Mutational landscape of gastrointestinal stromal tumors using next-generation sequencing of a 73-gene panel

Abstract

Purpose

Mutational analysis guides therapeutic decision-making for patients with advanced-stage gastrointestinal stromal tumors (GISTs).

Materials and methods

A total of 491 patients with GISTs were retrospectively included, and their genomic profiles were determined by targeted NGS of 73 or 1,021 gene panels.

Results

KIT mutations were identified in 84.7% (416/491) of patients with GISTs, and PDGFRA mutations were identified in 4.5% (22/491) of patients. Among the patients with KIT mutant-GISTs, most had KIT mutations in exon 11 (81.3%, 338/416). PDGFRA mutations were located mainly in exon 18 (77.3%, 17/22). In the remaining 11% of patients with GISTs without KIT or PDGFRA mutations (wild-type GISTs), BRAF and NF1 were the most commonly mutated genes. Compared with KIT/PDGFRA-mutant GISTs, wild-type GISTs were associated with younger age, a greater proportion of female patients, and lower levels of copy number variations. Notably, the concomitant alterations of KIT/PDGFRA-mutant or wild-type GISTs were similar when sequencing data from a 73-gene panel and a 1,021-gene panel were analyzed. The genetic landscape of patients with treatment-naïve GISTs at different locations was described by using a 73-gene panel, and compared with patients with gastric GISTs, patients with nongastric GISTs had higher levels of copy number variations but a lower proportion of PDGFRA mutations. Analysis of the genomic profiles before and after imatinib treatment revealed KIT T670I mutation, as well as ATM and JAK2 mutations, as potential underlying mechanisms of resistance.

Conclusion

Using a 73-gene panel, we characterized the molecular characteristics of GISTs and revealed a correlation with their clinical features. Moreover, KIT/PDGFRA-dependent and KIT/PDGFRA-independent mechanisms underlying resistance to imatinib were explored. Overall, our 73-gene panel is sufficient for clinical application in cases of GISTs.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12957-025-04085-6.

Introduction

Gastrointestinal stromal tumors (GISTs) originate from the gastrointestinal mucosa of Cajal interstitial cells and are the most common subtype of soft tissue sarcoma, accounting for 0.1–3.1% of all gastrointestinal malignancies [1]. In the past few decades, the incidence of gastrointestinal stromal tumors has been increasing, typically in middle-aged and elderly people; however, although the incidence rate is not significantly different between men and women, it is slightly higher in men [2]. GISTs mainly occur in the stomach (60%−70%), followed by the small intestine (20%−25%), colon and rectum (5%), and esophagus (< 1%) [3]. Approximately 85% of GISTs harbor mutations in the KIT or PDGFRA genes, whereas the remaining 15% of cases do not harbor these mutations and are termed wild-type GISTs [4]. Among KIT/PDGFRA-mutant GISTs, approximately 60% are driven by KIT exon 11 (insertion and/or deletion), which accounts for approximately 80% of all KIT mutations [5]. KIT exon 11 mutations occur throughout the gastrointestinal tract, with the most common site being the proximal stomach [6]. KIT exon 9 mutations account for approximately 10% of all KIT mutations [5] and arise in the small intestine, colon, or rectum, and few such GISTs have been reported in the stomach [7]. Furthermore, driver mutations of KIT exons 8, 13 or 17 are rare and occur mainly in the intestines [8]. KIT and PDGFRA mutations are common in GISTs [9]. PDGFRA D842V is the most common PDGFRA mutation and accounts for approximately 79% of all PDGFRA mutant-GISTs [10]. Almost all the GISTs in the patients with PDGFRA mutations originate in the stomach or omentum [11]. The remaining 15% of patients with wild-type GISTs have receptor tyrosine kinase gene fusions (NTRK3 or FGFR1, 1.0%), SDH-deficient (9.0%), NF1 (2.0%), PIK3CA (0.9%), BRAF (0.8%), RAS-mutant (0.4%), or others (0.9%) [8]. Notably, a vast majority of these oncogenic drivers have therapeutic potential [12]. However, the mutation spectrum of patients with GISTs has not been fully elucidated, and the identification of additional cancer genes may extend the molecular spectrum of GISTs and reveal potential therapeutic targets.

As a tyrosine kinase inhibitor, imatinib has become the first-line standard targeted therapy for advanced localized and metastatic GISTs with KIT or PDGFRA mutations [13]. More importantly, the treatment response and clinical benefit of imatinib are closely related to the molecular characteristics of the GIST [6]. Notably, most GISTs with KIT exon 11 mutations are highly sensitive to first-line therapy with imatinib, but progression-free survival may vary widely among patients [14]. Although PDGFRA D842V-mutant GISTs do not respond to imatinib, the natural history shows that recurrence-free survival is more favorable in patients with GISTs with any PDGFRA mutation than in those with KIT mutation [15]. Therefore, clarifying the molecular characteristics of GIST tissues is extremely important for the selection of treatment strategies.

With the development of NGS technology, the use of imatinib for targeting KIT has marked a new era in the treatment of GISTs, creating a precision treatment for all solid malignant tumors. However, GISTs become more difficult to treat by acquiring mutations that are resistant to imatinib treatment [16]. The major mechanism of resistance to imatinib involves secondary KIT mutations, including those in exons 13, 14, 17 or 18 [17]. To date, sunitinib, regorafenib and ripretinib have been approved by regulatory authorities for use as second-line, third-line, and fourth-line multikinase inhibitors after imatinib treatment fails [18]. Consequently, a comprehensive analysis of the mechanism responsible for resistance to imatinib treatment is necessary to overcome primary and secondary drug resistance.

In this retrospective study, we aimed to analyze the molecular characteristics of GISTs, especially the differences in molecular characteristics between gastric and nongastric GISTs and to assess the potential mechanism responsible for resistance to imatinib treatment.

Materials and methods

Patients and samples

This was a retrospective study, and patients were not enrolled consecutively. A total of 491 patients with GISTs who received commercial next-generation sequencing of 73 or 1,021 cancer-related gene panels (Geneplus-Beijing, Beijing, China) between January 2013 and September 2022 were included. The main inclusion criteria were as follows: (i) patients aged > 18 years with stage I-IV GISTs identified by pathological examination and (ii) patients with no malignant tumor history within the past 5 years. The exclusion criteria were as follows: (i) histology revealed that the tumor was not a GIST and (ii) samples were of insufficient quality. Clinicopathological data, including age, sex, tumor histology, smoking status, and treatment information, were collected from the patients’ medical records. Written informed consent was obtained from each patient. All procedures were conducted in adherence to the principles of the Helsinki Declaration.

DNA extraction and targeted next-generation sequencing

Tumor DNA was isolated from 5 μm thick formalin-fixed, paraffin-embedded (FFPE) tissue samples using commercial kits (Qiagen, Hilden, Germany). Blood samples were centrifuged for 10 min at 1600 g, and peripheral blood leukocytes (PBLs) were collected to extract germline genomic DNA. The supernatant was centrifuged at 16,000 g for 10 min to remove cell debris, transferred to microcentrifuge tubes and then stored at −80 °C. Cell-free DNA (cfDNA) was isolated using the QIAamp Circulating Nucleic Acid Kit (Qiagen, Hilden, Germany). Germline genomic DNA from PBLs was isolated using the QIAamp DNA Blood Mini Kit (Qiagen). The concentration and fragment distribution of cfDNA were measured using an Agilent 2100 Bioanalyzer (Agilent Technologies, Inc., Santa Clara, CA, USA). A KAPA Library Preparation Kit (Kapa Biosystems, Wilmington, MA, USA) was used to prepare the indexed Illumina NGS libraries. Custom-designed 1,021 or 73 cancer-related gene panels were used to hybridize the DNA libraries, and their selected regions and genes are listed in Tables S1 and S2. The hybridized libraries were sequenced using a 100-bp paired-end configuration on a DNBSEQ-T7RS sequencer (MGI Tech, Shenzhen, China).

Duplicated reads were marked and removed using the MarkDuplicates tool in Picard (version 4.0.4.0; Broad Institute) for tumor and germline genomic DNA. For ctDNA, duplicated reads were identified by UID and the position of template fragments to eliminate errors introduced by PCR or sequencing using realSeq (v3.1.0 Geneplus-Beijing, inhouse). After the removal of terminal adaptor sequences and low-quality reads, the remaining reads were mapped to the reference human genome (hg19) and aligned using the Burrows-Wheel Aligner (version 0.7.12-r1039) with default parameters. GATK (3.4–46-gbc02625) and MuTect2 (1.1.4) were used to call somatic single nucleotide variants (SNVs) and small insertions and deletions (InDels), respectively. Contra (2.0.8) was used to identify copy number variations. NCsv (in-house algorithm 0.2.3) was employed to detect structural variants (SVs). All candidate variants were manually confirmed by using the Integrative Genomics Viewer browser. Variants were filtered out if any of the following occurred: (a) the mutation was detected in at least 5 high-quality reads containing the particular base; (b) the mutation was not present in > 1% of the population in the 1000 Genomes Project (version 3) or dbSNP databases (The Single-Nucleotide Polymorphism Database, version dbSNP 137); (c) the mutation was not present in a local database of normal samples; (d) high-quality reads were selected with a Phred score ≥ 30, mapping quality ≥ 30, or a lack of paired-end read bias.

Statistical analysis

All statistical analyses were conducted using R (version 4.0.2). Graphs were drawn using R or GraphPad Prism (version 8.0.1). Unpaired Student’s t tests were used to compare the differences between two groups of normally distributed variables, whereas the Wilcoxon test was used to compare nonnormally distributed variables. Chi-square analysis or Fisher’s exact probability test was used for group comparisons. All tests were two-tailed, with a P value < 0.05 considered to indicate statistical significance.

Results

Molecular subtypes of 491 gists

A total of 491 GISTs were included in this study. The median age at diagnosis was 63.0 years (range, 31–91), with 53.6% (263/491) of patients being male and 46.4% (228/491) of patients being female. Among the 491 tumor tissues used for NGS, 69.2% (340/491) of the samples were from primary lesions, 8.8% (43/491) were from metastatic lesions, and 22.0% (108/491) were of unknown origin (Table S3). A 1,021-gene panel was submitted for NGS for 180 (36.7%, 180/491) patients, and a 73-gene panel was submitted for NGS for 311 (63.3%, 311/491) patients. Because the entire region of the 73-gene panel was covered in the 1,021-gene panel, we analyzed the sequencing data from the 73-gene panel of all the patients to elucidate the genetic landscape of the GISTs.

A total of 1298 somatic alterations were identified, with a mean of 2.6 alterations per sample. Among these alterations, 760 were substitutions/indels, with 66 amplifications, 438 deletions, and 11 fusions/rearrangements (Fig. 1a and b). As previously reported [9], the KIT and PDGFRA mutations were mutually exclusive. Moreover, KIT mutations were also mutually exclusive with FBXW7, PMS2, and CHEK2 mutations (Fig. 1c). KIT mutations were observed in 84.7% (416/491) of patients and mainly occurred in exon 11 (81.3%, 338/416), followed by exons 9 (13.7%, 57/416), 17 (7.5%, 31/416), 13 (5.3%, 22/416), 18 (1.2%, 5/416) and other rare regions (exons 1, 2, 3, 5, 8, 14, 15, 16, 20) (Fig. 1d and Figure S1). The juxtamembrane domain of KIT was detected as the most frequently mutated region (96.4%), followed by the adenosine triphosphate binding pocket (8.7%), the kinase domain activation loop (6.3%) and others (3.1%). Two KIT oncogenic types were identified, including missense and in-frame mutations, which are documented as pathogenic in the OncoKB database (https://www.oncokb.org). Across all KIT oncogenic mutations, A502_Y503dup was the most common mutation (11.6%), followed by V559D (9.2%), W557_K558del (8.3%), W557R (4.4%), D579del (3.1%), V560D (3.1%), N822K (3.1%), L576P (2.6%), V654A (2.4%) and other mutations (Fig. 1d and Figure S1). PDGFRA mutations were found in 4.5% (22/491) of the GISTs and were located mainly in the region of exon 18 (77.3%, 17/22), followed by exons 12 (13.6%, 3/22), 8 (4.5%, 1/22), and 4 (4.5%, 1/22) (Fig. 1e). The kinase domain activation loop of PDGFRA was detected as the most frequently mutated region (77.3%), followed by the juxtamembrane domain (13.6%) and others (9.1%). A total of two PDGFRA oncogenic types, missense and in-frame mutations, which are documented as pathogenic in the OncoKB database, were identified. Among all PDGFRA oncogenic mutations, D842V was the most common mutation (70.0%), followed by I843_D846del (15.0%), V561D (10%), and S566_E571delinsR (5.0%) (Fig. 1e). Apart from KIT and PDGFRA mutations, BRAF and NF1 mutations were the most prevalent in our cohort.

Fig. 1.

Molecular characterization of GISTs. (a) Co-mutation plot from next-generation sequencing of 491 GISTs, including patient demographics and clinical features. (b) Summary of gene variant classification in 491 GISTs. (c) Statistics of co-occurring and mutually exclusive gene pairs. Star/Spot indicates statistical significance (p values from Fisher’s test), * P < 0.01, · P < 0.05. (d) Mutation status of 416 KIT-mutant GISTs. (e) Mutation status of 22 PDGFRA-mutant GISTs

In our cohort, 54 (11%) cases that did not have KIT or PDGFRA mutations were considered wild-type GISTs. The genes most commonly altered in wild-type GISTs were BRAF (9%, 5/54) and NF1 (9%, 5/54), which are involved in the RAS/MAPK pathway (Figure S1). A comparison of the clinical features between KIT/PDGFRA-mutant and wild-type GISTs revealed that patients with wild-type GISTs were younger and were more likely to be female (Table S4), which is consistent with previous literature [19]. Furthermore, compared with wild-type GISTs, KIT/PDGFRA-mutant GISTs had higher proportions of MSH2 mutations (7.5% vs. 0%, P = 0.04) and RB1 mutations (11.0% vs. 0%, P = 0.007) but a lower proportion of BRAF mutations (0.0% vs. 9.3%, P = 0.001) (Fig. 2a). Additionally, KIT/PDGFRA-mutant GISTs had a higher prevalence of CDKN2A (17.4% vs. 3.7%, P = 0.005) and CHEK2 (15.1% vs. 3.7%, P = 0.02) copy number deletions (Fig. 2b).

Fig. 2.

Molecular characterization of KIT/PDGFRA-mutant GISTs and Wild-type GISTs. (a) Comparison of gene mutation frequency between KIT/PDGFRA-mutant GISTs (N = 437) and Wild-type GISTs (N = 54). (b) Comparison of copy number amplification/deletion frequency between KIT/PDGFRA-mutant GISTs and Wild-type GISTs. Horizontal dotted line: −log₂ (p value = 0.05), vertical dotted line: log₂ (odds ratio (OR) = 1). Left side of vertical dashed line: KIT/PDGFRA-mutant GISTs, right side of dashed line: Wild-type GISTs. Genes with relatively higher mutation frequencies are marked in red. Values in parentheses represent the ratio of mutation frequency within the groups. Fisher’s test was used for statistical analysis, and the P-value was calculated with a threshold of 0.05. SNVs: single nucleotide variations, CNVs: copy number variations

To determine whether a larger gene panel would provide a more comprehensive genetic landscape of GISTs, 1021 genes from 180 patients were analyzed, including 164 (91.1%, 164/180) with KIT/PDGFRA mutations and 16 (8.9%, 16/180) with KIT/PDGFRA mutations (wild-type) (Table S3). KIT, PDGFRA, TP53, and NF1 were the most commonly mutated genes in these patients, which is consistent with the sequencing data captured by the 73-gene panel (Figure S2a-d). Patients with KIT/PDGFRA mutations had a lower prevalence of CTNNB1 mutations (0.0% vs. 18.8%, P = 0.002) than those with KIT/PDGFRA wild-type (Figure S2e-f), although their clinical characteristics were similar (Table S5). Overall, the 1,021-gene panel did not provide more information on the molecular characteristics of GISTs.

As an exploratory part of our study, we conducted a detailed analysis of the clinical and genetic profiles of NF1-mutant tumors. The prevalence of PIK3CA (15.0% vs. 1.7%, P = 0.007) and DDR2 mutations (5.0% vs. 0.0%, P = 0.04) was greater in patients with NF1 mutations than in those with wild-type mutations (Figure S3a-S3b), although their clinical characteristics were similar (Table S8).

Clinical and mutational characteristics of patients with gastric and nongastric gists

Previous studies have revealed that genetic profiles differ greatly across patients with GISTs with different primary tumor locations [6]. We further investigated the molecular characteristics of treatment-naïve KIT/PDGFRA-mutant gastric and nongastric GISTs on the basis of our sequencing data. One hundred thirty-one patients with gastric GISTs and 91 patients with nongastric GISTs (small intestine, rectum, colon, and duodenum) were included in this exploratory investigation. The patients with gastric GISTs were younger than those with nongastric GISTs were (mean age at diagnosis 50.6 vs. 56.5 years, P = 0.02). All GISTs with PDGFRA mutations had a unique gastric location (100%, 12/12) (Table S6), which is in accordance with previous results [20]. Compared with patients with nongastric GISTs, patients with gastric GISTs had a higher prevalence of KIT exon 9 mutations (17.6% vs. 2.5%, P < 0.001) but a lower prevalence of KIT exon 11 mutations (76.9% vs. 94.1%, P < 0.001) (Table S6). We observed that compared with patients with gastric GISTs, patients with nongastric GISTs had higher proportions of RB1 mutations (16.5% vs. 2.3%, P = 0.009) and copy number deletions of RB1 (16.5% vs. 2.3%, P < 0.001), BRCA2 (13.2% vs. 1.5%, P = 0.001), MSH2 (8.8% vs. 0.8%, P = 0.004), MSH6 (7.7% vs. 0.8%, P = 0.009), ATM (6.6% vs. 0.0%, P = 0.004), EGFR (4.4% vs. 0.0%, P = 0.03), and EPCAM (4.4% vs. 0.0%, P = 0.03) (Fig. 3a and b).

Fig. 3.

Mutation analysis within gastric GISTs and nongastric GISTs. (a) Comparison of gene mutation frequency between gastric GISTs (N = 131) and nongastric GISTs (N = 91). (b) Comparison of copy number amplification/deletion frequency between gastric GISTs and nongastric GISTs. Horizontal dotted line: −log₂ (p value = 0.05), vertical dotted line: log₂ (odds ratio (OR) = 1). Left side of vertical dashed line: Nongastric GISTs, right side of dashed line: Gastric GISTs. Genes with relatively higher mutation frequencies are marked in red. Values in parentheses represent the ratio of mutation frequency within the groups. Fisher’s test was used for statistical analysis, and the P-value was calculated with a threshold of 0.05. SNVs: single nucleotide variations, CNVs: copy number variations

To further verify the clinical performance of the 73-gene panel, a 1021-gene panel from 101 treatment-naïve patients with KIT/PDGFRA-mutant GISTs was used to compare the differences in molecular characteristics between gastric and nongastric GISTs. Among them, 60 patients had gastric GISTs, and 41 had nongastric GISTs. Compared with those who had gastric GISTs, those who had nongastric GISTs had a higher prevalence of KIT exon 9 mutations (22.0% vs. 3.6%, P = 0.007) but a lower prevalence of KIT exon 11 mutations (73.2% vs. 94.6%, P < 0.001), which is consistent with the sequencing data captured by the 73-gene panel (Table S7). Collectively, it is possible and sufficient to determine molecular characteristics by using a 73-gene panel.

Mechanisms of resistance to imatinib treatment

The pre- and post-treatment genomes of 4 patients (P001, P002, P003, and P004) were analyzed for potential resistance to imatinib treatment. All four patients carried KIT mutations (Fig. 4). Only one patient still harbored KIT alterations after imatinib treatment. We first assessed KIT-dependent resistance to imatinib treatment. For patient P002 harboring KIT M552_E554del and receiving imatinib as a first-line treatment for 22 months, KIT T670I was identified at disease progression (Fig. 4).

Fig. 4.

Mutation analysis identifies known and novel acquired resistance patterns in 4 imatinib-resistant GISTs. Wheat color represents gene mutations that were detected in pre-treatment (T, tumor tissue samples)/post-treatment (C, ctDNA). Green color represents copy number deletion. Blue color represents copy number amplification. Red fonts represent imatinib-resistant GISTs carrying oncogenic mutations

To explore KIT-independent resistance to imatinib treatment, we analyzed the alterations acquired from three patients who did not have secondary alterations in KIT upon disease progression (PD). Mutations reported in OncoKB datasets as oncogenic mutations were analyzed in detail. Two oncogenic mutations were identified in two patients. ATM participates in the DNA damage repair (DDR) pathway, and its oncogenic mutation was identified in one patient (P001) who received first-line treatment with imatinib. Additionally, a JAK2 oncogenic mutation involved in RAS/MAPK pathway regulation was detected in one patient (P003), suggesting that RAS/MAPK pathway dysregulation might also play a role in resistance to KIT-targeted therapy.

Concordance between tumor tissues and ctdna

Next, we explored the clinical utility of ctDNA in GISTs. Patients who had no ctDNA alterations or variations of uncertain significance were excluded. Nine patients with KIT mutations whose paired tissue and plasma samples were subjected to NGS were included in this study. When we analyzed KIT mutations in ctDNA at baseline, we found 44.4% mutation concordance between ctDNA and matched tumor tissue (Fig. 5).

Fig. 5.

Concordance between tissues and ctDNA. Consistency was evaluated between the tumor tissue and paired plasma samples. Blue bars represent tumor tissue samples, yellow bars represent paired plasma samples 

Discussion

While it is well established that the mutational status of GISTs significantly influences therapeutic decisions, more comprehensive data on how specific mutation patterns impact long-term clinical outcomes and guide personalized treatment strategies are needed. It is urgent to fully understand the mutational landscape of GISTs. Our cohort of 491 patients with GISTs was retrospectively analyzed in detail to understand the molecular characteristics of GISTs. KIT and PDGFRA are the most commonly mutated genes in these patients, which is consistent with the findings of previous studies [20]. Research has indicated that KIT mutations are mutually exclusive with PDGFRA mutations in patients with GISTs [9]. While KIT and PDGFRA mutations are typically mutually exclusive in patients with GISTs, with each tumor usually harboring mutations in only one of these genes, there are rare instances where both mutations can coexist [21]. These findings highlight the complexity and heterogeneity of GISTs and underscore the importance of comprehensive genetic testing to identify all potential actionable mutations. Additionally, the mutual exclusivity of KIT with FBXW7, PMS2, and CHEK2 mutations in patients with GISTs suggests that these pathways may be functionally redundant or compensatory. These findings highlight the importance of comprehensive genomic profiling to identify the most relevant therapeutic targets for individual patients. Further research is needed to fully understand the clinical implications of these findings. The coalteration profiles of the patients with GISTs with or without KIT/PDGFRA mutations were studied. Our results revealed that patients with KIT/PDGFRA mutations were older and had higher incidences of RB1 and MSH2 mutations and CDKN2A and CHEK2 deletions than those with wild-type KIT/PDGFRA mutations did. Given that RB1 mutations and CDKN2A and CHEK2 deletions are correlated with a poor prognosis of GISTs [22–24] and that MSH2 mutations play a role in the occurrence and development of cancer [25], we suggest that patients with KIT/PDGFRA wild-type mutations exhibit weaker oncogenic features than those with KIT/PDGFRA mutations do. While the literature indicates that FGF3/FGF4 amplifications are predominantly found in KIT/PDGFRA wild-type GISTs [26], the lack of a significant difference in the incidence of these mutations between the patients with KIT/PDGFRA wild-type GISTs and the patients with KIT/PDGFRA-mutant GISTs suggests that these amplifications may not be exclusive to patients with KIT/PDGFRA wild-type GISTs. These findings have important implications for understanding the genetic landscape and potential therapeutic targets in patients with GISTs.

Additionally, in our cohort, NF1-mutant GISTs had a specific genetic profile. Our analysis highlights the need for a tailored approach for the clinical management of these tumors. Future studies should focus on further elucidating the genetic landscape of NF1-mutant GISTs and exploring targeted therapies for this distinct subgroup.

Interestingly, we have demonstrated that mutational signatures vary substantially across different anatomical locations, which is well known from the literature [6]. PDGFRA-mutant GISTs mainly occur in the stomach, whereas KIT-mutant GISTs could arise in the stomach, small intestine, or rectum, suggesting an innate and important biological difference between these driver mutations. Studies have revealed that the copy number variation of tumor-related genes strongly affects the occurrence and metastasis of tumors [27, 28]. Compared with patients with gastric GISTs, patients with nongastric GISTs were younger and had a higher prevalence of RB1 mutation and RB1, BRCA2, MSH2, MSH6, ATM, EGFR, and EPCAM deletion, indicating that gastric GISTs exhibit fewer malignant features than nongastric GISTs do. Overall, this research further emphasizes the importance of considering tumor localization in clinical decision-making. Mutation landscapes between different gene panel tests of GISTs were studied. Patients who underwent 1,021-gene panel tests were analyzed separately and showed concordance with the sequencing data captured by the 73-gene panel. This means that a less expensive 73-gene panel is more suitable for clinical applications in cases of GISTs when financial resources are limited.

We investigated the potential mechanisms of imatinib treatment resistance in patients with GISTs by comparing the genomic profiles before and after treatment. Consistent with previous reports, KIT T670I mutation, which is a secondary imatinib-resistant mutation, was detected after treatment [29]. KIT T670I oncogenic mutation was identified in one patient with PD, and the presence of these secondary mutations might indicate a more aggressive disease course and necessitate more intensive monitoring and the development of novel TKI therapies that specifically target secondary mutations to improve the clinical benefit for patients [30]. In summary, secondary mutations in patients with GISTs have significant clinical implications for treatment planning and resistance management. By identifying these mutations, clinicians can tailor therapies to improve patient outcomes and overcome resistance mechanisms. Two patients carried oncogenic mutations involved in the DNA damage repair (DDR) pathway or the RAS/MAPK pathway at PD. The participation of ATM in the DDR pathway, its oncogenic mutation, was identified in one patient who received first-line treatment with imatinib. The residual NSCLC tumor cells that survived targeted therapy and the subsequent repair of DNA damage are important reasons for drug resistance, and the combined use of EGFR inhibitors and ATM inhibitors can inhibit the growth of gefitinib-resistant tumor cells [31]. We speculated that ATM mutation might also play a role in imatinib treatment resistance. In addition, we identified a JAK2 oncogenic mutation in one patient who received first-line treatment with imatinib. Notably, JAK2 mutation might also indirectly lead to imatinib resistance by hyperactivation of the RAS/MAPK pathway ([32]– [33]). To overcome resistance mediated by these bypass pathways, further studies are needed to test the efficacy of imatinib therapy in treating patients with these mutations in in vivo and clinical trials.

Finally, we explored the consistency between the tissue and ctDNA detection of mutations. The consistency rate was 44.4% when KIT mutations were analyzed at baseline, which was slightly lower than the previously reported range of 56.0% to 100.0% ([34]– [35]) because the ctDNA detection rate was limited to patients with a limited tumor stage or who were not treated in our cohort.

This study has some limitations. First, this study is a retrospective study and lacks clinical data. Second, no SDH mutations were detected in our cohort because in 63.3% (311/491) of the patients, a 73-gene panel was used for uncovered SDH genes. Third, the number of patients used to evaluate the mechanism underlying the resistance to imatinib treatment was limited, and more research is needed to further confirm our findings. Finally, the number of patients with GISTs for whom both tissue NGS and plasma NGS were performed was limited; thus, verification by large prospective studies is needed.

Conclusion

In summary, our study demonstrated that a 73-gene targeted NGS panel captures the clinically actionable mutations in patients with newly diagnosed GIST, enabling risk stratification and genotype-guided therapy decisions. And large-scale prospective studies are needed to further test its clinical utility.

Authors’ contributions

### Conceived and designed the analysis: Chen R, Gao Z, Ye Y.### Collected the data and organized the database: Wu Y, Wang Chao, Wang B.### Contributed data or analysis tools: Wang C, Cheng B, Yang S, Yin H.### Performed the analysis: Wang C, Cheng B, Yang S, Yin H.### Wrote the paper: Wang C Yin H, Xiong Y.### All authors contributed to manuscript revision, read and approved the submitted version.

Funding

Declaration.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

All raw data supporting the article are available from the corresponding author upon reasonable request.

Ethics approval and consent to participate

Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

Competing interests

The authors declare no competing interests.