Discovery of cell active macrocyclic peptides with on-target inhibition of KRAS signaling
Sotorasib answered one KRAS mutation. This work engineered a peptide that blocks the others inside living cells — then found that the very feature carrying it through the membrane triggers histamine release, a liability the authors argue should give the whole field pause.
A closed macrocycle seated in a shallow pocket on a protein surface, its charged tails trailing away — the shape of the problem this work set out to solve.
In May 2021 the first drug to strike KRAS directly reached patients. Sotorasib, approved for non-small-cell lung cancers driven by the KRAS G12C mutation, ended four decades in which the most frequently mutated oncogene in human cancer had no approved therapy pointed at it. It was, by any measure, a landmark.
It was also narrow. Sotorasib and the covalent inhibitors that follow it depend on the cysteine that the G12C mutation puts at position 12 — a residue present in only 3.4% of colorectal cancers and 7.4% of non-small-cell lung cancers. The mutations that account for most KRAS-driven disease, G12D and G12V among them, offer no such handle. For those patients the target remains where it has always been: inside the cell, on a protein whose surface offers almost nowhere for a small molecule to hold.
A team drawn from MSD International in Singapore, Merck & Co. in the United States and Singapore’s Agency for Science, Technology and Research (A*STAR) set out to reach it another way — with a macrocyclic peptide. Their account, published in Chemical Science in November 2021, is in one sense a success: they built a molecule that blocks KRAS signalling inside living cells, and proved it was doing so on target. In another sense it is a warning, and the authors are explicit that the warning is the more broadly useful finding.
A promising molecule with two fatal habits
They began with KRpep-2d, a peptide identified by phage display by Sakamoto and colleagues. It binds KRAS G12D tightly — surface plasmon resonance put the dissociation constant at 1.7 nM — and it binds somewhere useful: a pocket near the Switch II region, from which it allosterically obstructs SOS1, the exchange factor that reloads KRAS with GTP.
But KRpep-2d was built in ways that suited a binding assay rather than a cell. Its ring is closed by a disulfide bond, which the reducing interior of a cell would be expected to cleave; in the presence of the reducing agent DTT, binding disappeared entirely. Its backbone is also readily cut by proteases. In the authors’ hands the parent peptide did not block ERK phosphorylation in cells at all.
The first repair was the ring itself. Most substitutes for the disulfide destroyed binding outright. What worked was a thioacetal bridge combined with an inverted stereocentre — linking a D-cysteine at position 5 to the cysteine at position 15. That peptide, MP-6483, bound with a TR-FRET EC50 of 172 nM, an eleven-fold improvement on the parent, and, importantly, held its affinity whether the environment was oxidising or reducing.
A co-crystal structure explained why the inversion mattered. With KRAS loaded with a non-hydrolysable GTP analogue, an analogue of the new peptide sat in the same pocket as the original, the two structures overlapping to within 0.51 Å outside the switch I region. The bond between D-Cys5 and Pro6 had flipped into a cis conformation, which let the macrocycle occupy the binding site despite the changed chirality. An alanine scan across the ring confirmed the residues doing the work — Pro6, Leu7, Ile9 and Asp12 — consistent with what had been reported for the parent.
Getting in, and staying intact
Binding is not blocking. To act on KRAS the peptide has to cross a membrane and survive long enough to matter, and KRpep-2d carries eight arginines on its termini precisely to help with the first problem. Those arginines were also being clipped off rapidly by proteases. Replacing all eight with their D-enantiomers gave MP-4090, a peptide that was measurably permeable and that, for the first time in the series, inhibited both ERK and AKT phosphorylation in AsPC-1 cells — a pancreatic line carrying KRAS G12D.
Its potency still decayed over time, pointing to a remaining soft spot inside the ring. An α-methyl scan found that methylation was tolerated at Ser10 without any loss of affinity while improving stability in cell homogenate. Folding that single change into MP-4090 produced MP-3995, whose activity held steady rather than fading: pERK EC50 values of 1.2 µM at one hour and 1.5 µM at eighteen, against 3.6 µM and 30.8 µM for its predecessor.
Proving the effect is real
Cationic, hydrophobic peptides are notorious for producing cell-based results that have nothing to do with the intended target — a problem this group had documented previously, and one they treat here as a design requirement rather than a caveat. Their controls are the most methodologically instructive part of the paper.
They built peptides as close to the active molecules as possible but unable to bind: an all-D version, and single stereo-inversions at the critical Ile9 and Asp12 positions. None bound, and none inhibited signalling, though they entered cells perfectly well. Lactate dehydrogenase release showed the active peptides were not simply perforating membranes. In A375 and SK-MEL-28 cells — driven by BRAF V600E, downstream of KRAS — the peptides did nothing, while small-molecule BRAF and MEK inhibitors behaved as expected. There was no effect on EGF receptor activation, and none on TNFα-stimulated NF-κB signalling.
Two experiments addressed engagement directly. In a cellular thermal shift assay, the active peptides stabilised KRAS in intact cells and the non-binders did not. And when a RAF RBD-CRD-eGFP reporter was introduced into AsPC-1 cells, its normal accumulation at the membrane was lost on treatment with the binders — the peptide was competing with the protein–protein interaction it was designed to disrupt.
Two ways of stopping the switch
The peptides appear to interfere on both sides of the KRAS cycle. Structural comparison suggested they should physically obstruct SOS, and that by pushing switch II and helix H2 outward they should also allosterically prevent RAF’s RAS-binding domain from docking. Both predictions were borne out: the improved analogues inhibited SOS-mediated nucleotide exchange, and blocked the KRAS–RBD interaction more effectively than the parent peptide.
The most pointed demonstration used a KRAS G12C/A59G double mutant, engineered to sit almost entirely in the active GTP-bound state. Sotorasib, which binds only the GDP state, lost its effect against it. MP-3995 did not — evidence that the peptide can still suppress signalling when the protein is pushed into its active conformation, which is precisely the escape route that cancers use to resist covalent G12C inhibitors.
Across a panel of lines, MP-3995 blocked ERK phosphorylation at low micromolar concentrations in KRAS G12C, G12V and G12D backgrounds — pan-KRAS activity, not mutation-specific — and inhibited proliferation in eight of thirteen KRAS-mutant lines, while leaving RAS-independent HEK293 cells alone.
The arginines give, and the arginines take
Then the series ran into something it could not engineer around. Cationic peptides can trigger mast cell degranulation without involving IgE, releasing histamine along with proteases, cytokines and other mediators — a reaction that in a patient can range from itching and hives to fatal anaphylaxis. Tested in an ex vivo rat mast cell assay, the arginine-rich analogues were potent activators of it.
The relationship was uncomfortably clean. Peptides carrying eight or six arginines released histamine at thresholds around 0.14 µM. Cutting to four raised the threshold to roughly 3.7 µM. Peptides with two arginines or fewer were free of the liability altogether — and also, by then, no longer permeable, and no longer active in cells. Cell entry and mast cell degranulation tracked inversely, and neither could be had without the other. Stereochemistry made no difference; it was the charge itself.
This observation should signal to researchers that cationic-mediated cell entry — an approach that has yet to succeed in the clinic despite a long history of attempts — carries significant therapy-limiting safety liabilities.
That sentence, from the authors’ own abstract, is aimed past their molecule and at a design strategy the field has leaned on for years.
What is left standing
The authors do not claim a drug. They are direct that the poly-arginine dependence makes these particular analogues difficult to progress toward the clinic, and that their attempts to remove it cost them the activity they had worked to build.
What survives is narrower but real. The molecules confirm that this epitope on KRAS — distinct from the switch II pocket that the covalent inhibitors exploit — can be engaged inside a living cell, with dual inhibition of both the GDP and GTP states, and across mutations that current approved therapy cannot reach. They offer templates for a next attempt, one whose central problem is now clearly defined: finding a way in that does not rely on positive charge.
The paper is also, quietly, an argument about evidence. Given how readily this class of molecule produces false positives, the authors advocate that such work routinely carry the full array of chemical and biological controls they applied here — and they note, in their discussion of related work on the same scaffold, how difficult results become to interpret when those controls are absent.
Source
Shuhui Lim, Nicolas Boyer, Nicole Boo, Chunhui Huang, Gireedhar Venkatachalam, Yu-Chi Angela Juang, Michael Garrigou, Hung Yi Kristal Kaan, Ruchia Duggal, Khong Ming Peh, Ahmad Sadruddin, Pooja Gopal, Tsz Ying Yuen, Simon Ng, Srinivasaraghavan Kannan, Christopher J. Brown, Chandra S. Verma, Peter Orth, Andrea Peier, Lan Ge, Xiang Yu, Bhavana Bhatt, Feifei Chen, Erjia Wang, Nianyu Jason Li, Raymond J. Gonzales, Alexander Stoeck, Brian Henry, Tomi K. Sawyer, David P. Lane, Charles W. Johannes, Kaustav Biswas and Anthony W. Partridge. “Discovery of cell active macrocyclic peptides with on-target inhibition of KRAS signaling.” Chemical Science, 2021, volume 12, pages 15975–15987. Published 25 November 2021 by the Royal Society of Chemistry under CC BY 4.0. DOI: 10.1039/d1sc05187c.
Affiliations: MSD International, Singapore; Merck & Co., Inc., Boston, Massachusetts, Kenilworth, New Jersey and West Point, Pennsylvania; Agency for Science, Technology and Research (A*STAR), Singapore. Corresponding authors: Charles W. Johannes, Kaustav Biswas and Anthony W. Partridge.
This article is an editorial summary prepared by the Maestro Chronicle. The research, results and conclusions described are those of the authors above and their institutions. Maestro Therapeutics was not involved in the work.