Crystal Packing Reveals a Potential Autoinhibited KRAS Dimer Interface and a Strategy for Small-Molecule Inhibition of RAS Signaling

Abstract

KRAS GTPases harbor oncogenic mutations in more than 25% of human tumors. KRAS is considered largely undruggable due to the lack of a suitable small-molecule binding site. Here, we report a unique crystal structure of His-tagged KRAS^G12D that reveals a remarkable conformational change. The Switch I loop of one His-KRAS^G12D structure extends into the Switch I/II pocket of another His-KRAS^G12D in an adjacent unit cell to create an elaborate interface that is reminiscent of high-affinity protein-protein complexes. We explore the contributions of amino acids at this interface using alanine-scanning studies with alchemical free energy perturbation (FEP) calculations based on explicit-solvent molecular dynamics simulations. Several interface amino acids were found to be hot spots as they contributed more than 1.5 kcal/mol to the protein-protein interaction. Computational analysis of the complex revealed the presence of two large binding pockets that possess physicochemical features typically found in pockets considered druggable. Small-molecule binding to these pockets may stabilize this autoinhibited structure of KRAS if it exists in cells to provide a new strategy to inhibit RAS signaling.

Graphical Abstract

INTRODUCTION

RAS is mutated in about a quarter of all human cancers. The gene encodes four identical proteins with more than 80% sequence identity: HRAS, NRAS, KRAS4A and KRAS4B. KRAS4B is the major KRAS isoform. Mutant KRAS is a driver in several lethal and incurable cancers such as pancreatic ductal adenocarcinoma [PDAC] (>90%), colorectal cancer (>40%), and lung adenocarcinomas (>30%) (1–3). The most common KRAS mutations are G12D, G12V, G12C, and G12R. In tumors with KRAS mutations, G12D is found in half of PDAC and colorectal tumors, and G12V is found in about a third of PDAC, colorectal and lung adenocarcinomas tumors (4). Despite significant progress in the development of G12C inhibitors with one candidate that has been FDA-approved (5), there exists no therapeutic agents targeting common KRAS mutations like G12D and G12V.

KRAS is a membrane-bound small GTPase that cycles between GDP-bound inactive state and a GTP-bound active state. Activated KRAS binds to effector proteins such as RAF to initiate a series of protein-protein interactions and phosphorylation events that culminate in transcriptional activity. For over 20 years, evidence has accumulated that KRAS GTPases dimerize or oligomerize at the cell surface. Inouye and co-workers found that in artificial lipid bilayers KRAS formed dimers in protein-fragmentation complementation assays and when dimerization of KRAS was forced in HEK293 cells, these dimers were capable of activating RAF (6). Further lipid bilayer studies combined with computational analysis have determined that RAS protein was capable of forming varying degrees of oligomers with approximately 70% of membrane-bound RAS remaining monomeric and 30% forming in oligomers of up to 6–8 proteins (7,8). The most definitive evidence for physiologically relevant KRAS dimerization was recently demonstrated using photo-activated localization microscopy (PALM) to reach single molecule sensitivity within live cells (9). They confirmed that KRAS localized to the membrane and formed dimers within live cells at physiological levels which propagated downstream activation.

The protein-protein interface of RAS dimerization is still unknown. Güldenhaupt used models based on crystal structures of KRAS and molecular dynamics simulations to investigate possible dimerization models (10). Their work produced one of the leading hypotheses that the activating KRAS dimer has an α4-α5 interaction. The physiological importance of the α4-α5 dimer has been supported by recent work, which have shown that disruption of the interface by inhibitors or single point mutations of residues along the α-helices prevents downstream signaling of KRAS (11–13). Recent modeling and mutagenesis experiments by Mysore and co-workers have shown that the KRAS dimer interface may be significantly larger than previously predicted and likely involves interactions between β4–α3, β5–α4, and β6–α5-α4 of a donor KRAS and α5, β6 of an acceptor KRAS (14). This dimer model is consistent with prior work as the interface overlaps with the known binding location of the KRAS dimer inhibitor, NS1 (11). Their work also supports the existence of long-suspected KRAS nanoclusters, also known as signalosomes, which are large complexes of KRAS, RAF, and other supporting proteins which can propagate a downstream signal.

Here, we report a 1.48-Å resolution crystal structure of His-tagged KRAS^{G12D, C118S} (herein referred to as His-KRAS^G12D). The high-resolution structure reveals a substantial conformational change of the Switch I loop that engages the Switch I/II pocket of a neighboring His-KRAS^G12D symmetry mate. Visual inspection of the crystal packing interface revealed features that are commonly found at protein-protein interaction interfaces. A search for binding pocket identified small-molecule binding sites that could provide a potential new strategy for inhibition of KRAS signaling. Free energy perturbation (FEP) calculations based on explicit-solvent molecular dynamics simulations was used to carry out an alanine scanning analysis of the interface to identify potential hot-spot residues.

RESULTS

Crystal Structure of His-KRAS with Unique Crystal Packing Interface.

Two crystals for GDP-bound His-KRAS^G12D were obtained and solved at 1.5 and 1.4 Å in an orthorhombic P2₁2₁2₁ and a trigonal R32:H space group respectively (Table 1 and Fig. 1A). Both crystals are solved with a single monomer in the asymmetric unit in the P2₁2₁2₁ space group. Inspection of the structures reveals that one of them shows a dramatic conformational change compared with the canonical structure of KRAS found in the other structure we solved. To get deeper insight into this conformational change, the two structures we solved were superimposed revealing a dramatic shift in the position of the Switch I loop of the His-KRAS^G12D structure (Fig. 1A). The loop experiences a large flip from its standard position as illustrated by the position of the amino acids His-27, Phe-28, and Asp-30. In the canonical KRAS structure, these amino acids are located close to the nucleotide. In the His-KRAS^G12D structure with the conformational change at Switch I, Phe-28 is located more than 13 Å away from the nucleotide.

Table 1.

Data-collection and refinement statistics

	His-KRASG12D, C118S	His-KRASG12D, C118S
Data collection
Wavelength (Å)	1.00003	1.00002
Space group	P21 21 21	R 3 2 :H
Cell dimensions
a, b, c (Å)	38.93 39.81 91.67	92.52 92.52 117.77
α, β, γ (°)	90.00 90.00 90.00	90.00 90.00 120.00
Resolution * (Å)	47.65 – 1.48 (1.51 – 1.48)	47.45 – 1.33 (1.35 – 1.33)
Rsym	0.069 (0.910)	0.101 (1.507)
Rmeas	0.075 (1.012)	0.107 (1.606)
Rpim	0.028 (0.431)	0.034 (0.549)
Total reflections	166551 (6027)	427341 (18654)
No. unique reflections	24873 (1155)	43647 (2229)
CC1/2	0.999 (0.668)	0.999 (0.686)
I/σ(I)	18.5 (1.6)	12.4 (1.2)
Completeness (%)	99.7 (96.0)	98.0 (100.0)
Multiplicity	6.7 (5.2)	9.8 (8.4)
Wilson B-factor	18.45	12.96
Refinement
Resolution (Å)	32.74 – 1.48	47.45 – 1.33
No. unique reflections	24790 (2385)	43647 (4427)
Rwork	0.1984 (0.2751)	0.1422 (0.2482)
Rfree	0.2356 (0.3276)	0.1665 (0.2550)
R.m.s.d values
Bond lengths (Å)	0.004	0.008
Bond angles (°)	0.82	1.132
No. atoms
Protein	1414	1515
Ligand	28	52
solvent	196	200
B-factors (Å2)
Protein	23.21	18.94
ligand	14.65	25.88
solvent	33.58	33.33
Ramachandran plot
Favored (%)	97.7	97.6
Allowed (%)	2.3	2.4
Clashscore	4.59	3.29
Rotamer outliers (%)	0.0	0.6
PDB code

^*Highest-resolution shell values are shown in parentheses.

Figure 1.

(A) Structure superposition of His-KRAS^G12D (grey cartoon) in the canonical conformation and His-KRAS^G12D (green cartoon) in the noncanonical form with bound GDP (stick representation). Switch I loop of noncanonical His-KRAS^G12D is peeled away from the protein surface and is bound into a symmetry mate; prominent residues shown in stick representation. Binding site residues on monomer 1 are highlighted in text. (B) Crystallographic assembly of His-KRAS^G12D with monomer 1 (green cartoon) donating its switch I loop (pink) to monomer 2 (surface representation) and receiving the switch I loop (pink) of monomer 3. GDP are shown in stick while the binding site magnesium ion is shown in sphere. Phe-28 of each monomer is highlighted in stick representation. (C) Simplified diagram showing all interacting residues of KRAS•KRAS complex. Positive residues are shown in blue, negative in red, neutral in green, aliphatic in grey, aromatic in pink and proline and glycine in orange. (D) Overview representation showing the number of amino acids in the contact zones and the types of interactions along with a table of interface statistics for monomer 1 and 2.

Interestingly, when His-KRAS^G12D from an adjacent symmetry mate is included, an elaborate protein-protein interaction interface is observed. In this crystal packing interface, the flipped out Switch I of one His-KRAS^G12D protein (Monomer 1) is bound to the Switch I/II pocket of His-KRAS^G12D from the adjacent unit cell (Monomer 2) [Fig. 1B]. A third His-KRAS^G12D (Monomer 3) can also be inserted showing its Switch I protrude into Monomer 1 (Fig. 1B). Even though the shifting of Switch I exposes the bound GDP nucleotide and its metal ligand, they both appear to be stable as the electron density for the bound GDP molecule is strong with full occupancy.

The binding interface between the monomers is extensive, involving nearly all the residues on Switch I and the Switch I/II pocket (Fig. 1B). These interactions include hydrophobic, polar, and non-polar interactions (Fig. 1C and D). For example, a hydrophobic interaction is found between Phe-28 of the Switch I loop that engages Tyr-71, Leu-56, and Leu-6 of the adjacent Monomer 2 Switch I/II pocket amino acids. Several salt-bridge interactions are also present, including His-27 of Monomer 1 with Asp-54 of Monomer 2, Arg-41 of Monomer 1 with Glu-63 of Monomer 2. There is also a salt bridge between Asp-30 of Monomer 1 with a His-tag His-4 of Monomer 2. This interaction is unlikely the driving force behind dimerization since we also solved the structure of His-KRAS^G12D in its canonical conformation (Fig. 1A). Other notable interactions include cation–π interaction (Phe-28 to Lys-5), π-π interaction (Phe-28 to Tyr-71) and several hydrogen bonds such as between Tyr-32 and Met-67.

Alanine Scanning with Free Energy Perturbation Alchemical Calculations.

Native protein-protein interfaces generally contain amino acids that contribute significantly to binding. These amino acids, also known as hot spots, enhance binding affinity typically by 1.4 kcal/mol or more (15–17). Generally, hot spots are identified by alanine scanning either using biophysical or computational methods. Here we resorted to alanine scanning using free energy perturbation (FEP) to determine the free energy change as a result of mutation of each amino acid to alanine at the crystal packing interface between two His-KRAS^G12D symmetry mates. FEP is a rigorous alchemical free energy calculation method based on explicit-solvent molecular dynamics simulations that is capable of determining free energy of binding within 1 kcal/mol of experiment (18,19). The method can determine the free energy change between two states by analysis of the atomistic simulations using an alchemical path. We used FEP+, which is the Schrodinger Inc. implementation of FEP (20).

FEP+ was used to carry out an alanine scan for the amino acids on the Switch I loop at the crystal packing interface (Fig. 2 and Figs. S1–S13). Two His-KRAS^G12D monomers from adjacent unit cells were used for these studies, and it is important to note that while one of the Switch I loop is bound to the Switch I/II pocket of the other His-KRAS^G12D monomer, the Switch I loop of the other monomer is also flipped out, but it is not in contact with another KRAS monomer. We kept the loop in this flipped out conformation during the FEP+ calculations. This is not ideal but considering that the simulations at each lambda value of the FEP calculations were relatively short, and that the loop is located away from the crystal packing protein-protein interface, keeping the second Switch I loop in its flipped conformation will unlikely have an impact on the FEP+ free energy changes at the interface.

Figure 2.

(A) His-KRAS^G12D monomer 1 (green cartoon) switch I loop (pink cartoon) is extended and bound into monomer 2 (yellow transparent surface with yellow cartoon). Binding site residues and GDP are shown in stick representation. (B) Molecular Dynamics Free Energy Perturbation of His-KRAS^G12D monomer 1. Mutational alanine scanning of binding site residues showing differences in FEP. (C) Tabulation of Monomer 1 FEP+ values (D) His-KRAS^G12D Monomer 1 (green cartoon) Switch I loop is extended and bound into Monomer 2 (yellow cartoon). Binding site residues and GDP are shown in stick representation. Binding site residues on Monomer 2 are highlighted in text. (E) Molecular dynamics FEP+ of His-KRAS^G12D Monomer 2. Mutational alanine scanning of binding site residues from Monomer 2 revealed differences in FEP+ (F) Tabulation of Monomer 1 FEP+ values.

Amino acids at the crystal packing interface (Fig. 2A) were mutated to alanine followed by FEP+ calculations. Mutation of four of the six amino acids resulted in an increase in the free energy change by more than 2 kcal/mol (Fig. 2B, C). The most significant contribution to the binding affinity comes from Phe-28 with a ΔΔG_FEP of 3.7 ± 0.4 kcal/mol (Fig. 2B and S1). Mutation of other aromatic residues Tyr-40 and Tyr-32 to Ala also led to substantial ΔΔG_FEP value increases of 2.8 ± 0.3 kcal/mol and 2.1 ± 0.4 kcal/mol, respectively (Fig. S2 and S3). Mutation of Val-29 or His-27 to alanine also led to destabilization of the interface with ΔΔG_FEP values of 1.5 ± 0.4 and 1.9 ± 0.4 kcal/mol, respectively (Fig. S4 and S5). While the contributions of Val-29 are mainly hydrophobic, His-27 engages Arg-54 of the neighboring His-KRAS^G12D monomer likely through pi-cation and hydrogen bonding interactions. Interestingly, mutation of Asp-38 resulted in no change in the free energy, even though it is engaged in a hydrogen bond with the hydroxyl moiety of Tyr-64 in the neighboring KRAS (Fig. S6).

We investigated the role of amino acids that are located in the Switch I/II pocket on Monomer 2 in the crystal packing interface (Fig. 2D). Mutation of the residues on the Switch I/II pocket in Monomer 2 revealed fewer hot spots, but substantially larger free energy changes (Fig. 2E and F). Lys-5 is the most important residue with a binding contribution of over 6 kcal/mol for ΔΔG_FEP upon mutation to alanine (Fig. 2E, F and S7). Another amino acid on Monomer 2 that contributes substantially is Asp-54, which led to a change of 4.0 ± 0.4 kcal/mol upon mutation to alanine (Fig. S8). Interestingly, hydrophobic residues on Monomer 2 do not contribute substantially to binding. For example, Tyr-71, Leu-56, and Tyr-64 do not contribute to the stabilization of the interface (Fig. S9, S10, and S11). In fact, Tyr-71 appears to be a slightly destabilizing amino acid as the change in the free energy is slightly favorable upon mutation. Met-67 and Ile-36 stabilize the interaction by about 1 kcal/mol (Fig. S12 and S13).

Binding Site Detection.

Schrodinger’s SiteMap module was used to identify potential small-molecule binding sites on the surface of His-KRAS^G12D in the flipped out obtained from adjacent unit cells. The program detected two possible binding sites with SiteScore values that are larger than 0.9 (Fig. 3). This suggests that these sites have physico-chemical features found in druggable binding pockets like the ATP pocket of kinases for example. The first SiteMap binding site is located in a pocket created by the large conformational change experienced by the Switch I loop (Fig. 3A). The site is delineated by the nucleotide, magnesium ion, as well as hydrophobic side chains of Tyr-40, Met-67, Tyr-71, Ile-21, and Ile-24.

Figure 3.

(A) Structure of two KRAS proteins from two different unit cells depicting the crystal packing interface along with a small-molecule binding pocket located on Monomer 1. The binding pocket is occupied by cyan spheres that are used by Schrodinger SiteMap to define the pocket. KRAS is shown in ribbon representation. The nucleotide is shown in capped-sticks representation color-coded by atom type, and the magnesium ion is depicted as a sphere in magenta. (B) a small-molecule binding pocket located at the crystal packing interface between Monomer 1 and Monomer 2 visualized in the same manner as previous small molecule site I. (C) Stereoview of His-KRAS^G12D superimposed onto the structure of KRAS bound to a small molecule (PDB code 6GJ8) located at the Switch I/II pocket. His-KRAS^G12D is shown in green ribbon, while KRAS with the bound compound is shown in yellow Connolly surface rendering. (D) Stereoview of KRAS in complex with RAF (PDB ID: 6XHB). The amino acids of the Switch I loop that would be flipped out in the noncanonical structure of His-KRAS^G12D are shown in green. RAF is shown in red ribbon rendering.

The second pocket is located at the crystal packing interface (Fig. 3B). On one side, the pocket is delineated by amino acids of KRAS Monomer 1 and its Switch I loop that has experienced the large conformational change described above. In fact, several amino acids that are on the Switch I loop are located at the periphery of the pocket: His-26, His-27, and Phe-28. Other amino acids on Monomer I that engage the pocket include Arg-41, Tyr-40, Ser-39, and Ile-24. On Monomer 2, there is an extensive array of hydrophobic residues with side chains pointing into the pocket, such as Tyr-32, Ile-36, Tyr-71, Met-67, Ala-59, and Tyr-64. There are also several charged Monomer 2 amino acids defining the pocket such as Asp-33, Gln-37, Arg-68, and Glu-63.

DISCUSSION

There is extensive evidence that KRAS can dimerize and oligomerize at the cell surface. In one study, a small molecule developed by Kessler and co-workers led to KRAS dimerization as later demonstrated by Tran and co-workers (21,22). The symmetric compound interacts with Glu-37, Asp-54, and Glu-3 of β-sheets 1, 2, and 3 on each of the two monomers of KRAS (PDB code 6GJ8). This created a β-sandwich interface that included two compounds and resulted in the formation of two salt bridge interactions between Lys-5 and Glu-3 across the two monomers. This dimer is supported by models that had been previously proposed by Nussinov and co-workers (23). In their study, they proposed four possible RAS dimer interfaces that were evaluated using molecular dynamics simulations of GTP-bound KRAS. Two of the dimers involved the α3 and α4 helices of KRAS with a protein-protein interface that was outside of the RAF effector binding site on KRAS. The other two dimer models proposed in the study overlapped with the RAF effector binding domain on KRAS; these models formed either a stable shifted β-sheet interface or a less stable β-sandwich, both of which are expected to block binding of RAF to KRAS.

In this work, we discovered an unusual structure of His-KRAS^G12D that shows dramatic conformational change of the Switch I loop. It is interesting that when two His-KRAS^G12D proteins from adjacent unit cells are visualized, the Switch I loop that experiences conformational change on one KRAS monomer is bound to the Switch I/II pocket of an adjacent KRAS symmetry mate. Interestingly, Phe-28 of Switch I occupies a pocket on the Switch I/II that has been shown to be the binding site of aromatic rings for fragments and small molecules that have been previously co-crystalized with KRAS by Fesik and others (24) [Fig. 3C].

While many of these crystal packing protein-protein interactions are generally considered artifacts, the dimer interface that we observe in this structure possesses many features that are commonly observed at protein-protein interaction interfaces. For example, amino acids that are common at protein-protein interfaces such as tyrosine (Tyr-32, Tyr-40, and Tyr-71), phenylalanine (Phe-28), lysine (Lys-5), and arginine (Arg-41) are present at the crystal packing interface of His-KRAS^G12D. The solvent-accessible area created by the interface is large. Finally, we were concerned that the His-Tag of KRAS may possibly be responsible for promoting the protein-protein interaction. However, we were able to solve the same His-KRAS^G12D structure that did not exhibit any crystal packing conformational change at Switch I.

We explore the His-KRAS^G12D crystal packing interface with alanine scanning using free energy perturbation (FEP) calculations. These calculations are based on explicit-solvent molecular dynamics simulations and can provide binding affinities within a kcal/mol of experimental values. Scanning amino acids on Monomer 1 with the flipped out Switch I loop showed that five residues contributed by more than 1.5 kcal/mol, and three more than 2 kcal/mol. The most significant contribution was unsurprisingly by Phe-28, which is observed to occupy a well-defined hydrophobic pocket of the Switch I/II pocket of KRAS. Interestingly, alanine scanning of interface amino acids on Monomer 2 led to substantially different results. Only two hot spots were identified, both involved in salt-bridge interactions, such as Lys-5 and Asp-54. None of the other amino acids, including Tyr-71 contributed significantly to the affinity at the interface.

The crystal packing interface between His-KRAS^G12D monomers that we observe in our structure overlaps with the known KRAS•RAF protein-protein interface (Fig. 3D). Furthermore, the structure will not be capable of binding to guanine exchange factors (GEFs) for activation to GTP-bound KRAS. This potential dimer structure will therefore trap KRAS into an inactive autoinhibited form. Therefore, stabilization of this structure, rather than its inhibition, could be an effective approach to develop small molecules that inhibit KRAS signaling. The presence of two well-defined pockets on our His-KRAS^G12D structure suggest a potential strategy for accomplishing this. Both pockets are large and possess physicochemical features that are found in druggable pocket as evidenced by a Schrodinger SiteScore greater than 0.9. A small molecule that binds to the first pocket located at the Switch I loop may trap KRAS into this conformation. Small-molecule binding to the second pocket located at the crystal packing interface may also be used to stabilize the interface and promote the autoinhibited complex. Others have already shown that small molecules that bind into the Switch I/II pocket for example can stabilize the formation of non-active KRAS dimers to prevent downstream signaling (25).

In summary, small molecules that can bind to the binding pocket we identified on our His-KRAS^G12D structure or at the crystal packing interface, may act as a molecular glue that could shift the equilibrium towards this unproductive state and inhibit KRAS^G12D signaling. One strategy to identify such small molecules is to use crystallography screens whereby the KRAS crystals are soaked with fragments and small molecules. Our work here provides a potential new strategy for inhibition of KRAS that involves stabilization of an autoinhibited potential complex rather than inhibition of protein-protein interactions.

METHODS

Protein Expression and Purification.

His-KRAS^G12D in pET-21a plasmid was transformed into E. coli BL-21(DE3) cells. Bacterial culture was grown in terrific broth (TB) until OD₆₀₀ of 0.6 was reached and induced with 0.5 mM Isopropyl β-D-1-thiogalactopyranoside (IPTG) at 37 °C for 3.5 h. Cells were harvested by centrifugation, re-solubilized in lysis buffer (40 mM HEPES pH 7.5, 500 mM NaCl, 5 mM MgCl₂, 10 mM Imidazole) and lysed by multiple passages through a micro-fluidizer. The lysate was clarified by centrifugation and loaded onto a HisTrap HP column (GE Healthcare, Boston, MA) pre-equilibrated in lysis buffer. The bound protein was washed extensively with lysis buffer and gradient-eluted with 500 mM imidazole in lysis buffer. The fractions consisting of His-KRAS^G12D were further purified using size exclusion chromatography on a Hiload Superdex 75 pg column (GE Healthcare, Boston, MA) in crystallization buffer (20 mM Tris pH = 8.0, 2 mM MgCl₂, 2 mM DTT). His-KRAS^G12D was also purified by size exclusion chromatography using a second crystallization buffer (25 mM Tris pH 8.5, 150 mM NaCl, 2 mM MgCl₂) on the same size column.

Crystallography.

His-KRAS^G12D crystals grew at 20°C using the hanging-drop vapor-diffusion method mixing 1:1 v/v of protein and mother liquor containing 0.2 M ammonium chloride and 20% w/v PEG 3350. His-tagged KRAS^G12D in space group R 3 2:H was crystalized using 1 uL of protein at 40 mg/mL and 1 uL of mother liquor containing 0.1 M Tris at pH 8, NaAc 0.2 M and PEG3350 30–34%. The crystals were harvested, cryo-protected in reservoir solution supplemented with 20% glycerol, and flash-cooled in liquid nitrogen. Diffraction data were collected at 100 K at the Beamline station 4.2.2 at the Advanced Light Source (Berkeley National Laboratory, CA) and were indexed, integrated, and scaled using XDS (26). Molecular replacement was performed using PHASER and PDB code 5US4 as search model. Successive cycles of automatic building in Autobuild (PHENIX) and manual building in Coot (27), as well as refinement with PHENIX (28) led to final models. MolProbity software (29) was used to assess the geometric quality of the models, and PyMol (version 2.5.1) was used to generate molecular images. All data collection statistics are shown in Table 1.

Alanine Scanning by Free Energy Perturbation.

The Schrodinger suite of programs (2021–4 release) was used to carry out alanine scanning with free energy perturbation (FEP+). The PDB file corresponding to structures of His-KRAS obtained from two-unit cells were prepared using the Protein Preparation module within Schrodinger Maestro. The protein was first pre-processed to assign correct bond orders, replace hydrogens, and create disulfide bonds. The resulting structure was subjected to diagnosis and analysis in the module, followed by hydrogen bond assignments using ProPKa. Finally, the structure was energy minimized and waters were deleted from the structure. Free energy perturbation calculations were carried out individually using the FEP+ Protein Mutation panel. The protein structure within the Project Table was selected and loaded in the FEP+ Protein Mutation Panel. For each calculation, an amino acid was replaced by alanine, and the calculation was carried out to determine the free energy change for that mutation using the FEP+ Protein Mutation Panel.

Small-Molecule Binding Site Identification.

Schrodinger’s SiteMap (Schrödinger, LLC, New York, NY, 2021) module was used to identify small-molecule binding sites using two KRAS proteins obtained from adjacent unit cells. The option to identify top-ranked potential receptor binding sites was selected. At least 15 points per reported site were required, and up to 5 site-point groupings was chosen. A more restrictive definition of hydrophobicity was used and the option to use fine grids was choses. The site maps were cropped at least 4 Å from the nearest site point. Images were generated using Schrodinger’s Maestro.