Expanding Addressable KRAS Mutations through the Structure- and Property-Based Design of Dual-State (GDP/GTP), Reversible Pan-KRAS Inhibitors

Abstract

Therapeutically targeting mutant KRAS represents a clinically validated approach for the treatment of solid tumors, including lung, colon, and pancreatic cancers. The approval of covalent KRAS^G12C inhibitors, such as sotorasib and adagrasib, has fueled intense interest in expanding KRAS-directed therapies to mutations beyond KRAS ^G12C, such as KRAS ^G12D, KRAS ^G12V, and KRAS ^G13D. Here, we describe the structure- and property-based design of reversible inhibitors of diverse oncogenic mutants of KRAS, leading to AM-2383, a pan-KRAS inhibitor that blocks signaling via both the GDP­(off)- and GTP­(on)-bound states of KRAS, while sparing the closely related RAS isoforms HRAS and NRAS. AM-2383 disrupts signaling downstream of KRAS, potently suppressing the growth of KRAS ^G12D and KRAS ^G12V tumor xenografts following oral administration. AM-2383 represents an important proof-of-concept that structural insights from prior covalent KRAS^G12C inhibitors can be leveraged in the design of efficacious and well-tolerated inhibitors of diverse KRAS mutations.

Introduction

Mutations in the RAS oncogene are the most common activating mutations in human cancers, occurring in ∼20% of human tumors. Of the three paralogs, KRAS, HRAS, and NRAS, comprising the RAS gene family, 85% of RAS-driven cancers harbor mutations in the KRAS isoform, with mutations occurring most frequently in solid tumors such as lung, colon, and pancreatic cancers. Among KRAS mutant tumors, 81% of all oncogenic mutations occur at codon 12, with the most common mutations being G12D (29%), G12V (23%), G12C (15%), G13D (7%), KRAS wild-type amplified (Amp.; 5%), and G12A (4%) (Figure ). ,

1.

Clinical KRAS mutational frequency across tumor types (e.g., NSCLC, CRC, PDAC, etc.), illustrating the significant unmet need for patients with non-KRAS ^G12C tumors, for which no targeted therapies have been approved.

KRAS functions as a molecular switch, regulating proliferation by alternating between a guanosine diphosphate (GDP)-bound inactive form and a guanosine triphosphate (GTP)-bound active form, which is capable of binding downstream effector proteins and eliciting a pro-proliferative response. Mutations in codon 12 (or alternatively codon 13 or 61) of KRAS impair the regulated cycling between these two states by disrupting binding with GTPase-activating proteins (GAPs), dramatically slowing GTP-hydrolysis rates and leading to the accumulation of the pro-proliferative GTP-bound state.

Despite being one of the first human oncogenes identified, more than four decades of effort failed to identify clinically useful inhibitors of the KRAS protein. Two features of KRAS contributed to this challenge: (1) KRAS binds to GDP and GTP with picomolar affinity, severely hindering efforts to develop nucleotide-competitive inhibitors, and (2) the KRAS protein appeared to lack other deep surface hydrophobic pockets, thwarting efforts to identify high-affinity allosteric inhibitors.

The identification of chemical matter capable of binding to an allosteric groove adjacent to codon 12 (the “switch II” (Sw-II) or “P2” pocket) in 2013 led to a resurgence in efforts to directly inhibit KRAS. , Significant efforts in our laboratories, leveraging covalent hit-finding approaches and structure- and property-guided optimization, led to the development of a highly potent and selective covalent KRAS^G12C inhibitor, sotorasib (LUMAKRAS), which was subsequently granted accelerated approval (US, May 2021) for the treatment of adult patients with KRAS ^G12C-mutated locally advanced or metastatic nonsmall cell lung cancer (NSCLC) and full approval (US, January 2025) for the treatment of adult patients with KRAS ^G12C-mutated metastatic colorectal cancer (mCRC) in combination with panitumumab (human monoclonal antiepidermal growth factor receptor (EGFR) antibody).

The clinical success of sotorasib (and the subsequently approved KRAS^G12C inhibitor, adagrasib) brought renewed attention to KRAS-driven cancers, leading to efforts toward the design of inhibitors capable of targeting other common oncogenic mutations such as KRAS^G12D, KRAS^G12V and KRAS^G13D. , While KRAS ^G12C mutations predominate in NSCLC, KRAS ^G12D and KRAS ^G12V are more frequent in pancreatic and colon cancers (Figure ). A targeted KRAS therapy capable of addressing one or more of these mutations would therefore significantly expand the patient population that could potentially benefit from such therapies.

A key challenge in developing inhibitors for other KRAS mutants was the lack of obvious covalently targetable residues in these more prevalent KRAS mutants. Covalent KRAS^G12C inhibitors such as sotorasib and adagrasib derive the majority of their potency from irreversible binding to cysteine (vide infra), and at the outset of our efforts, it was unclear how similar KRAS affinity might be engineered into what would likely be noncovalent inhibitors of other activating mutations. It should be noted that several preclinical efforts have subsequently been reported that attempt to target various non-cysteine KRAS mutants with a variety of novel reactive warheads. −

In developing a strategy to noncovalently target other codon 12 mutants of KRAS, we began by generating a model of sotorasib bound to the most prevalent KRAS mutant, KRAS^G12D. Computational replacement of Cys12 with aspartic acid in a structure of sotorasib bound to KRAS^G12C (PDB: 6OIM) indicated that the acrylamide warhead of sotorasib would sterically clash with most low-energy poses of the Asp12 side chain. Tantalizingly, however, one pose positioned the Asp12 carboxylate side chain within hydrogen bonding distance of the N4 piperazine nitrogen of sotorasib (Figure ). This observation encouraged us to consider exploring a hydrogen bonding interaction between a suitably designed amine-containing inhibitor scaffold and Asp12 to achieve both potency and selectivity toward KRAS^G12D. Such a design promised to potentially offset the potency loss incurred by removing the covalent interaction with Cys12 while also maintaining selectivity for KRAS^G12D (and sparing the inhibition of KRAS^WT), something we viewed as potentially essential from a tolerability perspective.

2.

Computational modeling indicated that KRAS^G12C inhibitors (e.g., sotorasib (PDB: 6OIM, depicted)) could serve as starting points for the design of KRAS^G12D inhibitors. Noncysteine codon 12 mutants such as KRAS^G12D (illustrated, orange) position their amino acid side chains in close proximity to the piperazine “warhead” of Sw-II pocket binders (cyan).

Results and Discussion

To explore this strategy, we prepared the des-acrylamide analogue of sotorasib (Figure ) and assessed its activity toward KRAS^G12D in a coupled nucleotide exchange assay, an assay which assessed a compound’s ability to block the activation of GDP-bound (inactive) KRAS via GTP exchange and subsequent effector protein binding. Disappointingly, whereas sotorasib potently inhibited KRAS^G12C (IC₅₀ = 0.014 μM) and was highly selective against KRAS^G12D and wild-type KRAS (KRAS^WT; IC₅₀ values >50 μM), des-acryloyl sotorasib lacked measurable activity toward KRAS^G12D (and other codon 12 variants), demonstrating that simple substitution of a hydrogen bonding interaction for a covalent bond to KRAS was insufficient to impart significant KRAS binding affinity.

3.

Initial efforts to replace a covalent bond to KRAS^G12C (sotorasib) with a hydrogen bonding interaction to KRAS^G12D (des-acryloyl sotorasib) led to a dramatic loss of biochemical activity, as assessed by a coupled nucleotide exchange assay.

Recognizing that sotorasib is a relatively low-affinity noncovalent binder to the Sw-II pocket of KRAS^G12C (K _I = 86 μM), we explored whether other Sw-II binding motifs imparted greater noncovalent binding affinity of KRASand might therefore serve as superior starting points for the design of KRAS^G12D-selective inhibitors.

Two principal binding modes have been reported for covalent KRAS^G12C inhibitors: the “His95-out” binding mode, exemplified by sotorasib, and the “His95-in” binding mode, exemplified by ARS-1620 and adagrasib (Figure ). These binding modes differ principally in the side chain rotamer populated by His95, with the “His95-out” conformer of KRAS opening a lipophilic subpocket (“cryptic pocket”) in the Sw-II binding pocket that is engaged by the isopropylpyridine ring of sotorasib (Figure A), and the “His95-in” conformer orienting His95 inward toward the Sw-II pocket, enabling hydrogen bonding with pyrimidine-based inhibitors (Figure B). Although both binding modes have led to potent and clinically efficacious KRAS^G12C inhibitors, a comparison of the dissociation constants (K _I ) derived from kinetic analysis (k _inact /K _I ) of the binding of sotorasib and ARS-1620 to KRAS^G12C reveals that ARS-1620 possess a K _I value nearly 3-fold lower than that of sotorasib, indicating that “His95-in” binding mode may impart meaningfully tighter noncovalent binding to KRAS^G12C that might aid in the design of high-affinity noncovalent KRAS inhibitors.

4.

Structural differences in the binding of sotorasib and ARS-1620 to the Sw-II pocket of KRAS impart ARS-1620 with greater noncovalent binding affinity. (A) The “His95-out” binding mode of sotorasib results in decreased noncovalent binding affinity to the Sw-II pocket of KRAS G12C (K _I = 86 μM) relative to the (B) “His95-in” binding mode of ARS-1620 (K _I = 31 μM), which benefits from greater binding pocket contact score (1.08 vs 0.97) and an energetically favorable hydrogen-bonding interaction with His95.

Building on this observation, we subsequently employed a “His95-in”-binding quinazoline scaffold in our search for high-potency noncovalent inhibitors of KRAS^G12D (Table ). Although des-acryloyl analogues of ARS-1620 showed only moderate activity in a KRAS^G12D coupled nucleotide exchange assay (c.f., compound 1, IC₅₀ = 5.71 μM), substitution of the fluorophenol ring of ARS-1620 with a 2-aminobenzothiazole ring provided encouraging initial activity, with compound 2 showing a CE IC₅₀ of 3.03 μM toward KRAS^G12D, with 3-fold selectivity toward KRAS^G12C (CE IC₅₀ = 10.7 μM).

1. Key Structural Changes in the Design of Potent, Reversible KRAS^G12D-Selective Inhibitors from Covalent KRAS^G12C Inhibitors .

^aSOS1-catalyzed GDP/GTP coupled exchange (CE) assay (AlphaScreen^TM with KRAS^G12D/C and c-RAF Ras binding domain, 2 h incubation).

^bAlphaScreen^TM with GTP-loaded KRAS^G12D and c-RAF Ras binding domain (RBD); 30 min incubation.

^cBasal p-ERK1/2 immunoassay (MSD) in AsPC-1 cells (KRAS^G12D), 2 h incubation. All data represent n ≥ 2.

^dBelow the sensitivity limit of assay.

Given its moderate molecular weight (415 Da), compound 2 appeared an attractive starting point for further optimization. As 2-substitution of similar quinoline scaffolds had provided enhanced potency for past KRAS^G12C inhibitors, we explored incorporation of an N-methyl prolinol substituent with the aim of engaging Glu62 in a hydrogen bonding interaction (cf. Figure B). Gratifyingly, the resulting compound (3) showed a > 35-fold increase in activity in the G12D coupled exchange assay (CE IC₅₀ = 0.083 μM). To further interrogate the functional activity of compound 3, we subsequently tested its ability to block effector protein (Raf) binding to activated (GTP-loaded) KRAS^G12D. Although inactive up to the highest concentration tested (RBD IC₅₀ > 50 μM), the analogous fluorinated aminobenzothiazole, compound 4, which showed a 5-fold improvement in G12D CE assay activity (IC₅₀ = 0.017 μM), showed the first measurable functional activity in this RBD assay (IC₅₀ = 44.3 μM).

Building on potency gains imparted by the prolinol unit in compounds 3 and 4, we subsequently incorporated the previously reported (2R,7S)-2-fluorotetrahydro-1H-pyrrolizine-7a­(5H)-methanol unit , (“butterfly amine”), into compound 5, further optimizing interactions with the switch-II loop of KRAS^G12D and leading to enhanced activity in both the CE assay (IC₅₀ = 0.018 μM) and RBD assay (IC₅₀ = 12.9 μM). As a result of these potency improvements, compound 5 was the first compound in this series to show single-digit micromolar activity in a cellular assay measuring the suppression of ERK phosphorylation (p-ERK IC₅₀ = 7.24 μM).

Models of compound 5 bound to KRAS suggested further opportunities for enhancing interactions between the piperazine ring and KRAS protein. Replacing the piperazine ring with a bridged piperazine ring resulted in compound 6, which dramatically enhanced potency in both the KRAS CE assay (IC₅₀ = < 0.002 μM; below assay detection limit) and RBD functional assay (IC₅₀ = 0.430 μM), additionally leading to significant improvements in cellular activity (p-ERK IC₅₀ = 0.11 μM). Further optimization of compound 6 to substitute the aminobenzothiazole ring with a cyanoaminobenzothiophene ring led to compound 7, which showed a further 16-fold increase in RBD potency (IC₅₀ = 0.027 μM) and an 8-fold enhancement in cellular activity (IC₅₀ = 0.014 μM). This potent cellular activity translated into strong inhibition of cellular proliferation in a 72 h CellTiter-Glo (CTG) assay (AsPC-1 IC₅₀ = 0.083 μM) and furthermore proved selective for AsPC-1 (KRAS ^G12D) versus SW620 (KRAS ^G12V, IC₅₀ = 0.75 μM) and MIA PaCa-2 (KRAS ^G12C, IC₅₀ = 0.71 μM).

Co-crystallization of compound 7 with GDP-KRAS^G12D revealed binding to the Sw-II pocket, as expected (Figure ). The potency-enhancing cyanoaminobenzothiophene engages in three hydrogen bonds with Asp69 and an adjacent Glu63 residue, and the butterfly amine forms a salt bridge interaction with Glu62 of the switch II loop. The butterfly amine engages in an edge-to-face interaction with His95, which itself is hydrogen-bonded to N1 of the quinazoline ring (“His95-in” binding mode). As anticipated, the bicyclic piperazine ring forms a salt bridge with Asp12 (as well as making an additional hydrogen bond to Gly60), thereby rationalizing the selectivity of compound 7 for KRAS^G12D versus KRAS^G12C or KRAS^G12V due to this energetically favorable hydrogen-bonding interaction with KRAS^G12D.

5.

X-ray cocrystal structure of compound 7 bound to GDP-KRAS^G12D (PDB: 9YMN; 1.25 Å resolution).

Disappointingly, despite promising intravenous (i.v.) PK properties (1 mg/kg dosing: CL = 2.9 L h^–1 kg^–1, t _1/2 = 7.1 h), oral administration of compound 7 in CD1 mice (5 mg/kg) resulted in no measurable bioavailability (F < 1%), likely due to low permeability (MDCK P_app = 0.06 μcms^–1) arising from its dibasic (and physiologically doubly ionized) structure. Subcutaneous (s.c.) dosing (30 mg/kg) achieved sustained plasma exposures in excess of cellular p-ERK IC_50,u values for 8 h, however tumor exposures fell short of cellular IC_50,u values due to high tissue binding (f _u = 0.0009), resulting in no significant tumor PD effects in an AsPC-1 (KRAS ^G12D) mouse xenograft model (Figure ).

6.

Subcutaneous (s.c.) dosing of compound 7 (30 mg/kg, single dose) fails to achieve cellular p-ERK IC_50,u coverage in an AsPC-1 (KRAS ^G12D) mouse xenograft model, thus eliciting no significant in vivo reduction in ERK phosphorylation. s.c. dosing: vehicle, compound 7; oral dosing: MEKi (mirdametinib, PD0325901). Plasma concentrations of compound 7 are indicated by black circles, tumor concentrations are indicated by red squares. Data are presented as percent of control versus vehicle ± SEM (n = 3/group).

Cognizant of the additional challenges that the polar cyanoaminobenzothiophene “tail” of compound 7 posed to membrane permeability and oral bioavailability, we also examined naphthol “tails” as potency enhancing motifs in the design of KRAS^G12D inhibitors (Table ). Compound 8, a close analog of aminobenzothiophene 7, provided encouragement for this strategy, providing excellent activity in the G12D CE assay (IC₅₀ < 0.005 μM) and submicromolar activity in our p-ERK assay (IC₅₀ = 0.290 μM). A 10-fold improvement in cellular activity (compound 9; p-ERK IC₅₀ = 0.031 μM) was observed by changing the regiochemistry of the bridging ring in the bicyclic piperazine. Gratifyingly, such potency was also largely preserved in a corresponding azaquinazoline analogue (10; p-ERK IC₅₀ = 0.110 μM), which both relieved restricted rotation about the biaryl bond to the naphthol tail (which had given rise to atropisomeric analogues in earlier compounds) while also meaningfully reducing lipophilicity (cLogP 4.6 → 3.0). Through fine-tuning of the substitution of the naphthol ring (i.e., installation of a C5-ethyl group and C6-fluoro substituent), biochemical activity was further optimized, leading to compound 11, which showed potent cellular activity (p-ERK IC₅₀ = 0.003 μM) along with enhanced activity in the GTP-state RBD assay (IC₅₀ = 0.160 μM). Subsequent to this work, Mirati Therapeutics and Array BioPharma published their leading efforts in developing the closely related KRAS^G12D inhibitor MRTX1133, efforts which similarly arrived at 6-azaquinazoline-based inhibitors such as 11 as preferred KRAS^G12D inhibitor motifs.

2. SAR of Naphthol-Containing KRAS^G12D-Selective Inhibitors .

^aSOS1-catalyzed GDP/GTP coupled exchange (CE) assay (AlphaScreen^TM with KRAS^G12D/C and c-RAF Ras binding domain, 2 h incubation).

^bAll analogues showed IC₅₀ < 0.005 μM in corresponding KRAS^G12D CE assays.

^cAlphaScreen^TM with GTP-loaded KRAS^G12D and c-RAF Ras binding domain (RBD); 30 min incubation.

^dBasal p-ERK1/2 immunoassay (MSD) in AsPC-1 cells (KRAS^G12D), 2 h incubation. All data represent n ≥ 2.

Despite the improved physicochemical profile of compound 11 vs compound 7 (HBD: 1 fewer, tPSA: −17.2 Ų, cLogD (pH 7.4) = 3.04), compound 11 displayed no significant oral bioavailability in CD1 mice (F < 1%; 5 mg/kg), again implicating the dibasic, bis-amine functionality of this compound as an impediment to gut permeability (MDCK P_app = 0.8 μcms^–1) and oral bioavailability. Intraperitoneal (i.p.) administration (30 mg/kg; i.v. CL = 5.8 L h^–1 kg^–1, t _1/2 = 3.2 h) of 11 in athymic nude mice, however, resulted in sustained plasma p-ERK IC_50,u coverage for 24 h (and IC_90,u coverage for 4 h) and 70% p-ERK inhibition in an AsPC-1 (KRAS ^G12D) tumor xenograft model 2 h postdosing, with sustained p-ERK suppression up to 24 h postdose (Figure ).

7.

Compound 11 (30 mg/kg, i.p., single dose) achieved cellular p-ERK IC_90,u coverage and resulted in sustained ERK1/2 phosphorylation in an AsPC-1 (KRAS ^G12D) mouse xenograft model. i.p. dosing: vehicle, compound 11; oral dosing: MEKi (mirdametinib, PD0325901). Plasma concentrations of compound 11 are indicated by black circles, tumor concentrations are indicated by red squares. Data are presented as percent of control versus vehicle ± SEM (n = 3/group). ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05 by one-way ANOVA followed Dunnett’s posthoc analysis.

Dibasic amine compounds were additionally found to sporadically demonstrate adverse local and systemic effects when dosed s.c. and i.p. (e.g., cutaneous necrosis and acute inflammation, animals becoming cold-to-the-touch, one unscheduled death). Although not mechanistically confirmed, these findings resembled those reported with other dibasic KRAS inhibitors that were attributed to MRGPRX2 agonism and consequent histamine release. ,

Taken together, the low oral bioavailability and sporadic tolerability issues associated with dibasic KRAS^G12D inhibitors encouraged us to explore removal of the Asp12-engaging piperazine unit to afford monobasic inhibitors that might show superior physicochemical and pharmacological properties. Although removal of the piperazine amine would preclude ion pairing interactions with Asp12, we hoped that appropriately placed hydrogen bond donors might allow us to preserve KRAS^G12D activity and selectivity (Figure ). ,

8.

Strategy to address pharmacokinetic limitations (low permeability and oral bioavailability) of compound 11: replacement of a charged Asp12-engaging moiety with a nonbasic hydrogen bond donor moiety.

To explore this strategy, we initially replaced the bicyclic piperazine of compound 11 with a piperidine ring, providing compound 12, which retained excellent potency in the G12D CE assay (IC₅₀ < 0.004 μM), but lost significant potency in the cellular p-ERK assay (IC₅₀ = 0.180 μM; Table ). Hydroxylation of the piperidine ring in an effort to engage Asp12 in a hydrogen bonding interaction (e.g., compounds 13–15) led to the identification of (R)-hydroxypiperidine 14, which moderately enhanced activity in the KRAS^G12D RBD assay and p-ERK cellular assay (IC₅₀ = 1.00 μM & 0.088 μM, respectively). Indicative of specific, directional interactions between the 3-hydroxypiperidyl group and the Sw-II pocket of KRAS^G12D, the epimeric (S)-hydroxypiperidine (15) was ∼13-fold less potent in the p-ERK assay (IC₅₀ = 1.10 μM).

3. SAR of Nonbasic Asp-12 Engaging “Headgroups” .

^aSOS1-catalyzed GDP/GTP coupled exchange (CE) assay (AlphaScreen^TM with KRAS^G12D/C and c-RAF Ras binding domain, 2 h incubation).

^bAll analogues showed IC₅₀ < 0.005 μM in corresponding KRAS^G12D CE assays.

^cAlphaScreen^TM with GTP-loaded KRAS^G12D and c-RAF Ras binding domain (RBD); 30 min incubation.

^dBasal p-ERK1/2 immunoassay (MSD) in AsPC-1 cells (KRAS^G12D), 2 h incubation. All data represent n ≥ 2.

Inspired by structural insights from prior efforts on KRAS^G12C inhibitors, we next explored the incorporation of additional lipophilic substituents on the hydroxypiperidine ring of compound 14 in an effort to enhance affinity for the Sw-II pocket and, thereby, inhibitory activity. Geminal methylation of compound 14 proved highly effective, leading to tertiary alcohol 16, which showed significant activity improvements across all assays, including both cellular activity rivaling that seen with earlier dibasic inhibitors (p-ERK IC₅₀ = 0.005 μM) as well as significant activity in the GTP-state RBD assay (IC₅₀ = 0.056 μM). Fluoromethylation of the geminal methyl group (compound 17) led to further small improvements in p-ERK activity (IC₅₀ = 0.002 μM) with a small loss in GTP-state potency (RBD IC₅₀ = 0.14 μM). Cyclization of the tertiary alcohol into a spirooxetane (18) demonstrated that a hydrogen bond acceptor was also well-tolerated as a piperidine substituent (p-ERK IC₅₀ = 0.003 μM), albeit with a moderate reduction in GTP-state activity (RBD IC₅₀ = 0.610 μM).

Further elaboration of the piperidine ring demonstrated that alternative display of a hydrogen bond acceptor (hydroxycyclobutyl analogue 19), larger spirocyclic substituents (oxazolidinone 20), and ring-expanded piperidines (oxazepanes 21–23) could all deliver excellent cellular activity (p-ERK IC₅₀ values = 0.005–0.064 μM). , In a second case of parallel optimization efforts, it should be noted that subsequent to the completion of this work, related optimization efforts (encompassing some of the same compounds) were also reported by researchers at Mirati Therapeutics.

To further assess the potential of potent KRAS^G12D inhibitors emerging from this effort, we collected mouse pharmacokinetic (PK) data for compounds 16–19, 21, and 23 (Table ). Compound 16 showed moderate clearance (CL = 3.0 L h^–1 kg^–1) and oral bioavailability (F = 12%; 5 mg/kg), with higher doses (100 mg/kg) leading to slight improvements in oral bioavailability (F = 28%), presumably due to saturation of clearance mechanisms (e.g., glucuronidation and efflux). Fluoromethylated analogue 17, in contrast, showed significantly reduced bioavailability (F = 4%). Although oxetanyl analogue 18potentially profiting from the elimination of a hydrogen-bond donorshowed 9% oral bioavailability, this compound proved sensitive to acidic media and showed partial oxetane ring-opening at pH 2, discouraging further investigation of this compound. Hydroxycyclobutyl analogue 19 showed improved clearance (CL = 0.9 L h^–1 kg^–1) and little change in oral bioavailability (F = 8%). Oxazepane 21 showed moderate clearance (CL = 1.9 L h^–1 kg^–1) and oral bioavailability (F = 7%), however the corresponding hydroxyoxazepane (23) suffered from reduced oral bioavailability (F = 3%) and increased clearance (CL = 4.4 L h^–1 kg^–1).

4. Mouse PK Properties of Potent Analogs from the AsPC-1 p-ERK Assay .

Cmpd	CL (L h–1 kg–1)	t 1/2 (h)	F (%)
16 (AM-2383)	3.0	4.3	12
17	2.4	2.2	4
18	1.6	2.7	9
19	0.9	3.8	8
21	1.9	2.0	7
23	4.4	2.1	3

^ai.v./p.o. dosing in CD1 mice (vehicle, i.v.: 1 mg/kg, DMSO; p.o.: 5 mg/kg, 1% Tween 80, 2% HPMC, 97% water, adjusted to pH 2 using methanesulfonic acid).

Although many of these analogues showed promising cellular activity and pharmacokinetic profiles, tertiary alcohol analogue 16 ultimately emerged as the most promising analogue for further follow-up, given its combination of cellular and GTP-state (RBD) activity, mouse half-life, and promising oral bioavailability. Accordingly, we next turned our attention to the detailed pharmacologic and pharmacokinetic characterization of compound 16, which we designated as AM-2383.

A cocrystal structure of AM-2383 with GDP-bound KRAS^G12D (Figure A) showed strong similarities between its binding mode and that of compound 7 (cf., Figure ). AM-2383 again bound in the Sw-II pocket, with the naphthol ring hydrogen bonded to Asp69, the butterfly amine unit hydrogen-bonded to Glu62, and the azaquinazoline N1 nitrogen hydrogen-bonded to His95. The azaquinazoline N6 nitrogen was found to engage Arg68 via a water-mediated hydrogen-bonding interaction. Interestingly, although the tertiary alcohol of AM-2383 was initially envisioned as a means of engaging Asp12 via a hydrogen-bonding interaction, this group was instead found to stabilize a high-energy water molecule (as predicted by Watermap analysis) , bound to Thr58 and Gly10 while simultaneously hydrogen-bonding to Tyr96.

9.

(A) X-ray cocrystal structure of compound AM-2383 bound to GDP-KRAS^G12D (PDB: 9YMO); (B) X-ray cocrystal structure of AM-2383 bound to GMPPcP-KRAS^G12D (PDB: 9YMP).

Co-crystallization of AM-2383 with GMPPcP-bound (a nonhydrolyzable GTP-surrogate) KRAS^G12D revealed only subtle differences in the inhibitor binding mode (Figure B). In particular, to accommodate the γ-phosphate of the nucleotide, AM-2383 was found to bind slightly deeper in the Sw-II pocket, and as a result, the water molecule previously bridging the N6 nitrogen of AM-2383 and Arg68 was displaced, and Arg68 was found to directly hydrogen-bond to AM-2383.

Given this lack of a direct hydrogen bonding interaction between AM-2383 and Asp12 of KRAS^G12D, we were interested to examine the selectivity of AM-2383 toward other KRAS mutants, particularly since early, piperazine-containing inhibitors that achieved good selectivity toward KRAS^G12D (presumably as a result of specific hydrogen-bonding with Asp12; see Figure ) nevertheless showed significant activity toward KRAS^G12C (see Table ). To assess KRAS isoform-selectivity, AM-2383 was profiled in a NanoBRET KRAS target engagement assay, where it was found to potently bind to a diverse set of common oncogenic KRAS mutants (IC₅₀ = 2–10 nM; Table ). AM-2383 was also found to potently bind to KRAS^WT.

5. AM-2383 KRAS Isoform Binding Profile as Assessed by NanoBRET Target Engagement Assay.

KRAS isoform	Nano BRET IC50 (nM)
WT	2
G12D	4
G12V	10
G12C	4
G12A	5
G12S	7
G13D	3
Q61H	2

^aFor assay details see Supporting Information.

Although AM-2383 showed high binding affinity toward diverse KRAS mutants, we were interested (though not surprised , ) to find that AM-2383 showed little activity toward related RAS isoforms (HRAS CE IC₅₀ = 0.320 (>80-fold selectivity); NRAS CE IC₅₀ = 1.60 μM (>400-fold selectivity); Table ). This selectivity could be readily rationalized based on the binding mode of AM-2383. As shown in Figure , the azaquinazoline core of AM-2383 engages in a key hydrogen-bonding interaction with His95. In HRAS and NRAS, this important stabilizing interaction is disrupted by the substitution of glutamine (HRAS) or leucine (NRAS) for histidine at position 95, disrupting Sw-II pocket binding and presumably accounting for the large loss of activity toward these RAS isoforms. This selective affinity of AM-2383 for KRAS WT vs H/NRAS WT was directly demonstrated by surface plasmon resonance (SPR; Table ), where AM-2383 was shown to bind to KRAS with a K_D value of 0.03 nM, HRAS with a K_D value of 139 nM, and NRAS with a K_D value of ∼2.6 μM.

6. AM-2383 Selectively Inhibits Coupled Exchange Activity in KRAS While Sparing HRAS and NRAS.

RAS Isoform	Residue 95	CE IC50 (μM)
KRASWT	His	<0.004
HRASWT	Gln	0.320
NRASWT	Leu	1.60

7. SPR Binding Analysis of AM-2383 across RAS Isoforms, Mutations, and Nucleotide Binding States .

RAS Protein and Bound Nucleotide	SPR KD (nM)
KRASWT | HRASWT | NRASWT (GDP)	0.03 | 139 | ∼ 2600
KRASG12D: GDP | GTP	0.01 | 2.2
KRASG12V: GDP | GTP	0.03 | 6.7

^aData collected at 10 °C.

SPR analysis also allowed for the direct interrogation of the binding affinity of AM-2383 to both the active (GTP-bound) and inactive (GDP-bound) states of KRAS. In the case of KRAS^G12D and KRAS^G12V, AM-2383 showed >200-fold higher affinity for GDP-bound protein than GTP-bound protein but showed roughly equal binding affinity (in either nucleotide bound-state) for KRAS^G12D and KRAS^G12V. Although AM-2383 demonstrated notably higher affinity for GDP-KRAS versus GTP-KRAS, in contrast to prior GDP-state-specific KRAS^G12C inhibitors (e.g., sotorasib and adagrasib), GTP-state affinities for AM-2383 are sufficiently high to allow for GTP-KRAS engagement to be achieved at pharmacologically relevant doses.

To determine how the binding profile and biochemical activity of AM-2383 translated into cellular, functional activity, we examined the effect of AM-2383 on ERK phosphorylation across a panel of cell lines bearing common KRAS mutations and in KRAS ^WT cells. As shown in Figure , AM-2383 potently inhibited ERK phosphorylation across common KRAS mutants (G12D, G12V, & G12C; IC₅₀ = 3–9 nM). Interestingly, although AM-2383 likewise potently inhibited ERK phosphorylation in KRAS ^WT cells (NCI-H3122), such cells showed only moderate maximal inhibition of phosphorylation (percentage-of-control (POC) ∼ 50%). In contrast, KRAS mutant cell lines uniformly showed maximal inhibition of phosphorylation of ≥ 75% POC.

10.

AM-2383 inhibits ERK phosphorylation across KRAS mutant cell lines and in KRAS wild-type cells.

Functional inhibition of ERK phosphorylation further translated into potent antiproliferative effects across KRAS mutant cell lines, as measured in a 72-h Cell Titer Glo assay (Figure ). Across KRAS mutant cell lines (n = 12), AM-2383 showed robust antiproliferative activity (avg. IC₅₀ = 30 nM). In contrast, in KRAS ^WT cell lines (n = 6) or an NRAS mutant cell line (Q61K; n = 1), no impact on proliferation was observed at concentrations of up to 10 μM, consistent with both the incomplete suppression of p-ERK signaling seen in KRAS ^WT cells (cf., Figure ) and with the ability of noninhibited HRAS and NRAS proteins to support cellular proliferation in the presence of KRAS signaling blockade. ,

11.

AM-2383 inhibits cellular proliferation across KRAS mutant cell lines but does not impact proliferation in KRAS ^WT or NRAS mutant cell lines.

Taken together, these in vitro profiling results demonstrated that replacing a direct, Asp12-engaging amine (in compounds 7 and 11) with a non-Asp12 hydrogen-bonding hydroxypiperidine led to a compound, AM-2383, which retained potent KRAS inhibitory activity but with a loss of KRAS^G12D-selectivity. The nonmutant specific (but HRAS & NRAS-sparing) binding profile of AM-2383 ultimately gave rise to a “pan-KRAS” inhibitory profile, leading to a potent, cellularly active inhibitor capable of binding with high affinity to both the GDP- and GTP-bound states of KRAS.

Encouraged by AM-2383’s pharmacological profile and promising mouse PK profile, we investigated its performance in multiple mouse xenograft studies (Figure ). In AsPC-1 (homozygous KRAS ^G12D mutant) tumor-bearing mice, oral administration of AM-2383 (100 mg/kg, QD) provided durable plasma and tumor coverage of p-ERK IC_90,u values and afforded robust inhibition of ERK phosphorylation for >8 h (Figure A). Administration of a second dose (100 mg/kg) of AM-2383 16 h after initial dosing (“BID” bars, in Figure A) demonstrated that robust p-ERK suppression could be maintained over 24-h period. Consistent with these observations, twice-daily (BID) oral dosing of AM-2383 in an AsPC-1 tumor xenograft model (Figure B) afforded 80% tumor growth inhibition (TGI) at a dose of 100 mg/kg, with significant TGI also observed with either QD dosing (100 mg/kg) or BID dosing (30 mg/kg).

12.

Studies were conducted in athymic nude mice bearing the indicated tumor xenografts (n = 3/group for PK/PD studies; n = 10/group for efficacy studies). Plasma concentrations of AM-2383 are shown as black circles, and tumor concentrations are shown as red squares. A MEK inhibitor (mirdametinib, PD0325901; red bar) was used as a positive control in all PK/PD studies. (A, C, E) AsPC-1 (KRAS ^G12D), Panc 04.03 (KRAS ^G12D), and SW620 (KRAS ^G12V)-bearing mice received a single oral dose of vehicle (black bar), AM-2383 at 100 mg/kg (blue bars), or a second oral dose of AM-2383 at 100 mg/kg (orange bars) given 16 h after the first dose. Tumors were harvested at the indicated time points and analyzed for p-ERK levels. Data are presented as percent of control versus vehicle ± SEM (n = 3/group). ****p < 0.0001; *** p < 0.001, ** p < 0.01, *p < 0.05 by one-way ANOVA followed Dunnett’s posthoc analysis. (B) Effect of AM-2383 on AsPC-1 tumor growth following QD or BID oral dosing ****p < 0.0001 by linear mixed-effects model followed by Dunnett’s test. (D) Effect of AM-2383 on Panc 04.03 tumor growth (100 mg/kg BID, oral dosing). ^# p < 0.0001, regression by paired t test. (F) Effect of AM-2383 on SW620 tumor growth (30, 100, or 300 mg/kg BID, oral dosing). ^**** p < 0.0001 by linear mixed-effects model followed by Dunnett’s test. No significant changes in body weight or unscheduled deaths were observed across all studies at doses of up to 300 mg/kg BID (data not shown).

In the KRAS ^G12D heterozygous setting, AM-2383 showed similar potency in Panc 04.03 cells as in AsPC-1 cells (i.e., p-ERK IC₅₀ = 0.0071 μM; CTG IC₅₀ = 0.024 μM). In a mouse xenograft, AM-2383 administration (100 mg/kg, PO, QD) resulted in durable p-ERK IC_90,u coverage and robust p-ERK suppression over for 8 h (Figure C). Twenty-four hour suppression of ERK signaling could be achieved through BID dosing (or QD dosing at 300 mg/kg). Consistent with this PD effect, an efficacy study with AM-2383 in Panc 04.03 tumor-bearing mice showed robust tumor regression (47%) at 100 mg/kg BID (Figure D).

Finally, we also examined the activity of AM-2383 in the homozygous KRAS ^G12V mutant setting, using the colon cancer-derived cell line, SW620. As KRAS ^G12V mutations represent the second most frequent oncogenic KRAS mutation (see Figure ), it was encouraging to see that AM-2383 showed similar potency in this setting as in early KRAS ^G12D cell lines (i.e., SW620 p-ERK IC₅₀ = 0.002 μM; SW620 CTG IC₅₀ = 0.012 μM). In a single-dose PD assay, AM-2383 (100 mg/kg, PO) achieved durable IC_50,u coverage and significant p-ERK inhibition for >8 h (Figure E), with 24 h suppression of ERK signaling again being achieved via BID dosing (or QD dosing at 300 mg/kg). Consistent with this impact on p-ERK signaling, an efficacy study with AM-2383 in SW620 tumor-bearing mice showed encouraging tumor growth inhibition (72% TGI) at 300 mg/kg BID (Figure F).

Chemistry

AM-2383 and related analogs were prepared according to the synthetic route outlined in Scheme . MOM-protected naphthol intermediate A was accessed through a six-step synthetic sequence from commercially available 7-fluoronaphthalene-1,3-diol, while the pyrido­[4,3-d]­pyrimidine intermediate B was accessed through a series of selective S_NAr reactions starting from commercially available 2,4,7-trichloro-8-fluoropyrido­[4,3-d]­pyrimidine. A Suzuki cross-coupling catalyzed by cataCXium A Pd G3 was used to efficiently append the naphthol motif (intermediate A) onto the pyrido­[4,3-d]­pyrimidine scaffold (intermediate B) in a high yield providing AM-2383 following MOM-deprotection.

1. Synthetic Route Used to Prepare AM-2383 .

^a Reagents and conditions: (A) (a) [RuCl₂(p-cymene)]₂, 2-bromoethynyl­(triisopropyl)­silane, KOAc, 1,4-dioxane; (b) MOMCl, DIPEA, DCM; (c) Tf₂O, DIPEA, DCM; (d) Pd­(dppf)­Cl₂, KOAc, B₂Pin₂, PhMe; (e) CsF, DMF; (f) Pd/C, H₂, THF. (B) (g) DIPEA, CH₃CN; (h) DIPEA, 1,4-dioxane; (C) (i) cataCXium A Pd G3, K₃PO₄, THF/water 10:1; (j) 4 M HCl in 1,4-dioxane.

Conclusions

In these studies, we describe the discovery of AM-2383, a potent oral pan-KRAS inhibitor, leveraging structure- and property-based design building on insights from earlier KRAS^G12C inhibitors. Initial structure-guided efforts to design KRAS^G12D-selective inhibitors through the incorporation of basic amine groups capable of selectively forming a salt bridge with Asp12 revealed the significant pharmacokinetic liabilities of dibasic KRAS inhibitors, leading us to adopt a property-based optimization strategy that ultimately led to monobasic inhibitors such as AM-2383. Although the removal of a specific salt bridging interaction with Asp12 resulted in nonmutant specific KRAS inhibitors, the intrinsic KRAS specificity of the azaquinazoline scaffold of AM-2383 led to the development of pan-KRASbut HRAS and NRAS-sparinginhibitors that proved well-tolerated in vitro and in vivo. Such monobasic pan-KRAS inhibitors also demonstrated significantly improved pharmacokinetic properties, ultimately delivering AM-2383, which showed robust antiproliferative effects across a set of KRAS ^G12D and KRAS ^G12V tumor xenograft models. These studies not only continued to illustrate the potential of pan-KRAS inhibitors in addressing diverse, common KRAS mutations not addressable by current KRAS^G12C inhibitors, − but also provided a useful starting point for the design of a clinical pan-KRAS inhibitor. Although AM-2383 provided useful insights into the pharmacology of pan-KRAS inhibitors, work on this molecule is not being pursued further in favor of other pan-KRAS inhibitor efforts, which will be reported in due course.

Experimental Section

General Synthetic Procedures

All materials were obtained from commercial suppliers and used without further purification unless otherwise noted. Anhydrous solvents were obtained from Sigma-Aldrich. Reactions involving air- or moisture-sensitive reagents were performed under a nitrogen or argon atmosphere. Silica gel chromatography was performed using prepacked silica gel cartridges (RediSep Rf, Teledyne ISCO). Reversed-phase HPLC purification was performed using Gilson (Middleton, WI) workstations. NMR spectra were acquired on Bruker Avance 400, 500, or 600 MHz spectrometers equipped with 5 mm BBFO probes. All compounds tested in in vitro and in vivo assays were >95% pure by HPLC analysis as determined by an Agilent 1100 or 1260 multiwavelength detector (215 nm detection) and an Advanced Materials Technology HALO C18 column (50 × 3.0 mm, 2.7 μm) at 40 °C with a 2.0 mL/min flow rate using a 5% to 95% gradient of acetonitrile/water with 0.1% trifluoroacetic acid over 1.5 min. Low-resolution MS data were obtained concurrently with UV chromatography using an Agilent G1956B LC MSD SL (6120B, 6130B, or 6140A) quadrupole MS in positive electrospray ionization mode. Resolved atropisomers were purified to >95% ee (or de) using a Thar 80, 200, or 350 preparative SFC instrument. Enantiomeric and diastereomeric excesses were determined by SFC (Waters Acquity UPC2 or Agilent 1260 Infinity analytical systems). The following abbreviations are used to designate NMR peak multiplicities: s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, quint = quintet, br = broad.

Synthesis of AM-2383. (A) Step a: 7-Fluoro-8-((triisopropylsilyl)­ethynyl)­naphthalene-1,3-diol. To a mixture of 7-fluoronaphthalene-1,3-diol (3.30 g, 18.52 mmol, PharmaBlock) and 2-bromoethynyl­(triisopropyl)­silane (5.81 g, 22.23 mmol, Enamine) in 1,4-dioxane (60 mL) was added KOAc (3.64 g, 37.05 mmol) and dichloro­(p-cymene) ruthenium­(II) dimer (1.13 g, 1.85 mmol, 0.1 eq, Alfa Aesar) in one portion at room temperature under nitrogen. The mixture was stirred at 110 °C for 2 h then cooled and poured into ice–water (60 mL). The mixture was extracted with ethyl acetate (3 × 50 mL), and the combined organic extracts were dried over anhydrous Na₂SO₄, filtered and concentrated. The crude residue was purified by column chromatography on silica gel, eluting with petroleum ether/ethyl acetate 8/1 to 1/1 to provide 7-fluoro-8-((triisopropylsilyl)­ethynyl)­naphthalene-1,3-diol (4.30 g, 11.27 mmol, 65%) as yellow solid. ESI [M + H]⁺: 359.2.

Step b: 7-Fluoro-3-(methoxymethoxy)-8-((triisopropylsilyl)­ethynyl)­naphthalen-1-ol. To a mixture of 7-fluoro-8-((triisopropylsilyl)­ethynyl)­naphthalene-1,3-diol (4.30 g, 11.99 mmol) in DCM (60 mL) was added Hünig’s base (4.65 g, 35.98 mmol, 6.27 mL). MOM-Cl (1.16 g, 14.39 mmol, 1.09 mL) was then added in portions at 0 °C under nitrogen. The mixture was stirred at room temperature for 12 h then poured into ice–water (60 mL) and stirred for 20 min. The mixture was extracted with ethyl acetate (3 × 60 mL), and the combined organic phase was washed with brine (2 × 60 mL), dried over anhydrous Na₂SO₄, filtered and concentrated. The residue was purified by column chromatography on silica gel, eluting with petroleum ether/ethyl acetate 4/1 to 1/1 to give 7-fluoro-3-(methoxymethoxy)-8-(2-triisopropylsilylethynyl)­naphthalen-1-ol (2.00 g, 4.97 mmol, 41%) as brown solid.

Step c: 7-Fluoro-3-(methoxymethoxy)-8-((triisopropylsilyl)­ethynyl)­naphthalen-1-yl trifluoromethanesulfonate. To a mixture of 7-fluoro-3-(methoxymethoxy)-8-(2-triisopropylsilylethynyl)­naphthalen-1-ol (2.50 g, 6.21 mmol) in DCM (30 mL) was added Hünig’s base (2.41 g, 18.63 mmol, 3.25 mL) in one portion at room temperature under nitrogen. The mixture was treated with Tf₂O (2.63 g, 9.32 mmol, 1.54 mL) in portions at −40 °C under nitrogen. The mixture was stirred at −40 °C for 1 h then the mixture was poured into ice–water (30 mL) and the aqueous phase was extracted with dichloromethane (3 × 30 mL). The combined organic extracts were dried over Na₂SO₄, filtered and concentrated in vacuo. The residue was purified by column chromatography on silica gel, eluting with petroleum ether/ethyl acetate 40/1 to 8/1 to provide 7-fluoro-3-(methoxymethoxy)-8-((triisopropylsilyl)­ethynyl)­naphthalen-1-yl trifluoromethanesulfonate (2.50 g, 4.68 mmol, 75%) as yellow oil.

Step d: ((2-Fluoro-6-(methoxymethoxy)-8-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)­naphthalen-1-yl)­ethynyl)­triisopropylsilane. To a mixture of 7-fluoro-3-(methoxymethoxy)-8-((triisopropylsilyl)­ethynyl)­naphthalen-1-yl trifluoromethanesulfonate (2.50 g, 4.68 mmol) in toluene (25 mL) was added KOAc (1.38 g, 14.03 mmol), bis­(pinacolato)­diboron (2.37 g, 9.35 mmol) and Pd­(dppf)­Cl₂ (0.34 g, 0.47 mmol) in one portion at room temperature under nitrogen. The mixture was stirred at 130 °C for 3 h then cooled and concentrated under reduced pressure then the residue was poured into ice–water (40 mL). The mixture was extracted with ethyl acetate (3 × 40 mL), and the combined organic phase was dried over Na₂SO₄, filtered and concentrated in vacuo. The residue was purified by column chromatography on silica gel, eluting with petroleum ether/ethyl acetate 40/1 to 10/1 to provide ((2-fluoro-6-(methoxymethoxy)-8-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)­naphthalen-1-yl)­ethynyl)­triisopropylsilane (1.20 g, 2.34 mmol, 50%) as yellow solid. ESI [M + H]⁺: 513.2.

Step e: 2-(8-Ethynyl-7-fluoro-3-(methoxymethoxy)­naphthalen-1-yl)-4,4,5,5-tetramethyl-1,3,2-dioxaborolane. To a mixture of ((2-fluoro-6-(methoxymethoxy)-8-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)­naphthalen-1-yl)­ethynyl)­triisopropylsilane (1.10 g, 2.15 mmol) in DMF (20 mL) was added CsF (1.96 g, 12.88 mmol) in one portion at room temperature under nitrogen. The mixture was stirred at 50 °C for 2 h then the mixture was poured into water (20 mL). The mixture was extracted with ethyl acetate (3 × 20 mL) and the combined organic phase was washed with brine (2 × 30 mL), dried over anhydrous Na₂SO₄, filtered and concentrated in vacuo to provide the crude 2-(8-ethynyl-7-fluoro-3-(methoxymethoxy)­naphthalen-1-yl)-4,4,5,5-tetramethyl-1,3,2-dioxaborolane (1.20 g, crude) as brown oil.

Step f: 2-(8-Ethyl-7-fluoro-3-(methoxymethoxy)­naphthalen-1-yl)-4,4,5,5-tetramethyl-1,3,2-dioxaborolane (Intermediate A). To a solution of 2-(8-ethynyl-7-fluoro-3-(methoxymethoxy)­naphthalen-1-yl)-4,4,5,5-tetramethyl-1,3,2-dioxaborolane (1.20 g, 3.37 mmol) in THF (10 mL) was added Pd/C (10 wt %, 20 mg) under argon. The suspension was degassed under vacuum and purged with hydrogen several times and stirred under hydrogen (15 psi) at room temperature for 1 h. The reaction mixture was filtered, and the filter cake was washed with THF (3 × 10 mL). The combined filtrate was concentrated under reduced pressure and the crude product was purified by column chromatography on silica gel, eluting with petroleum ether/ethyl acetate 25/1 to 10/1 to provide 2-(8-ethyl-7-fluoro-3-(methoxymethoxy)­naphthalen-1-yl)-4,4,5,5-tetramethyl-1,3,2-dioxaborolane (Intermediate A, 0.62 g, 1.69 mmol, 75% over two steps) as yellow solid. ESI [M + H]⁺: 361.1.

(B) Step g: ( R )-1-(2,7-Dichloro-8-fluoropyrido­[4,3- d ]­pyrimidin-4-yl)-3-methylpiperidin-3-ol. To a mixture of 2,4,7-trichloro-8-fluoropyrido­[4,3-d]­pyrimidine (25 g, 99 mmol, Sigma-Aldrich) in acetonitrile (500 mL) was added Hünig’s base (86 mL, 495 mmol) and (R)-3-methylpiperidin-3-ol hydrochloride (15 g, 99 mmol, Combi-Blocks). The mixture was stirred at 0 °C for 0.5 h under nitrogen then the reaction mixture was diluted with H₂O (1 L) and extracted with EtOAc (3 × 0.9 L). The combined organic layers were dried over Na₂SO₄, filtered and concentrated under reduced pressure. The crude residue was purified by column chromatography on silica gel, eluting with a gradient of 2–100% ethyl acetate/petroleum ether to provide (R)-1-(2,7-dichloro-8-fluoropyrido­[4,3-d]­pyrimidin-4-yl)-3-methylpiperidin-3-ol (28 g, 85 mmol, 85%) as yellow solid.

Step h: ( R )-1-(7-Chloro-8-fluoro-2-(((2 R,7a S )-2-fluorohexahydro-1 H -pyrrolizin-7a-yl)­methoxy)­pyrido­[4,3- d ]­pyrimidin-4-yl)-3-methylpiperidin-3-ol (Intermediate B). To a solution of (R)-1-(2,7-dichloro-8-fluoropyrido­[4,3-d]­pyrimidin-4-yl)-3-methylpiperidin-3-ol (20 g, 60 mmol, PharmaBlock) in 1,4-dioxane (300 mL) was added Hünig’s base (20 g, 151 mmol) and ((2R,7aS)-2-fluorotetrahydro-1H-pyrrolizin-7a­(5H)-yl)­methanol (14 g, 85 mmol, PharmaBlock). The resulting mixture was stirred at 100 °C for 12 h under nitrogen then the reaction mixture was diluted with H₂O (0.7 L) and extracted with EtOAc (3 × 0.5 L). The combined organic layers were dried over Na₂SO₄, filtered and concentrated and the crude residue was purified by column chromatography on silica gel, eluting with a gradient of 10–100% ethyl acetate/petroleum ether to provide (R)-1-(7-chloro-8-fluoro-2-(((2R,7aS)-2-fluorohexahydro-1H-pyrrolizin-7a-yl)­methoxy)­pyrido­[4,3-d]­pyrimidin-4-yl)-3-methylpiperidin-3-ol (Intermediate B). Batch 1:10.3 g with 98.5% purity by LCMS; Batch 2:3.3 g crude with 70% purity. ESI [M + H]⁺: 454.2, 456.2.

(C) Step i: ( R )-1-(7-(8-Ethyl-7-fluoro-3-(methoxymethoxy)­naphthalen-1-yl)-8-fluoro-2-(((2 R,7a S )-2-fluorotetrahydro-1 H -pyrrolizin-7a­(5 H )-yl)­methoxy)­pyrido­[4,3- d ]­pyrimidin-4-yl)-3-methylpiperidin-3-ol. To a solution of 2-(8-ethyl-7-fluoro-3-(methoxymethoxy)­naphthalen-1-yl)-4,4,5,5-tetramethyl-1,3,2-dioxaborolane (Intermediate A, 4.29 g, 11.90 mmol) and (R)-1-(7-chloro-8-fluoro-2-(((2R,7aS)-2-fluorotetrahydro-1H-pyrrolizin-7a­(5H)-yl)­methoxy)­pyrido­[4,3-d]­pyrimidin-4-yl)-3-methylpiperidin-3-ol (Intermediate B, 3.6 g, 7.93 mmol) in tetrahydrofuran (72 mL) and water (7.2 mL) were added potassium phosphate (5.05 g, 23.79 mmol) and cataCXium A Pd G3 (0.58 g, 0.79 mmol, Sigma-Aldrich). The reaction mixture was sparged with argon, sealed, and stirred at 70 °C for 16 h then the reaction mixture was concentrated and partitioned between water and ethyl acetate; the organic layer was concentrated. The crude product was purified by column chromatography on silica gel, eluting with 0–100% (3:1 ethyl acetate/ethanol +2% TEA) in heptane, to provide (R)-1-(7-(8-ethyl-7-fluoro-3-(methoxymethoxy)­naphthalen-1-yl)-8-fluoro-2-(((2R,7aS)-2-fluorotetrahydro-1H-pyrrolizin-7a­(5H)-yl)­methoxy)­pyrido­[4,3-d]­pyrimidin-4-yl)-3-methylpiperidin-3-ol (5.15 g, 7.90 mmol, quantitative yield) as yellow powder. ESI [M + H]⁺: 652.3.

Step j: ( R )-1-(7-(8-Ethyl-7-fluoro-3-hydroxynaphthalen-1-yl)-8-fluoro-2-(((2 R,7a S )-2-fluorotetrahydro-1 H -pyrrolizin-7a­(5 H )-yl)­methoxy)­pyrido­[4,3- d ]­pyrimidin-4-yl)-3-methylpiperidin-3-ol. To a solution of (R)-1-(7-(8-ethyl-7-fluoro-3-(methoxymethoxy)­naphthalen-1-yl)-8-fluoro-2-(((2R,7aS)-2-fluorotetrahydro-1H-pyrrolizin-7a­(5H)-yl)­methoxy)­pyrido­[4,3-d]­pyrimidin-4-yl)-3-methylpiperidin-3-ol (5.00 g, 7.67 mmol) in acetonitrile (38.4 mL) was added hydrogen chloride (4.0 M in 1,4-dioxane, 19.2 mL, 77 mmol). The reaction was stirred at room temperature for 30 min then MeOH (∼5 mL) was added, and the reaction mixture was concentrated. The crude product was purified by column chromatography on silica gel, eluting with 0–100% (3:1 ethyl acetate/ethanol +2% TEA) in heptane, to provide (R)-1-(7-(8-ethyl-7-fluoro-3-hydroxynaphthalen-1-yl)-8-fluoro-2-(((2R,7aS)-2-fluorotetrahydro-1H-pyrrolizin-7a­(5H)-yl)­methoxy)­pyrido­[4,3-d]­pyrimidin-4-yl)-3-methylpiperidin-3-ol (AM-2383, 3.36 g, 5.53 mmol, 72%) as yellow powder. ¹H NMR (400 MHz, MeOH-d ₄) δ ppm 9.22 (s, 1 H), 7.69 (dd, J = 8.9, 5.7 Hz, 1 H), 7.23–7.33 (m, 2 H), 7.08 (s, 1 H), 5.18–5.44 (m, 1 H), 4.52 (br d, J = 13.0 Hz, 1 H), 4.25–4.36 (m, 3 H), 3.59–3.70 (m, 1 H), 3.48–3.54 (m, 1 H), 3.15–3.27 (m, 3 H), 2.99–3.06 (m, 1 H), 2.49 (br dd, J = 14.0, 6.9 Hz, 1 H), 2.10–2.30 (m, 5 H), 1.76–2.05 (m, 8 H), 1.30 (d, J = 9.4 Hz, 3 H), 0.83 (q, J = 7.7 Hz, 3 H). ¹⁹F NMR (376 MHz, MeOH-d ₄) δ ppm 121.2 (s, 1 F), 139.1 (s, 1 F), 173.7 (s, 1 F). ESI [M + H]⁺: 608.2.

Glossary

ABBREVIATIONS

ADME

absorption, distribution, metabolism, and excretion

AUC

area-under-the-curve

BID

bis in die (twice daily)

c-RAF

rapidly accelerated fibrosarcoma kinase (cellular homologue)

CE

coupled nucleotide exchange

CTG

CellTiter-Glo

DCM

dichloromethane

ERK

extracellular signal-regulated kinase

FaSSGF

fasted-state simulated gastric fluid

FaSSIF

fasted-state simulated intestinal fluid

GDP

guanosine diphosphate

GTP

guanosine triphosphate

HBD

hydrogen bond donor

HRAS

Harvey rat sarcoma viral oncogene homologue

KRAS

Kirsten rat sarcoma viral oncogene homologue

MDCK

Madin-Darby Canine Kidney cells

NRAS

neuroblastoma RAS viral oncogene homologue

PBS

phosphate-buffered saline

p.o

per os (oral)

MDCK

Madin-Darby Canine Kidney cell

NSCLC

nonsmall cell lung cancer

PD

pharmacodynamic

PDB

protein data bank

PK

pharmacokinetic

p-ERK

phosphorylated extracellular signal-regulated kinase

QD

quaque die (once daily)

rac

racemic

SAR

structure–activity relationship

Sw-II

switch II pocket

SPR

surface plasmon resonance

MS

mass spectrometry

TGI

tumor growth inhibition

SOS1

son of sevenless homologue 1 protein

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jmedchem.6c01325.

• Experimental details for in vitro and in vivo assays, crystallographic data for compound 7 (PDB: 9YMN) and AM-2383 (PDB: 9YMO, 9YMP, 9YMQ, 9YMR), synthetic procedures for compounds 1–23, and associated analytical data (¹H and ¹⁹F NMR, spectra and LCMS data) (PDF)

• Molecular formula strings with associated assay data (CSV)

Manuscript prepared by R.P.W. and B.A.L. with contributions from all authors.

The authors declare the following competing financial interest(s): All authors are current or former employees of Amgen, Inc. and may hold stock in the same.

The manuscript is dedicated to Kexue Li who passionately supported this program for many years. You will live on in the memories and hearts of the scientists on the KRAS team.