Homology Modeling of Human Concentrative Nucleoside Transporters (hCNTs) and Validation by Virtual Screening and Experimental Testing to Identify Novel hCNT1 Inhibitors

Abstract

Objective The nucleoside transporter family is an emerging target for cancer, viral and cardiovascular diseases. Due to the difficulty in the expression, isolation and crystallization of membrane proteins, there is a lack of structural information on any of the mammalian and for that matter the human proteins. Thus the objective of this study was to build homology models for the three cloned concentrative nucleoside transporters hCNT1, hCNT2 and hCNT3 and validate them for screening towards the discovery of much needed inhibitors and probes. Methods The recently reported crystal structure of the Vibrio cholerae concentrative nucleoside transporter (vcCNT), has satisfactory similarity to the human CNT orthologues and was thus used as a template to build homology models of all three hCNTs. The Schrödinger modeling suite was used for the exercise. External validation of the homology models was carried out by docking a set of recently reported known hCNT1 nucleoside class inhibitors at the putative binding site using induced fit docking (IDF) methodology with the Glide docking program. Then, the hCNT1 homology model was subsequently used to conduct a virtual screening of a 360,000 compound library, and 172 compounds were obtained and biologically evaluated for hCNT 1, 2 and 3 inhibitory potency and selectivity. Results Good quality homology models were obtained for all three hCNTs as indicated by interrogation for various structural parameters and also external validated by docking of known inhibitors. The IDF docking results showed good correlations between IDF scores and inhibitory activities; particularly for hCNT1. From the top 0.1% of compounds ranked by virtual screening with the hCNT1 homology model, 172 compounds selected and tested for against hCNT1, hCNT2 and hCNT3, yielded 14 new inhibitors (hits) of (i.e., 8% success rate). The most active compound exhibited an IC50 of 9.05 μM, which shows a greater than 25-fold higher potency than phlorizin the standard CNT inhibitor (IC50 of 250 μM). Conclusion We successfully undertook homology modeling and validation for all human concentrative nucleoside transporters (hCNT 1, 2 and 3). The proof-of-concept that these models are promising for virtual screening to identify potent and selective inhibitors was also obtained using the hCNT1 model. Thus we identified a novel potent hCNT1 inhibitor that is more potent and more selective than the standard inhibitor phlorizin. The other hCNT1 hits also mostly exhibited selectivity. These homology models should be useful for virtual screening to identify novel hCNT inhibitors, as well as for optimization of hCNT inhibitors.

In Silico Screening of Isocitrate Lyase for Novel Anti-Buruli Ulcer Natural Products Originating from Africa

“…The GlideScore, which is used to rank resulting complexes after induced-fit docking, is an empirical scoring function that provides an estimate of the binding affinity between a ligand and a receptor. Lower GlideScores are mostly representative of reasonably good binding between a ligand and a receptor [ 80 , 81 , 82 ]. Therefore, the more negative the GlideScores, the more plausible the binding [ 83 ].…”

Homology Model and Docking-Based Virtual Screening for Ligands of Human Dyskerin as New Inhibitors of Telomerase for Cancer Treatment

“…The models obtained were used to carry out a DBVS assay. Then, the compounds were evaluated, obtaining as a result a molecule that has 25 times greater inhibitory ability than the commercial inhibitor [ 43 ]. Therefore, as shown by other authors, the bioinformatic path that we applied here is able to select molecules with the activity mentioned in our hypothesis.…”

Molecular Insight into Stereoselective ADME Characteristics of C20-24 Epimeric Epoxides of Protopanaxadiol by Docking Analysis

“…Then the crystal structure of macrolide glycosyltransferase was screened from PDB (ID: 2IYA) as the best template for modeling. The protein 2IYA displayed a sequence homology above 37% and the similarity to UGT1A8 could be reach up to 33% (Figure 2), indicating a reliable template for homology modeling in view of the rule of thumb that a sequence homology of 30% or above is adequate [27].…”

“…The optimal ligand conformations for docking were screened via Emodel module in light of the GlideScore, a mixture of interaction energy and parameter-based penalty functions that roughly represents binding free energy (∆G bind ) [25]. The ∆G bind that permits accurate analysis of contribution from each residue by decomposing ligand-protein interaction into different terms [26,27] was calculated to quantitatively assess and rank the ligand-protein binding modes. The docking results were clustered for further analysis, including binding type, the residues involved and some empirical scoring function.…”