Phase I Study of Rapid Alternation of Sunitinib and Regorafenib for the Treatment of Tyrosine Kinase Inhibitor Refractory Gastrointestinal Stromal Tumors Free

This article is featured in Highlights of This Issue, p. 7267

Mutational activation of KIT or PDGFRA receptor tyrosine kinases is the crucial transforming event in approximately 90% gastrointestinal stromal tumors (GIST; refs. 1, 2), and therapeutic inhibition of KIT/PDGFRA oncogenic signaling with first-line imatinib substantially improves outcomes in most patients (3). However, metastatic patients with initial clinical benefit will eventually progress, normally after a median of 20–24 months. Resistance to imatinib occurs in 80%–90% of patients with GIST with the expansion of heterogeneous subclones harboring different KIT secondary mutations (4, 5). Therefore, strategies aiming to target KIT activation after imatinib failure remain useful in GIST. However, response duration is relatively brief, 4–6 months, irrespective of the standard second- (sunitinib; ref. 6) and third-line (regorafenib; ref. 7) agent used. Other nonapproved tyrosine kinase inhibitors (TKI) with KIT inhibitory activity have shown similar clinical activity in phase II and III trials (8).

We have determined that each TKI with KIT-inhibitory activity has a drug-specific activity profile against a subset of the KIT secondary mutational spectrum, which constitutes the molecular basis for the modest clinical benefit observed after the onset of imatinib resistance (9). Remarkably, sunitinib and regorafenib have complementary activities against imatinib-resistant KIT oncoproteins. Namely, sunitinib inhibits mutations in the ATP-binding pocket, whereas regorafenib primarily inhibits mutations in the activation loop, the two regions of the KIT kinase domain in which resistance mutations typically cluster. Therefore, this evidence supports combination of TKIs with complementary activity to suppress heterogeneous cross-resistant subpopulations in imatinib-resistant GIST. To improve tolerability while reaching effective doses, we established the preclinical rationale for an innovative therapeutic combination strategy based on rapid alternation of sunitinib (3 days) with regorafenib (4 days; ref. 9).

In this phase I study we report, for the first time in cancer, safety and efficacy from a therapeutic combination based on rapid alternation of TKIs with complementary activity against resistant subpopulations. As part of the planned biomarker analysis, circulating tumor DNA (ctDNA) assessment was used to determine the spectrum of resistance in patients with TKI-refractory GIST, and to evaluate clonal dynamics in GIST with clinical benefit from the rapid alternation combination approach.

The study enrolled patients aged ≥ 18 years with histologically confirmed diagnosis of metastatic and/or unresectable GIST with prior failure to at least imatinib, sunitinib, and regorafenib (fourth-line and beyond). Patients had documented progressive disease, Eastern Cooperative Oncology Group (ECOG) performance status of 0–1, RECIST 1.1 measurable disease, and normal organ function (full eligibility criteria in Supplementary Data). Primary KIT/PDGFRA tumor tissue genotype by Sanger sequencing was known in all recruited patients. The study was approved by the institutional review board and was conducted in compliance with guidelines for Good Clinical Practice. Patients provided written informed consent before study participation.

This was a phase I/II, single-center, open-label, Simon two-stage, dose escalation, and dose expansion study to evaluate the safety and the preliminary efficacy of short cycles of sunitinib alternated with regorafenib in TKI-refractory GIST.

Treatment consisted of 3 days of once daily sunitinib followed by 4 days of once daily regorafenib, continuously in 28-day cycles. The starting dose level was sunitinib 37.5 mg/day and regorafenib 120 mg/day. Doses were escalated following a classical 3+3 design up to sunitinib 50 mg/day and regorafenib 160 mg/day or until MTD and recommended phase II dose (RP2D) was determined (further details in Supplementary Data). Once 12 subjects were enrolled, we evaluated an interim efficacy analysis following the operating characteristics of a Simon two-stage design. With a null hypothesis of 10% and alternative hypothesis if 30%, we had 80% power with 4% type 1 alpha. We needed at least 2 responses within the first 12 patients to continue enrollment through the second stage.

The primary objective of the study was to evaluate the safety and tolerability for sunitinib alternating with regorafenib in patients with unresectable and/or metastatic GIST with prior failure to at least all established therapies approved in GIST. Pharmacokinetics and efficacy assessments were evaluated as secondary objectives. ctDNA evaluation to determine the resistance profile in advanced GIST and the role of serial monitoring was an exploratory endpoint.

Pharmacokinetic blood samples were collected predose in all patients for sunitinib and sunitinib metabolite SU012662 on days 1, 15, 18, and 22; and for regorafenib and its metabolites (M2/5), predose and on days 4, 8, 11, 15, 18, and 22 (Supplementary Table S1).

Tumor assessments by CT scan were performed at baseline and every 8 weeks thereafter. Tumor response was evaluated according to RECIST 1.1.

Plasma was collected and stored at −80°C on cycle 1 days 1 (time 0) and 15, day 1 on subsequent cycles, and at the end of treatment. Cell-free DNA (cfDNA) was extracted by QIAamp Circulating Nucleic Acid Kit (Qiagen).

ddPCR reagents were ordered from Bio-Rad. Primer/probe mix for KIT-resistant mutations was custom-made by Life Technologies and are described in Supplementary Table S2. The following KIT-resistant mutations were analyzed: V654A (exon 13), T670I (exon 14), exon 17, and A829P (exon 18). Cell lines GIST430/654, GIST-T1/670, GIST48, and GIST-T1/829 were used as positive controls; characteristics have been described previously (9). Genomic DNA of the cell lines were extracted by QIAamp DNA Mini Kit (Qiagen). PCR reactions (25 μL) that comprised ddPCR Supermix for Probes, custom-made TaqMan primer/probe mix, and appropriate DNA template were prepared in a 96-well PCR plate and transferred to the automated droplet generator. A second 96-well PCR plate which held the emulsified droplets were amplified by PCR on an S1000 thermal cycler. Cycling conditions were: 10 minutes at 95°C followed by 40 cycles of a two-step thermal profile of 30 seconds at 94°C denaturation and 60 seconds at specific annealing temperature (57°C for V654A, 55°C for T670I, 62°C for A829P, and 57°C for pan- ex17 mutations). The 96-well plate with amplified PCR products was then read by a QX200 Droplet Reader (Bio-Rad).

cfDNA genomic libraries were prepared from 2.5 to 49.5 ng of cfDNA as described previously (10). Targeted capture was performed using a custom set of hybridization probes targeting 81 kb from 58 cancer-driver genes (Supplementary Table S3). TEC-seq libraries were sequenced using 100-bp paired-end runs on the Illumina HiSeq 2500. Primary processing of next-generation sequence data for analyses of sequence alterations in cfDNA samples was performed as described previously (10). Candidate somatic mutations, consisting of point mutations, small insertions, and deletions were identified using VariantDx (Personal Genome Diagnostics) across the targeted regions of interest (11). For the scope of this study, an alteration was considered a candidate somatic mutation in KIT when at least two distinct paired reads contained the mutation in the plasma and the number of distinct paired reads containing a particular mutation in the plasma was at least 0.05% of the total distinct read pairs. As tissue was not available for comparative sequencing analyses, we excluded from our ctDNA dynamics approach all detected variants that could be related to clonal hematopoietic expansion as described previously (12), including alterations in DNMT3A, FGFR3, GNAS, or TP53.

A total of 14 patients with TKI-refractory, metastatic GIST were enrolled into the study between August 2014 and September 2016. One patient showed rapid symptomatic deterioration after signing the informed consent and never received study drugs. A second patient rapidly deteriorated 10 days after treatment initiation due to clinical progression. The remaining 12 patients completed at least one cycle (28 days) of sunitinib alternating with regorafenib. Baseline demographics and disease characteristics are summarized in Table 1.

Thirteen of 14 patients were evaluable for safety and dose-limiting toxicity (DLT). Patients in the initial dose level completed the first treatment cycle without experiencing DLTs. Two DLTs were evidenced in 5 patients recruited in the second dose level, which consisted on sunitinib 37.5 mg/day for 3 days followed by regorafenib 160 mg/day for 4 days. Both patients showed refractory asymptomatic grade 3 hypophosphatemia despite intravenous replacement. Four more patients added in the first dose level did not show DLTs, and therefore the RP2D was sunitinib 37.5 mg/day for 3 days followed by regorafenib 120 mg/day for 4 days.

All the 13 evaluable patients experienced treatment-related adverse events (AE) of any grade, although the majority were grade 1–2 in severity. Most common treatment-related AEs are listed in Table 2 and were fatigue (92%), weight loss (62%), palmar-plantar erythrodysesthesia syndrome (PPES; 54%), diarrhea (54%), anorexia (38%), hoarseness (38%), hypertension (38%), and mucositis (23%). Four patients experienced grade 3 AEs, including PPES (n = 2) and hypertension (n = 2), all of them clinically manageable. No toxicity could be specifically attributed to one of both drugs. Seven patients (58%) required dose delay/modification, and 1 patient (8%) definitive treatment interruption and removal from the study due to recurrent G2 rash despite dose reduction and optimal supportive dermatologic support, which occurred outside the DLT window.

The 13 patients evaluable for toxicity were evaluable for radiographic response. No objective responses were observed by RECIST v1.1. criteria. Stable disease (SD) was observed in 4 patients. The clinical benefit rate at 8 weeks was 31% (n = 4). The median progression-free survival (mPFS) was 1.9 months [95% confidence interval (CI), 1.4–3.6 months]. A total of 3 patients died on trial or during follow-up. The median overall survival was 10.8 months (95% CI, 5.9–∞). The 6-month overall survival rate was 62% (95% CI, 7%–91%). Four patients (29%) continued on treatment after progression for a median of 3.6 months (range, 3.6–7.3 months) due to clinical benefit, as subjectively assessed by the investigator.

Trough levels of sunitinib and regorafenib were obtained across several timepoints (Supplementary Table S1). Including all patients and timepoints, the maximum concentration and steady state of sunitinib and regorafenib were below reported threshold target inhibition concentrations (Table 3) due to scarce accumulation of sunitinib, regorafenib, and their metabolites during the on/off drug periods (Supplementary Fig. S1; refs. 13, 14). Patients with SD had slightly higher median drug exposure to sunitinib (4.99 ng/mL) and regorafenib (1,879.17 ng/mL) than those progressing at the first reevaluation (3.99 ng/mL and 991.43 ng/mL, respectively). Median sunitinib and regorafenib concentrations at the time of the two DLTs (1.47 ng/mL and 1,198.35 ng/mL, respectively) did not differ from median drug concentration across all the timepoints per drug.

ctDNA was prospectively assessed by two complementary technologies, TEC-seq and ddPCR, to investigate the mechanisms responsible of resistance in heavily pretreated GIST patients, and to determine whether ctDNA monitoring can be a potential biomarker during treatment. Primary and secondary mutations in KIT were detected at any timepoint, respectively, in 8 of 9 (89%) and 7 of 9 (78%) patients with TEC-seq, while ddPCR identified secondary mutations in hotspot regions of KIT in 11 of 12 patients (92%; Table 4). Together, both technologies detected KIT-resistant mutations at any timepoint in all but 1 patient. Median mutant allele frequency (MAF%) for KIT secondary mutations at baseline draw ranged from 0.09% to 16.4% by TEC-seq (median 1.3%) and 1.7%–19.9% (median 5.3%) by ddPCR. As noted in Table 4, the majority of KIT secondary mutations identified were in the KIT activation loop. Remarkably, there was limited heterogeneity of mechanisms of resistance, with a median of one resistant mutation per patient (range 0–2). Mutations involving genes other than KIT were present in 8 of 10 patients (80%; Table 5), although a hotspot mutation in NRAS in 1 patient was the only known alteration potentially related to TKI resistance.

Clonal dynamics was further studied by TEC-seq in the 3 KIT-mutant patients achieving treatment benefit. In all cases, an early decrease in KIT primary and/or secondary mutations was observed during cycle 1 of treatment, followed by prolonged suppression until clone reemergence at the time of disease progression, thus predicting the course of the disease (Fig. 1A–C). In contrast, this pattern was not observed in a patient with GIST with rapid progression (Fig. 1D).

Polyclonal resistance in GIST entails that a single drug is insufficient to target all resistant KIT variants that drive tumor progression, thus leading to modest clinical benefit irrespective of the second- or third-line used (4, 6, 7). Combination of targeted therapies will potentially augment the magnitude and/or duration of response. However, enhanced toxicities are often seen in such combinations, thereby limiting the use of effective doses (15). We have recently shown that sunitinib and regorafenib share a complementary pattern against resistant mutations in KIT, and therefore combination of both drugs would widen the spectrum of secondary-resistant clones effectively targeted (9). We subsequently established the preclinical rationale for a therapeutic regimen involving rapid alternation of sunitinib (3 days) followed by regorafenib (4 days), aiming to reach effective doses while improving tolerability (9).

The primary endpoint of this study was to determine the safety of the combination. Treatment was tolerable across the first dose level (sunitinib 37.5 mg/day and regorafenib 120 mg/day). The two DLTs occurred in the second dose level (sunitinib 37.5 mg/day and regorafenib 160 mg/day) in form of grade 3 hypophosphatemia. Toxicities were common but manageable, being most of them grade 1 and 2. Adverse events lied within the expected profile for each TKI, with no new or enhanced toxicities resulting from the novel combination strategy. Therefore, rapid alternation of sunitinib and regorafenib appears to be feasible and tolerable; however, the low plasma drug concentrations suggest that clinically meaningful drug exposure was not achieved.

mPFS in this heavily pretreated population was 1.9 months, comparable with that of imatinib rechallenge (16) and superior to placebo in recent GIST phase III trials (6, 7, 16). In addition, 4 of 13 patients reached SD >8 weeks, and several patients remained on trial due to clinical benefit. Together, these data suggest partial KIT inhibition despite neither sunitinib nor regorafenib reached reported steady states (13, 14). Indeed, the 3 KIT-mutant patients achieving SD demonstrated a ctDNA dynamics indicative of early and sustained clonal suppression, including known KIT-resistant mutations. Therefore, drug concentrations obtained in plasma with this rapid alternation scheme likely reached the threshold of inhibition of imatinib-sensitive subpopulations, but were insufficient to suppress all imatinib-resistant clones due to the long time for both drugs (7–10 days) needed to obtain active plasma concentrations (13, 14). One single dose of imatinib is capable of extinguishing FDG-PET activity in 24 hours (17). However, reaching the steady state appears to be meaningful for sunitinib and regorafenib efficacy against clones with sensitive KIT-secondary mutants for each agent as monotherapy.

Current knowledge on resistance to KIT inhibition comes from studies mostly performed at the onset of imatinib failure (4, 5), and little is known after subsequent lines of treatment. We showed by both TEC-seq and ddPCR of plasma samples that KIT secondary mutations were significant drivers of progression in this heavily pretreated population, with a predominance for mutations in the activation loop of KIT, which is highly relevant for the design of future treatment strategies. Remarkably, we observed little heterogeneity of mechanisms of resistance, and mutants were at low allele frequencies despite the usual high tumor burden in this neoplasia. ctDNA monitoring using complementary technologies identified cfDNA mutations in all but 1 patient and predicted the course of the disease, thus holding the potential for treatment monitoring and therapy selection against emerging resistant subpopulations (9).

In conclusion, rapid drug alternation is an unprecedented strategy for TKI combination in cancer that might be effective when combining drugs with favorable pharmacokinetics, potentially allowing active doses while minimizing AEs. ctDNA studies demonstrate that KIT remains the critical driver in TKI-refractory GIST, and supports further studies for its implementation as a biomarker to guide treatment in patients with GIST.

C. Serrano is an employee/paid consultant for Deciphera and Blueprint, reports receiving commercial research grants from Deciphera, Bayer, and Pfizer, reports receiving speakers bureau honoraria from Bayer and Blueprint, and reports receiving other remuneration from Pharmamar, Pfizer, Bayer, Novartis, and Lilly. A.J. Wagner is an employee/paid consultant for Daiichi-Sankyo, Eli Lilly, Nanosphere, Five Prime, Deciphera, and Novartis, and reports receiving commercial research grants from Plexxikon, Daiichi-Sankyo, Karyopharm, Aadi Bioscience, Eli Lilly, and Five Prime. G.D. Demetri is an employee/paid consultant for Bayer, Novartis, Pfizer, EMD-Serono, Sanofi, Janssen Oncology, Ignyta, Roche/Genentech, LOXO Oncology, AbbVie, Mirati Therapeutics, ICON PLC, Epizyme, PharmaMar, Daiichi-Sankyo, WIRB-Copernicus Group, ZioPharm, Polaris, Blueprint Medicines, Merrimack, Translate Bio, G1 Therapeutics, CARIS Life Sciences, and ERASCA, reports receiving commercial research grants from Bayer, Novartis, Pfizer, Janssen Oncology, Ignyta, Roche/Genentech, Loxo Oncology, AbbVie, Epizyme, Adaptimmune, GlaxoSmithKline, and Daiichi-Sankyo, holds ownership interest (including patents) in Blueprint Medicines, Merrimack, Translate Bio, CARIS Life Sciences, and ERASCA, and is listed on a patent for imatinib in GIST. V.E. Velculescu is an employee/paid consultant for Personal Genome Diagnostics and Takeda, and holds ownership interest (including patents) in Personal Genome Diagnostics. C.P. Paweletz reports receiving speakers bureau honoraria from Bio-Rad, and is an unpaid consultant/advisory board member for Dropworks and Xsphera Biosciences. S. George is an employee/paid consultant for Blueprint Medicines, Deciphera Pharmaceuticals, Bayer, AstraZeneca, Eli Lilly, and Exelixis, and reports receiving other remuneration from Bayer. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C. Serrano, C.M. Barysauskas, A.J. Wagner, G.D. Demetri, V.E. Velculescu, J.A. Fletcher, S. George

Development of methodology: C. Serrano, Y. Kuang, C.M. Barysauskas, A. Mariño-Enríquez, G.D. Demetri, S. George

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Serrano, A. Leal, Y. Kuang, J.A. Morgan, O. Triplett, A.J. Wagner, G.D. Demetri, C.P. Paweletz, S. George

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Serrano, A. Leal, Y. Kuang, C.M. Barysauskas, J. Phallen, O. Triplett, A. Mariño-Enríquez, A.J. Wagner, G.D. Demetri, V.E. Velculescu, C.P. Paweletz, S. George

Writing, review, and/or revision of the manuscript: C. Serrano, A. Leal, Y. Kuang, J.A. Morgan, C.M. Barysauskas, A.J. Wagner, G.D. Demetri, V.E. Velculescu, C.P. Paweletz, J.A. Fletcher, S. George

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Serrano, A. Leal, C.M. Barysauskas, S. George

Study supervision: C. Serrano, G.D. Demetri, V.E. Velculescu, S. George

The protocol for this study was developed during the 15th edition of the ECCO-AACR-EORTC-ESMO Workshop on Methods in Clinical Cancer Research. The study was registered at ClinicalTrials.gov (identifier: NCT02164240). This work was supported in part by a Pfizer research grant (to S. George), a Bayer research grant (to S. George); ASCO Young Investigator Award (to C. Serrano), FERO Foundation award (to C. Serrano), FIS ISCIII (PI16/01371; to C. Serrano); Expect Miracles Foundation, Robert and Rene Belfer Foundation (to Y. Kuang and C.P. Paweletz); the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation (to G.D. Demetri and V.E. Velculescu), the Stand Up to Cancer–Dutch Cancer Society International Translational Cancer Research Dream Team Grant (SU2C-AACR-DT1415, to V.E. Velculescu, A. Leal, and J. Phallen), the Commonwealth Foundation (to V.E. Velculescu), the AACR-Janssen Cancer Interception Research Fellowship (to J. Phallen), the Mark Foundation for Cancer Research (to A. Leal and J. Phallen), and U.S. NIH (grants CA121113, CA006973, and CA180950, to V.E. Velculescu). Stand Up To Cancer is a program of the Entertainment Industry Foundation administered by the American Association for Cancer Research.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.