Identification of an in vivo orally active dual-binding protein-protein interaction inhibitor targeting TNFα through combined in silico/in vitro/in vivo screening

Abstract

TNFα is a homotrimeric pro-inflammatory cytokine, whose direct targeting by protein biotherapies has been an undeniable success for the treatment of chronic inflammatory diseases. Despite many efforts, no orally active drug targeting TNFα has been identified so far. In the present work, we identified through combined in silico/in vitro/in vivo approaches a TNFα direct inhibitor, compound 1, displaying nanomolar and micromolar range bindings to TNFα. Compound 1 inhibits the binding of TNFα with both its receptors TNFRI and TNFRII. Compound 1 inhibits the TNFα induced apoptosis on L929 cells and the TNFα induced NF-κB activation in HEK cells. In vivo, oral administration of compound 1 displays a significant protection in a murine TNFα-dependent hepatic shock model. This work illustrates the ability of low-cost combined in silico/in vitro/in vivo screening approaches to identify orally available small-molecules targeting challenging protein-protein interactions such as homotrimeric TNFα.

Figure 1

Flowchart of the screening protocol used in the study. 1st step. A collection of 700,000 drug-like commercially available compounds was screened in silico. Molecular docking was performed using Surflex-dock version 2.5. After visual inspection, a 1000 compound hit list was selected for experimental testing. 2 ^nd step: The inhibitory activity of the compounds composing the hit list was evaluated in vitro on human TNFα induced apoptosis on the L929 cell line. Top hit compounds displayed an IC₅₀ between 1 and 100 µM. 3 ^rd step: 2D/3D similarity search methods were used to identify analogues of the top hits identified after step 2. Up to 100 analogues were found per top hit with a Tanimoto similarity score >0.6. 4 ^th step: As in step 2, their inhibitory activity was evaluated in vitro on human TNFα induced apoptosis on the L929 cell line. The 10 best compounds after these 4 steps were selected as candidates for in vivo evaluation on a murine model. 5 ^th step: The in vivo evaluation of the candidates was performed in the TNFα-dependent hepatic shock model triggered with LPS/D-Galactosamine via force-feeding. After this step, 1 in vivo active compound was selected. 6 ^th Step: Using 2D/3D similarity search methods, we searched in our large compound collection for new analogues of this best compound identified after step 5. Up to 500 analogues were identified and purchased from the chemical supplier. As in step 2, their inhibitory activity was evaluated in vitro on human and murine TNFα. The 9 best compounds were evaluated in vivo in our murine hepatic shock assay by force-feeding as described in step 5. The best compound identified after the 6^th step is compound 1.

Figure 2

(a) Structure of compound 1. (b) Dissociation constants of TNFα/Compound 1. Determination of dissociation constants of TNFα/compound 1 complex, from gravimetric biosensor response, by using a “two-site binding” model. K _d1 = 4.79 ± 1.12 µM and K _d2 = 2.31 ± 1.03 nM. (c) Intrinsic Tryptophane Fluorescence. Intrinsic Tryptophan Fluorescence of 0.5 μM TNFα diluted in Phosphate Buffered Saline in the presence of DMSO alone, SPD304 (5–100 µM) in DMSO, compound 1 (5–100 µM) in DMSO or SPD304 (25 µM) + compound 1 (10–100 µM) in DMSO.

Figure 3

(a) Illustration of the top scoring binding mode of compound 1 with Surflex-dock in the TNFα binding site defined for the study (PDBID:2AZ5). (b) Alternate binding mode of compound 1 (green) predicted using Surflex-dock on the surface of TNFα dimer. Tryptophan residues are displayed in orange. (c) Secondary pocket on the structure of human dimeric TNFα. (d) Dual binding of compound 1 (yellow and green) predicted with Surflex-dock on the surface of the TNFα dimer co-cristallized with SPD304 displayed in purple (PDB ID: 2AZ5). Tryptophan residues are displayed in orange.

Figure 4

(a) Relative binding on TNFRI and TNFRII. Compound 1 inhibits the interaction between TNFα and its receptors TNFRI and TNFRII in a dose dependant manner. TNFRI IC₅₀ = 37 μM, TNFRII IC₅₀ = 31 μM. (b) Neutralization of TNFα and survival of L929 cells. Compound 1 inhibition of TNFα induced apoptosis in L929 cell line. Data represent neutralization of TNFα in presence of various concentrations of compound 1. IC₅₀ = 12 μM (c) Neutralization of TNFα and survival of HEK cells. Compound 1 inhibition of the TNFα signaling pathway on HEK cells transfected with a reporter gene under the control of NF-κB. Data represent the neutralization of TNFα in the presence of different concentrations of compound 1. IC₅₀ = 10 μM. (d) Inhibition of CXCL1 secretion in L929 supernatants in presence of various concentrations of compound 1 after stimulation with TNFα (5 ng/ml). (e) Inhibition of caspase 3 activity in presence of compound 1. (f) Inhibition of caspase 8 activity in presence of compound 1.

Figure 5

Effect of compound 1 in an in vivo murine model of LPS/D-Galactosamine induced shock. (a) Mice survival after force-feeding with different doses of compound 1 and an intraperitoneal injection of LPS/D-Galactosamine. Groups of eight mice were used. Values are mean ± s.e.m. (b) TNFα levels in mice sera three hours after the induction of the shock by LPS/D-Galactosamine. Mice were forced-fed with 5 mg compound 1 in DMSO. Groups of eight mice were used. Values are mean ± s.e.m. (c) Aspartate aminotransferase (AST) levels in mice sera eight hours after the induction of the shock. Mice were forced-fed with 5 mg compound 1 in DMSO. Values are mean ± s.e.m. (d) Alanine aminotransferase (ALT) levels in mice sera eight hours after the induction of the shock. Mice were forced-fed with 5 mg compound 1 in DMSO. Values are mean ± s.e.m. (e) Livers sections (H&E staining) from mice forced-fed with DMSO (control) or compound 1 or injected with etanercept and an intraperitoneal injection of LPS/D-Galactosamine. (f) Liver sections from mice forced-fed with DMSO (control) or compound 1 or injected with etanercept and an intraperitoneal injection of LPS/D-Galactosamine. Sections were incubated with an anti-cleaved caspase-3 antibody and revealed with an anti-rabbit antibody and DAB.