Research ArticleCancer

Discovery and pharmacological characterization of AZD3229, a potent KIT/PDGFRα inhibitor for treatment of gastrointestinal stromal tumors

See allHide authors and affiliations

Science Translational Medicine  29 Apr 2020:
Vol. 12, Issue 541, eaaz2481
DOI: 10.1126/scitranslmed.aaz2481

Getting the GIST of treatment

Gastrointestinal stromal tumor, or GIST, can be targeted by inhibiting KIT and platelet-derived growth factor α (PDGFRα). Unfortunately, the therapeutic effects do not always last, as the tumors develop resistance mutations that are not susceptible to existing drugs. Banks et al. report the design of AZD3229, a compound that combines selectivity for KIT and PDGFRα with broad activity against a variety of disease-causing mutations. Its selectivity decreases the risk of off-target toxicity, which has been a problem for multiple other KIT/PDGFRα inhibitors. AZD3229 showed promising activity and safety in a range of drug-resistant GIST mouse models, suggesting its potential for clinical translation.


Gastrointestinal stromal tumor (GIST) is the most common human sarcoma driven by mutations in KIT or platelet-derived growth factor α (PDGFRα). Although first-line treatment, imatinib, has revolutionized GIST treatment, drug resistance due to acquisition of secondary KIT/PDGFRα mutations develops in a majority of patients. Second- and third-line treatments, sunitinib and regorafenib, lack activity against a plethora of mutations in KIT/PDGFRα in GIST, with median time to disease progression of 4 to 6 months and inhibition of vascular endothelial growth factor receptor 2 (VEGFR2) causing high-grade hypertension. Patients with GIST have an unmet need for a well-tolerated drug that robustly inhibits a range of KIT/PDGFRα mutations. Here, we report the discovery and pharmacological characterization of AZD3229, a potent and selective small-molecule inhibitor of KIT and PDGFRα designed to inhibit a broad range of primary and imatinib-resistant secondary mutations seen in GIST. In engineered and GIST-derived cell lines, AZD3229 is 15 to 60 times more potent than imatinib in inhibiting KIT primary mutations and has low nanomolar activity against a wide spectrum of secondary mutations. AZD3229 causes durable inhibition of KIT signaling in patient-derived xenograft (PDX) models of GIST, leading to tumor regressions at doses that showed no changes in arterial blood pressure (BP) in rat telemetry studies. AZD3229 has a superior potency and selectivity profile to standard of care (SoC) agents—imatinib, sunitinib, and regorafenib, as well as investigational agents, avapritinib (BLU-285) and ripretinib (DCC-2618). AZD3229 has the potential to be a best-in-class inhibitor for clinically relevant KIT/PDGFRα mutations in GIST.


Gastrointestinal stromal tumors (GISTs) are soft tissue sarcomas of the gastrointestinal tract with estimated 3300 to 6000 new cases per year in the United States (1, 2). The pathobiology of GISTs is widely investigated, and about 70 to 80% of them are associated with gain-of-function mutations in the gene KIT (3). Under physiological conditions, binding of the KIT cognate ligand, stem cell factor, to the extracellular domain of the receptor results in receptor dimerization, intermolecular autophosphorylation of specific tyrosine residues, and kinase activation (4). Mutations in KIT result in ligand-independent phosphorylation of the receptor and activation of RAS/extracellular signal–regulated kinase (ERK) and phosphatidylinositol 3-kinase (PI3K)/AKT pathways, which play roles in cell differentiation, survival, and oncogenesis (5).

Imatinib, a small-molecule inhibitor of KIT and platelet-derived growth factor α (PDGFRα), is an effective frontline therapy for patients with advanced GIST (6, 7). Imatinib achieves partial responses or stable disease in nearly 80% of patients with GIST, with a median time to disease progression of 18 months (7). Unfortunately, most patients develop resistance to imatinib through acquisition of secondary mutations in the KIT kinase domain (8, 9). These mutations tend to be single amino acid substitutions in the adenosine 5′-triphosphate (ATP)–binding pocket (exons 13/14), interfering with drug binding, or in the activation loop (exons 17/18), stabilizing KIT in the active conformation and hindering drug interaction (9).

The subsequent treatments for patients who do not tolerate imatinib or whose tumors are imatinib resistant are multikinase inhibitors sunitinib (10) and regorafenib (11). Sunitinib is ineffective against KIT A-loop mutations, which account for 50% of imatinib-resistant mutations (12), and regorafenib treatment is only modestly beneficial (13). These drugs demonstrate low overall response rates (ORRs) of 7 and 4.5%, with median time to tumor progression of 6.2 and 4.8 months, respectively (10, 13). Moreover, tolerability of sunitinib and regorafenib, owing to their broader kinase inhibition profile, is challenging and often leads to dose modifications and drug holidays, compromising efficacy (11, 14). These agents show potent inhibition of vascular endothelial growth factor receptor 2 [VEGFR2; kinase insert domain receptor (KDR)], associated with high-grade hypertension, along with other toxicities (11, 14). Beyond these approved lines of treatment, patients with GIST have an unmet medical need.

Although most GISTs are attributed to mutations in KIT, 12 to 14% of patients have activating mutations in PDGFRα, D842V substitution being most common (15). Although imatinib is effective against PDGFRα primary mutations (V561D), the most common resistance mutation, D842V, is strongly resistant to inhibition by imatinib and sunitinib (16, 17). Patients with PDGFRα D842V–mutant GIST treated with conventional GIST tyrosine kinase inhibitors (TKIs) have low rates of clinical benefit, with median overall survival of about 1 year, compared with the median of 4 to 5 years for patients with KIT exon 11–mutant GIST (17).

Therefore, additional agents capable of overcoming primary and secondary drug resistance mutations in KIT/PDGFRα with superior selectivity profiles have potential therapeutic utility in drug-resistant GISTs. Here, we report the development and pharmacological characterization of AZD3229, a next-generation KIT/PDGFRα inhibitor, specifically designed to target a spectrum of primary and drug-resistant KIT/ PDGFRα mutations observed in GIST with a greater selectivity for KIT versus VEGFR2. While characterizing AZD3229, we have seen the development of other structurally distinct inhibitors of KIT/PDGFRα kinases, notably avapritinib (18) and ripretinib (19). We demonstrate that AZD3229 has superior KIT potency and selectivity relative to standard of care (SoC)/investigational agents in vitro and in mouse xenograft and human patient-derived xenograft (PDX) models of GIST. We believe that AZD3229 will demonstrate utility in patients with both KIT and PDGFRα mutations.


Development of AZD3229, a potent and selective inhibitor of KIT/PDGFRα

There is extensive evidence in the literature, emphasizing the importance of maintaining continuous suppression of KIT/PDGFRα kinase activity to delay disease progression and achieve optimal clinical outcomes in GIST (20). Dose interruptions and modifications, as in the case of sunitinib and regorafenib due to off-target activity, specifically against VEGFR2, result in the generation of resistant subclones, thereby compromising efficacy. We therefore designed a medicinal chemistry strategy for targeting KIT while avoiding inhibiting VEGFR2, which was challenging due to high homology between the two kinases (56% identity in the kinase domain) (21). Using structure-based drug design, we tried to optimize a single molecule with a pan-inhibitory activity for both KIT and PDGFRα mutations in primary and resistant disease with superior potency and selectivity to SoC agents (Fig. 1A). In brief, we used a series of phenoxyquinazoline- and quinoline-based inhibitors of the tyrosine kinase PDGFRα as scaffolds, with a view that inhibitors of PDGFRα might show inhibition of some KIT mutants due to the close homology between KIT and PDGFRα. On these PDGFRα scaffolds, we performed iterative medicinal chemistry optimization (21) to improve anti-KIT activity, selectivity, and pharmaceutical properties, ultimately arriving at the quinazoline acetamide, AZD3229 (Fig. 1B). AZD3229, containing a C7 methoxyethoxy group, has the optimal balance between potency against mutant KIT and VEGFR2 selectivity (21). To rationalize the selectivity profile of AZD3229, we obtained a high-resolution crystal structure of the KIT kinase domain in complex with AZD3229 (Protein Data Bank accession code: 6GQM) (21). AZD3229 binds to the hinge region of KIT via the quinazoline moiety, which also makes a water-mediated interaction with the gatekeeper residue (Thr670) of KIT. It also makes a water-mediated interaction from the amide to conserved lysine and glutamic acid residues (Lys623 and Glu640) on KIT, enhancing selectivity. The binding mode of AZD3229 is consistent with type 2 kinase inhibition, with the Asp-Phe-Gly (DFG)–out fold of the kinase placing Phe811 into a stacking pose with the central ring of AZD3229. Selectivity for KIT over VEGFR2 arises from two key motifs: first, the water-mediated gatekeeper interaction with Thr670 of KIT is not available to the hydrophobic valine residue of VEGFR2 (Val916), and second, the triazole motif, located in the DFG-out pocket, has a local dipole moment, which stabilizes bound waters in KIT to a greater extent than in VEGFR2 (Fig. 1C). Despite high homology between KIT and antitarget VEGFR2, we propose that subtle differences in the way each is able to accommodate and, in particular solvate, the liganded structures are key to the enhanced selectivity of the compound.

Fig. 1 Development of AZD3229, a selective inhibitor of KIT and PDGFRα.

(A) Target profile for the design of pan-KIT/PDGFRα inhibitor. (B) Chemical structure of AZD3229. (C) Cocrystal structure of AZD3229 bound to KIT. (D) Inhibition of cell viability by AZD3229 and SoC/investigational agents in Ba/F3 cells expressing receptor tyrosine kinases (RTKs) of interest. GI50 values (nM) represent means ± SD of three independent experiments (data file S1). (E) Kinome tree depiction of AZD3229 selectivity in Thermo Fisher Scientific kinase activity panel compared with GIST SoC/investigational agents, at 1 μM produced with Kinome Render (kinome tree illustrations reproduced courtesy of Cell Signaling Technologies Inc.). The big and small blue circles represent 100 and 15% inhibition, respectively.

During chemistry optimization, we relied heavily on cellular assays, because biochemical assays do not encompass the physiological complexity and regulatory circuits that modulate kinase function in the presence of the cellular milieu. A range of cellular models, including the Ba/F3 cell model (22) expressing the wild type and mutated versions of KIT and PDGFRα, was used to profile compounds from the medicinal chemistry effort. The Ba/F3 cellular model has been demonstrated in several studies to mimic clinical efficacy of KIT inhibitors (23, 24). AZD3229 and the SoC/investigational agents were profiled in viability assays of Ba/F3 cell lines expressing KIT, PDGFRα, or VEGFR2. AZD3229 demonstrated single-digit nanomolar potency for KIT [concentration of the compound causing 50% inhibition of cell growth (GI50), 1 nM] and PDGFRα (GI50, 3 nM), with a 400- to 1000-fold margin over VEGFR2 (GI50, 1378 nM), resulting in a broad therapeutic window. AZD3229 was more potent than SoC agents and investigational agents while maintaining selectivity over VEGFR2 (Fig. 1D). Sunitinib, regorafenib, and ripretinib inhibited VEGFR2 with various degrees of potency, as reported in (10, 11, 19). To ensure that the potent KIT activity of AZD3229 had not been achieved at the expense of general kinome selectivity, we assessed the selectivity of AZD3229 in a panel of 379 human kinases (Thermo Fisher Scientific) alongside SoC/investigational agents. Besides inhibiting KIT and KIT-related kinases, AZD3229 inhibited 11 additional kinases by >60% at 1 μM (table S1). Its overall kinase selectivity profile was similar to imatinib and superior to sunitinib, regorafenib, avapritinib, and ripretinib when tested in a kinase activity panel (Fig. 1E).

AZD3229 is more potent than imatinib and has broader activity than imatinib in inhibiting KIT primary mutations seen in GIST

Imatinib as a first-line treatment for GIST is efficacious against most primary KIT mutations with the exception of KIT exon 9 insertion. Seventy percent of all primary mutations in KIT affect the juxtamembrane region (exon 11), composed predominantly of in-frame deletions, substitutions, and insertions. We chose KIT exon 11 deletion as the primary mutation backbone in the Ba/F3 cell line for generation of all secondary mutations because two-thirds of GISTs harbor mutations in exon 11, specifically deletions (557–558), associated with malignant behavior (25). We also generated Ba/F3 cell lines with exon 11 substitution V560D and exon 9 insertion AY502-503. Exon 9 insertions account for 10 to 15% of KIT mutations in patients and show an innate resistance to imatinib and aggressive clinical behavior (26). AZD3229 was designed to inhibit all the primary mutations in KIT.

AZD3229 was profiled alongside imatinib and SoC/investigational agents in viability assays of Ba/F3 KIT exon 11 V560D, KIT exon 9 insertion, and KIT exon 11 deletion (557–558). AZD3229 inhibited growth of these Ba/F3 cell lines with GI50 values of 2.7, 3.1, and 0.7 nM, respectively, with 26- to 64-fold increase in potency relative to imatinib (Fig. 2A). AZD3229 was 2- to 10-fold more potent than sunitinib and 30- to 50-fold more potent than regorafenib. AZD3229 was also 8- to 110-fold more potent than avapritinib and ripretinib in inhibiting KIT primary mutations.

Fig. 2 AZD3229 is more potent and has broader activity than imatinib in inhibiting KIT primary mutations.

(A) Effect of AZD3229 or SoC/investigational agents on viability of Ba/F3 cells expressing KIT primary mutations (KIT V560D, KIT exon 9 insertion, and KIT exon 11 deletion). Data represent mean of three independent experiments ± SD. (B) Inhibition of KIT signaling in Ba/F3 cells expressing KIT exon 9 insertion upon treatment with AZD3229, imatinib, or sunitinib for 4 hours. (C) Inhibition of KIT signaling in GIST-T1 cells [KIT Ex11 Deletion (560–578)] after treatment with AZD3229 or imatinib for 4 hours. (D) Cell viability data comparing AZD3229 and SoC/investigational agents in GIST-T1 cells. Data represent mean of two or three independent experiments ± SD. (E) Effect of oral doses of AZD3229 on body weight of CB17 SCID mice. (F) Antitumor activity of AZD3229 in a KIT exon 11 deletion PDX model. (G) Tolerability of AZD3229 or imatinib in a KIT exon 11 deletion PDX model. (H) PD analysis of KIT signaling upon dosing with AZD3229 in a PDX model with KIT exon 11 deletion. Free compound exposure in micromolar is also shown. Compound is below detection limit at time 16 h (4 mg/kg) and 24 h (20 mg/kg).

AZD3229 inhibited KIT exon 9 insertion mutation with a GI50 of 3.1 nM and was 65, 60, and 10 times more potent than imatinib, avapritinib, and ripretinib, respectively, in inhibiting growth of the Ba/F3 KIT exon 9 insertion cells (Fig. 2A). AZD3229 also potently inhibited KIT phosphorylation in these cells with IC50 (concentration of the compound causing 50% inhibition of phosphorylation) of 1 nM and concomitant inhibition of signaling downstream of KIT, namely, inhibition of phosphorylation of ERK/AKT. Sunitinib showed activity in this cell line as expected, whereas imatinib did not inhibit KIT phosphorylation in this cell line up to 100 nM (Fig. 2B). AZD3229 was also more potent than ripretinib and avapritinib in inhibiting KIT signaling in this cell line (fig. S1).

We next investigated the potency of AZD3229 in GIST-T1, a physiologically relevant GIST cell line with a primary KIT exon 11 deletion (560–578) (27). In GIST-T1 cells, AZD3229 was more potent than SoC and investigational agents in inhibiting KIT signaling (Fig. 2C and fig. S2) and cell growth (Fig. 2D). We next tested the antitumor activity of AZD3229 in a GIST PDX model with KIT exon 11 deletion (K550fs) primary mutation. The target dose for in vivo study was established from a tolerability study in CB17 severe combined immunodeficient (SCID) mice. In this study, oral doses of AZD3229 ranging from 0.1 to 40 mg/kg, twice daily (BID) had no effect of the compound on the health or body weight of animals (Fig. 2E). AZD3229 at a single dose of 20 mg/kg reached a maximum concentration (protein unbound) of 395 nM in plasma of mice (fig. S3). This plasma concentration is well above the concentration required to inhibit several KIT primary mutations (Fig. 2A) and was chosen as a target dose for in vivo studies. AZD3229 at doses of 4 and 20 mg/kg resulted in robust tumor regressions in the PDX model with KIT exon 11 deletion (Fig. 2F). Imatinib dosed at 300 mg/kg once daily (QD) also showed tumor regression, consistent with clinical efficacy in KIT exon 11 mutant population. After the dosing was discontinued, we followed the mice for 258 days to monitor for regrowth. Mice on both doses of AZD3229 maintained tumor regressions for an extended period, even after the dosing was discontinued, suggesting cytotoxic nature of the drug candidate (Fig. 2F). Mice dosed with AZD3229 were healthy, with no change in body weight or signs of toxicity when monitored for up to 258 days after implant (Fig. 2G). In a pharmacodynamic (PD) study, a single dose of AZD3229 at 20 mg/kg resulted in potent inhibition (>90%) of KIT phosphorylation and signaling downstream of KIT, correlating with the antitumor activity reported above (Fig. 2H and fig. S4).

AZD3229 inhibits imatinib-resistant KIT ATP-binding pocket secondary mutations

We further investigated the potency of AZD3229 in inhibiting imatinib-resistant secondary mutations of KIT in ATP-binding pocket. AZD3229 was profiled in a panel of Ba/F3 cells expressing V654A or T670I (gatekeeper residue), the two most common ATP-binding mutations found in about 50% of patients with GIST in the resistant population after imatinib treatment (9). These mutations were generated in the background of several clinically relevant primary mutations. When tested in a viability assay, AZD3229 potently inhibited growth of these cells with nanomolar potency (4 to 23 nM), whereas imatinib remained completely inactive (Fig. 3A). AZD3229 was largely equipotent with sunitinib, which has shown clinical efficacy in this mutant population (28). In concordance with clinical data, regorafenib was not potent, especially for V654A mutation (29). AZD3229 was also several-fold more potent than avapritinib and ripretinib, the latter being moderately effective against V654A or T670I in the context of exon 11 primary mutation and least effective in the context of exon 9 insertion, with GI50 of 221 nM. AZD3229 was around 10 times more potent against this mutation (GI50, 23 nM). Avapritinib was not active against the ATP-binding pocket mutations (Fig. 3A), which is reflected in the clinical data of avapritinib (30). The signaling assay data in Ba/F3 KIT exon 11 deletion (557–558)/V654A cell line were similar to the viability assay data, with AZD3229 potently inhibiting KIT signaling in these cells, with single-digit nanomolar IC50, whereas imatinib remained inactive at 100 nM (Fig. 3B). Moreover, avapritinib had no effect on KIT signaling for up to 100 nM, and ripretinib was moderately active (Fig. 3B). Consistent with in vitro data, avapritinib did not show efficacy in Ba/F3 KIT exon 11 deletion (557–558)/V654A xenograft model at doses (10 and 30 mg/kg) equivalent to those being tested in the clinic (18, 30), whereas AZD3229 led to tumor regressions at both 5 and 20 mg/kg with no effect on body weight (Fig. 3C and data file S1). Sunitinib dosing of animals resulted in tumor regressions as expected (Fig. 3C). We further analyzed the effects of AZD3229 on growth and signaling in vitro and in vivo in a clinically relevant GIST cell line, GIST430/654 with KIT primary exon 11 deletion (560–578) and secondary ATP-binding pocket mutation, V654A. Whereas imatinib at 1000 nM was inactive in this cell line as expected, AZD3229 potently inhibited KIT signaling, with 10 nM of the compound completely inhibiting KIT signaling (Fig. 3D). Ripretinib showed some KIT signaling inhibition at 100 nM, whereas avapritinib was inactive in this cell line (fig. S5). In a cell viability assay, AZD3229 inhibited growth of GIST430/654 cells with GI50 of 10 nM. Sunitinib also potently inhibited the growth of this cell line as expected, whereas other SoC/investigational agents were several times less potent than AZD3229 (Fig. 3E). Decrease in cell viability by AZD3229 was accompanied by a corresponding increase in caspase-3 activity over a 3-day period in the GIST430/654 cell line (fig. S6). Time course of treatment of cells with AZD3229 shows a dose-dependent and sustained decrease in KIT signaling for up to 24 hours with a corresponding increase in the expression of the proapoptotic protein B-cell lymphoma 2 interacting mediator of cell death (BIM) and cleaved poly(adenosine 5′-diphosphate–ribose) polymerase (PARP), suggesting that cells are primed to undergo apoptosis upon treatment with AZD3229 (fig. S7). We further tested the antitumor activity of AZD3229 alongside SoC/investigational agents in a GIST430/654 GIST model in two studies (Fig. 3, F and G). Oral dosing of mice with AZD3229 led to tumor growth inhibition at the 0.5 mg/kg dose, and tumor stasis and regressions at the 5 and 20 mg/kg doses, respectively, with all doses being well tolerated (Fig. 3, F and G, and fig. S8). Dosing of sunitinib at 80 mg/kg showed tumor regressions, whereas imatinib was not efficacious as expected (Fig. 3, F and G). Ripretinib at 50 mg/kg was completely inactive in this model (Fig. 3F), and a higher dose of ripretinib could not be tested because animals were showing signs of body weight loss (fig. S8). Avapritinib tested at 10 and 30 mg/kg did not show activity in this model, but it was tolerated (Fig. 3G and data file S1), which is consistent with lack of efficacy in the clinic for avapritinib in this mutant population (30). When tested in a PD study, AZD3229 at a single dose of 4 or 20 mg/kg rapidly reduced KIT phosphorylation for over 8 hours, with a concomitant inhibition of signaling downstream of KIT at both 4 and 20 mg/kg doses, with a stronger inhibition of KIT signaling at the 20 mg/kg dose (Fig. 3H and fig. S9). A direct relationship was observed between the inhibition of KIT signaling and plasma concentration of AZD3229 (fig. S10). The effect of AZD3229 on inhibition of cell proliferation at the two dose levels was also confirmed by Ki67 staining of tumor sections after 4 days of dosing BID, where a decrease in proliferating cells was observed at both doses (4 and 20 mg/kg) (fig. S11).

Fig. 3 AZD3229 inhibits imatinib-resistant KIT ATP binding pocket secondary mutations.

(A) Effect of AZD3229 or SoC/investigational agents on cell viability in Ba/F3 cell lines expressing KIT ATP binding mutations. Data represent mean of three independent experiments ± SD. (B) AZD3229 is more potent than imatinib, ripretinib, and avapritinib in inhibiting KIT signaling in Ba/F3 KIT exon 11 deletion/V654A cells. (C) Antitumor activity of AZD3229 in a Ba/F3 KIT exon 11 deletion/V654A xenograft model. (D) AZD3229 inhibits KIT signaling in GIST430/654 cells [KIT exon 11 deletion (560–578)/V654A]. (E) GIST430/654 cells treated with SoC/investigational agents in a viability assay. GI50 values represent means ± SD of two (for avapritinib) or three (for all other compounds) independent experiments. (F) Antitumor activity of AZD3229 in a GIST model (GIST430/654) upon dosing with AZD3229, imatinib, sunitinib, or ripretinib. (G) Antitumor activity of AZD3229 in a GIST model and comparison with sunitinib or avapritinib. (H) Temporal and dose-dependent modulation of KIT signaling in a pharmacokinetic/pharmacodynamic (PK/PD) study in a GIST model after a single dose of AZD3229. Unbound plasma concentration (free exposure) of AZD3229 is also shown. (I) Antitumor activity of AZD3229 in a PDX model with KIT exon 11 deletion (557–558)/V654A. (J) PD study after dosing of AZD3229 (20 mg/kg, BID) for 4 days in a PDX model with KIT Ex 11 deletion (557–558)/V654A mutation.

We further tested the antitumor activity of AZD3229 in a GIST PDX model, derived from an imatinib-refractory lesion of a patient with KIT exon 11 deletion (557–558)/V654A mutation (fig. S12). Oral dosing of mice with 20 mg/kg of AZD3229 produced robust tumor regressions. Sunitinib also caused regressions in this model as expected (Fig. 3I). Analysis of tumor lysates collected after dosing with 20 mg/kg of AZD3229 for 4 days showed inhibition of KIT signaling, which was sustained for more than 24 hours with a simultaneous increase in the expression of BIM (Fig. 3J), suggesting an increase in apoptotic priming and cytotoxic activity, mediated by AZD3229 in this model. The cytotoxic nature of AZD3229 should prevent disease relapse in patients by maintaining continuous target suppression.

AZD3229 inhibits KIT A-loop mutations refractory to imatinib and sunitinib in GIST

The A-loop mutant population constitute about 50% of imatinib-resistant patients, and regorafenib is active against a few of the A-loop mutations (31), whereas sunitinib is inactive (32). AZD3229 was tested against a panel of Ba/F3 cell lines expressing various clinically relevant KIT A-loop mutations for inhibition of cell viability. These cell lines express the KIT resistance mutations D816H, D820A, N822K, Y823D, and A829P in the background of several KIT primary mutations. AZD3229 potently inhibited viability of all seven cell lines tested and was found to be more potent than any of the SoC/investigational agents, resulting in GI50 values of less than 20 nM for most of the mutants tested and a 6- to 38-fold greater potency than regorafenib, which shows efficacy in patients with KIT A-loop mutations (Fig. 4A).

Fig. 4 AZD3229 inhibits imatinib-resistant KIT A-loop mutations.

(A) Viability assay data in Ba/F3 lines expressing KIT A-loop mutations treated with AZD3229 or SoC agents/investigational agents. GI50 values represent means ± SD of three independent experiments. Antitumor activity of AZD3229 in Ba/F3 KIT exon 11 deletion (557–558)/D816H xenograft model (B) and in PDX model with KIT 642E/N822K mutation (C). (D) Antitumor activity of AZD3229 in PDX model with KIT exon 11 deletion (557–558)/Y823D. The body weight changes for this study are shown in (E).

D816H secondary mutation is one of the more difficult to treat A-loop mutations identified in imatinib-resistant patients (12). In an established Ba/F3 KIT exon 11 deletion/D816H xenograft model, AZD3229 dosed at 20 mg/kg led to robust tumor regression, whereas imatinib dosed at 300 mg/kg QD resulted in little inhibition of tumor growth (Fig. 4B) in concordance with clinical data. Although we used a higher dose of regorafenib (100 mg/kg) than the clinically equivalent dose of 30 mg/kg (18, 33), we observed greater efficacy with AZD3229 compared with regorafenib in this model. All the agents were well tolerated in this study (fig. S13).

The antitumor activity of AZD3229 in two different PDX models of GIST with KIT A-loop mutations, which are resistant to imatinib, was further explored. In the GIST PDX model (KIT K642E/N822K), AZD3229 at doses of 4 and 20 mg/kg resulted in dose-dependent tumor regressions, with the latter dose resulting in deeper and durable tumor regressions that were maintained for at least a month even after the dosing had stopped. Imatinib at 300 mg/kg remained completely inactive, and regorafenib was less efficacious than AZD3229 (Fig. 4C). All the doses were well tolerated in this study (fig. S14). The animals on this study were observed for a longer period (~90 days) to observe regrowth of tumor upon compound withdrawal. Mice that had regressed on 20 mg/kg of AZD3229 did not show signs of tumor regrowth for a month after the dosing had stopped, whereas mice dosed with regorafenib or a lower dose of AZD3229 (4 mg/kg) showed tumor regrowth much faster (Fig. 4C). In a PD study in this model, 20 mg/kg of AZD3229 led to a greater inhibition of KIT signaling at 4 and 8 hours of dosing (fig. S15), consistent with the efficacy data.

In addition, AZD3229 was tested in a second KIT A-loop PDX model, with KIT exon 11 deletion (557–558)/Y823D mutation, where AZD3229 at doses of 2 and 20 mg/kg resulted in dose-dependent inhibition of tumor volume, with the latter dose resulting in tumor regression (Fig. 4D). Ripretinib at a dose of 50 mg/kg resulted in tumor stasis in this model. While the doses of AZD3229 were well tolerated, ripretinib resulted in body weight loss at day 114, and animals were given a drug holiday for 2 days to recover. The regular dosing was continued on day 116 onward (Fig. 4E). We therefore could not increase the dose of ripretinib beyond 50 mg/kg. Although work published by Smith et al. (19) showed that the dose of 50 mg/kg is tolerated in mice, the differences in tolerability may be attributed to the route of drug administration, which was oral gavage in our studies versus drug administered by diet in the published study. PD study in this model demonstrated inhibition of KIT signaling up to 8 hours in tumors from animals dosed with both 2 and 20 mg/kg of AZD3229 (fig. S16).

AZD3229 inhibits common PDGFRα activating mutations in GIST

The cellular potency of AZD3229 was assessed in Ba/F3 cell lines harboring clinically relevant PDGFRα mutations, including primary mutation V561D and the resistant mutations, D842V and V561D/D842V. AZD3229 potently inhibited PDGFRα phosphorylation, with IC50 <10 nM in Ba/F3 PDGFRα V561D/D842V cells. Inhibition of PDGFRα phosphorylation also potently inhibited downstream signaling in these cells, whereas imatinib was several times less potent than AZD3229 in inhibiting PDGFRα signaling, in concordance with clinical data (Fig. 5A).

Fig. 5 AZD3229 inhibits PDGFRα activating mutations in GIST.

(A) Inhibition of PDGFRα signaling in Ba/F3 cells expressing PDGFRα V561D/D842V after 4-hour treatment with AZD3229 or imatinib. (B) Viability assay data in Ba/F3 cells expressing PDGFRα V561D, PDGFRα D842V, and PDGFRα V561D/D842V mutations. GI50 values represent means ± SD of two or three independent experiments.

Inhibition of cellular signaling by AZD3229 translated to potent inhibition of cell viability in PDGFRα V561D, PDGFRα D842V, and PDGFRα V561D/D842V cells with GI50 values of 0.2, 50, and 27 nM, respectively (Fig. 5B). Although SoC agents also inhibited the primary V561D mutation, AZD3229 effectively inhibited imatinib-resistant D842V and V561D/D842V mutations (representing the unmet need populations), being 10- and 25-fold more potent than imatinib (Fig. 5B). Among the investigational agents, avapritinib inhibited the PDGFRα resistance mutations more potently (two to three times) than AZD3229, in concordance with clinical data (30), and ripretinib was not potent.

AZD3229 has a good selectivity against VEGFR2 and does not alter BP in rats at efficacious exposures

One of the main aims for the design and development of AZD3229 was to avoid inhibition of VEGFR2, which would cause dose-limiting toxicity in patients and potentially result in the use of subefficacious doses and incomplete target suppression. Using a Ba/F3 VEGFR2 model, we demonstrated that AZD3229 does not inhibit VEGFR2 activity, whereas sunitinib, regorafenib, and ripretinib inhibited VEGFR2 with various degrees of potency (Fig. 1D). In addition, VEGFR2 inhibition was investigated using a physiologically relevant model, PAE-VEGFR2 (porcine aortic endothelial cells overexpressing VEGFR-2). PAE-VEGFR2 cells combine the sensitivity of a traditional human umbilical cord endothelial cell (HUVEC) assay with a higher throughput. The relationship between potent PAE-VEGFR2 activity and blood pressure (BP) elevation in rats and in patients has been previously quantified (34). When tested in the PAE-VEGFR2 assay, AZD3229 had an IC50 of 324 nM and was 51, 43, and 22 times less potent than sunitinib (IC50, 6.3 nM), regorafenib (IC50, 7.5 nM), and ripretinib (IC50, 15 nM), respectively (Fig. 6A).

Fig. 6 AZD3229 had no effect on arterial BP in rat telemetry studies.

(A) Effect of AZD3229, sunitinib, regorafenib, or ripretinib on inhibition of VEGFR2 phosphorylation in PAE-VEGFR2 cells. (B) Time course of changes in systolic BP in rats treated with vehicle on day 1 and AZD3229 (500 mg/kg) on days 2 and 3. Data are expressed as change compared to time-matched vehicle control, relative to time zero. The simulated unbound plasma concentration is also plotted against the right-hand axis. (C) A PK/PD model was used to simulate BP changes at predicted human efficacious doses. The graphs depict unbound concentration on the x axis and absolute BP change (systolic for AZD3229 and regorafenib or mean arterial pressure for sunitinib) on the y axis. The curves are from PK/PD modeling of rat telemetry data from in-house studies [AZD3229 and regorafenib (34)] or reported in the literature for sunitinib (36). Modeling processes have been reported in (34). The dotted lines refer to the predicted human Cmax for AZD3229 and reported human Cmax for regorafenib (37) and sunitinib (38).

To determine whether the lack of VEGFR2 activity for AZD3229 was reflected in an in vivo model in terms of lack of effect on BP elevation, we conducted a rat telemetry study. We were interested in seeing an exposure-response relationship between AZD3229 and BP to validate the model and therefore chose doses of 200 and 500 mg/kg to explore a potential exposure-response relationship. The dose of 500 mg/kg was the highest dose that could be tested in a short-term study to ensure that a cardiovascular effect was observed to enable modeling of the dose response. Treatment of rats with 200 or 500 mg/kg AZD3229 resulted in similar unbound maximum plasma concentrations (Cmax) of 499 and 583 nM, respectively. Consequently, both dose levels caused a similar increase in systolic BP of up to 11 mmHg from vehicle, which was observed between 2 and 7 hours postdose. Therefore, the time course of the BP and plasma pharmacokinetics (PK) of only the highest dose (500 mg/kg) are shown in Fig. 6B. The time course of the BP increase mirrored the early peak in plasma concentration of AZD3229, followed by rapid reduction in BP upon clearance of AZD3229 (Fig. 6B). Because of day-to-day variation, there were small transient inflections in BP after treatment with vehicle. After treatment with AZD3229, a clear sustained increase in BP was observed above this baseline variability. The PK/PD relationship for the effect of AZD3229 on BP in rats was modeled and was compared to the predicted human plasma concentration (Fig. 6C). The predicted efficacious exposure in humans was based on AZD3229 efficacy in several GIST and PDX models and PK/PD modeling of doses leading to >90% of KIT phosphorylation inhibition over the dosing interval, which resulted in maximum tumor growth inhibition in efficacy studies (35). On the basis of this information, the predicted efficacious human Cmax range for AZD3229 was between 0.004 and 0.139 μM for different models. At the predicted efficacious exposure, the PK/PD model predicted BP changes of less than 1 mmHg (Fig. 6C). Regorafenib was included in the study as a positive control for increase in BP, and previously published rat telemetry data for sunitinib (36) were also used for comparison. Regorafenib dosed daily at 16 mg/kg in rats (corresponding to the clinically efficacious dose) caused an increase in systolic BP that persisted for up to ~24 hours postdose (Fig. 6C) [maximum increase of 8 mmHg when averaged over an interval of 12 hours, data extracted from (34)]. Rat telemetry data for sunitinib showed BP elevation after the first administration of sunitinib at 50 mg/kg, and it remained increased through the end of the study (48 hours after the third dose). BPs were 20 to 30% greater in rats administered sunitinib at 50 mg/kg compared with control animals (Fig. 6C) [data extracted from (36)]. Figure 6C shows the rat unbound concentration–BP relationship contextualized to clinical exposure of these molecules. These simulated curves represent the steady-state rat concentration-BP response, removing the different delays between PK and effect, and may therefore give slightly different BP predictions from the observed findings (Fig. 6B) (34, 36).

AZD3229 has the best in class profile as a KIT/PDGFRα inhibitor for GIST

The preclinical models used in the present study have good translatability and have been predictive of clinical efficacy of SoC and investigational agents (23, 24). It was therefore possible to benchmark AZD3229 against SoC agents and investigational drugs avapritinib and ripretinib for clinically relevant KIT/PDGFRα mutations in these models. We demonstrate that the dose of AZD3229 in these models that results in robust antitumor activity (20 mg/kg) also leads to >90% of inhibition of KIT phosphorylation. On the basis of these data along with extensive PK/PD modeling of the efficacy data in several PDX models (35) and the PK properties of AZD3229 (21), the predicted human dose of AZD3229 is anchored at 34 mg BID, resulting in an unbound Cmax of 139 nM (35), which would provide adequate exposure to provide coverage for all the mutations tested in vivo and represented in the Ba/F3 cell panel. Figure 7 lists the GI90 values measured across the Ba/F3 KIT/PDGFRα cell panel for AZD3229 and the SoC/investigational drugs. The potency (GI90 values) of each compound for a specific mutation has been color coded to show whether the predicted clinical exposure (for AZD3229) or observed clinical trough concentrations for clinical agents is greater than GI90 for a specific mutation. Green indicates that the clinical trough concentrations of SoC/investigational agents or predicted trough concentrations for AZD3229 are greater than GI90 for a specific mutation. Gray indicates that the clinical concentration is equal to or less than GI90 for a specific mutation. This scale is arbitrary but is nevertheless useful to benchmark the potency of each compound relative to the clinical exposure. The pattern observed for the SoC agents fits with the known clinical activity in that imatinib is active against primary mutations (green) except for KIT exon 9 insertion (gray) but is dominated by gray for the secondary mutations. Sunitinib, on the other hand, delivers good exposure against secondary mutations in the ATP-binding pocket but does not do so for secondary mutations in the A-loop domain. Regorafenib and ripretinib provide onefold or greater exposure over GI90 broadly across the spectrum of most mutations, whereas avapritinib is not predicted to deliver exposure against the secondary mutations in the ATP-binding pocket. By contrast, AZD3229 (34 mg of dose predicted exposure) shows a greater uniformity of good potency across the whole spectrum of KIT mutations represented in this Ba/F3 panel. Therefore, assuming adequate exposure can be delivered in the clinic, AZD3229 should deliver an overall superior KIT/PDGFRα-inhibitory profile compared with other KIT inhibitors.

Fig. 7 AZD3229 is a pan-KIT and pan-PDGFRα inhibitor.

Representation of common GIST mutations showing functional locations and protein-unbound GI90 values (average of two or three independent experiments) for each mutation tested in a cell viability assay of Ba/F3 KIT- and PDGFRα-mutant cell lines comparing AZD3229, SoC agents, and investigational agents, ripretinib and avapritinib. Color coding is based on GI90 fold coverage (trough concentration/GI90) over predicted human dose for AZD3229 or recommended clinical doses for SoC and investigational agents. Green indicates ratio greater than 1, and gray indicates ratio less than or equal to 1. Trough concentration for ripretinib is variable ranging from 5 to 26 nM. Calculations are in data file S1.


After the approval of imatinib in GIST, there has been a paradigm shift in treatment options for patients with GIST from conventional chemotherapy to targeted therapies, improving the median survival from 18 months to 5 years with frontline imatinib. Despite this progress, it has become apparent that secondary resistance, characterized by disease progression after an initial objective response, develops in many patients (~80%) due to emergence of heterogeneous secondary mutations occurring in KIT (8, 9). Imatinib-resistant disease is characterized by intra- and intertumor heterogeneity. There have been reports of up to five different drug resistance mutations across portions of an individual lesion and up to seven secondary resistance mutations across multiple tumors in the same patient (39). Therefore, subsequent treatment with TKIs that have suboptimal activity for multiple mutations is challenging and results in drug-resistant cells that lead to disease relapse, especially upon drug interruptions.

Sunitinib and regorafenib as subsequent therapies are subefficacious, with an ORR of 7 and 4.5% and progression-free survival (PFS) of 6.2 and 4.8 months, respectively (10, 13). In a phase 3 trial of regorafenib, drug-related grade 3 or higher, hypertension was reported in 23.5% of patients, and hand-foot skin reaction occurred in 56.1% patients. Moreover, 82% of patients underwent dose reductions due to VEGFR2-driven hypertension (13, 40), compromising efficacy. Hence, there is a clear unmet need for a compound that potently inhibits imatinib-resistant mutations, with high selectivity necessary to limit pharmacologic failure.

Here, we describe the development and preclinical characterization of AZD3229, a potent KIT/PDGFRα inhibitor designed to overcome the limitations of approved and developmental compounds for GIST. Our chemical design strategy was aimed at developing a selective inhibitor of KIT and PDGFRα with the ability to inhibit a wide spectrum of clinically relevant mutations known in GIST, with a therapeutic margin to VEGFR2 inhibition. AZD3229 has good kinase family selectivity and a kinome profile similar to imatinib. We have comprehensively profiled this molecule in vitro and in vivo using more than 20 engineered clinically relevant cell lines and several PDX models. AZD3229 demonstrated exquisite potency compared with SoC agents against all mutants tested, suggesting potential clinical activity across a range of KIT and PDGFRα mutations. AZD3229 potently inhibits KIT exon 9 mutations, which should give it a clear advantage over imatinib in first-line patients. Potent and sustained activity of AZD3229 across KIT and PDGFRα mutations with a selective kinase inhibition profile could be key to deliver effective tumor cell eradication, minimizing the risk of residual cell persistence and delaying the emergence of resistance.

In this study, we also show preclinical data comparing AZD3229 to avapritinib and ripretinib, demonstrating superiority of AZD3229 over these compounds in inhibiting a range of KIT/PDGFRα-resistant mutations. According to our data, avapritinib is a potent KIT A-loop and PDGFRα inhibitor and clearly ineffective for the ATP-binding pocket mutations. This is evident from avapritinib’s phase I/II clinical data that have demonstrated lack of efficacy in patients with GIST with ATP-binding pocket mutations (30). On the basis of these data, a phase 3 trial testing the efficacy of avapritinib in second line is ongoing in patients with genotype-selected (V654A negative, T670I negative) tumors. Avapritinib was recently granted Food and Drug Administration (FDA) approval for the treatment of patients with GIST harboring PDGFRα exon 18 mutation, including D842V mutations, being the only agent approved in this segment of patients with GIST. Ripretinib, on the other hand, has broader activity than avapritinib but shows low potency for ATP-binding pocket mutations and PDGFRα mutations in our studies. Data from the phase 3 clinical study of ripretinib, INVICTUS, were recently reported. The data look encouraging, showing an ORR of 9.4% and a median PFS of 6.3 months with ripretinib versus 1 month with placebo and showing an improvement over regorafenib, which demonstrated a median PFS of 4.8 months (41). We believe that there is potential for improvement of clinical responses over these investigational agents and that AZD3229 has a profile that is promising. AZD3229 should provide an advantage for patients with GIST over SoC/investigational agents, including a superior selectivity profile for improved tolerability in the clinic. AZD3229 potently inhibits a broad spectrum of KIT and PDGFRα primary and secondary mutations, including those not targeted by imatinib (KIT exon 9 insertion). Together, the excellent biological activity and selectivity establish AZD3229 as a potential drug candidate worthy of clinical evaluation in patients with advanced GISTs. Despite the compelling data presented here, the real implications of our findings need to be tested in clinical settings to determine whether patients with GIST with KIT/PDGFRα benefit from the pan-inhibitory potential of AZD3229 and whether treatment with this potent inhibitor would delay or prevent the emergence of resistance mutations.


Study design

The purpose of this study was to design and characterize selectivity, in vitro potency, and in vivo efficacy of AZD3229 in KIT/PDGFRα-driven models of GIST. The study also compared SoC/investigational agents alongside AZD3229. The in vitro Ba/F3 or GIST cell line data have three biological replicates for the majority of data (data file S1). The sample size for animal experiments (n = 5 to 8) was based on the results of preliminary experiments, with exact numbers described below. Mice were randomly assigned to the treatment and control groups; investigators were not blinded during evaluation of the preclinical experiments.

Cell culture

The GIST430/654 cell line was licensed from J. Fletcher (Dana-Farber Cancer Institute) (9). The GIST-T1 cell line was purchased from Cosmobio. GIST430/654 and GIST-T1 cells were grown in Iscove’s modified Dulbecco’s medium (Life Technologies) and Dulbecco’s modified Eagle’s medium (Life Technologies), respectively, with 10% fetal bovine serum. Ba/F3 lines were obtained from Deutsche Sammlung von Mikroorganismen und Zellkulturen (German Collection of Microorganisms and Cell Cultures) and maintained in Roswell Park Memorial Institute medium (RPMI) (Life Technologies) with interleukin-3 (IL-3; 10 ng/ml) (Millipore Sigma). KIT/PDGFRα-transduced Ba/F3 lines were cultured in RPMI medium with puromycin (0.5 μg/ml; Life Technologies) without IL-3. Parental Ba/F3 cells were grown in RPMI, supplemented with IL-3 (10 ng/ml). PAE-VEGFR2 SV40 cells were grown in Ham’s F12 medium with puromycin (2 μg/ml).

Viability assays

Before compound testing, growth was measured for each cell line of interest to identify the optimal seeding densities in 384-well format. Optimal seeding density is chosen on the basis of linear growth over 72 hours. Optimal cell numbers range from 900 to 4500 cells per well for Ba/F3 cell lines, 1000 cells per well for GIST-T1, and 7500 cells per well for GIST430/654 cells. Cells were seeded into 384-well plates, treated with compound or dimethyl sulfoxide (DMSO; Millipore Sigma), and incubated for 72 hours at 37°C, followed by the addition of MTS [3-(4,5-Dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium] reagent (Promega). Absorbance readings were recorded at 490 nm. Results were analyzed using Genedata Screener software, and GI50 and GI90 was determined.


Cells were treated with compound for 4 hours and lysed with radioimmunoprecipitation assay buffer (Thermo Fisher Scientific). Protein (25 μg) was subjected to Western blotting by using primary antibodies and horseradish peroxidase–conjugated secondary antibodies (Cell Signaling Technology). The signal was visualized using enhanced chemiluminescence (Pierce) and imaged with the ImageQuant chemiluminescent system.


All animal studies were conducted under Institutional Animal Care and Use Committee (IACUC) guidelines established where the study was conducted. Female CB17 and nonobese diabetic (NOD)–SCID mice were purchased from Charles River Laboratories. Female NSG (NOD-SCID IL2Rgammanull) mice were purchased from the Jackson laboratory. Mice were housed under pathogen-free conditions in ventilated cages at our AAALAC (Association for the Assessment and Accreditation of Laboratory Animal Care)–accredited facility in Waltham, MA. Animal studies were conducted in accordance with the guidelines established by internal IACUC and reported following the ARRIVE (Animal Research: Reporting In Vivo Experiments) guidelines. Compounds were administered to mice by oral gavage. Growth inhibition was assessed by comparison of the differences in tumor volume between control and treated groups (n = 7 per group). Treatment started when mean tumor size reached about 150 mm3. Tumor volume was calculated (length being the longest diameter and width the corresponding perpendicular diameter) using the formula: length (mm) × width (mm)2/0.52.

Plasma and tumor for PK/PD assessments were collected at indicated time points (n = 3). Treatment started when mean tumor size reached about 300 mm3. Tumors were snap frozen in liquid nitrogen, manually dissociated, lysed with PhosphoSafe buffer (Novagen), and immunoblotted. Resulting bands on Western blots were quantified using ImageJ software and reported as % control [= sample quantification/(average of control sample quantification) × 100].

Cell line xenograft studies

Ba/F3 KIT exon 11 deletion (2 × 106) (557–558)/D816H or Ba/F3 KIT exon 11 deletion (5 × 106) (557–558)/V654A tumor cells were used for allograft studies in CB17 SCID mice. GIST430/654 tumor cells (5 × 106) were used in xenograft studies in NSG mice. Cells were suspended 1:1 in phosphate-buffered saline (PBS)/Matrigel, and 0.1 ml was injected subcutaneously in the right flank of mice. Tumor volumes and body weight were recorded twice weekly for the duration of study.

PDX studies

PDX models were purchased from CrownBio as dissociated single cell suspensions. A cell bank was created through serial passaging of tumors. Cells were thawed and resuspended, and then, 5 × 104 to 1 × 105 viable cells were injected in PBS/Matrigel (1:1) into the right flank of female NOD-SCID mice. Models include HGiXF-105 (GS5108) (KIT exon 11 deletion 557–558/Y823D), HGiXF-106 (GS11331) (KIT exon 11 deletion 557–558/V654A), HGiXF-107 (GS5106) (KIT K642E/N822K), and HGiXF-108 (GS11327) (KIT exon 11 deletion K550fs). NOD-SCID mice have been shown to have a high incidence of pro–T cell lymphoma (76%) by 6 months of age (42), with more females affected than males, and we lost 5% of animals in short-term efficacy studies among both vehicle-treated and compound-treated animals.

Rat telemetry study

Rat telemetry was performed as previously described (33). Briefly, eight male Han Wistar rats were implanted with HD-S11 transmitters (Data Sciences International) under isoflurane anesthesia at Charles River Laboratories and allowed to recover for at least 2 weeks. Animals were transferred to the AstraZeneca facility at around 2 to 3 months old and allowed a minimum acclimatization period of 7 days. The study was conducted over a 3-week period: Week 1 consisted of three consecutive daily doses of vehicle (0.5% HPMC, 0.1% Tween 80). Weeks 2 and 3 consisted of a single dose of vehicle on day 1 followed by two consecutive daily AZD3229 doses on days 2 and 3 of either 200 mg/kg (in week 2) or 500 mg/kg (in week 3). Animals were dosed in an ascending dose design, receiving all doses of vehicle and compound, so randomization to treatment groups was not required. Rats weighed 354 to 416 g at the time of treatment. Blood samples (obtained via microsampling: 32 μl) were taken at 2 hours after the first dose and 2 and 26 hours after the second dose to determine plasma exposure. These exposure values were used alongside PK data from other studies to simulate the plasma PK profile. Animals were routinely monitored for general welfare on each day of dosing by visual observation of animals in their home cages. Systolic BP was obtained as a series of 1-min averages, which were subsequently used to generate a moving average over a 60-min period.

Statistical analysis

All statistical analyses were performed using Prism software version 8.0 (GraphPad). The information about statistical details and methods is indicated in the figure legends, Results, or Materials and Methods. Original numerical data are provided in data file S1. Results for in vivo preclinical efficacy studies were expressed as geo means ± SEM. Results for in vivo preclinical PD studies were expressed as means ± SD. Only two groups were compared at any given time, allowing a t test to be used to identify statistical significance of differences between groups. For all statistical tests, P < 0.05 was used to denote statistical significance.


Materials and Methods

Fig. S1. AZD3229 inhibits KIT exon 9 insertion more potently than ripretinib and avapritinib.

Fig. S2. AZD3229 is more potent than ripretinib and avapritinib in inhibiting KIT exon 11 deletion mutation.

Fig. S3. Plasma PK of AZD3229 in mice.

Fig. S4. AZD3229 PD analysis in KIT exon 11 deletion PDX model.

Fig. S5. Comparison of AZD3229, ripretinib, and avapritinib in inhibiting KIT exon 11 deletion/V654A in GIST430/654 cells.

Fig. S6. Induction of apoptosis by AZD3229 in GIST430/654 cells.

Fig. S7. Time- and dose-dependent inhibition of KIT signaling by AZD3229 in GIST430/654 cells.

Fig. S8. AZD3229 dosing in a GIST430/654 model does not alter body weight in mice.

Fig. S9. AZD3229 PD analysis in a GIST430/654 model.

Fig. S10. Plasma concentrations of AZD3229 correlate with inhibition of KIT signaling by AZD3229 in a GIST430/654 model.

Fig. S11. AZD3229 dosing decreases proliferating cells in GIST430/654 tumors.

Fig. S12. KIT exon 11 deletion/V654A PDX model is resistant to imatinib.

Fig. S13. AZD3229, imatinib, and sunitinib are well tolerated in SCID mice bearing Ba/F3 KIT exon 11 deletion/D816H tumors.

Fig. S14. AZD3229 dosing in KIT K642E/N822K PDX model is well tolerated.

Fig. S15. AZD3229 PD analysis in KIT A-loop PDX model (KIT K642E/N822K).

Fig. S16. AZD3229 PD analysis in KIT A-loop PDX model (KIT exon 11 deletion/Y823D).

Table S1. Biochemical enzyme profiling of AZD3229 across the Thermo Fisher Scientific kinome panel.

Data file S1. Original data.


Acknowledgments: We thank K. Jacques for generating lentiviral constructs for KIT, W. Nissink for helping to generate the kinome trees using compound selectivity data, L. Bao for assistance in generating data to support PD assessment, and M. Zinda for helpful insight during the project. Funding: The studies reported in the paper were funded by AstraZeneca. Author contributions: R.A., S.G., and S.F. conceived the project; R.A. and S.G. supervised the study; R.A., E. Banks, and M.G. wrote the manuscript with comments from all authors; E. Banks, D.B., and H.W. performed cellular assays; M.G., C.R., E. Barry, and C.B. designed and performed the in vivo xenograft studies; J.G.K. and D.W. contributed to the medicinal chemistry effort; V.P.R. and R.D.O.J. conducted modeling studies; T.C., A.R.H., J.T.M., and O.A. conducted telemetry study; M.J.P. provided structural biology and computational chemistry support; R.O. provided in vitro enzymatic assay support; D.L. performed IHC studies; bioscience support was provided by L.D. and project support was provided by W.S. and S.C. Competing interests: E. Banks, M.G., D.B., V.P.R., J.G.K., C.B., H.W., J.T.M., T.C., O.A., D.L., A.R.H., C.R., L.D., M.J.P., S.C., R.D.O.J., W.S., S.F., and R.A. are employees and/or shareholders of AstraZeneca. The authors indicate that they have no other conflicts of interest. Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials.

Stay Connected to Science Translational Medicine

Navigate This Article