Discovery of Next-Generation Tropomyosin Receptor Kinase (TRK) Inhibitors for Combatting Resistance Associated with Protein Mutation

Aberrant signaling from tropomyosin receptor kinases (TRKs) has been identied as the oncogenic driver in a large variety of cancers, suggesting that inhibition of TRK may be an attractive strategy of attack for anticancer therapeutics. Despite the encouraging therapeutic response to the rst-generation TRK inhibitor Larotrectinib (1), the emergence of drug-resistant secondary mutations within its ATP-binding pocket cause relapse in cancer patients. A next-generation inhibitor that overcomes multiple Larotrectinib-resistant mutations is still a major “unmet clinical need”. In this study, the progressive exploration of the structure-activity relationship (SAR) through scaffold hopping, exibilization, and uoroalkoxy substitutions produced modications on the reported compound 1 that led to the discovery of a superior derivative 7g. Compound 7g is a novel, orally available TRK inhibitor that showed excellent in vitro potency on a panel of Larotrectinib-resistant mutants both in biochemical and cellular assays. Compound 7g also exhibited improved in vivo antitumor ecacy in TRK wild type and mutant fusion-driven tumor xenograft models compared to the ecacy of Larotrectinib and Selitrectinib (3). These encouraging results suggest that compound 7g is a promising drug candidate to pursue for development of a TRK inhibitor to overcome clinically acquired resistance to Larotrectinib. reaction was initiated by addition of 80 µL of the NADPH regenerating system to 320 µL of each incubation mixture. The nal incubation conditions achieved in 400 µL are: 0.5 mg/mL human liver microsomes, 1 µM test compounds, 1.3 mM NADPH, 3.3 mM glucose 6 phosphate, 0.6 U/mL glucose 6 phosphate dehydrogenase. The mixtures were incubated in a 37°C water bath with gentle shaking. A 100 µL aliquot of each mixture was removed at 10, 30, 90 minutes to a clean 96-well plate which contains 400 µL quench reagent to precipitate proteins, and centrifuged (5000 ×g, 15 min). 80 µL of supernatant are taken into 96-well assay plates pre-added with 160 µL ultrapure water, and then analyzed by LC-MS/MS.


Introduction
Tropomyosin receptor kinases (TRKs), also known as neurotrophic factor receptor kinase, are a family of three distinct isoforms, TRKA, TRKB, and TRKC, which are encoded by NTRK1, NTRK2, and NTRK3 genes, respectively 1 . Upon binding to their cognate neurotrophin ligands (NGF with TRKA, BDNF and NT-4/5 with TRKB, and NT-3 with TRKC), the TRK proteins dimerize and become autophosphorylated. This leads to the activation of downstream signaling cascades consisting of PLCγ-PKC, PI3K-AKT, and RAS-MAPK pathways, that support the crucial processes of cell survival, growth, proliferation, and differentiation in the nervous system 2 .
Since the rst identi cation of NTRK1 as a proto-oncogene in colon cancer in 1982 3 , aberrant TRK signaling due to somatic gene mutations, splice variants, overexpression, and gene fusions has been shown to be oncogenic driver in a large variety of cancers 4 . Among these, NTRK fusions are the most common tumorigenic forms 5 . The chimeric protein tropomyosin 3 (TPM3)-TRKA that results from the fusion of the N-terminal of the normally expressed protein TPM3 to the TRKA kinase domain was rst found in human colorectal cancer 3,6 . This fusion is similar to the EML4-ALK gene fusion seen in lung cancer 7 . Subsequently, multiple constitutively activated NTRK fusion forms (e.g. ETV6-TRKC, QKI-TRKB, and TPR-TRKA) have been found and established as oncogenic drivers in lung cancers (prevalence of 3.3%) 8, 9 , colorectal cancers (2.2%) 8 , glioblastomas (2.5%) 10 , pediatric gliomas (7.1%) 10 , thyroid cancers (16.7%) 11,12 , and mammary analogue secretory carcinoma (> 90%) 13 . Recently, an "age-and tumoragnostic" therapy, also well-known as a basket trial, was performed in patients with tumors harbouring the identical NTRK fusion partner, has reshaped the landscape of molecular targeted cancer therapies and represents an important milestone in precision medicine 14 . Thus, targeting TRK is an attractive therapeutic strategy for a wide variety of cancers.

Currently, TRK-targeted small molecule kinase inhibitors (TRKIs), including rst-generation inhibitors
Larotrectinib (1) 15 and Entrectinib (2) 16 (Fig. 1), were approved by U.S. Food and Drug Administration (FDA) in 2018 and 2019, respectively, for treatment of adult and pediatric patients with solid tumors that evidencing a NTRK gene fusion. Both of these achieved striking clinical e cacy in NTRK-fusion-positive tumors, showing objective response rates (ORR) of 57-79% 17,18 . However, despite the initial encouraging therapeutic response to rst-generation TRK inhibitors, duration of the response is invariably limited by progressively acquired resistance 19,20 . The primary mechanisms for acquired resistance are found to be the secondary mutations occurring at the ATP-binding site of TRK kinase. These have been shown to involve three major regions: the solvent front (e.g., TRKA G595R and a homologous TRKC G623R mutation), an activation loop xDFG motif (e.g., TRKA G667C and a homologous TRKC G696A ), and the so-called oncogene gatekeeper residue (e.g., TRKA F589L ) 4 . Data from molecular modeling studies of these residue substitutions indicate that the resulting steric hindrance may have negative effects on the binding of Larotrectinib and Entrectinib 21 . To address the acquired resistance to prior TRK kinase inhibition, two second-generation macrocycle TRK inhibitors were developed (shown in Fig. 1): Selitrectinib (3, from Bayer/Loxo Oncology) and Repotrectinib (4, from Turning Point Therapeutics). Encouragingly, the recent phase I/II clinical results for 3 and 4 demonstrate promising prospects to overcome the solvent front mutation-mediated acquired resistance, and may provide an effective therapeutic option for patients with tumors that progressed on larotrectinib 20,22 . Structurally, the conformationally constrained macrocyclic inhibitors 3 and 4 can well accommodate the bulky TRKA G595R mutant at the solvent front without any steric clashes. However, recent clinical ndings indicate that therapeutic e cacy of both drugs can be compromised by the TRKA G667C mutant in xDFG motif, as the resulting steric hindrance between the mutant and the restricted uorinated aromatic moieties in 3 and 4 reduce binding a nity of drugs to protein 23 . To date, there are no approved drugs for treatment of patients who had developed resistance to 1 and 2 resulting from TRKs kinase secondary mutations. This situation still represents an existing "unmet clinical need". Therefore, the exploration of next-generation TRKIs with potent inhibitory activities toward multiple TRKs mutants are urgently needed.Here, guided by structure-based drug design (SBDD) 24,25 , we produced a structural modi cation of the reported compound 1 by scaffold hopping on the solvent front interaction region ("head" moiety marked in blue in structure 1, Fig. 2B), followed by exibilization and uoroalkoxy substitution of phenylpyrrolidine ("tail" moiety marked in red in 1), leading to the discovery of the superior derivative 7g. This compound showed excellent potency on a panel of Larotrectinib-resistant mutants both in biochemical and in cellular assays with reasonable pharmacokinetic (PK) properties in rats and dogs, in addition to improved in vivo antitumor e cacy in TRK wild type fusion-driven and TRK mutant fusion-driven tumor xenograft models. Thus, 7g appears to be a promising drug candidate for development as a TRK inhibitor that may overcome clinically acquired resistance to Larotrectinib.

Results
Structure-based Optimization and Structure − Activity Relationship (SAR) Exploration. In 2015, the G595R and G667C substitutions in TRKA kinase domain were rst reported to be associated with acquired resistance to 2 in a patient with LMNA-TRKA rearranged colorectal cancer 19 . Then, a novel gatekeeper F589L substitution in TRKA was identi ed as a well-characterized mechanism of resistance to 1 17 . Based on the X-ray structure of TRKA (PDB ID: 4AOJ) 26 , we performed a molecular docking modelling study to show positions of the three forms of TRKA mutants (Fig. 2a). Our study revealed that the substitutions that represent the G595R, G667C and F589L mutations are located in the solvent front, DFG, and gatekeeper regions, respectively. In order to deliver next-generation TRK inhibitors with the potential to overcome the resistance to 1, it was necessary to develop insight into the structural basis of the interaction of 1 with wide-type TRK and its three mutants in the above regions. Therefore, we preliminarily carried out a molecular modeling study that incorporated TRKA WT and 1 (Fig. 2c). Results illustrate that 1 have the following structural features: (1) The (S)-3-hydroxypyrrolidine-1-carboxamide moiety "head" is oriented toward the solvent front, which then forms a crucial hydrogen bond (H-bond) with the carbonyl of Asp596. guanidyl side chain of arginine or the thiol side chain of cysteine) causes the steric clash, highlighted with a red arrow, and weakens the binding a nity of ligand to protein. There is also steric hindrance between the mutated Leu589 locus of TRKA F589L and pyrrole ring in the "Tail" moiety ( Fig. 2f). In light of the above structural analysis, we initially imagined that the replacement of the bulky "head" moiety with a smaller substituent may avoid this steric clash and relieve stress at the Arg595 residue. To test this hypothesis, we proposed a scaffold hopping strategy to investigate a small set of novel TRK inhibitors 5 ( Table 1). This compounds feature diverse substituents shown in block C with both the pyrazolo[1,5a]pyrimidine framework and the phenylpyrrolidine component left intact.
The kinase inhibitory activities of newly synthesized compounds 5a-f against TRKA wild type and different mutant forms of TRKA including TRKA G595R , TRKA G667C , and TRKA F589L , were evaluated using a validated uorescence resonance energy transfer (FRET)-based Z'-Lyte kinase assay 27 . Compound 3 was served as positive control and compound 1 was used for a direct comparison. As shown in Table 1, the 3hydroxypyrrolidine-1-carboxamido group in block C of structure 1 was replaced with a variety of smaller substituents, including 2-oxoimidazolidin 5a, furan 5b, isoxazole 5c, pyrazole 5d, benzene 5e, and pyridine 5f. As expected, compound 5a (IC 50 of 28.6 nM) exhibited a 3.8-fold improvement in potency against TRKA G595R relative to reference compound 1. However, 5a did not improve the inhibitory activities against TRKA WT , TRKA G667C and TRKA F589L . Interestingly, the ve-membered aromatic heterocyclic substitutes in Block C had different impacts on inhibitory potency. For instance, the pyrazole 5d exhibited excellent potency against TRKA WT . With an IC 50 value of 0.2 nM, it was 427-fold more potent than furan 5b (IC 50 of 85.4 nM), and 1491-fold more potent than isoxazole 5c (IC 50 of 298.1 nM). Importantly, 5d achieved inhibitory activity at low nanomolar concentrations against a panel of TRKA mutants (G595R, G667C, and F589L), having IC 50 values ranging from 1.4 to 4.1 nM. This inhibitions were approximately 1-5-and 8-78-fold lower than that exhibited by reference drugs 3 and 1, respectively. Although introduction of the six-membered-ring groups in 5e (IC 50 of 3.0 nM) and 5f (IC 50 of 2.8 nM) exhibited strong inhibition of TRKA G595R , they were less active against TRKA WT , TRKA G667C , and TRKA F589L relative to compound 5d. Because the pyrazole ring is an optimal group for block C to avoid the resistance-related steric clash, and it is also a highly acceptable fragment for interacting with the solvent front region in other receptor tyrosine kinases (RTKs) 28,29 , we selected compound 5d as the new lead for further structural optimization.
As shown in Table 2, an array of R 1 , R 2 , and R 3 substituted pyrazolyl groups were introduced into the C3 position of the pyrazolo[1,5-a]pyrimidine core. It was found that monomethyl modi cations of 5d yielded compounds 5g (IC 50 of 0.4-9.0 nM) and 5h (0.6-6.4 nM) that displayed comparable inhibitory potency to that of compound 5d against TRKA WT , TRKA G595R , TRKA G667C and TRKA F589L . The dimethyl substituted pyrazole derivatives 5i and 5j showed a slight decrease in inhibitory potency compared to the monomethyl derivative 5g. For example, compound 5j showed an IC 50 values toward TRKA WT and TRKA G595R of 5.0 nM and 5.9 nM, respectively; this is about 12-fold less active than 5g. Subsequently, the replacements of the N-methyl group in the N-R 3 position with slightly larger substituents, such as hydroxyethyl (5k), methoxyethyl (5l), or cyclopropyl (5m) groups, achieved high potency against TRKA WT/mutants. Speci cally, the introduction of cyclopropyl group (5m) at the N-R 3 position of pyrazole led to overall subnanomolar activities against TRKA WT , TRKA G595R and TRKA G667C . However, in every practical sense, the binding a nity of the compounds to TRKA F589L had not improved compared to that of the three other forms of TRKA kinase. Furthermore, the preliminary metabolic stabilities of compounds in human liver microsome were evaluated (Table 2). Unfortunately, all selected compounds (5d, 5g, 5k and 5m) showed high clearance with a half-life of between 3.0-44.6 min.
Further structural modi cations were therefore completed to improve TRKA F589L inhibitory potency and drug-likeness. A better understanding of the subtle differences in the binding of 5m to TRKA WT and TRKA F589L could pave the way for further rational molecular design. We thus undertook additional in silico molecular modeling to explore the protein-ligand interactions in the region of the gatekeeper residue. We discovered that compound 5m adopts a nearly planar con guration that allows it to t well into the ATP-binding pocket of TRKA WT . The pyrrole ring moiety of 5m is orthogonal to the plane of gatekeeper residue Phe589, and localizes within a distance of 4.0 Å to form favorable hydrophobic interaction (Fig. 3a). However, in the mutant TRKA F589L , the free rotation of Leu589 might be limited by the rigid pyrrole moiety due to the closer distance between them (2.3 Å, Fig. 3b). This constraint is believed to contribute to its 12-fold potency loss. On the basis of the presumed structural rigidity of the pyrrole con guration, we envisage that the replacement of the pyrrole moiety by a group that introduces greater exibility and releases this limitation might be a feasible strategy. To accomplish this, the pyrrole moiety in compound 5m was subjected to decyclization; this process yielded the benzylamine analogue 6a (Scheme 2).
The design goal of the next stage is to improve TRKA G667C inhibitory potency, while maintaining its effectiveness against TRKA WT , TRKA G595R and TRKA F589L as well as to maintain the metabolic stability of 6d. To guide further drug design efforts and rationalize the decreased potency of benzylamine derivatives (Fig. 4a), computational modeling of 5d and 6b with TRKA G667C was performed. The model suggests that the phenylpyrrolyl moiety of 5d is directed toward the back pocket and forms satisfying hydrophobic interactions with residues Phe589, Lys544, Phe521, and Cys667. Predictably, the exibilization of pyrrole that yields 6b may have led to loss of those hydrophobic interactions, that are likely responsible for the 6.4-fold reduction of inhibitory activity against the TRKA G667C mutant (5d vs 6b). We thus suggest it is necessary to create additional interactions to compensate for the loss of binding a nity. A further understanding of the binding mode of 6b-TRKA G667C indicates that there is a large un lled space between the 2-F atom in the benzylamino fragment and a loop containing residues Glu518, Gly519, and Ala520 (highlighted with a red arrow, Fig. 4b). We envision that lling the space with a substituent bearing a terminal H-bond acceptor to interact with this loop will be a useful adjustment. Therefore, we decided to screen compounds containing a set of uoroalkoxy substituents at the C2position of the benzylamino fragment.
As summarized in Table 4, the uoromethoxy-substituted compound 7a exhibited good potency against TRKA G667C with an IC 50 value of 6.1 nM, which is 4.1-fold more potent than parent compound 6b.
Moreover, 7a displays no obvious decrements to the inhibitory potency against other types of TRK kinase.
Surprisingly, compound 7b, substituted at the R 4  To interpret the SAR, molecular docking studies were constructed for the most potent compound 7g. As shown in Fig. 5A, 7g is accommodated in the active site of G595R-mutated TRKA as a "U-shaped" con guration, and makes the same backbone hinge contact as 1 due to the overlapping positions of pyrazolo[1,5-a]pyrimidine in both of them. Compared to the rigid hydroxypyrrolidine ring of 1, the exible hydroxyethyl group of 7g not only indulges the free rotation of residue Arg595 without causing steric clash but also forms an H-bond with the NH of Asp596. Moreover, there is an additional H-bond between di uoroethoxy in 7g and NH of Glu518 that plays a key role in enhancing binding a nity, and supports our predictions. In the TRKA G667C mutant (Fig. 5B), close to the solvent front region, the hydroxyl moiety of compound 7g forms two H-bonds with the carbonyl of Arg595 and the NH of Asp596, whereas a single weaker H-bond was observed for compound 1. There is also an additional H-bond between the di uoroethoxy moiety of 7g and Glu518, which might drag the exible benzylamine fragment away from the residue C667 to avoid steric hindrance. However, the restricted phenylpyrrole ring in 1 is more likely to hinder the free rotation of the residue Cys667 than the phenyl-di uoroethoxy group of 7g. The combination of the above led us to adopt 7g as the lead candidate, particularly because of its stronger binding a nity with TRKA G667C as compared to 1.
Cellular Antiproliferative Assay. In order to evaluate the suppressive effects on cellular growth of the new TRK inhibitor, we tested the most potent compound 7g as well as the second-generation TRK inhibitor respectively. In comparison, 7g achieved a better antitumor response in these two TRK mutant-dependent tumor models. In addition, no mortality and no signi cant loss of body weight occurred in any of the 7gtreated groups (Fig. S1).
Chemistry. The syntheses of 5a-m are illustrated in Scheme 1. The commercially available (R)-2-(2,5di uorophenyl)pyrrolidine hydrochloride 8 was reacted with 5-chloropyrazolo[1,5-a]pyrimidine to provide 9 under DIPEA in n-Butanol. This was followed by iodination with NIS and led to the key intermediate 10 35 . Compound 5a was prepared by a copper-catalyzed amination reaction of 10 with imidazolidin-2-one.
Suzuki coupling of 10 with boronic acid pinacol ester derivatives 11 provided 5b-m. The synthetic route of 6a-e was similar to that described for 5b-m, as depicted in Scheme 2. Brie y, the key intermediate 14 was

Discussion
Clinically acquired resistance caused by secondary mutations within the ATP-binding pocket represents a common mechanism and a major "unmet clinical need" in targeted cancer therapy. Similar to mutations found in ALK 36, 37 and RET 38 , the most frequent secondary mutations in TRKs appear at the solvent front TRKA G595R and TRKC G623R , the xDFG motif TRKA G667C and TRKC G696A , and the gatekeeper TRKA F589L .

Mutations at these loci confer confer resistance to FDA-approved rst-generation TRK inhibitors
Larotrectinib (1) 20 and Entrectinib (2) 19 . Although the second-generation inhibitors Selitrectinib (3) and Repotrectinib (4)  at 15 mg/kg), TMP3-TRKA G595R (TGI of 61% at 30 mg/kg), and EVT6-TRKC G623R (TGI of 88% at 30 mg/kg).All 7g tumor inhibition results were found to be superior to those produced by Larotrectinib (TGI of 70% at 15 mg/kg) and Selitrectinib (TGI of 30% and 73% at 30 mg/kg, respectively). Collectively, the potential for the novel TRK inhibitor 7g to overcome the common TRK mutations that contribute to Larotrectinib resistance suggests that it is a promising drug candidate for further development.

Methods
General Chemistry. Unless otherwise noted, reagents and solvents used in experiments were purchased from commercial sources and used without further puri cation. Flash chromatography was performed triplet; q, quartet; m, multiplet; dd, doublet of doublet. The purities of all target compounds were determined to be > 95% by highperformance liquid chromatography (HPLC). HPLC conditions were as follows: Gemini C18 column at room temperature, 4.6 cm × 150 cm, 5 µm, 10 − 90% acetonitrile (0.05% TFA)/water (0.05% TFA), 10 min run; ow rate, 1 mL/min; UV detection λ = 214 nm, 254, and 280 nm. MS spectra were obtained on an agilent technologies 6120 quadrupole LC/MS (ESI). High-resolution mass spectra (HR-MS) were obtained on an Agilent 6224 TOF LC/MS (USA). Yields were of puri ed compounds and were not optimized.
In vitro enzymatic activity assay. The inhibitory activities of the compounds against native TRK and TRK mutants (Invitrogen) were determined using a uorescence resonance energy transfer (FRET)-based Z'-Lyte kinase assay assay system following to the manufacturer's instructions.