Chemical acylation of an acquired serine suppresses oncogenic signaling of K-Ras(G12S)

Drugs that directly impede the function of driver oncogenes offer exceptional efficacy and a therapeutic window. The recently approved mutant selective small-molecule cysteine-reactive covalent inhibitor of the G12C mutant of K-Ras, sotorasib, provides a case in point. KRAS is the most frequently mutated proto-oncogene in human cancer, yet despite success targeting the G12C allele, targeted therapy for other hotspot mutants of KRAS has not been described. Here we report the discovery of small molecules that covalently target a G12S somatic mutation in K-Ras and suppress its oncogenic signaling. We show that these molecules are active in cells expressing K-Ras(G12S) but spare the wild-type protein. Our results provide a path to targeting a second somatic mutation in the oncogene KRAS by overcoming the weak nucleophilicity of an acquired serine residue. The chemistry we describe may serve as a basis for the selective targeting of other unactivated serines.

We next asked whether the G12S mutation hampers the rate of GTP hydrolysis, a common biochemical mechanism that confers functional activation and extended pro-growth signaling to K-Ras hotspot mutants 5 . We measured single-turnover GTP hydrolysis by K-Ras(G12S) using a purine nucleoside phosphorylase-coupled assay that monitors free phosphate formation (Fig. 1d). Compared with wild-type K-Ras, K-Ras(G12S) showed a diminished intrinsic GTP hydrolysis rate, and importantly, was insensitive to GTPase-activating protein NF1-mediated acceleration.
A β-lactone ligand acylates the mutant serine. The residual intrinsic hydrolytic activity of GTP suggests that the GDP-bound state may constitute a substantial population of cellular K-Ras(G12S) and may be targetable by small-molecule ligands, especially those that irreversibly engage the protein through covalent bond formation. To design compounds targeting K-Ras(G12S), we drew lessons from the successful drug discovery efforts directed against K-Ras (G12C) 9,11,15,40 , as well as a family of β-lactone-containing natural products (for example, salinosporamide A, omuralide) that inhibit the 20S proteasome by forming a covalent bond with the catalytic threonine (Thr1) 34,35 . We chose to target the S-IIP because it proves to be a privileged drug-binding pocket for K-Ras and offers direct access to the mutant residues at codon 12.
To assess whether G12Si-1 affects nucleotide cycling of K-Ras, we performed nucleotide-exchange experiments by monitoring the is an oncogenic driver with intrinsic GtPase activity. a, Immunoblot of Ba/F3 cells expressing wild-type K-ras (WT), K-ras(G12S) or K-ras(G12C). IL-3 was removed from the culture media 10 min before cells were lysed and analyzed. Data are representative of two experiments using independently generated Ba/F3 transductants. For gel source data, see Supplementary Information. b, Growth of Ba/F3 transductants in the absence of IL-3 (n = 3). Data are representative of two experiments using independently generated Ba/F3 transductants. error bars represent s.d. of three technical replicates. c, X-ray crystal structure of K-ras(G12S)•GDP. Insets depict the 2F o -F c map for the mutant serine and GDP (1.0σ, gray mesh) and the superimposed structures K-ras(G12S)•GDP (this structure, gray) and K-ras(G12C)•GDP (PDB: 4L8G, yellow). For X-ray crystallography data collection and refinement statistics, see Supplementary Table 1. d, Intrinsic and NF1-mediated single-turnover GTPase activity of wild-type K-ras and K-ras(G12S) (n = 3). Data points are plotted as the mean. error bars represent s.d. and are plotted as dashed lines above and below the data points.
exchange of a fluorescent GDP analog (BODIPY-GDP) for unlabeled GDP in the presence of the guanine nucleotide-exchange factor Son of Sevenless (Sos) or the metal chelator EDTA (Fig. 2d,e). As seen with S-IIP ligands of K-Ras(G12C), G12Si-1 blocked Sos-catalyzed exchange and decreased the rate of EDTA promoted exchange.
β-Lactones are highly strained electrophiles with an estimated strain energy of 22.7 kcal mol −1 (ref. 41 ). We hypothesized that adduct formation between K-Ras(G12S)•GDP and G12Si-1 involved a nucleophilic attack from the mutant serine residue to the carbonyl group of the β-lactone in G12Si-1 and subsequent ring opening (Fig. 3a). To test this hypothesis, we solved a 2.0-Å crystal structure of the K-Ras(G12S)•GDP•G12Si-1 complex (Fig. 3b). Comparison of this structure with previously reported co-crystal structures of S-IIP ligands with K-Ras(G12C) shows that G12Si-1 binds in the S-IIP and adopts an orientation similar to the G12C ligands (r.m.s. deviation 0.273 Å; Fig. 3c,d and Extended Data Fig. 3). Well-defined electron density confirmed that Ser12 was acylated by G12Si-1 (Fig. 3b, inset), giving rise to a protein-drug complex mediated by an ester linkage in which the strained β-lactone ester is replaced by an unstrained ester. We also observed that the carbonyl oxygen of this ester group is engaged in a hydrogen bond with Lys16, and that the secondary alcohol resulting from the β-lactone opening formed a hydrogen bond with the backbone carbonyl of glycine 10 ( Fig. 3b, inset). A water molecule was also observed bridging the same secondary alcohol and the backbone N-H of glycine 10. Hydrogen bonding between Lys16 and the carbonyl group on acrylamide electrophiles has been proposed to greatly enhance the reactivity for K-Ras(G12C) inhibitors 42 . These anchoring interactions explain the drastic difference in activity between the closely related compounds G12Si-1 and G12Si-2, because the latter not only has a misaligned electrophile, but also lacks the correct geometry to form either of these hydrogen bonds.
We asked whether these optimized β-lactones allow targeted inhibition of K-Ras(G12S) in genetically characterized cancer cell lines. We treated A549 cells (homozygous KRAS p.G12S mutation) with 10 µM adagrasib, G12Si-3, G12Si-4 or G12Si-5 for 2 h and monitored phospho-AKT and phospho-ERK levels. We also took advantage of a mutant selective antibody that recognizes Ras(G12S) and does not react with covalently modified K-Ras(G12S) protein (see Supplementary Figs. 2 and 3 for its validation) and used it in concert with Raf-RBD pulldown to assess the intracellular level of GTP-bound K-Ras(G12S). Although neither adagrasib nor G12Si-3 had any observable effects on these markers, treatment by G12Si-4 decreased the level of Ras•GTP, and treatment by G12Si-5 led to nearly complete loss of Ras•GTP and concomitant inhibition of phospho-ERK (Fig. 4c). These changes were accompanied by direct adduct formation with cellular K-Ras(G12S), as seen in the upward shift of the K-Ras band in the anti-Ras blot and the disappearance of the signal in the anti-Ras(G12S) blot. Inhibition of Ras signaling by G12Si-5 in A549 cells was dose-dependent, with an apparent half-maximum inhibitory concentration (IC 50 ) of approximately 3 µM under these treatment conditions (Fig. 4d). A terminal alkyne-containing analog of G12Si-5 (Probe 6; Extended Data Fig. 6a,b) enabled us to use click chemistry to enrich the covalently modified proteins in A549 cells and analyze them using LC-MS/ MS. Using competition proteomics, we found K-Ras(G12S) to be a selective target of G12Si-5 with high stoichiometric engagement (Extended Data Fig. 6c,d and Supplementary Dataset 1). We also observed high stoichiometric engagement for three other proteins (PLD3, TRMT61A and TRMT6), which merit future investigation because they could represent targets that are either specific to this chemical scaffold or common for this class of electrophile, similar to the general reactivity of FAM213A and AHR with clinically approved cysteine-targeting acrylamide electrophiles 44 .
We next examined four cell lines with confirmed KRAS p.G12S mutations (see Supplementary Fig. 1 for sequencing data). In each of these cell lines, treatment with G12Si-5 led to a reduction in the phospho-ERK level and gel-mobility shift of the K-Ras protein (Fig. 4e). Meanwhile, G12Si-5 did not perturb the signaling in A375 cells (wild-type KRAS) or SW1990 cells (homozygous KRAS p.G12D), and showed only weak inhibition of phospho-Erk in H358 cells (heterozygous KRAS p.G12C), an effect we ascribed to possible ring opening of the β-lactone by the nucleophilic cysteine.
Although these data confirmed the specificity of G12Si-5 against the KRAS G12S allele, we reasoned that the chemoselectivity encoded by the β-lactone group could afford a therapeutic window between cells expressing the mutant K-Ras(G12S) and those expressing wild-type K-Ras. To test this, we compared the effects of G12Si-5 on the proliferation of Ba/F3:K-Ras(G12S) cells (without IL-3) and parental Ba/F3 cells (with IL-3). G12Si-5 preferentially inhibited the growth of the K-Ras(G12S)-transduced cells with an IC 50 of 2.4 µM and, to a lesser extent, that of the parental cells with an IC 50 of 12.5 µM (Fig. 4f). Notably, the non-G12S-targeting compound adagrasib did not display such selectivity and caused complete cell death at 10 µM for both cell lines. Consistent with this result, treatment of Ba/F3:K-Ras(G12S) cells with compound G12Si-5 led to a dose-dependent reduction in phospho-ERK levels (Extended Data Fig. 7a,b). G12Si-5 also inhibited the growth of G12Si-1 Adagrasib Sotorasib ARS1620 . For X-ray crystallography data collection and refinement statistics, see Supplementary Table 1. KMS20, a patient-derived myeloma cell line with a homozygous KRAS p.G12S mutation, with an IC 50 of 7.5 µM (Extended Data Fig. 8a-c). We did not observe growth inhibition of A549 or LS123 cells below the toxic threshold of 10 µM, two cell lines reported to have low KRAS dependency 45-47 (Extended Data Fig. 8). Although the fivefold selectivity observed with G12Si-5 merits extensive optimization, our data provide the first example of selective targeting of the K-Ras(G12S) mutant using small-molecule agents.

Discussion
The discovery that the mutant cysteine in K-Ras(G12C) can be exploited by small-molecule electrophiles has fueled renewed efforts to develop agents targeting K-Ras, culminating in the recent U.S. Food and Drug Administration approval of sotorasib with five additional drug candidates under clinical investigation. Nevertheless, therapeutic strategies based on covalent targeting of other hotspot mutants of K-Ras have not been reported. Compared with cysteine, the acquired serine residue in K-Ras(G12S) has much weaker nucleophilicity and does not react with electrophiles tailored for K-Ras(G12C). Inspired by a family of threonine-targeting natural products containing the four-membered strained β-lactone, we sought to target the G12S allele of K-Ras. Using structure-guided chemical design, we identified β-lactone as a privileged electrophile for the acquired serine in K-Ras(G12S) and synthesized S-IIP ligands that rapidly undergo covalent engagement with this mutant residue. Similar to the case of K-Ras(G12C) inhibitors,  even though our compounds do not bind to the GTP-bound form of K-Ras(G12S), its intrinsic GTPase activity is sufficient to support the complete irreversible trapping of this mutant allele and allele-specific suppression of oncogenic cellular signaling.
Natural and synthetic β-lactones 48 are known to undergo ring opening following nucleophilic attack from catalytically active serine or threonine residues in enzymes (for example, omuralide/20S proteasome, lipstatin/pancreatic lipase and palmostatin B/acyl protein thioesterase). Yet, to the best of our knowledge, such reactivity has not been observed with noncatalytic serines, including the acquired serines in mutant proteins. Our work demonstrates that chemical acylation of a noncatalytic serine can be achieved using β-lactone electrophiles 49,50 . Although we focused on K-Ras(G12S) in this study, we are optimistic that our strategy adds to the toolbox of serine-targeting electrophiles and may be adopted for other targets with ligand-accessible serine residues.

online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41589-022-01065-9. Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. were plated in six-well plates (3 × 10 5 per ml, 2 ml). The next day, cells were transfected with 2.5 µg of pMSCV plasmid using Lipofectamine 3000 following the manufacturer's instructions. The cells were incubated for 66 h, and the virus-containing supernatants were then collected and passed through a 0.22-µm syringe filter. The harvested virus was used immediately for spinfection of Ba/F3 cells or stored at -80 °C.

Open Access
Generation of stable Ba/F3 transductants. One milliliter of MSCV-containing supernatant (vide supra) was added to one well of a six-well plate containing 1 × 10 6 Ba/F3 cells in 1 ml of media comprised of 60% RMPI 1640, 40% heat-inactivated FBS, 10 ng of mouse IL-3 and 4 µg of polybrene. Cells were spinfected by centrifugation at 2,000g for 90 min at room temperature and then placed in the incubator for 24 h. After 1 d, cells were diluted into 10 ml culture medium (RPMI 1640 + 10% heat-inactivated FBS, 10 ng ml −1 mouse IL-3) and recovered for a second day after spinfection. On the third day after spinfection, cells were pelleted at 500g for 5 min and resuspended in 10 ml of selection medium (RPMI 1640 + 10% heat-inactivated FBS, 10 ng ml −1 mouse IL-3, 1.25 µg ml −1 puromycin). Cells were maintained under puromycin selection for 4-7 d, splitting as required to maintain a density of <2 × 10 6 cells per ml. After 7 d, cells were pelleted, washed once with culture medium free of IL-3 (RPMI 1640 + 10% heat-inactivated FBS) and pelleted again before resuspending at 2-4 × 10 5 cells per ml in culture medium free of IL-3. Cells were maintained under these conditions for 7 d, passaging as needed to maintain a density of <2 × 10 6 cells per ml. Growth was monitored (Countess II Cell Counter) over these 7 d to confirm that an IL-3-independent population has been achieved.

Detection of covalent modification of K-Ras proteins by mass spectrometry.
Test compounds were prepared as 100× stock solutions in DMSO. K-Ras proteins were diluted with SEC buffer (20 mM HEPES 7.5, 150 mM NaCl, 1 mM MgCl 2 ) to 400 nM or 1 µM. In a typical reaction, 0.5 µl of 100× compound stock was mixed with 50 µl of diluted K-Ras protein, and the resulting mixture was incubated for the desired length of time. The extent of modification was assessed by electrospray MS using a Waters Xevo G2-XS system equipped with an Acquity UPLC BEH C4 1.7 µm column. The mobile phase was a linear gradient of 5%-95% acetonitrile/ water + 0.05% formic acid. For kinetic measurements, a 2× compound solution was first prepared in SEC buffer, which was then mixed with 400 nM K-Ras(G12S) protein at 1:1 (v/v) ratio. Injection time stamps were used to calculate elapsed time.
Cell viability assay. Cells were seeded into 96-well white flat-bottom plates (1,000 cells per well) (Corning) and incubated overnight. Cells were treated with the indicated compounds in a nine-point threefold dilution series (100 μl final volume) and incubated for 72 h. Cell viability was assessed using a commercial CellTiter-Glo (CTG) luminescence-based assay (Promega). Briefly, the 96-well plates were equilibrated to room temperature before the addition of diluted CTG reagent (100 μl) (1:4 CTG reagent/PBS). Plates were placed on an orbital shaker for 30 min before recording luminescence using a Spark 20 M (Tecan) plate reader.
In-gel fluorescence. One day before treatment, A549 cells were plated in six-well tissue culture plates (3 × 10 5 cells per well). The next day, cells were pretreated with DMSO or G12Si-5 (10 µM, 0.1% final DMSO) at 37 °C for 1 h. Probes (DMSO, 6 or 7) were then added to the culture media as a 1,000× DMSO solution. The cells were incubated for an additional 1 h at 37 °C. At the end of the treatment period, cells were placed on ice and the media were removed by aspiration. Cells were washed once with ice-cold PBS (1 ml), scraped with a spatula and pelleted by centrifugation (500g, 5 min). Cells were lysed in Co-IP lysis buffer supplemented with protease and phosphatase inhibitors (cOmplete and phosSTOP, Roche) on ice for 10 min. Lysates were clarified by high-speed centrifugation (19,000g, 10 min). Lysate concentrations were determined using a BCA protein assay (Thermo Fisher Scientific) and adjusted to 2 mg ml −1 with additional Co-IP lysis buffer. For each treatment condition, 20 µl of lysate was mixed with 5 µl of freshly prepared 5× TAMRA-N 3 Click Master Mix. The reactions were allowed to proceed in the dark for 1.5 h, then 5 µl of 6× SDS loading buffer was added to the reaction mixture. Fifteen microliters of the resulting mixture was resolved by SDS-PAGE on a Novex 4%-12% Bis-Tris gel (Invitrogen) in MES running buffer (Invitrogen) at 200 V for 60 min. The gel was washed successively with gel-fixation buffer (40% methanol, 10% acetic acid, 2 × 1 h), 50% methanol (1 h) and water (1 h), then scanned on a Typhoon FLA 9000 scanner (532 nm, LPG filter).
Liquid chromatography tandem mass spectrometry analysis of proteins covalently modified by Probe 6. One day before treatment, A549 cells were added to 15-cm tissue culture dishes (4 × 10 6 cells per dish). The next day, cells were treated with 10 µM G12Si-5 (competition) or DMSO (no competition) for 1 h. Probe 6 was then added to the culture media and cells were incubated for an additional 1 h. Each treatment condition was performed in biological duplicate. Cells were then placed on ice, and the media removed by aspiration. Cells were washed once with ice-cold PBS (10 ml), scraped with a spatula and pelleted by centrifugation (500g, 5 min). Cells were lysed in Co-IP lysis buffer supplemented with protease and phosphatase inhibitors (cOmplete and phosSTOP; Roche) on ice for 10 min. Lysates were clarified by high-speed centrifugation (19,000g, 10 min).
Concentrations of lysates were determined using a BCA protein assay (Thermo Fisher Scientific) and adjusted to 2 mg ml −1 with additional Co-IP lysis buffer. Lysate (500 µl) was mixed with 80 µl of high-capacity neutravidin agarose beads (Pierce, 50% v/v suspension, washed with Co-IP lysis buffer), and the mixture was incubated overnight at 4 °C with constant end-to-end rotation to deplete the endogenously biotinylated proteins. The resin was removed by filtration and the lysate was mixed with 125 µl of freshly prepared 5× Biotin-N 3 Click Master Mix. The reaction was allowed to proceed at 23 °C for 1.5 h. Then, 80 µl of high-capacity neutravidin agarose beads (Pierce; 50% v/v suspension, washed with Co-IP lysis buffer) were added, and the mixture was incubated at 4 °C for 1 h with constant end-to-end rotation. The beads were washed successively with: (1) PBS + 1% NP-40 + 0.1% SDS, three times, 23 °C (2) Freshly prepared 6 M urea in PBS, three times, 4 °C (3) Ice-cold PBS, three times, 4 °C (4) Digestion buffer (20 mM HEPES, 2 mM CaCl 2 , pH 8.0), twice, 4 °C.
Each wash was performed by spinning down the resin (1,500g, 5 min), removing the supernatant, resuspending the resin in 0.5 ml of wash buffer, and rotating the resin-buffer mixture for 10 min at the specified temperature. Beads were moved to new tubes after the first, fourth and seventh washes.
After the washes, beads were resuspended in 80 µl of digestion buffer and mixed with 2 µl of 200 mM DTT. The mixture was heated at 56 °C for 30 min. After cooling to 23 °C, 4 µl of 200 mM iodoacetamide was added. After 15 min at 23 °C, 2.1 µl of 200 mM DTT was added. After a further 5 min, 500 ng of trypsin was added to the mixture, and the samples were incubated at 37 °C overnight with constant agitation (1,000 r.p.m., orbital shaker). The tryptic peptides were labeled with TMTsixplex Isobaric Label Reagent Set (Thermo Fisher Scientific, catalog no. 90061) following the manufacturer's instructions. After completion of the labeling reaction and quenching with 0.5% hydroxylamine, the peptide mixtures were pooled in a single tube, and 60 µl of 10% formic acid was added to acidify the mixture. The acidified mixture was dried by vacuum centrifugation (Genevac). The residue, which contains labeled peptides, salts and quenched TMT reagents, was resuspended in 500 µl of 5% acetonitrile/water + 0.1% formic acid. The sample was loaded onto a Waters Oasis HLB column (Waters, catalog no. 186000383; 1 ml, 30 mg sorbent), the column was washed with 0.1% formic acid (3 × 1 ml), and the peptides were eluted with 70% acetonitrile/water + 0.1% formic acid (2 × 200 µl). The eluted peptides were dried under vacuum (Genevac). The dried peptides were fractionated using a Pierce High pH Reverse-Phase Peptide Fractionation Kit (Thermo Fisher Scientific, catalog no. 84868) following the manufacturer's instructions. Peptide fractions were dried individually and resuspended in 100 µl of 0.1% formic acid. Peptides were resolved on an Easy-Spray nano-HPLC column (Thermo Fisher Scientific, catalog no. ES800A; 150 mm length, 3 µl particle size, 100-Å particle size) over a 180-min gradient of 2%-37% acetonitrile/water + 0.1% formic acid and analyzed by a Orbitrap Eclipse tribrid mass spectrometer (precursor range: 400-1,600 m/z, charge state 2-6, MS1 Orbitrap resolution 120,000, max injection time 50 ms, RF lens 30%; dynamic exclusion: 60 s; MS2 ion trap precursor isolation window 0.7 m/z; CID collision energy: 35%; real-time search was enabled for MS3 fragmentation using SwissProt Homo sapiens reference proteome (2021.06.20), constant modification of carbamidomethyl (Cys), TMT6plex (Lys and N-terminal) and variable modification of oxidation (Met), synchronous precursor selection: 10 precursors, precursor isolation window: 3 m/z, HCD collision energy: 55%). Peptides were searched against the SwissProt Homo sapiens reference proteome (2021.06.20), with the modification of the K-Ras protein sequence to reflect the homozygous p.G12S mutation, using MaxQuant (v.2.0.3.1, https://www.maxquant.org/). Quantification was performed with reporter ions in MS3 using unique peptides only. Fold-enrichment was calculated as the ratio of the geometric means of peptide intensities of three biological replicates between the two conditions. P values were calculated using unpaired, two-tailed Student's t-test.

Recombinant protein expression and purification. K-Ras (wild-type), K-Ras (G12S) and K-Ras (G12S)
Cyslight. DNA sequences encoding human K-Ras (wild-type, amino acids 1-169), human K-Ras (G12S, amino acids 1-169), human K-Ras G12S Cyslight (G12S/C51S/C80L/C118S, amino acids 1-169) and human NF1-GRD (amino acids 1,203-1,530) were codon optimized, synthesized by Twist Biosciences and cloned into pJExpress411 vector using the Gibson assembly method 51 . The resulting constructs contain a N-terminal 6× His tag and a TEV cleavage site (ENLYFQG). The full amino acid sequences can be found in Supplementary Fig. 4. The proteins were expressed and purified following previously reported protocols 15 Briefly, chemically competent BL21(DE3) cells were transformed with the corresponding plasmid and grown on LB agar plates containing 50 µg ml −1 kanamycin. A single colony was used to inoculate a culture at 37 °C, 220 r.p.m. in terrific broth containing 50 µg ml −1 kanamycin. When the optical density reached 0.6, the culture temperature was reduced to 20 °C, and protein expression was induced by the addition of IPTG to 1 mM. After 16 h at 20 °C, cells were pelleted by centrifugation (6,500g, 10 min) and lysed in lysis buffer (20 mM Tris 8.0, 500 mM NaCl, 5 mM imidazole) with a high-pressure homogenizer (Microfluidics). The lysate was clarified by high-speed centrifugation (19,000g, 15 min) and the supernatant was used in subsequent purification by immobilized metal-affinity chromatography. His-TEV tagged protein was captured with Co-TALON resin (Clontech, Takara Bio; 2 ml slurry per liter of culture) at 4 °C for 1 h with constant end-to-end mixing. The loaded beads were then washed with lysis buffer (50 ml per liter of culture) and the protein was eluted with elution buffer (20 mM Tris 8.0, 300 mM NaCl, 300 mM imidazole). To this protein solution was added His-tagged TEV protease (0.05 mg TEV per mg of Ras protein) and GDP (1 mg per mg of Ras protein), and the mixture was dialyzed against TEV cleavage buffer (20 mM Tris 8.0, 300 mM NaCl, 1 mM EDTA, 1 mM DTT) at 4 °C using a 10K molecular weight cutoff dialysis cassette until LC-MS analysis showed full cleavage (typically 16-24 h). MgCl 2 was added to a final concentration of 5 mM, and the mixture was incubated with 1 ml of Ni-NTA (Qiagen) beads at 4 °C for 1 h to remove TEV protease, any residual His-tagged proteins and peptides. The protein solution was diluted 1:10 v/v with 20 mM Tris 8.0 and further purified using anion-exchange chromatography (HiTrapQ column, GE Healthcare Life Sciences) with a NaCl gradient of 50 to 500 mM in 20 mM Tris 8.0. Nucleotide loading was performed by mixing the ion exchange-purified protein with an excess of GDP (5 mg per liter of culture) or GppNHp (5 mg per liter of culture) and 5 mM EDTA at 23 °C for 30 min. The reaction was stopped by the addition of MgCl 2 to 10 mM. For GppNHp, an additional calf intestine phosphatase treatment was performed as follows to ensure high homogeneity of the loaded nucleotide. The protein buffer was exchanged for phosphatase buffer (32 mM Tris 8.0, 200 mM ammonium sulfate, 0.1 mM ZnCl 2 ) with a HiTrap desalting column (GE Healthcare Life Sciences). To the buffer-exchanged protein solutions, GppNHp was added to 5 mg ml −1 , and calf intestine phosphatase (NEB) was added to 10 U ml −1 . The reaction mixture was incubated on ice for 1 h, and MgCl 2 was added to a final concentration of 20 mM. After nucleotide loading, the protein was concentrated using a 10K molecular weight cutoff centrifugal concentrator (Amicon-15, Millipore) to 20 mg ml −1 and purified by size-exclusion chromatography on a Superdex 75 10/300 GL column (GE Healthcare Life Sciences). Fractions containing pure biotinylated Ras protein were pooled, concentrated to 20 mg ml −1 and stored at -78 °C. In our hands, this protocol gives a typical yield of 5-15 mg per liter of culture.
NF1-GRD. DNA sequence encoding human NF1-GRD (amino acids 1,203-1,530) was codon optimized, synthesized by Twist Biosciences and cloned into pJExpress411 vector using the Gibson assembly method. The resulting construct contains n N-terminal 6× His tag and a TEV cleavage site (ENLYFQG). Protein was expressed and purified using the identical protocol as for the Ras proteins, except that the ion exchange and nucleotide loading steps were omitted. In our hands, this protocol gives a typical yield of 5-15 mg per liter of culture.
Sos cat . The catalytic domain of Sos (residues 466-1,049, Sos cat ) was expressed and purified following a published protocol. 52 Crystallization. K-Ras(G12S) Cyslight (G12S/C51S/C80L/C118S) bound by GDP purified by size-exclusion chromatography was diluted to 100 µM in reaction buffer (20 mM HEPES 7.5, 150 mM NaCl, 1 mM MgCl 2 ). For co-crystal structures, G12Si-1 or G12Si-5 was added as a 10 mM solution in DMSO to a final concentration of 200 µM. The mixture was allowed to stand at 23 °C until LC-MS analysis of the reaction mixture showed full conversion to a single covalent adduct. The reaction mixture was purified by size-exclusion chromatography (Superdex75, 20 mM HEPES 7.5, 150 mM NaCl, 1 mM MgCl 2 ) and concentrated to 20 mg ml −1 . For crystallization, 0.1 µl of the protein was mixed with 0.1 µl of well buffer. Crystals were grown at 20 °C in a 96-well plate using the hanging-drop vapor diffusion method. Maximal crystal growth was achieved after 7 d. The crystals were transferred to a cryoprotectant solution (same composition as the well solution with 20% glycerol) and flash-frozen in liquid nitrogen. The well buffers for the crystals reported in this study are as follows: X-Ray data collection and structure determination. The dataset was collected at the Advanced Light Source beamline 8.2.2 with X-ray at a wavelength of 0.999907 Å. The dataset was indexed and integrated using iMosflm 53 , scaled with Aimless 54 , and solved by molecular replacement using Phaser 55 in CCP4 software suite 56 . The crystal structure of GDP-bound K-Ras(G12C)-MRTX849 adduct (PDB code: 6USZ) was used as the initial model. The structure was manually refined with Coot 57 and PHENIX 58 . Data collection and refinement statistics are listed in Supplementary Table 1.