AncPhore: A versatile tool for anchor pharmacophore steered drug discovery with applications in discovery of new inhibitors targeting metallo- β-lactamases and indoleamine/tryptophan 2,3-dioxygenases

Abstract

We herein describe AncPhore, a versatile tool for drug discovery, which is characterized by pharmacophore feature analysis and anchor pharmacophore (i.e., most important pharmacophore features) steered molecular fitting and virtual screening. Comparative analyses of numerous protein-ligand complexes using AncPhore revealed that anchor pharmacophore features are biologically important, commonly associated with protein conservative characteristics, and have significant contributions to the binding affinity. Performance evaluation of AncPhore showed that it had substantially improved prediction ability on different types of target proteins including metalloenzymes by considering the specific contributions and diversity of anchor pharmacophore features. To demonstrate the practicability of AncPhore, we screened commercially available chemical compounds and discovered a set of structurally diverse inhibitors for clinically relevant metallo-β-lactamases (MBLs); of them, 4 and 6 manifested potent inhibitory activity to VIM-2, NDM-1 and IMP-1 MBLs. Crystallographic analyses of VIM-2:4 complex revealed the precise inhibition mode of 4 with VIM-2, highly consistent with the defined anchor pharmacophore features. Besides, we also identified new hit compounds by using AncPhore for indoleamine/tryptophan 2,3-dioxygenases (IDO/TDO), another class of clinically relevant metalloenzymes. This work reveals anchor pharmacophore as a valuable concept for target-centered drug discovery and illustrates the potential of AncPhore to efficiently identify new inhibitors for different types of protein targets.

Keywords: AMPC, asian mouse phenotyping consortium; AP, anchor pharmacophore; AR, aromatic ring; AUC, area under the curve; Anchor pharmacophore; BACE1, beta-secretase 1; BRD4, bromodomain-containing protein 4; CA, carbonic anhydrase; CA2, carbonic anhydrase 2; CDK2, cyclin-dependent kinase 2; CTS, cathepsins; CV, covalent bonding; CatK, cathepsin K; EF, enrichment factor; EX, exclusion volume; GA, genetic algorithm; HA, hydrogen-bond acceptor; HD, hydrogen-bond donor; HIV-P, human immunodeficiency virus protease; HIV1-P, human immunodeficiency virus type 1 protease; HY, hydrophobic; IDO1, indoleamine 2,3-dioxygenase 1; IMP, imipenemase; Indoleamine 2,3-dioxygenase; LE, ligand efficiency; MAPK14, mitogen-activated protein kinase 14; MB, metal coordination; MBL, metallo-β-lactamase; MIC, minimum inhibitory concentration; MMP, matrix metalloproteinase; MMP13, matrix metallopeptidase 13; Metallo-β-lactamase; Metalloenzyme; NDM, new delhi MBL; NE, negatively charged center; NP, without anchor pharmacophore features; PO, positively charged center; RMSD, root mean square deviation; ROC curve, receiver operating characteristic curve; ROCK1, rho-associated protein kinase 1; RT, reverse transcriptase; RTK, receptor tyrosine kinase; SBL, serine beta lactamase; SSEL, secondary structure element length; STK, serine threonine kinase; TDO, tryptophan 2,3-dioxygenase; TDSS, torsion-driving systematic search; TNKS2, tankyrase 2; Tryptophan 2,3-dioxygenase; VEGFR2, vascular endothelial growth factor receptor 2; VIM, verona integron-encoded MBL; Virtual screening.

Graphical abstract

Figure 1

Anchor pharmacophore features are conservative in target protein family and great contributions to the binding affinity. (A) Analysis of number of pharmacophore features versus binding affinity of >15,000 complex structures revealed a certain degree of relevance between them. (B) Anchor pharmacophore features are commonly conservative within the family of target proteins, as observed for HIV-P, RTK, STK, SBL, MMP, CTS, MBL, and CA family proteins. (C) Comparison of the binding affinity of the protein–ligand complexes with/without anchor pharmacophore features (AP/NP) indicated the features are of significant importance to achieve the high binding affinity.

Figure 2

The activity and mode-of-action of new MBL inhibitors identified by AncPhore. (A) The catalytic mechanism of class B1 di-zinc MBLs. (B) Defined anchor pharmacophore features for B1 MBLs. (C) The inhibitory activity of 4, 6, and 9 with VIM-2, NDM-1, and IMP-1 and their fitting modes with the pharmacophore models. (D) The inhibitory activity for 4, 6, and 9 with VIM-2 at three different concentrations of Zn(II) (0, 1, and 100 μmol/L) revealed their difference in the potential of zinc ion chelation in solution. (E) Crystallographic analysis revealed the binding mode of 4 with VIM-2; the mF_o-DF_c electron density (OMIT maps) around 4 (blue mesh, contoured to 3σ) and view from a crystal structure of the VIM-2:4 complex (PDB code 7CHV). (F) Superimposition of VIM-2:4 and VIM-1:hydrolyzed meropenem (PDB code 5N5I) revealed the importance of anchor pharmacophore features and their intrinsic connection with the substrate binding nature.

Figure 3

The inhibitory activity and binding features of hit compounds 16, 19, and 20 with IDO1/TDO. (A) The catalytic mechanisms of IDO1/TDO mediated oxidative cleavage of the 2,3-indole position of Trp. (B) A metal coordination feature is defined as the anchor pharmacophore feature. (C) The fitting modes of 16, 19, and 20 with the pharmacophore models and their IC₅₀ curves with IDO1/TDO. (D)‒(F) UV–Vis absorption spectra of IDO1/TDO in complex with 16, 19, or 20, revealing that 16 and 19 are likely to bind with heme-iron of IDO1, while 20 does not involve the heme binding for both IDO1 and TDO.

Figure 4

Definition of HA and HD. ‘D’ represents hydrogen bond donor, ‘A’ refers to hydrogen bond acceptor, ‘R’ is the root atom of ‘D’ or ‘A’; $d_{\text{D}-\text{A}}$ is the distance from ‘D’ to ‘A’, ${\theta }_1$ and ${\theta }_2$ represent the angles associated with root atoms. Hydrogen bonds are calculated by four categories according to the number of the root atoms.

Figure 5

Definition of MB and XB. (A) ‘M’ is the active site metal ion(s) in the protein, ‘L’ refers to the ligand atom (O/N/S/F/Se) that is positioned to coordinate with metal ions, and ‘R’ is the root atom of ‘L’; $d_{\text{L}-\text{M}}$ is the distance from ‘L’ to ‘M’, $\theta$ is the directional bond angle, and the purple arrow represents its direction. Metal coordination bonds are calculated by two categories according to the number of ‘L’ root atoms. (B) ‘X’ is the halogen atom (I/Br/Cl) in ligand, ‘A’ refers to the halogen bond acceptor (O/S/N) that has at least one lone pair, and ‘C’ is the carbon atom. $d_{\text{X}-\text{A}}$ is the distance from ‘X’ to ‘A’, R_X and R_A are the van der Waals radii of X and A, respectively. ${\theta }_1$ is the angle of the ‘A’ relative to the C‒X bond, and ${\theta }_2$ is the angle of the ‘X’ relative to the A–C bond. The brown arrow represents its direction.

Figure 6

Definition of AR and CR. (A) An aromatic ring is represented by the mass of the ring and the normal vector perpendicular to the ring plane; AR is calculated by face-to-face and edge-to-face catalogues. (B) CR is calculated by the aromatic ring and the cation center in appropriate distance and angle.