A Small Molecule Selected from a DNA-Encoded Library of Natural Products That Binds to TNF-α and Attenuates Inflammation In Vivo

Abstract

Tumor necrosis factor α (TNF-α) inhibitors have shown great success in the treatment of autoimmune diseases. However, to date, approved drugs targeting TNF-α are restricted to biological macromolecules, largely due to the difficulties in using small molecules for pharmaceutical intervention of protein-protein interactions. Herein the power of a natural product-enriched DNA-encoded library (nDEL) is exploited to identify small molecules that interfere with the protein-protein interaction between TNF-α and the cognate receptor. Initially, to select molecules capable of binding to TNF-α , "late-stage" DNA modification method is applied to construct an nDEL library consisted of 400 sterically diverse natural products and pharmaceutically active chemicals. Several natural products, including kaempferol, identified not only show direct interaction with TNF-α, but also lead to the blockage of TNF-α/TNFR1 interaction. Significantly, kaempferol attenuates the TNF-α signaling in cells and reduces the 12-O-tetradecanoylphorbol-13-acetateinduced ear inflammation in mice. Structure-activity-relationship analyses demonstrate the importance of substitution groups at C-3, C-7, and C-4' of kaempferol. The nDEL hit, kaempferol, represents a novel chemical scaffold capable of specifically recognizing TNF-α and blocking its signal transduction, a promising starting point for the development of a small molecule TNF-α inhibitor for use in the clinical setting.

Keywords: DNA-encoded library; flavonoid; inflammation; kaempferol; tumor necrosis factor α.

Figure 1

Library panning and hit identification of nDELs targeting hTNF‐α. A) Workflow of competitive panning of nDEL library. SPD304, a known small molecule binder to TNF‐α, serves as the competition eluent (red pentagon). B,C) Fingerprint plots for the nDEL screenings enriched by hTNF‐α coated or empty magbeads, respectively, in which y‐axis represents enrichment‐folds, while x‐axis represents sequence counts; red dashed lines are the cut‐off values for hits selection and red dots represent enriched nDELs. D) Chemical structures (upper panel) and summary table (lower panel) of confirmed nDEL hits, in which four compounds, including kumatakenin B (KB), kaempferol (Kae), gancaonin I (GCN), and moxifloxacin (MHCl), showed measurable K _D,app values, and the other four compounds, including flumequine (FMQ), semilicoisoflavone B (SFB), gentisic acid (GA), and glycyrol (GCR), showed weak binding. The “+” sign in the lower panel table represents a weak binding signal in SPR sensorgram that is too weak to quantitate. E) Apparent K _D,app values of the four potent nDEL hits. All results are shown as means ± SD.

Figure 2

SAR analysis of flavonoid analogues. A) Chemical structures (upper panel) and K _D,app measurements (lower panel) of apigenin (APN), galangin (GLG), kaempferide (KMF), and 4',5‐dihydroxyflavone (DHF). B) Competitive affinity measurement of flavonoids to hTNF‐α in the presence of hTNFR1‐ECD‐HRP by ELISA assay (n = 6). C) Summary table of apparent binding affinity of flavonoids to hTNF‐α in the absence (K _D,app) and presence (IC ₅₀) of hTNFR1‐ECD‐HRP. All results are shown as means ± SD.

Figure 3

Effects of nDEL hits on cellular TNF‐α signaling. A) Morphology images of L929 cells upon treatment with actinomycin D (Control), hTNF‐α (TNF‐α), hTNF‐α and adalimumab (TNF‐α+mAb), hTNF‐α and KB (TNF‐α+KB), hTNF‐α and Kae (TNF‐α+Kae), and hTNF‐α and MHCl (TNF‐α+MHCl), respectively. B) Dose‐dependent rescue of hTNF‐α induced cell death of L929 by KB, Kae, and MHCl (n = 6). C) Western‐blot analyses of cellular caspase‐3 and cleaved‐caspase‐3. D) Western‐blot analyses of cellular IκBα. Relative intensity of cleaved caspase‐3 for each testing compound was quantitated by normalizing against band intensity of the internal β‐actin and subtracting band intensity of background. Results are shown as means ± SD (n = 6).

Figure 4

Anti‐inflammatory effect of Kae in TPA induced skin edema mouse model. A) Representative images (200 × magnification) of H&E stained mice ear sections from four cohort groups, in which Indo represents the positive control, indomethacin; Blank represents the untreated group (without TPA and compound treatment); TPA+Vehicle represents the group treated with 2.4 µg TPA and neat DMSO (vehicle); TPA+Indo represents the group with 2.4 µg TPA stimulation and 0.5 mg Indo treatment; and TPA+Kae 0.5 represents the group with 2.4 µg TPA stimulation and 0.5 mg Kae treatment. B) Dose‐dependent efficacy of Kae against TPA‐induced ear edema in mice ear (n = 4). Normalized ear thickness and ear weight of mice in different cohort groups are recorded and compared. For each mouse, normalization was carried out by subtracting the thickness and weight of the other untreated mouse ear from those of the treated ear. Kae 0.5, Kae 0.2, and Kae 0.1 represent different Kae dosages of 0.5, 0.2, and 0.1 mg, respectively. C) Representative FACS histograms. D) Quantitative FACS results showing the comparison of neutrophil amounts in peripheral blood of experimental mice (n = 4). E) mRNA expression of TNF‐α, IL‐1α, and CXCL2 in experimental mice ear determined by RT‐PCR (n = 4). All results were shown as means ± SD. p‐values are calculated using one‐way ANOVA with Bonferroni correction, **p < 0.01, ***p < 0.001, ****p < 0.0001.