RET solvent front mutations are vulnerable to vepafestinib

We employed rational chemical design to develop a potent and selective RET inhibitor and identified vepafestinib, a small molecule that is structurally distinct from existing RET inhibitors18,32. The alkyne moiety of vepafestinib (4-amino-N-[4-(methoxymethyl)phenyl]-7-(1-methylcyclopropyl)-6-[3-(morpholin-4-yl)prop-1-yn-1-yl]-7H-pyrrolo[2,3-d]pyrimidine-5-carboxamide) located in the 6-position on the 7H-pyrrolo[2,3-d]pyrimidine-5-carboxamide part of the structural core, resulted in a highly unique derivative in kinase inhibitors (Fig. 1a). Vepafestinib potently inhibited recombinant WT RET kinase at subnanomolar concentrations, similar to half-maximum inhibitory concentration (IC50) values obtained with selpercatinib or pralsetinib (IC50 values (nM): vepafestinib, 0.33 ± 0.01; pralsetinib, 0.31 ± 0.01; selpercatinib, 0.13 ± 0.03; vandetanib, 6.2 ± 0.8). A single concentration of 23 nM vepafestinib was tested on a panel of 255 recombinant kinases. RET was the only kinase inhibited by >50% (Fig. 1b and Supplementary Table 1a). Selpercatinib (22 nM) and pralsetinib (17 nM) were less specific, inhibiting three (including KDR (kinase insert domain receptor)) and 11 kinases by >50%, respectively (Extended Data Fig. 1a,b). These results were confirmed in dose–response studies of 14 kinases, in which KDR (also known as vascular endothelial growth factor receptor 2) was potently inhibited by selpercatinib (IC50 = 14 nM) and pralsetinib (IC50 = 35 nM) (Supplementary Table 1b). We also tested the RET and SRC family inhibitor TPX-0046 (enbezotinib, 26 nM) against a similar panel of kinases and found that TPX-0046 is an MKI, inhibiting 39 kinases by >50% (Extended Data Fig. 1c and Supplementary Table 2a). Targets of TPX-0046 included the kinases TRKA-C, FGFR1–FGFR4, most SRC family members, ACK and TXK (Supplementary Table 2a). The IC50 for inhibition of RETWT by TPX-0046 was 0.26 ± 0.02 nM.

Fig. 1: Structure and biochemical characterization of vepafestinib (TAS0953/HM06).

a, Chemical structure of vepafestinib. b, Kinase selectivity profile of vepafestinib across 255 kinases. Enzyme activities were assessed in the presence of 23 nM vepafestinib, which is approximately 70-fold higher than the IC50 for inhibition of RETWT. Only one kinase (RET) was inhibited by >50% and is shown as a blue circle on the kinome tree. TK, tyrosine kinase; TKL, tyrosine kinase-like; CAMK, calcium/calmodulin-dependent protein kinase; STE, homologs of yeast sterile 7, sterile 11 and sterile 20 kinases; CK1, casein kinase 1; CMGC, cyclin-dependent kinases, mitogen-activated protein kinases, glycogen synthase kinases and cell division control protein-like kinases; AGC, protein kinase A, protein kinase G and protein kinase C families. c, GI50 (50% growth inhibition) values of vepafestinib, in comparison to other RET inhibitors on proliferation of Ba/F3 cells expressing KIF5B–RETWT or KIF5B–RET harboring mutations in the solvent front of the kinase domain (G810R, G810S or G810C) or the gatekeeper domain (V804L or V804M). Data represent the mean ± s.d. of three independent experiments. d, Effect of vepafestinib on phosphorylation of RET and downstream signals in Ba/F3 cells expressing KIF5B–RETWT, KIF5B–RETG810R, KIF5B–RETG810S or KIF5B–RETG810C. Cells expressing KIF5B–RETWT, KIF5B–RETG810R, KIF5B–RETG810S or KIF5B–RETG810C were treated with the indicated concentrations of each drug for 1 h before preparation of cell extracts for western blotting. Representative immunoblots from two independent experiments are shown. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a loading control. p, phosphorylated.

Source data

The cellular potencies of RET inhibitors against RET fusions and mutations, including RETV804L, RETV804M, RETG810R, RETG810S and RETG810C were evaluated using engineered Ba/F3 cells (Fig. 1c). Vepafestinib inhibited growth of Ba/F3 cells expressing KIF5B–RETWT or KIF5B–RET mutants (V804M, V804L, G810R, G810S, G810C) (Fig. 1c). By contrast, growth of Ba/F3 cells expressing KIF5B–RETG810R, KIF5B–RETG810S or KIF5B–RETG810C was less sensitive to selpercatinib and pralsetinib than that of cells expressing RETWT, RETV804M or RETV804L as previously reported19,21. Vandetanib was less potent than the RET-selective inhibitors. Consistent with cell viability data, phosphorylation of RET and ERK were blocked by vepafestinib in Ba/F3 KIF5B–RETWT cells (Fig. 1d). Of note, vepafestinib suppressed phosphorylation of RETG810R, RETG810S and RETG810C with near-complete inhibition at 100 nM (Fig. 1d). TPX-0046 inhibited phosphorylation of the RETG810R, RETG810S and RETG810C mutants, with RETG810R being the least sensitive (IC50 values, RETWT, 21.9 nM; RETG810R, 108 nM) (Supplementary Table 2b). Selpercatinib and pralsetinib did not block phosphorylation of RETG810R, RETG810S or RETG810C (Fig. 1d).

Crystal structure of RET complexed with selective inhibitors

The crystal structure of the RET kinase domain complexed with a vepafestinib analog, TAS compound 1 (TAS-C1) (Fig. 2a), was successfully solved at 1.64 Å. TAS-C1 was used because attempts to crystalize RET with vepafestinib were unsuccessful. Imposition of vepafestinib upon the TAS-C1–RET co-crystal structure showed substantial overlap of the two small molecules, suggesting that the data obtained with TAS-C1 could be extended to vepafestinib (Extended Data Fig. 2a,b). We also solved the crystal structures of RET complexed with selpercatinib and pralsetinib at 2.75 Å and 2.31 Å, respectively, in concordance with recently reported co-crystal structural data33. The pyrimidine ring in TAS-C1 forms hydrogen bonds with both E805 and A807 in the hinge region (Fig. 2b). In addition, nitrogen atoms in the pyrazole moiety in TAS-C1 forms hydrogen bonds with E775 and D892. At the opposite side, the cyclopropyl group occupies a hydrophobic environment, surrounded by L730, G731, F735, V738 and L881 (Extended Data Fig. 3a). The flexibility of the amide bond in TAS-C1 seems to be affected less sterically by the bulky substitutions of gatekeeper positions (V804) (Extended Data Fig. 3b). The methylpyrazole moiety of TAS-C1 is positioned in the pocket of the neighboring amino acids E775, L779, L802 and V804 (Fig. 2c). By contrast, the terminal moieties of the structures in selpercatinib and pralsetinib are inserted into another pocket surrounded by M759, L760, E768 and L772 (Fig. 2d). Additionally, TAS-C1 is positioned some distance away from the direction of the glycine side chain of the solvent front position 810, but selpercatinib and pralsetinib are closer (Fig. 2c,d). These findings indicate that substitution of glycine at codon 810 with other bulky residues is likely to establish steric hindrance for selpercatinib and pralsetinib but not for vepafestinib. This likely contributes to maintaining biological potency of vepafestinib toward RETG810 mutations.

Fig. 2: X-ray crystallography of RET complexed with RET-selective inhibitors.

a, Chemical structure of TAS-C1. b, X-ray structure of RET complexed with TAS-C1. c, View from the solvent front area in the co-crystal structure of RET with TAS-C1. d, Overlay of co-crystal structures of selpercatinib and pralsetinib bound to RET. The viewpoint is the same as in c. The binding compounds are shown as stick models, with yellow (TAS-C1), cyan (selpercatinib) and magenta (pralsetinib) representing each RET inhibitor. e, Positions of the amino acid residues where mutagenesis was performed for in-cell western assays are shown in the co-crystal structure of RET with TAS-C1, overlaid with selpercatinib and pralsetinib. f, IC50 values calculated from in-cell western assays of Jump-In GripTite HEK293 cells transiently expressing WT or mutant KIF5B–RET. Cells were treated with the indicated compounds for 1 h. The assay was performed in triplicate, and mean IC50 values are represented with the color codes shown at the bottom.

Further analysis of the X-ray crystal structure revealed that there are roughly two clustering selpercatinib–RET or pralsetinib–RET complexes and the TAS-C1–RET complex in the point of the inserted area by the terminal moiety of these drugs. To assess the biological effects of these structural differences, we established a panel of RET mutations by substituting amino acids at positions close to the interaction site of each drug. We surmised that substitutions of amino acids that are in close proximity to a RET inhibitor when bound to the kinase may induce resistance to the respective drug. We identified nine residues in RET (E732, G736, K737, M759, L760, E768, L772, K808, G810) that have side chains or main chains within 4 Å of both selpercatinib and pralsetinib (Supplementary Table 3) and anticipated that substitution of these amino acids might influence binding of selpercatinib and pralsetinib but not vepafestinib. We also selected one residue (I788) with a side chain within 4 Å of TAS-C1 and hypothesized that substitutions at this site might reduce vepafestinib activity. Although two other residues (L730, Y806) are located within 4 Å of the three drugs, these residues form direct or indirect interactions with selpercatinib or pralsetinib. The positions of the 12 amino acids in the co-crystal structure of RET with the three drugs are shown in Fig. 2e. Subsequently, we established 15 potential mutations in the selected positions. Substituted amino acids were selected to generate previously reported RET mutations19,33,34,35,36,37,38,39 and/or to be larger or more charged than the original residue, which could affect RET–compound binding. Vepafestinib inhibited phosphorylation of RETWT and most of the RET mutants (non-solvent front) with similar IC50 values (Fig. 2f). By contrast, phosphorylation of several RET mutants (L730Q, L730R, G736A, L760Q) was refractory to selpercatinib and pralsetinib compared to RETWT phosphorylation. As predicted, RETI788N conferred resistance to vepafestinib. Importantly, all RETG810 mutations remained vulnerable to vepafestinib. Although the RETG810C mutant appeared about threefold less sensitive than RETWT, our data from Ba/F3 cells (Fig. 1c,d) imply that overcoming the RETG810C mutation with vepafestinib is likely. All RETG810 mutations conferred decreased sensitivity to selpercatinib and pralsetinib (Fig. 2f) but resulted in sensitivity to TPX-0046 (Supplementary Table 2b). Further docking studies indicate that vepafestinib, pralsetinib and selpercatinib are likely to be type 1 inhibitors, based on predicted binding modes (Extended Data Fig. 3c).

Vepafestinib blocks growth and signal transduction

Serum-starved cells were treated with 5, 50 or 500 nM inhibitor for 2 h, and then protein phosphorylation levels were examined (Fig. 3a). Exposure of LUAD-0002AS1 (NSCLC, KIF5B–RET), ECLC5B (NSCLC, tripartite motif-containing 33 (TRIM33)–RET) and TT cells (medullary thyroid carcinoma, RETC634W) to vepafestinib resulted in efficient downregulation of RET phosphorylation at Y905 and Y1062 and downstream effectors. Near-complete inhibition of phosphorylation was achieved with 50 nM vepafestinib, similar to results with selpercatinib and pralsetinib. Vandetanib was less effective. We performed additional dose–response western blotting studies using lower inhibitor concentrations. Immunoblots were quantitated by densitometry, and the EC50 for phosphorylation inhibition was estimated (Extended Data Fig. 4). We confirmed that vepafestinib was as effective as selpercatinib and pralsetinib at inactivating RET signaling in LUAD-0002AS1 (Extended Data Fig. 4a) and TT (Extended Data Fig. 4b) cells. Quantitation of immunoblots is shown in Extended Data Fig. 4c,d.

Fig. 3: Vepafestinib inhibits transmission of signals and blocks growth of cells with RET alterations.

a, LUAD-0002AS1, ECLC5B and TT cells were serum starved for 24 h and then treated with the indicated concentrations of vepafestinib (TAS0953/HM06), selpercatinib, pralsetinib or vandetanib for 2 h. Following treatment, whole-cell extracts were prepared and subjected to western blotting analysis. Representative immunoblots from two independent experiments are shown. GAPDH was used as a loading control. RSK, ribosomal protein S6 kinase; S6RP, S6 ribosomal protein. b,c, Cells were plated in 96-well plates and treated for 96 h with the inhibitors shown. The number of viable cells was assessed using alamarBlue. b, Viability curves for control HBEC cells (HBECp53-EV) and HBEC cells with the CCDC6-RET fusion (HBECp53-RET) are shown at the left. Results are the mean ± s.e.m. of four independent experiments. Data were analyzed by non-linear regression, and IC50 values were estimated by curve fitting. A heatmap of the IC50 values is shown on the right. Missing values indicate that the experiment was not done. c, Viability curves for LUAD-0002AS1 (n = 3), ECLC5B (n = 3) and TT (n = 5) cells. Results are mean ± s.e.m. Each condition was assayed in triplicate for all viability studies.

Source data

Next, we examined the efficacy of vepafestinib in blocking growth of 12 tumor cell lines (patient -derived and isogenic) that are models of RET fusions or RET mutations found in NSCLC and thyroid cancers and three nontumor cell lines. Vepafestinib effectively inhibited growth of HBECp53-RET (CCDC6-RET fusion; IC50 = 60 nM) but had little effect on the isogenic control HBECp53-EV cells at concentrations <1,000 nM (IC50 = 7,905 nM) (Fig. 3b). This result was comparable to those obtained with pralsetinib and selpercatinib (Extended Data Fig. 5a,b). Similarly, vepafestinib inhibited growth of LUAD-0002AS1 cells (Fig. 3c and Extended Data Fig. 5b) and Ba/F3 cells expressing RET fusions (KIF5B–RET, CCDC6–RET, CCDC6–RETS904F)40 or the RETM918T mutation (Extended Data Fig. 5c). Vepafestinib was more effective at inhibiting growth of all tumor cell lines than vandetanib and as effective as selpercatinib and pralsetinib (Fig. 3b, right, Fig. 3c and Extended Data Fig. 5). No RET inhibitor showed preference toward any of the three RET fusions in our study. The nontumor cholangiocyte cell line MMNK1 was more sensitive to selpercatinib, pralsetinib and vandetanib than to vepafestinib (Extended Data Fig. 5b).

Vepafestinib modulates growth and survival pathways

To gain further insight into the mechanism by which vepafestinib inhibited growth, we assessed expression of markers of cell cycle progression and apoptosis in cells treated with inhibitors. In LUAD-0002AS1 cells, vepafestinib caused almost complete inhibition of RET, AKT, S6, ERK1 and ERK2 phosphorylation after 6 h of treatment, and this was maintained for up to 24 h (Fig. 4a). Similar results were obtained with TT cells. Sustained treatment of LUAD-0002AS1 and TT cells with vepafestinib and other RET-selective inhibitors resulted in downregulation of the cell cycle regulator cyclin D1 and increased expression of the cell cycle inhibitor p27. Treatment of LUAD-0002AS1 cells (p53 mutant) with vepafestinib resulted in downregulation of the cell cycle inhibitor p21; however, the opposite was observed in TT cells (p53 WT). Expression of apoptosis markers such as cleaved PARP (c-PARP), BIM and PUMA was induced in all cell lines by 6 h. The results obtained with vepafestinib were similar to those obtained with selpercatinib and pralsetinib. Vandetanib was less effective at blocking expression of cyclin D1 and increasing expression of cell cycle inhibitors and pro-apoptotic proteins (Fig. 4a). Exposure to vepafestinib resulted in dose-dependent increases in caspase 3 and 7 activity in the five lung cancer cells tested (Fig. 4b, LUAD-0002AS1, TT, ECLC5B; Extended Data Fig. 6, LC-2/ad, LUAD-0087AS2). The degree of caspase 3 and 7 stimulation by vepafestinib was similar to that observed with selpercatinib and pralsetinib treatment.

Fig. 4: Vepafestinib modulates expression of cell cycle and apoptosis markers.

a, LUAD-0002AS1 and TT cells were serum-staved for 24 h and then treated with 100 nM vepafestinib (TAS0953/HM06), selpercatinib, pralsetinib or vandetanib for 24 h. Following treatment, whole-cell extracts were prepared and subjected to western blotting analysis. Representative immunoblots from two independent experiments are shown. GAPDH was used as a sample-processing control. b, Cells were treated with the indicated RET inhibitors for 48 h before measuring caspase 3 and 7 enzymatic activity in cell homogenates. Results represent the mean ± s.d. of two independent experiments in which each condition was assayed in triplicate.

Source data

Vepafestinib blocks growth of RET fusion models in vivo

We next examined vepafestinib efficacy in vivo. Mice implanted with NIH-3T3-RET (NIH-3T3 cells expressing CCDC6-RET fusion complementary DNA), ECLC5B or LC-2/ad (CCDC6-RET) cells, or LUAD-0057AS1 (CCDC6-RET) patient-derived xenograft (PDX) tumors were treated with various dosages of vepafestinib, or vandetanib or cabozantinib (Fig. 5). Cabozantinib was used as a control drug for LUAD-0057AS1 cells, as the PDX model was derived from tumor tissue of a patient with poor response to cabozantinib. Tumor growth is shown on the left; area under the curve (AUC) analysis is shown in the middle; the percent change in individual tumor volume from baseline is shown on the right (Fig. 5). Administration of vepafestinib resulted in a dose-dependent decrease in growth of NIH-3T3-RET xenograft tumors (Fig. 5a, left), with all dosages of vepafestinib tested resulting in a significant reduction in tumor volume (Fig. 5a, middle). There was no statistically significant reduction in animal weight for any of the treatments (Fig. 5a, right). Similarly, vepafestinib treatment resulted in a significant reduction in LC-2/ad tumor growth, with substantial tumor regression observed with the 50 mg per kg twice daily (BID) dosage (Extended Data Fig. 7a). There was no statistically significant reduction in animal weight with any vepafestinib dosage (Extended Data Fig. 7b). Vepafestinib treatment caused significant reductions in ECLC5B xenograft tumor growth (Fig. 5b, left), with 50 mg per kg BID and 100 mg per kg once daily (QD) dosing resulting in 100% ± 0% and 90.3% ± 4% tumor regression, respectively. Vandetanib treatment inhibited tumor growth significantly (P < 0.0001), with all tumors shrinking (Fig. 5a, left). However, vandetanib-treated animals showed significant weight loss (P = 0.01) and were killed early. No dosage of vepafestinib had any adverse effect on animal health or animal weight (P > 0.05) (Extended Data Fig. 8a). Treatment of mice bearing LUAD-0057AS1 PDX tumors with vepafestinib also resulted in significant reductions in tumor volume (Fig. 5c, left). Tumors shrank by 44% ± 3% and 48% ± 1% when treated with 50 mg per kg BID or 100 mg per kg QD vepafestinib, respectively. As expected in this model, cabozantinib slowed growth but did not lead to any tumor shrinkage at a dosage that has been shown to completely inhibit growth of RET fusion-driven xenograft tumors (30 mg per kg QD)41, while vandetanib and vepafestinib treatment caused substantial tumor regression (Fig. 5c, middle and right). Vandetanib (50 mg per kg QD) caused a significant reduction in animal weight (P = 0.0015) (Extended Data Fig. 8b). No dosage of vepafestinib or the other RET-selective inhibitors had any adverse effect on animal health or animal weight (P > 0.05) (Extended Data Fig. 8). These results suggest that vepafestinib is effective at reducing tumor growth, including in a model that was refractory to cabozantinib.

Fig. 5: Efficacy of vepafestinib in RET fusion-dependent disease models in vivo.

Cell lines (NIH-3T3 expressing CCDC6–RET, ECLC5) or PDX tumors were implanted into subcutaneous flanks of female mice and treated as indicated. a, NIH-3T3-RET xenograft (athymic nude mice). b, ECLC5 xenograft (NOD–SCID gamma (NSG) mice). c, LUAD-0057AS1 PDX. a–c, Left, time course of treatment. Data represent mean ± s.e.m. There were five (NIH-3T3-RET and ECLC5 xenografts) or eight (LUAD-0057AS1) animals per group. a–c, Middle, AUC analysis of tumor growth. Data represent mean ± s.e.m. of n = 12 (NIH-3T3-RET), n = 32–44 (ECLC5) or n = 46–49 (LUAD-0057AS1) values per group. a, Right, animal weight. b,c, Right, percent change in the volume of individual tumors at the end of the study. Mean ± s.e.m. are shown. The volume of tumors in all treatment groups in each model was significantly lower than that of the respective vehicle-treated groups (P < 0.0001). P values for statistical significance are shown for other comparisons (ANOVA with Dunnett’s multiple-comparison test). All tests were two sided.

Source data

We expanded our efficacy studies to include two additional NSCLC PDX models with RET fusions. We compared vepafestinib to selpercatinib and pralsetinib, both of which have been shown to inhibit growth of RET fusion-driven tumors in vivo at dosages of 10 mg per kg BID or less17,18. Vepafestinib treatment also caused significant reductions in tumor growth in LUAD-0087AS2 PDX (Fig. 6a) and LUAD-0077AS1 PDX (Fig. 6b) models. None of the RET-selective inhibitors caused any change in animal health or weight (P > 0.05) (Extended Data Fig. 8c–i). In a Ba/F3 KIF5B–RET allograft tumor model, 50 mg per kg BID vepafestinib was as efficacious as 30 mg per kg selpercatinib and 60 mg per kg pralsetinib in reducing tumor burden (Extended Data Fig. 9).

Fig. 6: Efficacy of vepafestinib compared to other RET-selective inhibitors in PDX models.

a, LUAD-0087AS2 PDX. b, LUAD-0077AS1 PDX. a,b, Left, time course of treatment. Data represent mean ± s.e.m. There were five mice in each group in both models. a,b, Middle, AUC analysis of tumor growth. Data represent mean ± s.e.m. of n = 56 (LUAD-0087AS2) or n = 32 (LUAD-0077AS1) values per group. a,b, Right, percent change in the volume of individual tumors at the end of the study. Mean ± s.e.m. are shown. Each group consisted of five animals. The volume of tumors in all treatment groups in each model was significantly lower than that of the respective vehicle-treated groups (P < 0.0001). P values for significance are shown for other comparisons (ANOVA with Dunnett’s multiple-comparison test). All tests were two sided.

Source data

RETG810R in vivo models remain susceptible to vepafestinib

To address vepafestinib potency against RETG810R in vivo, we examined the ability of the drug to block growth of Ba/F3 KIF5B–RETWT or Ba/F3 KIF5B–RETG810R allograft tumors. Treatment of Ba/F3 KIF5B–RETWT allograft tumors with vepafestinib (12.5, 25, 50 mg per kg BID) resulted in dose-dependent inhibition of tumor growth (Fig. 7a) without any body weight changes (Extended Data Fig. 8e). To assess target engagement in vivo, tumor-bearing animals were given a single dose of vepafestinib (50 mg per kg), and then tumors were extracted at various time points. Western blot analysis showed that vepafestinib completely inhibited phospho-RET and phospho-ERK for at least 8 h after drug administration (Fig. 7b). At an equivalent dosage (10 mg per kg BID), vepafestinib was more effective than selpercatinib and pralsetinib at slowing growth of Ba/F3 KIF5B–RETG810R allograft tumors (Fig. 7c). The identical dosage of selpercatinib and pralsetinib, however, caused substantial reduction in growth of Ba/F3 KIF5B–RETWT tumors (Extended Data Fig. 9a,b). Administration of 50 mg per kg BID vepafestinib had a significant anti-tumor effect on Ba/F3 KIF5B–RETG810R tumors without any animal body weight changes (Fig. 7d and Extended Data Fig. 8f,g). Consistent with the anti-tumor activity, vepafestinib completely inhibited RETG810R phosphorylation in tumors treated with doses of 10 mg per kg and 30 mg per kg (Fig. 7e). Although the highest dosage of selpercatinib and pralsetinib (30 mg per kg BID) showed moderate anti-tumor effect against Ba/F3 KIF5B–RETG810R allograft tumors (Fig. 7d), there was not a commensurate decrease in phosphorylation of the RETG810R mutant (Fig. 7e), suggesting that these effects may be due to off-target effects.

Fig. 7: Anti-tumor activity of vepafestinib against KIF5B–RETG810R-driven allograft tumors.

a, Animals bearing Ba/F3 KIF5B–RETWT allograft tumors were treated with vehicle (n = 6) or the indicated dosages of vepafestinib (n = 6). b, Animals bearing Ba/F3 KIF5B–RETWT tumors were treated with a single dose of 50 mg per kg vepafestinib, and then tumors were collected at the indicated time points after inhibitor administration for western blotting analysis. Representative immunoblots on which two tumors from each condition were examined are shown. c,d, Mice bearing Ba/F3 KIF5B–RETG810R xenograft tumors were administered vepafestinib (n = 5), selpercatinib (n = 5), pralsetinib (n = 5) or vehicle (n = 5) orally at the indicated dosages BID for 14 d (days 1–14) after grouping. e, Mice bearing Ba/F3 KIF5B–RETG810R allograft tumors were administered 10 or 30 mg per kg vepafestinib, selpercatinib or pralsetinib, and then tumors were collected 1 h later for western blot analysis. Representative immunoblots on which two tumors from each condition were examined are shown. Tumor volume for each dosing group was measured and shown as mean ± s.e.m. Statistical analysis was performed using Dunnett’s test (vehicle versus vepafestinib, selpercatinib or pralsetinib) or Tukey’s test (vepafestinib versus selpercatinib or pralsetinib), and P values are shown. All tests were two sided. GAPDH was used as a loading control in b,e.

Source data

Vepafestinib exhibits high CNS availability

We designed vepafestinib to have enhanced blood–brain barrier (BBB) penetration and retention. Here, we assessed pharmacological and pharmacokinetic properties of vepafestinib, including membrane permeability, transport by efflux transporters and brain penetrance. The key pharmacological characteristics of vepafestinib, selpercatinib and pralsetinib are illustrated in Fig. 8a. The three RET inhibitors showed excellent membrane permeability but different susceptibility to efflux transporters. MDR1 (P-glycoprotein; P-gp) and breast cancer resistance protein (BCRP) are two major efflux transporters expressed at the BBB, where they prevent entry of many endogenous substances and chemicals into the CNS42. Vepafestinib showed low net flux ratio for P-gp and BCRP (Fig. 8a). By contrast, selpercatinib and pralsetinib were higher affinity substrates for P-gp; selpercatinib also showed slight substrate susceptibility for BCRP. Substances with Kp,uu,brain value > 0.3 in mice are regarded as favorable brain-penetrating agents43,44. Vepafestinib showed relatively high Kp,brain and Kp,uu,brain values in mice (1.8 and 1.3, respectively), while the values for selpercatinib and pralsetinib were <0.3 in mice. We also examined the same parameters for TPX-0046 and found that this compound was a substrate for P-gp and is expected to have poor BBB permeability based on its Kp,uu,brain of 0.077 (Supplementary Table 2c). These results indicate that vepafestinib concentrations in the brain would be better maintained than those of selpercatinib, pralsetinib and TPX-0046.

Fig. 8: Vepafestinib is more effective than selpercatinib at penetrating the brain and blocking intracranial tumor growth.

a,b, Pharmacokinetic properties. a, *Apparent permeability coefficient (Papp) values were calculated as the mean of Papp values in the apical-to-basal direction in mock-transfected LLC-PK1 cells. †,‡Total (Kp,brain) and unbound (Kp,uu,brain) brain/plasma concentration ratios were calculated based on total and unbound concentrations in plasma and brain at 0.5 h or 1 h after oral administration of each agent to male BALB/c mice dosed with 50 mg per kg drug. Unbound fractions in plasma (fu,plasma) and brain (fu,brain) were obtained by the equilibrium dialysis method with plasma and brain homogenate. §,¥Net flux ratio (NFR) values for MDR1 (P-gp) and BCRP were obtained from transcellular transport assays using control or MDR1-expressing LLC-PK1 cells and control and BCRP-expressing MDCK II cells. b, Single-dose vepafestinib (3 mg per kg, 10 mg per kg or 50 mg per kg) was administered orally to male Han Wistar rats at time = 0 min (n = 12 per dosing group). Following equilibration, samples were collected at the indicated time points, and vepafestinib concentrations were then determined. Data for all dosages are shown in Extended Data Fig. 10. Data represent mean ± s.e.m. (n = 4 independent measurements in four animals). c, NIH-3T3 CCDC6-RET cells harboring a luciferase reporter were implanted intracranially into nude mice and treated with vehicle or 50 mg per kg vepafestinib BID. Treatment started 5 d after implantation. Bioluminescence images of animals 13 d after implantation are shown (left). Survival curves of each group are shown after implantation (n = 10, vehicle group; n = 7, vepafestinib group) (right). There was a significant difference in survival between the vehicle group and the vepafestinib group (P = 0.0016, log-rank test). d, ECLC5 cells labeled with a luciferase reporter were implanted intracranially into NSG mice and treated with vehicle, selpercatinib (10 mg per kg) or vepafestinib (50 mg per kg) BID. Treatment started 10 d after implantation. There were six animals in each group. d, Bioluminescence images of animals are shown for the last day when all animals were alive in the three groups (43 d after implantation) and at 92 d after implantation for the two treatment arms. e, Luciferase signals were quantified and are shown (left). Data represent mean ± s.e.m. (n = 6 per group). AUC analysis was performed for the selpercatinib and vepafestinib groups (middle, Brown–Forsythe and Welch ANOVA tests). For AUC, data represent mean ± s.e.m. of n = 100 (vepafestinib) or n = 65 (selpercatinib) values. Survival curves are shown for animals after treatment began (right). Treatment with selpercatinib (P = 0.0008, log-rank test) and vepafestinib (P = 0.0008, log-rank test) increased survival relative to the vehicle. However, animals treated with vepafestinib had longer survival (P = 0.001, log-rank test). All statistical tests were two sided.

Source data

We characterized the pharmacokinetics of vepafestinib in the prefrontal cortex (PFC), cerebrospinal fluid (CSF) and plasma of freely moving adult male Han Wistar rats following single-dose oral administration at 3, 10 and 50 mg per kg (Fig. 8b and Extended Data Fig. 10a,b). Once equilibrium was achieved between the compartments, the ratio of the observed concentrations of vepafestinib in microdialysates from the PFC, CSF and plasma-free fraction was close to 1:1:1. The concentrations were maintained from 2 h to 6.5 h after vepafestinib administration (up to 8 h for CSF) (Fig.