Probing the KRas Switch II Groove by Fluorine NMR Spectroscopy

While there has been recent success in the development of KRasG12C inhibitors, unmet needs for selective inhibitors of KRasG12D and the remaining oncogenic KRas proteins remain. Here, we applied trifluoromethyl-containing ligands of KRas proteins as competitive probe ligands to assay the occupancy of the switch II pocket by 19F NMR spectroscopy. Structure–activity-relationship studies of probe ligands increased the sensitivity of the assay and identified structures that differentially detected each nucleotide state of KRasG12D. These differences in selectivity, combined with the high resolution of 19F NMR spectroscopy, enabled this method to be expanded to assay both nucleotide states of the protein simultaneously.


■ INTRODUCTION
The proto-oncogene KRAS is among the most frequently mutated genes in human cancers, with mutations found in approximately 14% of patient samples. 1 Its protein product (KRas) is a membrane-localized small GTPase responsible for cell growth and proliferation signaling through its effectors RAF and PI3K. 2 Transforming mutations in KRAS (most commonly at G12, G13, or Q61) result in an increased cellular proportion of active GTP-bound KRas, typically by impairing the hydrolysis reaction of GTP.
Recent drug discovery campaigns relying on covalent chemistry have resulted in the first FDA-approved KRAS inhibitor sotorasib (AMG510), which selectively targets KRas proteins containing a G12C mutation. 3−12 Sotorasib has weak reversible affinity to a cryptic pocket on the protein, termed the switch II-pocket (SIIP), and relies on an irreversible covalent reaction with the mutant cysteine for its affinity and selectivity. In contrast, adagrasib (MRTX849) bears a weaker electrophile but possesses stronger reversible affinity to the same SIIP binding site (K i = 4 μM). 13 Further structure−activity relationship (SAR) studies of the adagrasib scaffold led to MRTX1133, the first cell-active KRas G12D SIIP inhibitor, in which the noncovalent interactions were optimized to achieve subpicomolar binding affinity. 14 Despite these recent successes, selective inhibitors for the protein products of many RASfamily oncogenes are still unknown, and occupancy probes which can report on ligand binding to the GDP-OFF and GTP-ON states of Ras proteins are limited.
Protein-observed nuclear magnetic resonance (NMR) spectroscopy has proven to be a powerful tool in drug discovery generally and in the study of KRas protein−ligand interactions specifically. 15−17 Furthermore, the two nucleotide states of KRas can be resolved by protein-observed NMR spectroscopy, enabling nucleotide-cycling reactions and nucleotide-state-specific binding to be directly observed in mixed samples containing both GDP and GTP. 18−21 While information-rich, protein-observed experiments are burdened by the requirements of high protein concentrations, long acquisition times, and isotopic labels. Ligand-observed NMR spectroscopy alleviates these burdens; well-validated "probe" ligands can be applied to assay properties of the protein at lower concentrations, with faster acquisition times, and without the need for isotopic labeling.
In this study, we applied trifluoromethyl-containing ligands to the KRas switch II groove as probes to assay KRas proteins in both nucleotide states by 19 F NMR spectroscopy. Onedimensional 19 F Carr−Purcell−Meiboom−Gill (CPMG1D) experiments were used to detect changes in the probes' rates of transverse relaxation resulting from binding to the KRas protein. These probes were applied to assay competitive binding by SIIP-targeted inhibitors such as MRTX849, MRTX1133, and the cyclic peptide KD2.

■ RESULTS AND DISCUSSION
Reversible Binding to the Switch II Groove Is Observed by NMR Spectroscopy. We considered that a suitable probe structure to assay the switch II pocket of KRas proteins by ligand-observed NMR spectroscopy would have the following characteristics: (1) short residence time and weak affinity (K D 10 −3 to 10 −5 M) to enable averaging of bound and unbound populations, (2) affinity for each of the two nucleotide states, and (3) an appropriate NMR handle to enable sensitive detection of the probe molecule under dilute conditions (≤100 μM). A previously reported disulfidetethering screen against a KRas mutant containing an engineered cysteine (M72C) yielded a fragment (2C07, 1; Figure 1A) that we considered likely to serve as a starting point for ligand design to meet these requirements. 22 This fragment occupied a site known as the switch II groove (SIIG), lying along a shallow lipophilic channel formed between α3 and SII. This disulfide fragment (1) and an acrylamide derived from it (2) reacted with KRas M72C in both its inactive state (bound to GDP) and its active state (bound to GNP, a nonhydrolyzable GTP analog). Furthermore, the trifluoromethylpyrazole moiety in these compounds is an ideal structure for ligand-observed 19 F NMR techniques; the three equivalent fluorines are observed as a strong singlet without any significant J-coupling due to the absence of nearby spin-active nuclei. The lack of biological fluorines greatly simplifies the analysis of 19 F NMR spectra of protein−ligand mixtures when compared to 1 H NMR spectra. 23 However, the reversible affinities of these compounds to proteins lacking the M72C mutation were unconfirmed.
Acrylamide 2 has very low aqueous solubility, and initial investigations of its reversible affinity were hindered by the formation of aggregates. To improve the solubility of 2, the solvent-exposed methyl ether was extended into a polar solubilizing tail. The acrylamide was removed because oncogenic KRas proteins do not possess the engineered M72C mutation. These modifications resulted in compounds 3 and 4 ( Figure 1A). Binding of 3 or 4 uniformly to 15 N-labeled KRas WT and KRas G12D GDP 1−169 was observed by HSQC NMR spectroscopy, and perturbed peaks were mapped onto the previously determined structure of KRas GDP ( Figure 1B, Figure S1, PDB 4LPK). The largest of these perturbations are in the SII and α3 regions, which is consistent with expectations based on the structure of the KRas M72C -1 GDP complex (PDB 5VBM). 22 Smaller perturbations were observed in the P-loop and in switch I (SI)�outside the expected binding site�likely due to conformational change in these dynamic regions. However, the reversible affinities of 3 and 4 were very weak; a 1 mM concentration of either ligand did not saturate the binding site ( Figure S1C).
Ligand-Observed CPMG NMR Spectroscopy Detects Oncogenic KRas Mutant Proteins. We then sought to determine whether these SIIG-binders could serve as probe ligands to detect KRas proteins by 19 F NMR spectroscopy. One-dimensional CPMG experiments [90-(τ-180-τ) n ], in which a spin−echo with delay τ is looped n times, attenuate signal intensity as a function of the transverse relaxation rate (R 2 ) and total spin echo time (2·τ·n). Prior studies have shown that reversible binding of proteins to fluorine-containing small molecules induces an increase in the observed 19 F R 2 at long values of τ, which maximize the exchange contribution to R 2 . 24,25 Applying a 160 ms (τ = 20 ms, n = 4) CPMG filter to the 19 F NMR spectrum of 4 (100 μM) in the presence of KRas GDP 1−169 resulted in a decrease in the measured integral (I) compared to the same spectrum acquired in the absence of protein (I 0 ) ( Figure 1C,D). However, these experiments required a relatively high concentration (20 μM) of protein to significantly reduce the integral of 4. Further modifications to the probe structure were required to improve its binding affinity and the sensitivity of the 19 F CPMG NMR assay.
To improve the sensitivity of this method for oncogenic KRas proteins and to explore the relationship between structure and nucleotide-state specificity, we synthesized a series of derivatives of 3 and 4 with varied structure at the aryl ring expected to be buried deepest within the pocket, and these derivatives were evaluated by CPMG NMR spectroscopy against both nucleotide states of KRas G12D (6 μM; Figure  2A,B). From this series, five compounds (9, 12, 14, 15, and 17) showed increased sensitivity compared to 3 and 4. Compound 9, an isomer of 4 in which the hydroxy group is attached at C-4 of the aryl ring, specifically detected the GDP state of KRas G12D . Compounds 12, bearing a methyl group at C-2, and 17, in which the phenyl ring is replaced by a naphthyl, detected both nucleotide states with similar sensitivity. Probe 9 relaxed more slowly in the absence of protein (R 2,free = 2.4 Hz) than did compounds containing the C-5 hydroxy group (R 2,free = 6.3 Hz for 4, 4.4 Hz for 12, and 5.0 Hz for 17), increasing the sensitivity and dynamic range of integral measurements with this probe ( Figure S2). Furthermore, probe 9 was confirmed to be a tighter binder to the SIIG of U− 15 N KRas G12D GDP by HSQC NMR compared to 4 ( Figure S3).
The probe and protein concentrations were varied to determine the dependence on each binding partner. The normalized integrals (I/I 0 ) of the probes decreased exponentially with increasing protein concentrations ( Figure 2C, Figure  S2C,D). The normalized integrals increased with increasing probe concentrations, and K D values for the probe−protein  ACS Chemical Biology pubs.acs.org/acschemicalbiology Letters binding were extracted by fitting the data to eq 1 ( Figure 2D, SI). Probe 9 bound the GDP state more strongly than did probes 12 or 17 (9K D = 0.20 mM); however, we did not observe a clear correlation between these three probes' binding affinities and sensitivities to detect KRas G12D proteins. Probes 9, 12, and 17 were evaluated against both the inactive GDP and active GNP nucleotide states of a panel of KRas proteins (three common mutations at G12 and three common mutations at Q61) to determine whether this method can be generalized across the protein products of the most frequent KRAS oncogenes (Figure 3). Probe 9 most sensitively detected the GDP state of KRas G12D and more weakly detected the remaining oncogene products excepting KRas Q61R GNP ( Figure 3A). Probes 12 and 17 detected both nucleotide states of all oncogene products tested, albeit with weaker sensitivity for KRas Q61L GDP and KRas Q61R GNP than for the others ( Figure 3B,C). The closely related protein HRas 1−166 contains only one residue difference within the SIIG binding site (Q95 in HRas vs H95 in KRas). HRas was only weakly detected by this method; however, the results with HRas Q95H closely matched those obtained with KRas WT . Similarly, a recently identified mutation that confers resistance to adagrasib (Y96C) greatly weakened detection of KRas G12D . 26

SIIP-Targeted Inhibitors Competitively Displace SIIG-Targeted 19 F NMR Probes.
Having identified probe ligand structures that detected low micromolar concentrations of KRas proteins, we next sought to determine whether these probes could assay competitive reversible binding of ligands within or adjacent to the SIIG. Since the SIIG and SIIP are overlapping binding sites, we expected binding to either of them to be mutually exclusive ( Figure S4A). We measured the effect of MRTX849 on the normalized integral of probe 9 (100 μM) in the presence of KRas G12D GDP (2 μM; Figure 4A,B). MRTX849 displaced probe 9, and the calculated fraction occupancy data were fit to eqs 3 and 4 (SI) to extract the K i of MRTX849 (2.9 μM). This result agrees well with the affinity expected based on kinetic data from its reaction with KRas G12C . 13 In contrast, two G12C-targeted inhibitors based on different scaffolds (AMG510 and JDQ443) did not occupy the SIIP of KRas G12D when tested at 60 μM ( Figure S4B,C).
Recognition of the two nucleotide states of KRas proteins is a key property of inhibitors, relating both to an inhibitor's mechanism of action and its ability to access constitutively active proteins. We sought to determine whether this 19 F NMR method could be extended to assay the nucleotide state of KRas G12D and the nucleotide-state specificity of SIIP inhibitors. Since their resonances are resolved, and neither ligand saturates the binding site, the transverse relaxation rates of 9 and 12 could be measured simultaneously in a mixture. We tested whether this combination of probes could discriminate between the inactive GDP state and the active and inactive conformations of the GTP state. The transverse relaxation rate of probe 9 exhibited a strong dependence on the nucleotide state and conformation of KRas G12D ; R 2,9 varied over 6 Hz between samples containing 6 μM of either KRas G12D GDP, KRas G12D GNP, KRas G12D/T35S GNP (state 1 inactive conformation), or KRas G12D GNP + RAF1-RBD 52−131 (state 2 active conformation; Figure S4D). Meanwhile, R 2,12 varied relatively little (<2 Hz) among the same samples.
From these data, equations to calculate the individual concentrations of the GDP and GNP states in a mixture were determined (eqs 5 and 6, SI, Figure S4D,E). Addition of the moderate-affinity binder MRTX849 (60 μM, GDP-state K i = 2.9 μM) selectively occupied the GDP state of the protein ( Figure 4C). The high-affinity binder MRTX1133 (GDP-state K i < 1 pM) occupied both nucleotide states when added in excess (15 μM). However, substoichiometric quantities (2, 4, and 6 μM) of MRTX1133 selectively occupied the GDP-state, consistent with the previously reported high GDP-state selectivity of a closely related structure. 21 In contrast to the small molecule SIIP binders, the cyclic peptide KD2 preferentially occupied the active GNP state. 27

■ CONCLUSIONS
We have synthesized trifluoromethyl-containing ligands that bind to the SIIG of KRas proteins in both nucleotide states, and we have shown that these compounds serve as 19 F NMR spectroscopy probes for a variety of oncogenic KRas proteins. SAR studies identified modifications in the probe structure to improve the sensitivity and nucleotide-state selectivity of the assay. The probes were stoichiometrically competed by SIIPtargeted inhibitors, enabling their use to quantify the occupancy of the SIIP.
While many biochemical methods to assay competitive binding between two ligands are known, few allow simultaneous interrogation of two closely related proteins in a single sample, such as the two nucleotide states of KRas. Knowledge of the nucleotide-state selectivity of SIIP inhibitors is likely important to understand their cellular engagement of constitutively activated KRas proteins, but current methods to directly query this selectivity are lacking. The method and probe ligands reported in this work enable the approximation

ACS Chemical Biology
pubs.acs.org/acschemicalbiology Letters of the occupancy of both nucleotide states from a single sample, allowing an inhibitor's nucleotide-state selectivity to be assayed in a competitive manner. The probe ligands described in this work can be used to detect KRas proteins at low micromolar concentrations with experiment times under 10 min. This sensitivity is sufficient to assay competitive binding, but the concentration of protein places a lower limit on quantifiable K i values of competitors. The throughput and sensitivity of this method could both be further improved with continued efforts toward probe ligand design.
■ ASSOCIATED CONTENT
Supplementary data and figures from HSQC and CPMG NMR experiments, parameters from nonlinear regressions, descriptions of data analysis and equations, experimental methods, chemical synthesis procedures and characterization data, and NMR spectra (PDF)