Antitarget Selectivity and Tolerability of Novel Pyrrolo[2,3-d]pyrimidine RET Inhibitors

Abstract

The selective inhibition of RET kinase as a treatment for relevant cancer types including lung adenocarcinoma has garnered considerable interest in recent years and prompted a variety of efforts toward the discovery of small-molecule therapeutics. Hits uncovered via the analysis of archival kinase data ultimately led to the identification of a promising pyrrolo[2,3-d]pyrimidine scaffold. The optimization of this pyrrolo[2,3-d]pyrimidine core resulted in compound 1, which demonstrated potent in vitro RET kinase inhibition and robust in vivo efficacy in RET-driven tumor xenografts upon multiday dosing in mice. The administration of 1 was well-tolerated at established efficacious doses (10 and 30 mg/kg, po, qd), and plasma exposure levels indicated a minimal risk of KDR or hERG inhibition in vivo, as evaluated by Miles assay and free plasma concentrations, respectively.

The RET (REarranged during Transfection) gene, first described by Takahashi,¹ encodes an RTK (transmembrane Receptor Tyrosine Kinase) that plays a key role in neuronal signaling as the canonical receptor for ligands of the GDNF (Glial cell-line Derived Neurotrophic Factor) family. Consequently, downstream signaling via RET activation has been demonstrated to be critical for early-stage peripheral nervous system development.^2,3 Gain-of-function mutations of the proto-oncogene, however, have long been recognized as potential drivers in a range of malignancies, including a relatively small subset (1–2%) of non-small cell lung cancer (NSCLC) cases.⁴ As such, the therapeutic deployment of small-molecule inhibitors to address RET oncogene-addiction in the context of lung adenocarcinoma has garnered considerable interest.⁵⁻⁸ Early efforts in this area have focused on the assessment of multikinase RET inhibitors in the clinic, but this approach has suffered from instances of dose-limiting toxicity attributed to the inhibition of non-RET kinases, most notably KDR (Kinase insert Domain Receptor).⁹ Current efforts in this field have therefore focused on the discovery of novel RET inhibitors with broad kinase selectivity to minimize the risk of off-target toxicities.

Late-stage clinical reports and subsequent approval of the treatment of RET-driven cancers with the selective inhibitors pralsetinib (BLU-667)^10,11 and selpercatinib (LOXO-292)^12,13 are promising breakthroughs and have reinforced the notion that the potent and selective inhibition of acquired RET kinase fusions can effect improved clinical outcomes in relevant oncogenic cohorts.^14,15 Our preliminary efforts to devise selective RET inhibitors resulted in the discovery of a highly potent pyrazolo[1,5-a]pyrimidine compound with considerable selectivity against antitarget KDR, brain penetration, and robust efficacy in vivo. Ultimately, though, poor tolerability proximal to the efficacious dose precluded further preclinical evaluation, as has been outlined in a previous disclosure.¹⁶ Our subsequent investigation of alternative scaffolds, in the hope of improving upon in vivo compound tolerability thresholds, resulted in focused efforts on a pyrrolo[2,3-d]pyrimidine core substructure, which is described herein.

The decision to pursue a pyrrolo[2,3-d]pyrimidine scaffold was initially informed by Gini coefficient analysis¹⁷ performed on historical enzymatic and Ba/F3¹⁸ kinase data, which resulted in several intriguing hits, including 2 and 3 (Table 1).¹⁹ Additional factors, including favorable ligand efficiency and LipE of these pyrrolo[2,3-d]pyrimidine hits, reinforced our interest in this scaffold and substantiated the assessment of additional archival analogs. During the course of this evaluation, our internal collection revealed several additional pyrrolo[2,3-d]pyrimidine RET kinase inhibitors including piperazine-containing compound 4. We found 4 to be particularly appealing because it demonstrated both broad selectivity across a panel of tyrosine kinases (Table 1) and favorable physicochemical properties. Although 4 demonstrated only moderate RET potency (Table 1), the activity of structurally similar hit compounds 2 and 3 gave us confidence that further cell-based RET inhibitory activity could be subsequently introduced during the course of wet chemistry efforts.

Table 1. Cellular Activity and Tyrosine Kinase Selectivity of Archival Hit Compounds^a.

compound	2	3	4
WTb,c	1.3	1.6	>10
KIF5B-RETc	0.011	0.007	0.24
KIF5B-RETV804Mc	2.1	1.1	>10
RET	0.051	0.0040	0.92
BCR-ABL	0.043	0.058	4.2
NPM-ALK	0.37	0.87	>8.5
BLK	0.57	0.14	3.6
BMX	6.9	10	>11
FGFR1	2.1	0.99	>11
FGFR2	0.027	0.19	>2.9
FGFR3	0.49	1.1	>9.4
FGFR4	1.1	2.4	>11
FGR2	0.37	0.022	2.6
FLT1	0.11	0.14	>6.7
FLT3	4.4	0.97	>11
FMS	0.96	0.48	>11
IGF1R	2.8	2.7	>11
INSR	1.6	1.4	>8.2
JAK2	5.1	6.6	>11
KDR	0.15	0.058	>7.8
KIT	0.042	0.043	3.4
LCK	0.13	0.062	3.5
LYN	0.13	0.17	5.0
MER	5.5	4.9	>11
MET	5.8	6.0	>11
PDGFRα	0.34	0.18	>10
PDGFRβ	0.42	0.27	>11
RON	6.0	7.0	>11
ROS	NTd	0.042	>11
SRC	0.063	0.12	2.4
SYK	2.4	4.9	>11
TIE1	4.4	2.5	>11
TRKA	1.5	1.2	9.3
TRKB	1.3	1.4	8.5
TRKC	0.92	0.51	>8.1
TYRO3	2.2	3.1	>11
ZAP70	2.1	2.9	>11

^aProliferation assays run in 1536 well format, employing Ba/F3 cells transfected with Tel fusions of the indicated kinase, and IC₅₀ values reported in μM, unless otherwise noted.

^bParental Ba/F3 cells utilized.

^cAssay run in 384 well format.

^dNot tested.

Using 4 as a basis template for structure–activity relationship (SAR) investigations, we first examined substitution at R₁. Our initial objectives focused on both improving RET inhibition in cellular assays and finding an appropriate phenol isostere due to the well-recognized reactive and phase II metabolism concerns for the moiety. A library of R₁ derivatives was synthesized via Suzuki coupling of the appropriate aryl boronic acid/ester with the intermediate (cis)-1-(4-(−3-(4-amino-5-bromo-7H-pyrrolo[2,3-d]pyrimidin-7-yl)cyclobutyl)piperazin-1-yl)ethanone (Scheme 1), and the key findings are summarized in Table 2. A cursory exploration of R₁ phenol trajectory found meta substitution to be ideal for both enzymatic and cell-based RET activity, akin to the phenolic compounds that surfaced from screening efforts (Table 1). A survey of halogen substitution of the phenol ring found chloro-substitution at the four-position to be particularly advantageous, and gratifyingly, this boost in potency was also translated to the corresponding aniline compound 10 (Table 2). Furthermore, the transition from phenol to aniline also led to increased selectivity versus KDR, as evidenced by the matched pair 8/10. These initial findings at R₁ were found to be consistent with an X-ray cocrystal structure of 10 with RET kinase, which indicated H bonds between the R₁ aniline and kinase residues E775 (αC-helix) and S891 (adjacent to the DFG motif). Additional hydrophobic interaction of the R₁ 4-chloro moiety (10) with the nonpolar carbon chain of amphipathic residue K758 (<4 Å, Figure 1) also was supportive of our preliminary SAR (Table 2).

Scheme 1. Synthetic Route for R₁ Analogs.

NaBH(OAc)₃, DCM.

TFA, DCM.

2-(4,6-Dichloropyrimidin-5-yl)acetaldehyde, iPr₂NEt, EtOH.

Br₂, AcOH.

NH₄OH, THF.

R₁B(OH)₂ or R₁B(pin), Na₂CO₃, K₃PO₄, SiliaCat DPP-Pd, dioxane.

Table 2. Impact of R₁ Modification on RET Potency^a.

^aIC₅₀ values reported in μM.

^bBiochemical (aa 658–1072) HTRF-based assay.

^cCell-based proliferation assays.

Figure 1.

X-ray cocrystal structure of 10 and RET kinase reveals interaction of R₁ NH₂ with kinase residues E775 (3.3 Å) and S891 (2.8 Å).

The increased cellular potency achieved through replacement of the phenol moiety at R₁ also advantageously coincided with an improvement in the pharmacokinetics (PK) of the scaffold. When comparing compounds 4 and 10 in PK studies, the latter was found to possess much improved oral plasma exposure, clearance, and bioavailability (Table 3). The suitable PK profile of 10 subsequently allowed for its rapid evaluation in mouse efficacy models, where it was hoped that sufficient in vivo RET-inhibitory exposures could be achieved. We also desired to quickly benchmark the scaffold in terms of tolerability, as poor tolerability ultimately doomed our previous efforts on a pyrazolo[1,5-a]pyrimidine series.¹⁶ As such, mice were dosed orally with 10 in a transgenic RIE KIF5B-RET xenograft tumor model at 10, 30, and 60 mg/kg once daily for 14 days. Gratifyingly, robust tumor regression was observed in the 60 mg/kg cohort, with corresponding inhibition/stasis demonstrated at lower doses. All doses were well-tolerated, with no adverse events or body weight loss observed (Supporting Information). Subsequent western blot analysis of intratumoral tissue collected on day 14 of the study confirmed target engagement, where a significant inhibition of RET phosphorylation was observed at a 5 h postdose time point for the 60 mg/kg study group, although the activity substantially declined by 24 h (Supporting Information). Ultimately, the promising results of this study gave us confidence to move forward and screen additional pyrrolo[2,3-d]pyrimidine analogs for efficacy in relevant murine xenograft models.

Table 3. PK Parameters of Key Compounds^a.

^aPK studies run in CD-1 mice with formulation of 1 mg/mL in PEG300/D5W, 3:1 for iv and in 0.5% MC/Tween80 for po, unless otherwise noted.

^bFormulation of 2.5 mg/mL in PEG300/D5W, 3:1 for both iv and po in Balb/C mice.

^cBalb/C mice used.

^d20 mg/kg of compound HCl salt dosed at 2 mg/mL in Wistar mice.

^eFormulated in 10% Eudragit, 0.5% Tween80 in 50 mM acetate buffer in Balb/C mice.

^f2 mg/mL used.

Resulting efforts to improve upon compound 10 primarily focused on lowering the in vivo efficacious dose. The first manner in which we sought to accomplish this was through strategic efforts to further increase the RET potency. Satisfied, for the time being, with prior modification at R₁ (vide supra), we turned our attention to the evaluation of R₂. The 10/RET kinase cocrystal structure (Figure 1) informed a detailed understanding of the R₂-region topography and suggested that more wide-ranging changes may be tolerated to improve the potency. Molecular docking served to guide these synthetic efforts, where four-, five-, and six-membered aliphatic rings flanked by terminal hydrophilicity abutting the solvent front were thought to be well-suited for RET inhibitory activity. More tailored attempts to create additional discrete polar contacts seemed to offer only a slim chance of success, however, as RET-specific residues in this region appeared to be poorly accessible. Second, we sought to lower the in vivo efficacious dose of this scaffold through the identification and elimination of R₂ metabolic soft spots. Metabolic identification (MetID) analysis of selected analogs streamlined the prosecution of the R₂-region SAR, with the objective of achieving reduced clearance and increased oral exposures. Subsequent testing of these R₂ analogs in vitro and in vivo validated our efforts, as cell-based RET potency was successfully increased by two- to three-fold (Table 4), and CL reduced from 15 to <2 mL/min/kg (Table 3), respectively.

Table 4. Optimization of in Vivo Efficacy and Antitarget Windows^a.

	Ba/F3			LC2/ad	hERG			in vivo mouse efficacyb
Cmpd	KIF5B-RET	Tel-KDR	WT	CCDC6-RET	bindingc	MPCd	PPBe	Eff Dosef	Eff AUCg	Eff Cmaxg	Eff AUCfreeh	Eff Cmax,freeh
10	0.044	0.84	4.6	0.046	19	7.9	64.4	60	115	19	41	6.8
11	0.015	0.66	2.6	0.035	25	5.3	96.9	10	50	8.3	1.6	0.26
12	0.020	0.44	4.1	0.11	9.9	3.0	99.58	3	106	7.4	0.45	0.031
1	0.022	0.80	8.1	0.051	>30	6.0	99.34	10	136	8.4	0.90	0.055

^aIC₅₀ values reported in μM for cell-based proliferation and hERG assays.

^bEvaluated in multiday RIE KIF5B-RET xenograft model.

^cAssessed with ³H-dofetilide.

^dManual patch clamp.

^eMouse plasma protein binding percentage. For cases >99%, 20% diluted plasma and rapid equilibrium dialysis utilized with calculation by Kalvass equation.

^fOral dose (mg/kg, qd) at which tumor regression was observed.

^gArea under curve (AUC) (h*μM) and C_max (μM) observed at regression dose after days 14, 14, 4, and 13 for compounds 10, 11, 12, and 1, respectively.

^hUnbound fraction of AUC (h*μM) and C_max (μM) at regression dose.

As a result of the improved RET cellular potency and oral exposures achieved, we felt that several of our advanced R₂ analogs were well-suited for further evaluation in murine xenograft models (Table 4). When compound 11 was assessed for efficacy and tolerability in the aforementioned 14-day KIF5B-RET xenograft model, its significant RET potency and low clearance translated to significant tumor regression at 10 mg/kg, with growth inhibition observed at 1 and 3 mg/kg. With respect to tolerability, mice dosed at 30 mg/kg required a drug holiday following significant body weight loss (BWL); however, once dosing resumed on day 9, no further adverse events were noted during in-life observation or upon necropsy. Similarly, the in vivo analysis of compound 12, studied in mice for 12 days with oral dosing qd at 1, 3, and 10 mg/kg, revealed significant tumor regression at 3 mg/kg. Although, at 10 mg/kg, body weights initially dropped <10% but gradually stabilized over time with ongoing dosing. Whereas the origin of the documented BWL remained unclear at this point, its manageability and the robust low-dose efficacy observed with 11 and 12 made us cautiously optimistic that further widening of the therapeutic index could be achieved with other scaffold analogs.

During the course of our optimization efforts, we discovered compound 1, a derivative of 12 where R₁ fluorination served to modestly reduce activity against antitargets KDR and hERG while retaining RET potency (Table 4). The PK properties were similar to those of 12, where a prolonged half-life, low clearance, and good bioavailability (Table 3) translated to high levels of compound exposure in multiday in vivo mouse xenograft models (Table 4, Supporting Information). For instance, when compound 1 was examined in a 16 day RIE KIF5B-RET mechanistic mouse xenograft model at 1, 3, 10, 30, and 60 mg/kg, exposures were generally dose proportional, and analysis on day 13 versus day 1 suggested that only minor compound accumulation occurred (Supporting Information). With regard to compound efficacy in this study, tumor inhibition, stasis, and regression were observed in the 1, 3, and 10 mg/kg cohorts, respectively (Figure 2). The tolerability for compound 1 was found to be significantly improved relative to that for other scaffold analogs (vide supra), as oral dosing of 30 mg/kg also demonstrated robust efficacy with no adverse effects. However, in the 60 mg/kg arm of the study, significant BWL among mice was observed, where subsequent necropsy revealed substantial gastrointestinal toxicity. The slightly lower exposures on day 1 for the 60 mg/kg group (AUC: 369 h*μM, Cmax: 19 μM) than on day 13 for the well-tolerated 30 mg/kg dose cohort (AUC: 409 h*μM, Cmax: 25 μM) seemed to be indicative of a relatively steep and acutely sensitive dose-toxicity paradigm, akin to compounds 11 and 12 (vide supra). Nevertheless, this widened therapeutic window marked a key achievement for the project and warranted the advancement of 1 to further profiling and more advanced mouse xenograft models.

Figure 2.

In a 16 day study, the administration of compound 1 (po, qd, MC/Tween80, suspension) resulted in significant tumor regression at well-tolerated doses of 10 and 30 mg/kg in RIE KIF5B-RET xenograft mice.

To gain an understanding of the PK/PD (pharmacodynamics) relationship for compound 1, we conducted a time-course analysis in RIE KIF5B-RET tumor-bearing mice with a single administration of the efficacious dose (10 mg/kg). As depicted in Figure 3, a clear relationship between the in vivo exposure of 1 and the intratumoral pRET inhibition, as measured by western blot, was found over the course of 48 h. In addition, 1 was analyzed in the context of a 21-day disease-relevant CCDC6-RET LC-2 tumor tissue model, which resulted in tumor regression at 10 and 30 mg/kg, with the compound being well tolerated (Figure 4). Following the cessation of treatment, mice were monitored for several weeks, and tumor regrowth was found to be heterogeneous but generally weak in the case of fully regressed tumors (Supporting Information).

Figure 3.

Intratumoral RET target engagement relative to plasma exposure in RIE KIF5B-RET xenograft mice at tumor regression dose (10 mg/kg, po, MC/Tween80, suspension) following a single dose of compound 1.

Figure 4.

Multiday treatment of LC-2/ad tumor-bearing mice with 1 (21 day, po, qd, MC/Tween80, suspension) is well tolerated and leads to significant regression at 10 and 30 mg/kg.

Now confident that well-tolerated robust tumor regression could be achieved with 1, we sought to examine how safety windows against specific antitargets might be impacted by efficacious levels of in vivo compound exposure. With regard to KDR, a well-recognized side effect of its inhibition in vivo is vascular toxicity mediated by disruption of the VEGF signaling pathway. Therefore, we hoped to determine if efficacious exposures of 1 would only minimally engage KDR in vivo and not result in the significant interruption of VEGF-mediated vascular effects. As such, a Miles assay²⁰ was performed at the maximum tolerated dose of 30 mg/kg to establish if in vitro cell-based RET/KDR selectivity for 1 (36-fold, Table 4) would be translated in vivo. As can be seen in Figure 5, compound 1 does not significantly impact the vascular permeability in non-tumor-bearing mice as compared with the multi-RTK inhibitor cabozantinib (at 14 day RIE KIF5B-RET model tumor stasis dose of 60 mg/kg, Supporting Information).

Figure 5.

Representative images of non-tumor-bearing mice (four or five per group) demonstrate that 1 does not significantly impact VEGF-driven Evan’s Blue Dye vascular permeability in the Miles assay over the course of 24 h.

Furthermore, despite hERG manual patch clamp data being <8 μM across the late-stage lead compounds in Table 4, plasma exposure data gleaned from efficacy studies gave us confidence that we could effectively engage RET in vivo without leading to significant hERG inhibition. The calculation of maximum free plasma concentrations for compounds at their efficacious dose (where tumor regression was observed in mouse RIE KIF5B-RET efficacy studies) established that the modification of R₂ led to progressively lower free concentrations, which were ultimately orders of magnitude below hERG IC₅₀ values from both binding and manual patch clamp assays in the case of 1 (Table 4). Proportionally wide hERG safety margins for 1 were also found at the tolerated dose of 30 mg/kg (AUC_free: 2.7 h*μM, C_max,free: 0.16 μM), stemming from its reasonably linear PK profile and minimal compound accumulation.

To summarize, efforts to optimize a pyrrolo[2,3-d]pyrimidine scaffold successfully demonstrated a correlation between in vitro cell-based kinase inhibition and in vivo murine model efficacy, as evidenced by robust RET fusion tumor regression and VEGF/KDR-driven vascular permeability. Furthermore, multiday studies with compound 1 proved that good tolerability could be achieved, with substantial tumor regression being observed at both 10 and 30 mg/kg in both KIF5B-RET and LC-2/ad xenograft mouse models. Calculations involving the free plasma exposure of compound 1 in these murine efficacy studies also suggested that other known off-target toxicity risks, such as hERG inhibition, should be minimal. The balanced composite profile of compound 1 ultimately warranted its advancement to a range of preclinical studies, including, but not limited to, dose formulation, rat toxicology, and pharmacokinetics in nonrodent species.

Glossary

Abbreviations

RET

rearranged during transfection

RTK

receptor tyrosine kinase

GDNF

glial cell-line derived neurotrophic factor

NSCLC

non-small cell lung cancer

KDR

kinase insert domain receptor

SAR

structure–activity relationship

RIE

rat intestinal epithelial

BWL

body weight loss

VEGF

vascular endothelial growth factor

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.1c00450.

• Experimental procedures and characterization of all compounds; in vitro assay descriptions; cellular kinase panel activity for compounds 4, 10, 11, 12, and 1; PK data for compounds 4, 10, 11, 12, and 1; in vivo efficacy/tolerability study design and results for compounds 10, 11, 12, and 1; PK/PD study design and results for compound 1; Miles assay study design and representative images for compound 1 (PDF)

Accession Codes

PDB accession code: 7RUN.

The authors declare no competing financial interest.