scores. Protein/ligand interactions in the highest scoring pose for each compound docked into the DPP-4 binding site using AutoDock are illustrated in Supplementary Figure S1.

As mentioned, the vildagliptin-DPP-4 cocrystal structure (PDB 3W2T) was used to select and define the binding region for this study. In AutoDock, a cube (grid points 52, 72, 50) was created around the reference ligand (vildagliptin) to dock vHTS compounds. The interacting residues from within this defined binding region of DPP-4 in the highest scoring docked pose of each compound are largely similar, as shown in Supplementary Figure S1. These interactions are dominated by hydrophobic contacts (shown in dotted red lines) supported by hydrogen bonds (green dotted lines), which likely strengthen binding.

3.3. Molecular dynamics

To overcome the lack of solvation in docking procedures, MD simulations with explicit waters are frequently applied as a subsequent assessment in the course of computational drug discovery (Alonso et al., 2006; Sixto-López et al., 2017). With present computational resources, MD is typically performed with a simulation time of 20 ns in this context (Gurung et al., 2016; Dolatkhah et al., 2017). In addition to the evaluation of conformational changes in the protein and ligand in the presence of solvent and the potential contribution of solvent-mediated interactions, MD simulations also allow the calculation of improved binding energies by using Poisson/Boltzmann surface area continuum solvation (MM/PBSA) (Brown and Muchmore, 2007; Valiev et al., 2008). Using the highest scored docked poses of the 13 compounds (identified based on the vildagliptin binding site; PDB 3W2T) as the starting conformations, MD simulations and MM/PBSA binding energy calculations were performed for 20 ns. A pre-equilibration MD was run for 100 ps and 1 ns under an NPT ensemble to compensate the effect of the added solvent. For comparison, the structure of the DPP-4:vildagliptin complex (PDB 3W2T) (Mathieu and Degrande, 2008) was subjected to the same process.

3.3.1. Conformational stability

The conformational flexibility of each ligand was evaluated by monitoring the RMSD (Cau et al., 2015) of its structure for the duration of the MD simulation (see Fig. S2a–c). The Cα-RMSD data indicate that all compounds with the exception of R2.5 adopted a stable conformation after ∼8 ns. Compound R2.5 underwent a conformational change between 9 and 12 ns. The standard deviations of Cα-RMSD values of the 13 compounds are in the range of 0.057–0.556; for comparison, the RMSD standard deviation of the vildagliptin simulation was ±0.11 nm. Equilibrium sampling (stable Cα-RMSD) was achieved in the last 10 ns of the main 20 ns simulation, as shown in Figure 2a–c. Earlier studies of protein/ligand binding suggest that a stable RMSD is necessary to study structural and chemical metrics (e.g., hydrogen bonds and binding energy) (Abdel-Hamid and McCluskey, 2014). Overall, the RMSD fluctuations are in agreement with those observed in other studies (Perola and Charifson, 2004).

FIG. 2.

Appraisal of ligand dynamics and ligand/protein hydrogen bonding in 20 ns MD simulations of solvated DPP-4:ligand complexes for the 13 compounds identified in virtual screening as well as vildagliptin. (a–c) RMSD of ligand coordinates. (d) Percentage of simulation frames with at least one hydrogen bond between ligand and protein. MD, molecular dynamics; RMSD, root mean square deviation.

3.3.2. Hydrogen bonds

The formation of hydrogen bonds between the ligand and DPP-4 is an important contributor to the stability of the resultant complex. Analysis of the MD trajectories for the occurrence of at least one hydrogen bond per frame reveals that the 13 identified compounds fall into three different groups (Fig. 2d). R1.1, R1.3, R1.6, R2.1, R2.2, and R2.5 showed hydrogen bonding between 25% and 48% of the time (i.e., 25%–48% of all frames have at least one hydrogen bond) and thus compare well with the behavior of the known inhibitor vildagliptin (hydrogen bonds present 50% of the time). A second group comprising R1.4 and R3.1 hardly engages in hydrogen bonds at all (less than 1% of the time). However, the third group of compounds (R1.2, R1.5, R2.3, R2.4, and R3.2) outperforms the control vildagliptin in this context and the compounds engage in hydrogen bonds with DPP-4 between 70% and 100% of the time.

3.3.3. MM/PBSA binding energy and identification of residues involved in ligand binding

The binding energy for each of the 13 compounds identified in virtual screening was calculated using the Poisson/Boltzmann surface area continuum solvation for the full simulation time of 20 ns (Supplementary Fig. S2). The mean binding energy and its individual terms, derived from the trajectories for each compound, are summarized in Table 2.

Table 2.
Average MMPBSA Binding Energies (kcal mol⁻¹) of 13 Dipeptidyl Peptidase-4-Compound Complexes Over 20 ns Simulations

Complex van der Waal Electro. Polar solvation Nonpolar solvation (SASA) Total binding energy

R1.1 −26.25 −25.22 30.20 −3.04 −24.32

R1.2 −33.51 −13.32 30.17 −3.61 −20.27

R1.3 −26.47 −17.89 32.70 −3.20 −14.87

R1.4 −25.25 −8.52 17.24 −2.82 −19.36

R1.5 −27.69 −26.63 34.34 −3.33 −23.31

R1.6 −28.34 −13.67 27.63 −2.97 −17.36

R2.1 −32.04 −20.69 29.98 −3.66 −26.41

R2.2 −24.55 −8.13 24.11 −3.25 −11.83

R2.3 −34.72 −32.62 51.43 −4.06 −19.98

R2.4 −30.40 −9.67 30.62 −3.71 −13.23

R2.5 −22.47 −12.26 25.95 −3.05 −11.83

R3.1 −21.16 −259.77 88.72 −2.94 −195.35

R3.2 −24.17 −226.76 64.47 −3.39 −180.78

Vildagliptin −20.19 −19.38 19.38 −2.41 −22.61

Complex	van der Waal	Electro.	Polar solvation	Nonpolar solvation (SASA)	Total binding energy
R1.1	−26.25	−25.22	30.20	−3.04	−24.32
R1.2	−33.51	−13.32	30.17	−3.61	−20.27
R1.3	−26.47	−17.89	32.70	−3.20	−14.87
R1.4	−25.25	−8.52	17.24	−2.82	−19.36
R1.5	−27.69	−26.63	34.34	−3.33	−23.31
R1.6	−28.34	−13.67	27.63	−2.97	−17.36
R2.1	−32.04	−20.69	29.98	−3.66	−26.41
R2.2	−24.55	−8.13	24.11	−3.25	−11.83
R2.3	−34.72	−32.62	51.43	−4.06	−19.98
R2.4	−30.40	−9.67	30.62	−3.71	−13.23
R2.5	−22.47	−12.26	25.95	−3.05	−11.83
R3.1	−21.16	−259.77	88.72	−2.94	−195.35
R3.2	−24.17	−226.76	64.47	−3.39	−180.78
Vildagliptin	−20.19	−19.38	19.38	−2.41	−22.61

SASA, solvent accessible surface area.

The binding energies of compounds identified by DPP-4 inhibitor-derived filters (Run 1, Run 2) range between −12 and −26 kcal mol⁻¹ and are thus similar to that of the known inhibitor vildagliptin (−23 kcal mol⁻¹). In contrast, the binding energies of the two compounds identified by pharmacophore search (Run 3) are significantly higher (−181 and −195 kcal mol⁻¹), owing to an electrostatic interaction. Furthermore, evaluation of the contributions of individual residues to the binding energy identified those that most prominently engage with individual ligands. The 10 residues with largest contributions to the estimated binding energy of each ligand in the MD simulations are summarized in Table 3. This analysis shows that all 13 compounds and vildagliptin interact with glutamate-206, and that residues contributing to ligand binding are mainly located in the N-terminal recognition region (Glu204, Glu205, Glu206) and the S2 substrate binding subsite (Phe357, Tyr547) or belong to the catalytic triad (Ser630, Asp708, His740), while in R2.4, one residue from the S1 substrate binding subsite is also involved (Tyr666). For the two compounds identified via pharmacophore-based screening (Run 3), all residues with the largest contributions to MM/PBSA energy are charged, as listed in Table 2. This explains the correspondingly high electrostatic terms (−181 and −195 kcal mol⁻¹).

Table 3.

Top 10 Residues as per MMPBSA Binding Energy Contributions for 13 Compounds Complexed with Dipeptidyl Peptidase-4 from 20 ns Simulations

Complex	Residues
R1.1	E205, E206, Y547, D663, R560, K554, W629, E204, D709, D708
R1.2	R560, Y547, W629, K554, K512, S552, E206, E205, R471, R356
R1.3	W629, E206, H740, G741, H704, K554, K122, R560, D663, R471
R1.4	W629, E206, H740, G741, H704, K554, K122, R560, D663, R471
R1.5	E205, E206, Y547, K554, W629, D663, E204, D709, D739, S630
R1.6	E205, E547, E206, R560, F357, S552, E204, E408, K554, E361
R2.1	E205, E206, R560, W629, Y547, D663, E204, D709, N562, K512
R2.2	R560, K512, R429, K554, W629, N562, H740, V558, R356, L57
R2.3	E206, E205, F357, W629, D663, Y547, E668, K122, D302, D200
R2.4	Y547, R560, K554, E206, E205, D663, Y666, K512, A654, Y631
R2.5	R560, E205, Y547, R471, E206, E204, D739, D708, L57, D709
R3.1	E205, E206, E204, D663, E361, D302, D709, E408, D556, D200
R3.2	E205, E206, D663, D709, E204, D708, D739, D200, D302, E668
Vildagliptin^a	E205, R560, E206, Y547, D663, D708, K554, D709, W629, R429

Known DPP-4 inhibitor.

3.4. Resourced compounds

Following screening and validation of compounds identified using the in silico pipeline discussed above, the selected compounds were resourced from commercial vendors for experimental testing. Where compounds were not available from the commercial vendor, the closest possible analogs were sought. Analogs were selected based on chemical similarity (computed using Tanimoto scoring) with a cutoff value ≥70%. No valid analog was available for R1.2, and thus, only 12 of the 13 compounds were obtained for experimental validation, as shown in Figure 3 under the title “resourced compound.”

FIG. 3.

Overview of compounds screened from all vHTS runs R1, R2, R3 and the resourced analogs 1–12 used for the exit{in vitro} enzyme assay. Percentage similarity (Tanimoto coefficient) is mentioned for each compound; no match was found for R1.2 at the 70% cutoff Tanimoto similarity score. vHTS, virtual high-throughput screening.

Of those purchased, five compounds were identical to the original vHTS hits and are shown in Figure 3 as 1, 5, 6, 11, and 12. Compound 3 has 90% similarity to the original hit, while compounds 7 and 9 are 82% and 83% similar in structure to the vHTS hits, respectively. Compounds 3, 7, and 9 have highly planar topologies of the aromatic rings, which are similar to the corresponding screening hits. Compounds 4 and 8 have ≈ 70% similarity to the original hits; in both cases, this is fragment similarity in that one structural fragment is common to both the resourced and screening hit compounds. However, compound 4 is bulkier than its parent hit compound R1.5, but possesses a terminal chlorobenzene group that matches the original. Similarly, the structure of compound 8 matches the “bottom half” fragment of its corresponding parent compound (R3.2) through a terminal NO₂ group. Compounds 2 and 10 both have 70% similarity to their parent hit compounds and the structures of both resemble those of the corresponding screening hits in the orientations of their terminal groups and aromatic rings.

The purchased compounds that differed from their corresponding screening hits were subjected to 20 ns MD simulations (compounds 2, 3, 4, 7, 8, 9, and 10; see Fig. 2) identical to those used for the original hits. Again, the RMSDs of the structure and hydrogen bonding of each bound ligand were monitored for eh duration of each trajectory, as shown in Figure 4. Moreover, Supplementary Figure S3 shows the fluctuations of temperature and pressure for NVT and NPT ensembles in the 100 ns equilibrium phase and the total potential energy curves of each 20 ns production phase. These compounds showed relatively lower stability during the 20 ns simulation than their screening hit counterparts. Compound 2 showed the highest structural stability and formed a stable complex with DPP-4 after 5 ns. Compounds 3, 4, 7, and 8 reached stable conformations after 10 ns, while compound 9 continued to fluctuate after 15 ns. In addition to RMSD, Figure 4b shows the hydrogen bond percentage throughout the trajectory, indicating that 9 had highest hydrogen bond content, which potentially explains its conformational change later in the simulation, as a means of maximizing hydrogen bonding. All compounds formed hydrogen bonds, although compound 3 showed the lowest hydrogen bond percentage; this is in agreement with its parent hit compound, which had a correspondingly low hydrogen bond score.

FIG. 4.

MD behavior of 12 purchased molecules that showed less than 100% similarity to the original hit compound under 20 ns explicit solvent conditions. (a) RMSD of ligand with respect to protein/ligand complex. (b) Percentage of simulation frames having at least one hydrogen bond.

3.5. In vitro enzyme assay

Following in silico studies, the screening hits (or analogs thereof) were tested for binding to DPP-4 using an in vitro enzyme inhibition assay from the DPP-IV Drug Discovery Kit (Enzo Life Sciences). In this study, the inhibition of DPP-4 by five original hit compounds (1, 5, 6, 11, and 12) and seven hit analogs (2, 3, 4, 7, 8, 9, and 10) was measured and compared with the inhibitory activity of a control inhibitor (P32/98). As the recommended control inhibitor concentration for this kit is 10 μM, all compounds were first tested at this concentration.

As shown in Figure 5a, 2 and 4 had the greatest inhibitory effects and lowered the activity of DPP-4 by 23% and 22%, respectively; the corresponding decreases in DPP-4 activity in the presence of 3 or 7 were 18% and 14%, respectively. None of the other compounds showed any inhibition of DPP-4. The inhibitory activity of 2, 3, 4, and 7 (shown as black bars in Fig. 5a) was further tested at a compound concentration of 100 μM to probe the likely maximum inhibition, as shown in Figure 5b. In this study, 4 had the greatest effect and decreased the activity of DPP-4 by 53%; it can thus be considered a potential lead for downstream inhibitor development. Compound 4 comprises two major fragments, as shown in Figure 5c. These were later separated and individually docked into the proposed DPP-4 binding site. Fragment 1 is the bulkier of the two and contains a spirochromanone motif and an overall topology similar to those of anagliptin, linagliptin, and sitagliptin. Two of the docked poses of fragment 1 (poses 1 and 6) overlap with the docked complexes of DPP-4 with anagliptin and sitagliptin (respectively). A further four poses (poses 2–5) partially overlap with the docked DPP-4:anagliptin complex (as shown in Fig. 5d).

FIG. 5.

In vitro testing of screened compound at (a) 10 μM and (b) 100 μM with their mean relative activity scores and corresponding standard error bar. Compounds that showed relatively high performance at 10 μM (shown in black bars) were tested at 100 μM for comparison with data reported by Guasch et al. (2012). (c) Two possible fragments of the most effective inhibitor at 100 μM. (d) 6 docking poses of Fragment 1 from 6 and its overlapping conformations with cocrystal structures of sitagliptin and anagliptin with DPP-4.

Based on the structure of fragment 1 of most active molecule 4, spirochromanone-based compounds with similar topology were synthesized in-house, as shown in Figure 6a. Initially, these five molecules were examined using the docking and MD techniques described earlier. The hydrogen bond content of the resulting 20 ns simulations is shown in Figure 6b; while all compounds formed hydrogen bonds, compound 13 showed the highest hydrogen bond percentage. Analysis of the RMSD values of each compound's structure during simulation revealed that compound 16 was the least unstable as shown in Figure 6c; the other compounds stabilized after 10 ns.

FIG. 6.

Spirochromanone compounds. (a) Structures of the five spirochromanone compounds. (b) Hydrogen bond percentage of these compounds during 20 ns simulation. (c) RMSD pattern of five spirochromanone compounds during 20 ns simulation. (d) Relative activity of the synthesized spirochromanone compounds at 100 μM with standard error bar. The compound that has the greatest inhibition of DPP-4 (3) is shown as a black bar.

The temperature and pressure of NVT and NPT ensembles in the 100 ps equilibrium phase and the total potential energy curves of the 20 ns simulation phase are shown in Supplementary Figure 3. These five compounds were tested in vitro using against DPP-4 at 100 μM and one compound (13) showed 53% inhibition. The % inhibition of each compound is given in Figure 6d. A recent study by Guasch et al. (2012) also implemented pharmacophore-based screening of the ZINC database to identify DPP-4 inhibitors. It is intriguing that the two inhibitors identified in Round 3 here were not also found in the previous study, although this may be due to updates to the database since the earlier work and improvements to the search algorithms. Nevertheless, when tested using the same DPP-4 assay kit (Enzo Life Science product supplied by Sapphire Bioscience, Australia), the two best leads identified in 2012 reduced DPP-4 activity by 55% and 70% at 150 and 100 μM, respectively. Thus, while the two compounds identified here remain within the micromolar inhibition range, they are in line with previous leads identified for rational drug discovery with DPP-4. Furthermore, the compounds tested in the assay were selected based on commercial availability; many are analogs of the original vHTS hits, which, if available for testing, could potentially have greater activity.

Overall, this study identified two compounds capable of reducing the activity of DPP-4 by 50% at 100 μM. Both possess novel spirochromanone centers not previously seen in DPP-4 inhibitors, although they show similar binding patterns to anagliptin and sitagliptin. These molecules have potential as lead compounds for the development of analogs with greater inhibitory activity against DPP-4.

4. Conclusions

This study demonstrates the efficacy of computer-aided drug design for the discovery of novel inhibitors of DPP-4, a prime target for treating type 2 diabetes and cardiovascular diseases. The present DPP-4 inhibitors were used as reference molecules to guide the structure-based screening approach. A comprehensive library of 4 million compounds was screened in silico using physicochemical filters based on the proposed binding region of DPP-4 to capture potential ligands. Similarly, the ZINC database was searched using a pharmacophore based on the cocrystal structures of DPP-4 with known inhibitors; the resultant 2000 compounds were used as a library for further virtual screening to identify those with the greatest structural complementary to the proposed binding region of DPP-4. The hit compounds were then docked in the binding site and subjected to MD simulations; all assumed stable conformations within the binding site of DPP-4 within 20 ns. MMPBSA calculations were used to predict the stability and binding affinity of hit molecules. This complete pipeline identified 13 compounds with potential binding affinity for DPP-4. As none of these compounds is more than 40% similar in structure (per Tanimoto coefficient) to the others or any known DPP-4 inhibitor, a novel scaffold for DPP-4 inhibition has been identified. The ligand having the highest scoring docked pose of all 13 compounds was found to have negative binding energy using both the AutoDock and INVENTUS docking packages.

Although they use two different algorithms, both packages produced binding scores that are well-correlated (Pearson coefficient 0.84), which supports the reliability of the predicted binding modes of these molecules. The average MMPBSA binding energies of the docked complexes for 20 ns simulations fall within the acceptable range of −26.41 to −11.88 kcal mol⁻¹. The complexes were also assessed for the role of the protein, and the main residues involved in protein/ligand interactions were also detected. Many of these residues are already known to contribute to interactions between DPP-4 and existing inhibitors. The hit compounds (or the closest structural analogs available) were purchased and tested for DPP-4 inhibition in enzyme assays. One molecule from the set of compounds reduced DPP-4 activity to 53% at 100 μM. With a predicted binding pose that is similar to those of know inhibitors (anagliptin and sitagliptin) but a novel scaffold, this compound comprises two structural elements/fragments, one of which contains spirochromanone-type center. To investigate this scaffold further, five additional compounds with similar topology were synthesized and tested both in silico and in enzyme inhibition assays. In this study, a second compound was found to reduce DPP-4 activity to 53%.

In summary, this study has used a rapid in silico screening technique to identify two active compounds, containing a novel spirochromanone scaffold, that are able to reduce DPP-4 activity by 50%. As they are structurally distinct from existing DPP-4 inhibitors/binders, both have the potential to serve as leads for medicinal chemistry approaches to enhance and refine their activity to achieve greater inhibition and, potentially, develop novel treatments for DPP-4-associated disorders.

Footnotes

Acknowledgments

A.M. acknowledges the Department of Biotechnology (DBT), India, for the award of an Indo-Australian Gold fellowship (BT/IACBGF/03/03/2015) and support from the Institute for Integrated and Intelligent Systems, Griffith University. M.C. is supported by PhD scholarships from The Equity Trustees and the Australian Government Research Training Program.

Author Disclosure Statement

The authors declare there are no competing financial interests.

Supplementary Material

References

Abdel-Hamid

M.K.

, and McCluskey

2014. In silico docking, molecular dynamics and binding energy insights into the bolinaquinone-clathrin terminal domain binding site. Molecules, 19, 6609–6622.

Agrawal

, Jain

, and Narayan Dikshit

2012. Ligand-based pharmacophore detection, screening of potential gliptins and docking studies to get effective antidiabetic agents. Comb. Chem. High Throughput Screen. 15, 849–876.

Ali

, and Fonseca

2013. Saxagliptin overview: Special focus on safety and adverse effects. Expert Opin. Drug Saf. 12, 103–109.

Alonso

, Bliznyuk

A.A.

, and Gready

J.E.

2006. Combining docking and molecular dynamic simulations in drug design. Med. Res. Rev. 26, 531–568.

Amaravadhi

, Baek

, and Sup Yoon

2014. Revisiting de novo drug design: Receptor based pharmacophore screening. Curr. Top. Med. Chem. 14, 1890–1898.

Arulmozhi

, and Portha

2006. GLP-1 based therapy for type 2 diabetes. Eur. J. Pharm. Sci. 28, 96–108.

Augeri

D.J.

, Robl

J.A.

, Betebenner

D.A.

, et al. 2005. Discovery and preclinical profile of saxagliptin (BMS-477118): A highly potent, long-acting, orally active dipeptidyl peptidase IV inhibitor for the treatment of type 2 diabetes. J. Med. Chem. 48, 5025–5037.

Bajusz

, Rácz

, and Héberger

2015. Why is Tanimoto index an appropriate choice for fingerprint-based similarity calculations?. J. Cheminform. 7, 20.

Baldi

, and Nasr

2010. When is chemical similarity significant? The statistical distribution of chemical similarity scores and its extreme values. J. Chem. Inf. Model. 50, 1205–1222.

10.

Barnett

2006. DPP-4 inhibitors and their potential role in the management of type 2 diabetes. Int. J. Clin. Pract. 60, 1454–1470.

11.

Berendsen

H.J.

, van der Spoel

, and van Drunen

1995. Gromacs: A message-passing parallel molecular dynamics implementation. Comput. Phys. Commun. 91, 43–56.

12.

Berger

J.P.

, SinhaRoy

, Pocai

, et al. 2018. A comparative study of the binding properties, dipeptidyl peptidase-4 (DPP-4) inhibitory activity and glucose-lowering efficacy of the DPP-4 inhibitors alogliptin, linagliptin, saxagliptin, sitagliptin and vildagliptin in mice. Endocrinol. Diabetes Metab. 1, e00002.

13.

Biftu

, Sinha-Roy

, Chen

, et al. 2014. Omarigliptin (MK-3102): A novel long-acting DPP-4 inhibitor for once-weekly treatment of type 2 diabetes. J. Med. Chem. 57, 3205–3212.

14.

Brown

S.P.

, and Muchmore

S.W.

2007. Rapid estimation of relative protein- ligand binding affinities using a high-throughput version of MM-PBSA. J. Chem. Inf. Model. 47, 1493–1503.

15.

Cau

, Fiorillo

, Mori

, et al. 2015. Molecular dynamics simulations and structural analysis of giardia duodenalis 14-3-3 protein–protein interactions. J. Chem. Inf. Model. 55, 2611–2622.

16.

Chen

, Magliano

D.J.

, and Zimmet

P.Z.

2012. The worldwide epidemiology of type 2 diabetes mellitus—Present and future perspectives. Nat. Rev. Endocrinol. 8, 228.

17.

Chinda

, Palee

, Surinkaew

, et al. 2013. Cardioprotective effect of dipeptidyl peptidase-4 inhibitor during ischemia–reperfusion injury. Int. J. Cardiol. 167, 451–457.

18.

Deng

, Forli

, He

, et al. 2014. Distinguishing binders from false positives by free energy calculations: Fragment screening against the flap site of HIV protease. J. Phys. Chem. B, 119, 976–988.

19.

DeVries

J.H.

, and Rosenstock

2017. DPP-4 inhibitor–related pancreatitis: Rare but real!. Diabetes Care, 40, 161–163.

20.

Dicker

2011. DPP-4 inhibitors: Impact on glycemic control and cardiovascular risk factors. Diabetes Care, 34, S276–S278.

21.

Dolatkhah

, Javanshir

, Sadr

A. S.

, et al. 2017. Synthesis, molecular docking, molecular dynamics studies, and biological evaluation of 4 h-chromone-1, 2, 3, 4-tetrahydropyrimidine-5-carboxylate derivatives as potential antileukemic agents. J. Chem. Inf. Model. 57, 1246–1257.

22.

Eckhardt

, Klein

, Nar

, et al. 2015. Discovery of linagliptin for the treatment of type 2 diabetes mellitus, 129–156. In Successful Drug Discovery. Eds: Fischer

, and Rotella

D.P.

Wiley-VCH Verlag GmbH & Co. KGaA Weinheim, Germany.

23.

Eckhardt

, Langkopf

, Mark

, et al. 2007. 8-(3-(r)-aminopiperidin-1-yl)-7-but-2-ynyl-3-methyl-1-(4-methyl-quinazolin-2-ylmethyl)-3, 7-dihydropurine-2, 6-dione (bi 1356), a highly potent, selective, long-acting, and orally bioavailable DPP-4 inhibitor for the treatment of type 2 diabetes. J. Med. Chem. 50, 6450–6453.

24.

Edmondson

S.D.

, Mastracchio

, Cox

J.M.

, et al. 2009. Aminopiperidine-fused imidazoles as dipeptidyl peptidase-IV inhibitors. Bioorg. Med. Chem. Lett. 19, 4097–4101.

25.

Feng

, Zhang

, Wallace

M.B.

, et al. 2007. Discovery of alogliptin: A potent, selective, bioavailable, and efficacious inhibitor of dipeptidyl peptidase IV. J. Med. Chem. 50, 2297–2300.

26.

Ferrannini

, and DeFronzo

R.A.

2015. Impact of glucose-lowering drugs on cardiovascular disease in type 2 diabetes. Eur. Heart J. 36, 2288–2296.

27.

Fogolari

, Brigo

, and Molinari

2002. The poisson–boltzmann equation for biomolecular electrostatics: A tool for structural biology. J. Mol. Recognit. 15, 377–392.

28.

Gao

, Yang

, and Zhu

2010. Pharmacophore based drug design approach as a practical process in drug discovery. Curr. Comput. Aided Drug Des. 6, 37–49.

29.

Gerich

J.E.

1998. The genetic basis of type 2 diabetes mellitus: Impaired insulin secretion versus impaired insulin sensitivity. Endocr. Rev. 19, 491–503.

30.

Gilbert

R.E.

, and Krum

2015. Heart failure in diabetes: Effects of anti-hyperglycaemic drug therapy. Lancet, 385, 2107–2117.

31.

Guasch

, Sala

, Ojeda

M.J.

, et al. 2012. Identification of novel human dipeptidyl peptidase-IV inhibitors of natural origin (part II): In silico prediction in antidiabetic extracts. PLoS One, 7, e44972.

32.

Gupta

, Gandhimathi

, Sharma

, et al. Pardock: An all atom energy based Monte Carlo docking protocol for protein-ligand complexes. Protein Pept. Lett. 14, 632–646.

33.

Gurung

A.B.

, Bhattacharjee

, and Ali

M.A.

2016. Exploring the physicochemical profile and the binding patterns of selected novel anticancer Himalayan plant derived active compounds with macromolecular targets. Infor. Med. Unlocked, 5, 1–14.

34.

Hausenloy

D.J.

, Whittington

H.J.

, Wynne

A.M.

, et al. 2013. Dipeptidyl peptidase-4 inhibitors and GLP-1 reduce myocardial infarct size in a glucose-dependent manner. Cardiovasc. Diabetol. 12, 154.

35.

Himsworth

H.P.

1936. Diabetes mellitus: Its differentiation into insulin-sensitive and insulin-insensitive types. Lancet, 227, 127–130.

36.

Ihlenfeldt

W.-D.

, Voigt

J.H.

, Bienfait

, Oellien

, et al. 2002. Enhanced CACTVS browser of the open NCI database. J. Chem. Inf. Comput. Sci. 42, 46–57.

37.

Irwin

J.J.

, Sterling

, Mysinger

M.M.

, et al. 2012. Zinc: A free tool to discover chemistry for biology. J. Chem. Inf. Model. 52, 1757–1768.

38.

Jain

, and Jayaram

2005. An all atom energy based computational protocol for predicting binding affinities of protein–ligand complexes. FEBS Lett. 579, 6659–6666.

39.

Joost

H.-G.

2011. The genetic basis of obesity and type 2 diabetes: Lessons from the New Zealand obese mouse, a polygenic model of the metabolic syndrome, 1–11. Eds: Meyerhof

, Ulrike

, and Joost

H.-G.

In Sensory and Metabolic Control of Energy Balance. Springer-Verlag; Berlin, Heidelberg.

40.

Kato

, Oka

, Murase

, et al. 2011. Discovery and pharmacological characterization of n-[2-({2-[(2S)-2-cyanopyrrolidin-1-yl]-2-oxoethyl} amino)-2-methylpropyl]-2-methylpyrazolo [1, 5-a] pyrimidine-6-carboxamide hydrochloride (anagliptin hydrochloride salt) as a potent and selective DPP-IV inhibitor. Bioorg. Med. Chem. 19, 7221–7227.

41.

Kim

M.-K.

, Chae

Y.N.

, Kim

H.D.

, et al. 2012. DA-1229, a novel and potent DPP4 inhibitor, improves insulin resistance and delays the onset of diabetes. Life Sci. 90, 21–29.

42.

Koes

D.R.

, and Camacho

C.J.

2012. ZINCpharmer: Pharmacophore search of the ZINC database. Nucleic Acids Res. 40, W409–W414.

43.

Kollman

P.A.

, Massova

, Reyes

, et al. 2000. Calculating structures and free energies of complex molecules: Combining molecular mechanics and continuum models. Acc. Chem. Res. 33, 889–897.

44.

Kubota

, Takano

, Wang

, et al. 2016. DPP-4 inhibition has beneficial effects on the heart after myocardial infarction. J. Mol. Cell. Cardiol. 91, 72–80.

45.

Kumari

, Kumar

, Consortium

O.S.D

.D., et al. 2014. g_mmpbsa a GROMACS tool for highthroughput MMPBSA calculations. J. Chem. Inf. Model. 54, 1951–1962.

46.

Ligueros-Saylan

, Foley

, Schweizer

, et al. 2010. An assessment of adverse effects of vildagliptin versus comparators on the liver, the pancreas, the immune system, the skin and in patients with impaired renal function from a large pooled database of phase II and III clinical trials. Diabetes Obes. Metab. 12, 495–509.

47.

Lusis

A.J.

2000. Atherosclerosis. Nature, 40, 233–241.

48.

Mathers

C.D.

, and Loncar

2006. Projections of global mortality and burden of disease from 2002 to 2030. PLoS Med. 3, e442.

49.

Mathieu

, and Degrande

2008. Vildagliptin: A new oral treatment for type 2 diabetes mellitus. Vasc. Health Risk Manag. 4, 1349.

50.

Metzler

W.J.

, Yanchunas

, Weigelt

, et al. 2008. Involvement of DPP-IV catalytic residues in enzyme–saxagliptin complex formation. Protein Sci. 17, 240–250.

51.

Moro

, Braiuca

, Deflorian

, et al. 2005. Combined target-based and ligand-based drug design approach as a tool to define a novel 3D-pharmacophore model of human A3 adenosine receptor antagonists: Pyrazolo [4, 3-e] 1, 2, 4-triazolo [1, 5-c] pyrimidine derivatives as a key study. J. Med. Chem. 48, 152–162.

52.

Morris

G.M.

, Huey

, Lindstrom

, et al. 2009. AutoDock4 and AutoDockTools4: Automated docking with selective receptor flexibility. J. Comput. Chem. 30, 2785–2791.

53.

Mukherjee

, and Jayaram

2013. A rapid identification of hit molecules for target proteins via physico-chemical descriptors. Phys. Chem. Chem. Phys. 15, 9107–9116.

54.

Nabeno

, Akahoshi

, Kishida

, et al. 2013. A comparative study of the binding modes of recently launched dipeptidyl peptidase IV inhibitors in the active site. Biochem. Biophys. Res. Commun. 434, 191–196.

55.

Owens

2003. Chris Lipinski discusses life and chemistry after the rule of five. Drug Discov. Today, 8, 12–16.

56.

Patlak

2002. New weapons to combat an ancient disease: Treating diabetes. FASEB J. 16, 1853e.

57.

Perola

, and Charifson

P.S.

2004. Conformational analysis of drug-like molecules bound to proteins: An extensive study of ligand reorganization upon binding. J. Med. Chem. 47, 2499–2510.

58.

Proud

2006. Regulation of protein synthesis by insulin. Biochem. Soc. Trans. 34(Pt 2), 213–216

59.

Remm

, Franz

W.-M.

, and Brenner

2015. Gliptins and their target dipeptidyl peptidase 4: Implications for the treatment of vascular disease. Eur. Heart J. Cardiovasc. Pharmacother. 2, 185–193.

60.

Robert

, Rydel

T.J.

, Storek

M.J.

, et al. 2009. Dicamba monooxygenase: Structural insights into a dynamic Rieske oxygenase that catalyzes an exocyclic monooxygenation. J. Mol. Biol. 392, 481–497.

61.

Röhrborn

, Wronkowitz

, and Eckel

2015. DPP4 in diabetes. Front. Immunol. 6, 386.

62.

Salvo

, Moore

, Arnaud

, et al. 2016. Addition of dipeptidyl peptidase-4 inhibitors to sulphonylureas and risk of hypoglycaemia: Systematic review and meta-analysis. BMJ, 353, i2231.

63.

Schneidman-Duhovny

, Dror

, Inbar

, et al. 2008. Deterministic pharmacophore detection via multiple flexible alignment of drug-like molecules. J. Comput. Biol. 15, 737–754.

64.

Schulz-Gasch

, and Stahl

2004. Scoring functions for protein–ligand interactions: A critical perspective. Drug Discov. Today Technol, 1, 231–239.

65.

Shirakawa

, Okuyama

, Kyohara

, et al. 2016. DPP-4 inhibition improves early mortality, β cell function, and adipose tissue inflammation in db/db mice fed a diet containing sucrose and linoleic acid. Diabetol. Metab. Syndr. 8, 16.

66.

Silva Júnior

W.S.D.

, Godoy-Matos

A.F.D.

, and Kraemer-Aguiar

L.G.

2015. Dipeptidyl peptidase 4: A new link between diabetes mellitus and atherosclerosis?. Biomed. Res. Int. 2015.

67.

Singh

T.P.

, Vangaveti

V N.

, and Malabu

U.H.

2015. Dipeptidyl peptidase-4 inhibitors and their potential role in the management of atherosclerosis—A review. Diabetes Metab. Syndr. 9, 223–229.

68.

Sixto-López

, Bello

, Rodríguez-Fonseca

R.A.

, et al. 2017. Searching the conformational complexity and binding properties of HDAC6 through docking and molecular dynamic simulations. J. Biomol. Struct. Dyn. 35, 2794–2814.

69.

Subramaniam

, Mehrotra

, and Gupta

2008. Virtual high throughput screening (vHTS)—A perspective. Bioinformation, 3, 14.

70.

Tang

, Majeti

, Sudom

, et al. 2012. Inhibitory antibody against dipeptidyl peptidase IV improves glucose tolerance in vivo. J. Biol. Chem. M112.

71.

Tattersall

R.B.

2010. The history of diabetes mellitus 1–22. In Textbook of Diabetes. Eds: Hott

R.I.G.

, Cockram

C.S.

, Flyvbjerg

, and Goldstein

B.J.

Wiley-Backwell; New Jersey.

72.

Thoma

, Löffler

, Stihle

, et al. 2003. Structural basis of proline-specific exopeptidase activity as observed in human dipeptidyl peptidase-IV. Structure, 11, 947–959.

73.

Thornberry

N.A.

, and Weber

A.E.

2007. Discovery of januvia™ (sitagliptin), a selective dipeptidyl peptidase IV inhibitor for the treatment of type 2 diabetes. Curr. Top. Med. Chem. 7, 557–568.

74.

Valiev

, Bylaska

E.J.

, Dupuis

, et al. 2008. Combined quantum mechanical and molecular mechanics studies of the electron-transfer reactions involving carbon tetrachloride in solution. J. Phys. Chem. A, 112, 2713–2720.

75.

Wallace

A.C.

, Laskowski

R.A.

, and Thornton

J.M.

1995. Ligplot: A program to generate schematic diagrams of protein-ligand interactions. Protein Eng. Des. Sel. 8, 127–134.

76.

Watanabe

Y.S.

, Yasuda

, Kojima

, et al. 2015. Anagliptin, a potent dipeptidyl peptidase IV inhibitor: Its single-crystal structure and enzyme interactions. J. Enzyme Inhib. Med. Chem. 30, 981–988.

77.

Yoshida

, Akahoshi

, Sakashita

, et al. 2012a. Discovery and preclinical profile of teneligliptin (3-[(2s, 4s)-4-[4-(3-methyl-1-phenyl-1h-pyrazol-5-yl) piperazin-1-yl] pyrrolidin-2-ylcarbonyl] thiazolidine): A highly potent, selective, long-lasting and orally active dipeptidyl peptidase IV inhibitor for the treatment of type 2 diabetes. Bioorg. Med. Chem. 20, 5705–5719.

78.

Yoshida

, Akahoshi

, Sakashita

, et al. 2012b. Fused bicyclic heteroarylpiperazine-substituted l-prolylthiazolidines as highly potent DPP-4 inhibitors lacking the electrophilic nitrile group. Bioorg. Med. Chem. 20, 5033–5041.

79.

Zhong

, Gong

, Goud

, et al. 2015. Recent advances in dipeptidyl-peptidase-4 inhibition therapy: Lessons from the bench and clinical trials. J. Diabetes Res. 2015.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.01 MB

0.82 MB

0.39 MB

0.24 MB

0.21 MB

0.41 MB

1.28 MB

Screening features	Run 1		Run 2
Screening features	Minimum	Maximum	Minimum	Maximum
HBD	1	3	1	3
HBA	5	10	5	10
logP	2	4	2	4
Aromatic atoms	5	15	5	15
No. of atoms	40	50	40	50
Carboxyl	NF	NF	0	10
Hydroxyl	NF	NF	0	1
Phosphate/sulfate	NF	NF	0	0
Amide	NF	NF	0	10
Ketones/aldehyde	NF	NF	1	4
Ester/ether	NF	NF	0	10
Binding cutoff	−5 kcal mol⁻¹