A prospective multicenter phase II study on the efficacy and safety of dasatinib in the treatment of metastatic gastrointestinal stromal tumors failed by imatinib and sunitinib and analysis of NGS in peripheral blood

Abstract Aim Dasatinib is a small molecule tyrosine kinase inhibitor with multiple targets including kit, PDGFR, and SRC. This prospective study evaluated the efficacy and safety of dasatinib as third‐line treatment for gastrointestinal stromal tumors (GIST). Methods The study enrolled adult patients (≥18 years of age) with histologically confirmed unresectable and/or metastatic GIST whose disease progressed despite imatinib and sunitinib therapy. Dasatinib (50 mg twice daily) was given orally for 1 week and escalated to 70 mg twice daily orally. The primary endpoint was to the 3‐month progression‐free survival (PFS) rate. Blood samples were acquired before dasatinib therapy for examination of gene mutations by next‐generation sequencing (NGS). Results From May 2016 to June 2018, 58 patients from 9 Chinese medical centers were enrolled in this study. The 3‐month PFS rate was 53.4% and the median overall survival (OS) was 14.0 months. Neither primary nor secondary gene mutations predicted the efficacy of dasatinib. Wild‐type GIST patients had longer PFS (5.5 months). The most common adverse events were anemia, proteinuria, fatigue, neutropenia, and diarrhea. The concordance of KIT/PDGFRA mutation was 61.9% between tissue and peripheral blood samples and additional KIT mutations were detected in the peripheral blood samples in 28.6% of the patients. Some SNV and CNV such as ATRX, TP53, TEKT4, STK11, SDHC, and CDKN2C related to tumor signaling pathways were detected. Patients with TP53 mutations and SDHC and TMEM127 gene copy number loss had longer OS. Conclusion Dasatinib has modest antitumor activity with tolerable toxicities in patients with metastatic GISTs who have failed imatinib and sunitinib therapy.


| INTRODUCTION
Gastrointestinal stromal tumor (GIST) is the most common mesenchymal tumor of the digestive tract. Advanced GISTs including unresectable, recurrent or metastatic are initially managed by imatinib; however, GIST patients may have primary resistance to imatinib due to mutations in KIT or PDGFRA rendering the patients less responsive to imatinib. In addition, resistance to imatinib may emerge due to secondary mutations in GIST patients in whom GISTs progress after an initial response to imatinib. 1 Although sunitinib and regorafenib as inhibitors of KIT, PDGFRA, and vascular endothelial growth factor receptor kinases have been approved for second-and third-line treatment, 2,3 their efficacy is not satisfactory, and the number of therapeutic drugs is still insufficient.
Dasatinib is a small molecule, adenosine triphosphate competitive inhibitor of KIT, PDGFR, and the protooncogene tyrosine-protein kinase Src (SRC) family. 4 The drug has demonstrated antitumor activities as first-and second-line treatment of GISTs. 5,6 Trent et al have shown in a phase II study of dasatinib for patients with imatinib-resistant GIST that 21% patients (10/47) had a progression-free survival (PFS) >6 months and an overall survival (OS) of 19 months. In a single-arm clinical trial of 50 patients with advanced GISTs resistant to imatinib, Shuetze et al found that patients with mutations in KIT or PDGFRA showed varied response to dasatinib. However, the relationship between secondary mutations of KIT, the change of downstream pathway status, and the efficacy of dasatinib has not been analyzed and explained in these studies.
Therefore, we carried out this prospective, multi-center study to evaluate the efficacy and safety of dasatinib in the third-line treatment of metastatic GISTs. Meanwhile, we performed next-generation sequencing (NGS) in peripheral blood to analyze gene variations of KIT downstream signaling pathway and the mechanism of drug resistance.

| Patient selection
The study enrolled adult patients (≥18 years of age) with histologically confirmed unresectable and/or metastatic GIST whose disease progressed despite imatinib and sunitinib therapy. The main inclusion criteria were (a) at least one measurable GIST lesion at baseline and (b) an Eastern Cooperative Oncology Group (ECOG) performance status score ≤ 2. We excluded patients who had received tyrosine kinase inhibitors (TKI) within 21 days of initiation of dasatinib therapy. Patients with impaired cardiac function, brain metastasis, or uncontrollable gastrointestinal bleeding were also excluded.

| Study design
This open-label, multi-center, and single-arm phase II study evaluated the efficacy of dasatinib for patients with metastatic GIST who had failed imatinib and sunitinib therapy. The primary endpoint was PFS at 3 months of dasatinib therapy. The secondary endpoints included PFS at 6 months, overall response rate (ORR), disease control rate (DCR), OS, and safety. Blood samples before dasatinib therapy were acquired for examination by NGS.
The study protocol adhered to the SPIRIT statement 7 and all procedures performed in this study involving human participants were approved by Beijing Cancer Hospital Ethics Committee. Written informed consent was obtained from the patients before study entry. The study was conducted according to the Declaration of Helsinki and the reporting of the study adhered to the CONSORT statement. 8 The clinical trial is registered with Clini caltr ial.gov (NCT02776878). between tissue and peripheral blood samples and additional KIT mutations were detected in the peripheral blood samples in 28.6% of the patients . Some SNV and CNV   such as ATRX, TP53, TEKT4, STK11,

| Study drug administration
Dasatinib (50 mg twice daily) was given orally for 1 week. If tolerable, the dosage of dasatinib was increased to 70 mg twice daily orally until tumor progression, appearance of unacceptable toxicities, or occurrence of death. Each dasatinib treatment cycle lasted 1 month.

| Efficacy and safety
Tumor assessment was evaluated by local radiological review according to the Response Evaluation Criteria in Solid Tumors (RECIST v1.1). All the tumor lesions were investigated using computed tomography at baseline within 2 weeks of enrollment before initiating therapy, and then every 6 weeks until disease progression. Physical examination, blood cell counts, blood biochemistry, and urine analysis were carried out at baseline and on day 1 of each cycle. Adverse events (AEs) were recorded according to Common Terminology Criteria for Adverse Events (CTCAE), version 5.0.

| Gene mutation examination
All the patients received KIT/PDGFRA genotype analysis with GIST tissue before and after imatinib resistance. Genomic DNAs were extracted from paraffin-embedded tumor specimens using E.Z.N.A. FFPE DNA Kit (Lot. D3399-1, OMEGA) according to the manufacturer's instructions. DNA fragments were aligned with exons 9, 11, 13, and 17 of KIT, and exons 12 and 18 of PDGFRA, and amplified by polymerase chain reaction (PCR) using specific primers. Sequencing results were analyzed with Chromas software. For patients with secondary resistance, exons 14 and 18 of KIT were also included.
Blood samples before dasatinib therapy were acquired for NGS by GENECAST Technology Co., Ltd.

| Library construction and sequencing
Buffy coat and plasma were separated by centrifugation at 1600 g for 10 minutes at room temperature. Genomic DNA and cfDNA were extracted from buffy coat and plasma, respectively. Genomic DNA was sheared into 150-200 bp fragments for DNA libraries construction. DNA libraries were captured with a targeted sequencing panel of 1406 genes that included major tumor-related genes.

| Identification of somatic mutations
VarDict and FreeBayes tools were used for somatic mutation calling with the following filters: (a) location in intergenic regions or intronic regions; (b) synonymous single nucleotide variations (SNVs); (c) allele frequency >0.002 in the database exac03; (d) allele frequency <0.01 in the plasma sample; (e) strand bias mutations in the reads; and (f) support reads <5.

| Copy number variation (CNV) calling and copy number instability (CNI) calculation
We used CNV kit (v0.9.2) to call CNV of plasma DNA samples. Paired blood samples were used as the baseline. A copy number ≥2.5 and ≤1.5 (in plasma-free DNA) was used to categorize altered regions as CNV gains (amplification) and copy number losses (deletions), respectively. After correction for GC content and length of the target region using proprietary algorithms for each region, the read counts were transformed into log2 ratios and converted into Z-score based on Gaussian transformations vs a normal control group (n = 30). The target regions that satisfied the Z-score greater than the 95th percentile plus two times absolute standard deviation (SD) of the normal control group were retained, and these Z-scores were summed as the CNI score.

| Quality control
VarDict, a software for detecting PCR artifacts, was used for SNV calling. Strict filter standards were introduced to ensure validation: allele frequency should be less than 0.002 in the database exac and greater than 0.01 in the plasma sample; furthermore, support reads should be greater than 5, and the mutations should be without strand bias.

| Statistical considerations
Based on the GRID trial, 3 the 3-month PFS rate was 11% in patients who received placebo after imatinib and sunitinib failure. Assuming a 3-month PFS rate of 50% with dasatinib treatment, a default rate of 10%, and a one-sided level of significance of 0.05, an overall sample size of 57 subjects was required to achieve 80% power at a 0.05% significance level in a two-sided log-rank test.
All statistical analyses were undertaken using SPSS 19.0 (SPSS Inc.). PFS and OS curves were constructed by the Kaplan-Meier method and compared using a log-rank test. In order to adjust for confounding variables, Cox proportional hazards models were used to estimate the simultaneous effects of prognostic factors on PFS. Frequency and percentage descriptions were used for categorical variables and chi-squared test was conducted to compare the incidence of different events. If the theoretical frequency was lower than 1, Fisher's exact test was conducted.

| Patient demographic and baseline characteristics
From May 2016 to June 2018, 58 patients from 9 medical centers across China were enrolled in this study. The median age of the patients at the time of study entry was 55 years (range, 24-77 years). In all, 43 patients (74.1%) were men, and 35 patients (60.3%) had ECOG performance status score of 0-1. GIST progressed in all the patients after imatinib and sunitinib treatment failure. Among them, 54 patients completed at least one imaging assessment. All the patients underwent KIT/PDGFRA genotyping. In total, 39 patients were examined for secondary gene mutations in GIST tissues from cytoreductive surgery or biopsy of progressing lesion after imatinib treatment failure. In addition, 22 peripheral blood samples before dasatinib treatment were obtained to examine gene mutations in circulating tumor DNA (Table 1).

| Efficacy measures according to patient genotypes
In two patients who achieved PR, one had a gastric GIST with KIT exon 11 V560D mutation and secondary exon 17 N820Y mutation, and the other had jejunum GIST with primary KIT exon 17 N822K mutation. In four patients with PDGFRA D842V mutation, two patients had SD. A patient with secondary KIT exon 17 N822K mutation had a PFS of 15.5 months, and was still alive with OS reaching 27 months.

| Safety
The dose of dasatinib was not escalated to 70 mg twice daily in 3 (5.2%) patients due to grade 2 fatigue. In addition,

| Concordance between liquid biopsy and tissue sequencing
NGS was undertaken of GIST tissues and blood samples in 21 patients and yielded identical results from the tissue and blood samples in six cases. Additional KIT mutations were detected in peripheral blood samples in six other cases including two cases with wild-type GIST. However, no KIT or PDGFRA mutations were detected in peripheral blood in eight cases with gene mutation in tissue examination including exon 11 point mutation, exon 11 deletion, exon 9 duplication, and D842V point mutation ( Table 2). Five of the eight cases had a low tumor burden.
Of the 21 patients, no secondary Kit mutation was detected either in the GIST tissue or peripheral blood samples

| The relationship of NGS results and efficacy of dasatinib
There was no statistical difference in SNV or CNV including Kit mutation between patients with PFS < 3 and ≥3 months. However, APC and/or TP53 mutation was only detected in patients with PFS ≥ 3 months. On the contrary, SUFU and TEKT4 mutation were found exclusively in patients with PFS < 3 months. The only one patient with SRC mutation had a PFS of 4.5 months. Moreover, patients with OS < 12 months had a significantly higher proportion of TP53 mutations than patients with OS > 12 months than (P = .012). In addition, patients with SDHC and TMEM127 copy number loss had longer OS (P = .035, .006) compared with patients with normal copy number.

| DISCUSSION
In phase I study and II clinical studies, dasatinib showed good antitumor activity, but had no obvious advantage over imatinib and sunitinib. 5,6 The inhibition by dasatinib of GISTs harboring KIT exon 17 mutation and SRC has led to more expectations for the drug as third-line treatment. In this study, the 3-month PFS rate was 53.4% in GIST patients receiving dasatinib therapy. However, in the study, patients received image examination every 6 weeks and some patients underwent imaging evaluation later than the planned 12 weeks after the start of dasatinib therapy. Therefore, the 3-month PFS rate could be lower than 53.4%. In order to evaluate the PFS of dasatinib therapy more accurately, we analyzed the 6-month PFS rate. Compared with the current standard thirdline treatment with regorafenib, dasatinib achieved a similar DCR, but both the 3-and 6-month PFS rates and even OS were lower. 3 Dasatinib showed moderate antitumor activity but failed to show the potential to surpass regorafenib as third-line treatment of GIST. In safety analysis, the main TEAEs were anemia, proteinuria, fatigue, and diarrhea. Dose reductions and treatment interruptions occurred in a few patients, but TEAEs were tolerable. However, we need to be vigilant that a small number of patients happened gastrointestinal bleeding.
Dasatinib is a small molecule, adenosine triphosphate competitive inhibitor of KIT, PDGFR, and SRC family of kinases. Therefore, it is very important to evaluate the correlation between genotyping, especially secondary gene mutations, and the efficacy of dasatinib. In vitro studies have shown that dasatinib was less active against KIT activation loop mutant isoforms than against juxtamembrane domain mutant KIT except for D816Y mutation. 9 Activated loop mutation is just the main type of KIT secondary mutation. However, unfortunately, neither primary nor secondary gene mutations were found to predict the efficacy of F I G U R E 3 Single nucleotide variation (SNV) changes in the blood samples of 21 gastrointestinal stromal tumors patients dasatinib in this study. The results of subgroup analysis in this study are consistent with those in vitro studies. This suggests that compared with the selective inhibition by sunitinib of tumors with secondary exon 13 V654A mutation 10 and by regorafenib of tumors with secondary exon 17 mutation, 11 dasatinib as third-line treatment has no advantageous selectivity for KIT secondary mutation types. However, it was found that wild-type GIST patients seemed to have longer PFS, showing that dasatinib as third-line treatment may be more effective for this subgroup of patients. In addition, GIST with PDGFRA D842V mutation is considered to be resistant to almost all existing TKIs. 12 In previous study, avapritinib has shown excellent efficacy for GISTs harboring D842V mutation. 13 In a second-line study, dasatinib was used to treat GIST patients with D842V mutation and PR was obtained in two cases. 6 In the current study, 50% of D842V mutated GIST patients had SD after dasatinib therapy. Dasatinib could still be used as an alternative drug for the treatment of GIST patients with D842V mutation before avapritinib is approved.
In addition, effective treatment of wild-type GIST suggested that inhibition of SRC or downstream signaling pathways of KIT could play another important role in the treatment of drug-resistant GIST with dasatinib. In the second-line treatment study of dasatinib, SRC-positive patients showed better efficacy. 6 In this study, no immunohistochemical staining of SRC was performed, but SRC mutation was found in peripheral blood sample of a patient by NGS. The PFS of this patient reached 4.5 months, which is better than the overall level.
In this study, 21 patients underwent NGS of peripheral blood samples to explore the role of activation of other pathways in resistance to GIST besides secondary KIT gene mutation and whether other gene mutations could predict the efficacy of dasatinib treatment. In the 21 cases, KIT gene mutation still accounted for the main proportion of mutations detected in the blood by NGS. At the same time, we detected some other SNVs such as ATRX, JUN, PTCH1, TP53, TEKT4, FOXP1, and SUFU and CNVs such as STK11, SDHC, and CDKN2C related to tumor signaling pathways. In the peripheral blood of eight patients without secondary gene mutation, we found that SNV and CNV mutations were mainly concentrated in the RAS pathway and the PI3K pathway. In previous studies, RAS mutation may be one of the pathogenesis of wild type GIST, 14,15 and activation of the PI3K pathway in drug-resistant GIST was thought to be associated with secondary drug resistance. 16,17 Therefore, we speculate that the activations of these signaling pathways in blood may be other mechanisms of drug resistance besides KIT secondary mutations. In in vitro studies, dasatinib showed moderate effect on the PI3K/AKT pathway by inhibition of SRC. 18 In this study, the 3-month PFS of patients treated with dasatinib who had no gene mutation suggests that dasatinib may effective in the inhibition of KIT downstream signaling pathways.
We also analyzed the predictive power of peripheral blood gene variations on the efficacy of dasatinib. An interesting finding is that TP53 mutations occur only in patients with longer PFS, while patients with TP53 mutations had longer OS, which is inconsistent with previous results that TP53 was considered a negative prognostic factor. 19,20 It was found that dasatinib combined with doxorubicin played a stronger role in chondrosarcoma with TP53 mutation. 21 In addition, GIST patients with SDHC and TMEM127 copy number loss had longer OS. However, the natural history of SDH deficient tumors is rather indolent, with longer disease control time than other tumors with different mutations. Meanwhile, the results were from a small sample size. Therefore, we cannot draw the conclusion that SDHC and TMEM127 copy number loss are related to the efficacy of dasatinib treatment.
Compared with NSG using tumor tissues, one of the main advantages of NSG using liquid biopsy is that it can overcome the heterogeneity of tumors and detect other mutant genes which are not found in a single tissue sample. This is also confirmed in this study. In 21 patients, the concordance of KIT/PDGFRA between tissue and peripheral blood samples was 61.9%. In 28.6% of the patients, new KIT mutations were detected in the peripheral blood, including secondary gene mutations that were not detected by tissue examination, which is very helpful to guide drug therapy. However, KIT/PDGFRA mutations were not detected in peripheral blood of eight patients, and five of them had a low tumor load. The failure to detect the mutations may be related to the insufficient content of GIST circulating tumor DNA in the peripheral blood. In previous studies, it has been proved that it is difficult to detect circulating tumor DNA in GIST with too low a tumor load. 22 In the GRID trial, the concordance rate was 84% between plasma and tissue for detection of primary KIT mutations and Beads Emulsion Amplification Magnetics technology was less sensitive for the detection of primary KIT exon 11 mutations in plasma DNA. 23 Therefore, at this stage, liquid biopsy could not replace tumor tissue sequencing for GIS T genotype analysis.
In summary, dasatinib shows moderate activity in metastatic GIST, and the toxicities are tolerable. Dasatinib could offer a therapeutic option for patients who cannot take regorafenib, especially those with wild-type GIST or D842V mutation GIST. Our NGS data suggest that dasatinib may also suppress signaling pathways apart from KIT, which warrants further study.