Abstract
K-Ras is the most commonly mutated oncogene in human cancer. The recently approved non-small cell lung cancer drugs sotorasib and adagrasib covalently capture an acquired cysteine in K-Ras-G12C mutation and lock it in a signaling-incompetent state. However, covalent inhibition of G12D, the most frequent K-Ras mutation particularly prevalent in pancreatic ductal adenocarcinoma, has remained elusive due to the lack of aspartate-targeting chemistry. Here we present a set of malolactone-based electrophiles that exploit ring strain to crosslink K-Ras-G12D at the mutant aspartate to form stable covalent complexes. Structural insights from X-ray crystallography and exploitation of the stereoelectronic requirements for attack of the electrophile allowed development of a substituted malolactone that resisted attack by aqueous buffer but rapidly crosslinked with the aspartate-12 of K-Ras in both GDP and GTP state. The GTP-state targeting allowed effective suppression of downstream signaling, and selective inhibition of K-Ras-G12D-driven cancer cell proliferation in vitro and xenograft growth in mice.
Main
Oncogenic mutations of Ras are among the most common genetic alterations in human cancer, with an estimated disease burden of >3 million new patients per year worldwide1. Despite widespread appreciation of the importance of Ras in cancer, direct binding ligands that block downstream signaling were not reported until 2013 (ref. 2) due to the lack of obvious drug-binding pockets in the protein. The only clinically approved K-Ras inhibitors are completely mutant specific as they rely on covalent recognition of the highly nucleophilic somatic cysteine residue of K-Ras-G12C. Recent preclinical reports3,4 of noncovalent K-Ras binding inhibitors have emerged that lack exclusive mutant specificity and exhibit varying degrees of biochemical preference for mutant K-Ras over the wild type. The challenge in covalent targeting the most common single variant in K-Ras-driven tumors (G12D) is that the carboxylic acid side chains of Asp and Glu are among the least nucleophilic heteroatom containing amino acid (a.a.) side chains. A further challenge for electrophiles capable of reacting with weak nucleophiles (Asp/Glu) is the need to withstand 55 M water and hypernucleophiles such as glutathione (GSH) in biological buffers. No suitable electrophiles for selectively targeting acidic a.a. side chains in aspartic or glutamic acids at physiological pH have been reported5,6,7,8 despite advances9 in targeting neutral or basic ones such as lysine10,11,12, tyrosine13, serine14,15, arginine16, methionine17 and histidine18,19,20. In this Article, we report an electrophile that alkylates the mutant aspartate in K-Ras-G12D in both GDP- and GTP-bound states, blocks effector interactions and inhibits the growth of K-Ras-G12D-driven cell lines.
Results
A malolactone covalently modifies mutant Aps12
To covalently target the mutant aspartate, we focused on electrophiles designed to exploit strain release upon carboxylate attack of small three- and four-membered heterocyclic ring systems. Our21 and others’22 exploration of electrophiles from the literature showed modest levels of K-Ras-G12D crosslinking exhibited by three-membered rings—NH-aziridine and epoxide electrophiles covalently modified K-Ras-G12D by 40–60% over the course of 24 h (Extended Data Fig. 1a,b). Further optimization proved unsuccessful due to the limited number of chemically stable modifications possible in three-membered rings. A recent report23 demonstrated that structurally similar epoxides could label recombinant K-Ras-G12D at extremely high concentration (1 mM), which precluded its use as a targeted covalent inhibitor in cells. A covalent K-Ras-G12D inhibitor based on cyclophilin recruitment has been reported in a meeting abstract8.
The four-membered ring electrophile β-lactone, found in both natural products and synthetic drugs14,24,25, was attractive because the parent β-propiolactone, a potent acylating and alkylating reagent26,27,28,29,30,31,32, modified K-Ras-G12D at multiple sites including the target Asp12 residue (Extended Data Fig. 2a) in a dose-dependent manner (Extended Data Fig. 2b). However, the [4.2.0]-fused β-lactones that covalently acylated a sibling K-Ras-G12S mutation14 were not reactive with G12D (Supplementary Fig. 1a), requiring exploration of another means to activate the nucleophilic attack of the β-lactone by Asp12.
Ambident electrophile β-lactones react with carboxylates with a preference at the β-carbon via an alkylation pathway at the β-carbon25,33 (cf. alcohols such as the serine side chain prefer acylation via carbonyl attack). The alkylation trajectory in the [4.2.0]-fused β-lactones at the bridgehead carbon was not accessible to Codon12 side chains (Supplementary Fig. 1b). We reasoned that switching the fused ring system to a simple carbonyl bridge would (1) geometrically enable the β-lactone electrophile to be poised for SN2 attack from the P-Loop Asp12, (2) activate the alkylation pathway by carbonyl π*-orbital participation and (3) potentially provide further hydrogen bonding activation by the neighboring Lys16 shown to be important in activating of K-Ras-G12C acrylamide electrophiles34.
A series of recently reported high-affinity noncovalent small molecules35,36, including MRTX1133 (refs. 3,37), which binds to K-Ras-G12D and wild-type (WT) K-Ras38 provided the necessary scaffold for presenting the electrophilic fragment toward the Asp12. We prepared (RS)-G12Di-1 (1) as a racemate by coupling (±)-β-carboxyl-β-propiolactone, also known as malolactone, to the bicyclic piperazine group of a Switch-II Pocket (S-IIP) ligand scaffold using the carbonyl bridge design (Fig. 1a). Using whole-protein mass spectrometry (MS), we assessed the reactions between recombinant K-Ras proteins and 10 µM (RS)-G12Di-1 at 23 °C (Fig. 1b,c). (RS)-G12Di-1 reacted rapidly with K-Ras-G12D•GDP with a half-life of 99 s (95% confidence interval 83–118 s) and fully modified K-Ras-G12D•GDP in less than 15 min (Fig. 1d). By contrast, (RS)-G12Di-1 achieved only 6.5 ± 0.3% modification of K-Ras-G13D•GDP and no detectable modification of K-Ras-WT•GDP after 1 h. Interestingly, (RS)-G12Di-1 showed markedly reduced reactivity with K-Ras-G12E•GDP, a nonnatural mutant that positions its carboxylate nucleophile (Glu) further from the backbone than that in G12D, and K-Ras-G12S•GDP. Although less potent, (RS)-G12Di-1 was also active against the GppNHp-bound K-Ras-G12D. This is distinct from S-IIP K-Ras-G12C inhibitors, which exclusively recognize the GDP-bound state2,34,39,40,41,42, whereas GTP-state recognition has so far only been observed for compounds with a different binding mechanism8,43. A recently reported K-Ras-G12C inhibitor dependent on cyclophilin recruitment is capable of GTP-state recognition and showed faster in-cell signaling inhibition kinetics43. The covalent modification of Asp12 by (RS)-G12Di-1 stabilized both K-Ras-G12D•GDP and K-Ras-G12D•GppNHp toward thermal denaturation (ΔTm = +10.3 °C and +2.5 °C, respectively). The adduct between K-Ras-G12D and (RS)-G12Di-1 was stable over pH 4.5–7.5, and did not degrade in the presence of 1 vol% of dithiothreitol (DTT) or hydrazine at 23 °C showing its intrinsic resistance to nucleophiles in the bulk solution (Extended Data Fig. 3).
The ability of (RS)-G12Di-1 to engage the GppNHp-bound state of K-Ras-G12D prompted us to ask whether such a covalent modification could disrupt the interaction with effector proteins such as Raf. Using a time-resolved fluorescence energy transfer assay44, we found that (RS)-G12Di-1 inhibited the interaction between Raf-RBD and K-Ras-G12D, but not WT K-Ras, in a dose-dependent fashion (Fig. 1e). We also measured the interaction between immobilized Raf-RBD with fully labeled K-Ras-G12D•GppNHp•(RS)-G12Di-1 using biolayer interferometry. Compared to unmodified K-Ras-G12D•GppNHp, the K-Ras-G12D•GppNHp•(RS)-G12Di-1 complex showed substantially decreased binding to Raf-RBD (Fig. 1f). Such a feature may be particularly advantageous for the G12D mutant, as its severely impaired GTP hydrolysis45 renders an abundant and persistent K-Ras population in the active GTP-state in the cell.
Mechanism- and structure-guided ligand evolution
To better understand the covalent reaction in the S-IIP of KRAS (G12D) between (RS)-G12Di-1 and Asp12, we solved a 1.7-Å crystal structure of the K-Ras-G12D•GDP•(RS)-G12Di-1 adduct (Fig. 1g). We observed (RS)-G12Di-1 in the S-IIP pocket adopting a conformation similar to that seen for K-Ras-G12C inhibitors (Supplementary Fig. 2), with clear electron density for the covalent ester bond between Asp12 and the compound as well as the free carboxyl group resulting from the ring opening. This high-resolution structure also allowed us to assign the stereochemistry of the adduct as S at the β-carbon (Fig. 1h,i). Because we obtained the co-crystal using racemic (RS)-G12Di-1, this S stereochemistry could in theory result from an SN2 attack on the β-carbon of R enantiomer of G12Di-1 or an attack on the carbonyl of the S enantiomer of G12Di-1 followed by an acyl transfer (Extended Data Fig. 4a). To distinguish between these two possibilities, we prepared pure R and S enantiomers of a structural analog. Monitoring the reaction rate with K-Ras-G12D•GDP showed that the R enantiomer (R)-G12Di-2 (2) was significantly more reactive toward K-Ras-G12D•GDP (Extended Data Fig. 4b). Furthermore, we observed the same R » S Cα-enantioselectivity (that is, C2 in malolactone system) in the context of K-Ras-G12D labeling across a panel of strain-release electrophiles such as epoxides and aziridines that can only react via an SN2 mechanism (Extended Data Fig. 1). Such enantioselectivity favoring a direct SN2 ring-opening mechanism aligns with literature observations with β-lactones25,33. The predominant SN2 preference of ring-opening reaction by Asp12 is distinct from the preference for attack by water, that is, hydrolysis, which often takes place at the carbonyl in near-neutral buffers46.
Despite being a potent covalent ligand of recombinant K-Ras-G12D, (RS)-G12Di-1 was not stable in aqueous buffers at pH 7.4, precluding its use in cellular assays. We reasoned that its stability could be improved by attaching substituents at either or both sides of the α-carbon. Substitutions at the α-carbon will block the trajectory of incoming water nucleophiles from at least one side of the β-lactone ring (Fig. 2a, lower path). The same modification of the electrophile was predicted to have little impact on the SN2 attack by Asp12 because of the pseudo-staggered conformation of the lowest unoccupied molecular orbital of the C–O bond and the α-carbon steric hindrance (Fig. 2a, upper path)….
