Covalent JNK inhibitors?

Stebbins et al. reported novel JNK inhibitors that block JNK binding to the scaffolding protein JIP1 (1). The most potent compound, BI-78D3, was postulated to bind noncovalently to JNK at the JIP1 site. The binding data and structure of BI-78D3 suggest, however, that BI-78D3 may instead act through covalent modification of JNK at Cys163. (Cys163 is referred to as Cys162 in ref. 1. See www.uniprot.org/uniprot/P45983).

First, these novel JNK inhibitors exhibit a puzzlingly sharp structure–activity relationship (see Table 1 in ref. 1). BI-83C11, in which the BI-78D3 thiazole is replaced by thiofuran, is completely inactive; it is unclear how this modest change leads to such a dramatic potency loss. Yet, mutation of Arg127, proposed to hydrogen bond to BI-78D3, to alanine reduced potency only 2-fold. Equally surprising, removal of the fused dioxane ring (BI-83C9) hardly affected potency, despite the proposal that the dioxane is the most tightly bound part of BI-78D3, deep in the JIP1 hydrophobic pocket (Fig. 2 in ref. 1). Second, ITC data of BI-78D3 binding to wild type and R127A JNK clearly indicate nonreversible binding (Fig. 3 in ref. 1). Only the C163S mutant appears to bind BI-78D3 noncovalently (K_d ≈50 μM). The binding enthalpies are also discordant, with R127A exhibiting the largest ΔH. Third, BI-78D3 is poised to undergo nucleophilic substitution with loss of the highly acidic 2-mercapto-5-nitrothiazole. BI-78D3 may also be prone to hydroquinone-like oxidation/elimination to form a potent electrophile, 4-(2,3-dihydrobenzol[b][1,4]dioxin-6-yl)-4H-1,2,4-triazoline-3,5-dione (2).

Thus, I would suggest that rather than acting noncovalently, BI-78D3 may instead form a covalent adduct with JNK, likely through Cys163, either by direct displacement of 2-mercapto-5-nitrothiazole or by addition to a BI-78D3 oxidation product.