Blocking C-terminal processing of KRAS4b via a direct covalent attack on the CaaX-box cysteine

Abstract

RAS is the most frequently mutated oncogene in cancer. RAS proteins show high sequence similarities in their G-domains but are significantly different in their C-terminal hypervariable regions (HVR). These regions interact with the cell membrane via lipid anchors that result from posttranslational modifications (PTM) of cysteine residues. KRAS4b is unique as it has only one cysteine that undergoes PTM, C185. Small molecule covalent modification of C185 would block any form of prenylation and subsequently inhibit attachment of KRAS4b to the cell membrane, blocking its biological activity. We translated this concept to the discovery and development of disulfide tethering screen hits into irreversible covalent modifiers of C185. These compounds inhibited proliferation of KRAS4b-driven mouse embryonic fibroblasts, but not cells driven by N-myristoylated KRAS4b that harbor a C185S mutation and are not dependent on C185 prenylation. Top-down proteomics was used to confirm target engagement in cells. These compounds bind in a pocket formed when the HVR folds back between helix 3 and 4 in the G-domain (HVR-α3-α4). This interaction can happen in the absence of small molecules as predicted by molecular dynamics simulations and is stabilized in the presence of C185 binders as confirmed by small-angle X-ray scattering and solution NMR. NOESY-HSQC, an NMR approach that measures internuclear distances of 6 Å or less, and structure analysis identified the critical residues and interactions that define the HVR-α3-α4 pocket. Further development of compounds that bind to this pocket could be the basis of a new approach to targeting KRAS cancers.

Keywords: C185; HVR; KRAS; KRAS4b; covalent inhibitor.

Fig. 1.

Discovery and development of C185 covalent inhibitors. (A) Disulfide tethering hit led to discovery of compound 1. Left Panel: Mass spectra of disulfide (compound 1) tethering reaction with KRAS4b 1–188 protein. Unmodified FLAG KRAS 4b G12V is represented as 24,823 Da peak. Peak labeled as fragment adduct is FLAG KRAS4b G12V protein bearing mass addition of 355 Da corresponding to the tethered fragment. Right panel: disulfide hit 1 did not react with KRAS protein when C185 was mutated to serine (C185S). (B) MALDI-TOF MS spectra obtained after Glu-C digest of KRAS4b (1–188). Top panel: control protein, bottom panel: protein modified with compound 2 that is the covalent, vinyl sulfonamide analogue of tethering hit 1. Peak m/z 2696 indicates the 372 Da mass addition to C-terminal C185. The C185 modification was the only one detected on the protein, other cysteine residues were not affected. (C) Covalent inhibitor 2 decreased proliferation and KRAS protein level in KRAS-dependent MEF cells. Compound 2 reduced proliferation of MEF cells driven by mutant KRAS4b, while MEFs rescued by myristoylated KRAS4b G12D/C185S were much less affected (72-h treatment). (D) Treatment with compound 2 decreased KRAS4b protein level in KRAS mutant MEF cells, but not in cells driven by myristoylated KRAS G12D/C185S. Compound 2 reduced levels of KRAS4b G12C and KRAS4b G12D, but not Myr-KRAS G12D/C185S. 72 h treatment with concentrations indicated on the image; Vinculin was used as a loading control.

Fig. 2.

Effects of compound 3 on KRAS4b but not NRAS, and C185 target engagement in cells. (A–C) GFP-KRAS4b G12D and GFP-NRAS G12D localization in HeLa cells. GFP-KRAS4b G12D and GFP-NRAS G12D localization was monitored in doxycycline-inducible HeLa cell lines by colocalization with concanavalin A at the plasma membrane. HeLa clonal cell lines expressing GFP-KRAS4b G12D (A), or GFP-NRAS G12D (B) were incubated with DMSO or 10 µM Compound 3 (Fig. 2C) for 24 h. KRAS4b G12D-expressing cells show a decrease in KRAS4b G12D localization at the membrane compared to the DMSO control. Cell nuclei were stained with Hoechst (blue) and plasma membranes were stained with Concanavalin A (red). (Scale bar, 5 µm.) (D) A significant difference in membrane localization of KRAS4b was observed using Pearson coefficient (mean ± SE, n = 25 for each condition). ****P < 0.0001 (E) Compound 3 decreased the amount of KRAS in MEF cells driven by KRAS4b G12V, but not in MEF cells driven by NRAS G12V. Treatment with 20 µM compound 3 for 48 h decreased KRAS4b levels in KRAS G12V MEFs. WCL – whole cell lysate. (F–I) Visualization of the compound 3-bound KRAS4b proteoform by high-resolution top–down MS, employing a selected ion monitoring (SIM) method targeting the 24+ charge state of the intact protein. (F) KRAS4b proteoforms isolated from DMSO-treated cells; (G) KRAS4b proteoforms isolated from compound 3-treated cells; (H) KRAS4b proteoforms isolated from cells treated with compound 3, and a dual farnesyl- and geranylgeranyl-transferase inhibitor, L-778,123. The unprocessed KRAS4b proteoform bearing the + 491.17 Da mass shift of bound compound 3 is indicated in purple, along with the processed and farnesylated, or canonical, KRAS4b G12D proteoform (PFR 249921). Loss of C-terminal carboxymethylation (PFR 249922) is indicated by the label “-COOMe”, while asterisks indicate oxidation products of electrospray ionization or other unrelated species. Inverted purple triangles indicate KRAS4b proteoforms believed to bear intermediates or metabolic products of compound 3, with calculated mass shifts as follows: a, + 156.98 Da; b, + 221.96 Da; c, + 300.96 Da; d, + 397.99 Da from the mass of the canonical proteoform. (I) Localization of KRAS-bound compound 3 to the targeted Cys residue by partial proteolytic digest and subsequent middle–down MS. Example identified peptides are shown, demonstrating that the +491.17 Da mass shift of compound 3 could be localized only to Cys 185, and not to any other Cys residue within the KRAS4b sequence.

Fig. 3.

MD simulation and SAXS data identify KRAS4b HVR interactions with the G-domain. (A) The site maps of a tricomponent pocket formed by HVR interacting with helix3 (α3) and helix4 (α4) in the G-domain of KRAS4b. (B) Interactions of KRAS4b HVR with the α3/α4 site (HHH, red line) is energetically favorable over interaction of KRAS4b HVR with α3/Switch2 site (HHS, blue line). (C) Among four major RAS isoforms, KRAS4b exhibits unique behavior of HVR interacting with the G-domain, that is not seen for other major RAS proteins. HVR of KRAS4b only interacts with the G-domain of the protein in the region between α3/α4. (D) MALDI-TOF MS analysis of covalent labeling to HVR cysteine residues in RAS proteins with selected compounds. KRAS4b (1–188) shows highest degree of covalent modification to C185, and NRAS (1–189) shows the lowest. Interestingly, the high level of labeling to C185 in KRAS4b-G-domain/NRAS-HVR chimera suggests KRAS4b HVR dictates interaction with the G-domain. (E) Structure of compound 4 and RMSF comparison of KRAS4b HVR without ligand (black), and with compound 4 covalently attached to C185 (red). Presence of the ligand stabilizes the HVR interaction with the G-domain, especially residues 182–187. (F) Three-dimensional molecular envelopes reconstructed from SAXS data. Thirty-two independent DAMMIN reconstructions were run using GNOM output from three KRAS constructs: KRAS4b-1–169 (green), the full-length protein KRAS4b(1–188) (rose), and KRAS4b(1–188) with C185 covalently modified with compound 4 (purple). The dummy atom models were then aligned, averaged, and filtered using DAMAVER. The corresponding crystallographic structures fitted to the ab initio envelopes are shown in two orientations.

Fig. 4.

Heteronuclear NMR for binding site identification of compound 4. (A) 2D ¹H-¹⁵N HSQC spectra comparison of the G-domain of KRAS4b (1–169) and the full-length protein (1–188) indicates differences in chemical shifts resulting from the interaction of the HVR with the G-domain. Residues eliciting effective CSP (SI Appendix, Fig. S4 B and C) are highlighted by single-letter amino acid sequence followed by its sequence number. CSP residue names superscripted with letter “G” belong to the G-domain. Unassigned crosspeaks in black do not belong to the G-domain residues. (B) Ensemble of 20 solution structures of KRAS4b (1–188). AA residues 174–188 of HVR at C-terminus are found conformationally flexible. (C) Detection of intramolecular NOE contacts between the aa residues in the HVR and the catalytic G-domain in KRAS4b (1–188) with compound 4 covalently attached to C185, as determined in the ¹⁵N-edited-NOESY-HSQC and ¹³C-edited NOESY-HSQC spectra at 25 °C (See SI Appendix, Fig. S4 G and H for additional such NOEs detected between the HVR and the G-domain). Top panels: amide protons (H^N) of M188 and its conformer show NOE with amide proton (H^N) of residue F82 and with aromatic ring protons (H^δ and H^ζ) rendered blue. Bottom panels, amide proton (H^N) of E91 shows NOE with side-chain methylene proton (Hγ¹²) of residue I187. Magnified view on left represents spectral regions in black boxes. None of the NOEs overlap with proton chemical shifts of compound 4. No such NOEs were observed in the samples treated with DMSO (in absence of compound 4), as seen in panels shown in red. (D) Representative solution structure of KRAS4b (1–188) upon inclusion of 7 intramolecular NOEs that are observed between the HVR residues I187 and M188, and G-domain residues F82, E91, and D92; these residues are shown in red. Rendered orange are the residues that show significant CSP (deduced from HSQC overlay of KRAS4b (1–188) treated with DMSO vs tethered with compound 4) in the tricomponent binding region (HVR-α3-α4).