Outcome of Patients with Platelet-Derived Growth Factor Receptor Alpha–Mutated Gastrointestinal Stromal Tumors in the Tyrosine Kinase Inhibitor Era

,

In patients with advanced disease, PDGFRA mutations seem to be less frequent (1%-3%; refs.8, 9), probably as a result of the better prognosis associated with this mutation in patients with localized disease (10).In the advanced setting, the type of KIT mutation (exon 9 vs. exon 11) has been shown to influence the probability of response and the duration of progression-free survival (PFS; refs.7,11,12).Furthermore, evidence suggests that starting with a higher dose of imatinib (800 mg daily) upfront may benefit patients with exon 9-mutant GISTs.Sensitivity of PDGFRA-mutant GISTs to the tyrosine kinase inhibitor imatinib has been assessed in vitro (5) and in vivo in clinical trials (7,8,11,13), but studies have shown conflicting results, and only the D842V mutant has been consistently shown to be imatinib-resistant.Activity of imatinib in patients with PDGFRA-mutant GISTs has previously been reported but data are scarce because of the rarity of this subgroup of patients.Our goal in this study was to better assess the in vivo activity of tyrosine kinase inhibitors on a larger cohort of patients with PDGFRA-mutant GISTs treated with imatinib for advanced disease.

Patients and Methods
We conducted a retrospective survey among referral centers and collaborative groups with experience in conducting clinical studies in sarcoma and/or GISTs in Europe.These institutions/investigators were members of the European Organisation for Research and Treatment of Cancer (EORTC) Soft Tissue and Bone Sarcoma Group (STBSG), the French Sarcoma Group (GSF-GETO), the Italian Sarcoma Group (ISG), or the Spanish Sarcoma Group (GEIS).This study was approved by the ethics committee in Lyon, France (Comit e de Protection des Personne Lyon Sud-Est IV).In each institution's database, data were collected for all patients with PDGFRA-mutant GISTs treated with imatinib for advanced disease.Patients who had received prior imatinib in the adjuvant setting were excluded because the primary objective of this study was to assess primary sensitivity of PDGFRA-mutant GISTs to imatinib, and it was felt that prior imatinib given in the adjuvant setting may promote secondary resistance.Most of the samples (32 of 58) were analyzed using a similar technique: genomic DNA from the tumor tissues was extracted from paraffinembedded 10-mm sections using microdissection technique to reduce contamination with nonneoplastic tissue or from slides showing more than 80% tumor cells.Exons 9, 11, 13, and 17 of KIT and exons 12, 14, and 18 of PDGFRA were amplified by PCR, and amplicons were analyzed for mutations by a combination of D-HPLC prescreening (Transgenomic WAVE DHPLC system; Transgenomic, Ltd.) and bidirectional sequencing, as previously described (11,14,15).In one center, screening was done using capillary electrophoresis followed by sequencing of PCR amplified exons.Data were extracted from individual patients' files and analyzed.Because only expert centers were involved, there was no central review of either radiology or pathology of retrieved cases.All centers conducted routine follow-up imaging studies [computed tomographic (CT) scan or MRI] at 3-month intervals.In EORTC study 62005, CT scan was done after 2, 4, and 6 months of treatment and every 3 months thereafter (16).In EORTC study 62001, imaging studies were done every 8 weeks (17).Patients and tumor characteristics were described using the median and range for continuous variables and percentages with 95% confidence interval (CI) for categorical variables.Response was assessed per-investigator (or trial data) using Response Evaluation Criteria in Solid Tumors (RECIST) 1.0 (18) and described as a response rate (RR) defined as the percentage of patients with partial or complete response (CR).Overall survival (OS) was defined as the time from the date imatinib was started to the date of death.PFS was calculated as the time from the date imatinib was started to the date of disease progression or death whichever occurred first.Survival times were plotted using the Kaplan-Meier method.We compared different groups of PDGFRA mutants: PDGFRA exon 12 mutants, PDGFRA non-D842V-exon 18 mutants, and PDGFRA-D842V mutants.RRs between subgroups were compared with the c 2 and the Fisher exact test where appropriate, and survivals were compared using the log-rank test.All statistical analyses were conducted using the SPSS 12.0 package.

Patients
Institutional databases from 12 institutions and comprising a total of 3,510 patients with genotyped GISTs were searched for patients with PDGFRA-mutant GISTs.PDGFRA mutations were found in 382 patients (11%), of which 44 (1.2%) had advanced disease.Two patients were excluded from further analysis: one patient because he had received adjuvant imatinib before relapse and one patient because he was never treated due to rapid clinical deterioration.Data of 16 additional patients (28% of the patients in this study) were retrieved from EORTC database (studies EORTC

Translational Relevance
Gastrointestinal stromal tumors are driven by mutually exclusive activating mutations of KIT or plateletderived growth factor receptor-alpha (PDGFRA), which makes them sensitive to inhibition of these tyrosine kinases by small-molecule tyrosine kinase inhibitors such as imatinib and sunitinib.Previous studies have shown that the type of KIT mutations may impact patient management.In this study, we confirm that as described in previous studies, when PDGFRa is the mutated kinase in gastrointestinal stromal tumors, the location of the mutation in the PDGFRA gene also impacts sensitivity to currently available tyrosine kinase inhibitors, whereas some mutations, such as PDGFRA-D842V, seem to be clearly imatinib-and sunitinib-resistant. 62001 and 62005) in which genotype information was available for 465 patients with advanced GISTs.Overall, 58 patients from 12 European centers and from 2 EORTC studies were identified for this study.Their main characteristics are described in Table 1.The date of starting imatinib ranged from January 2001 to November 2010 (10 years).The median age at initiation of imatinib for advanced disease was 61 years (range, 19-83).Fifty-six patients (97%) had metastatic disease.Two patients had locally advanced or recurrent disease.Eighteen patients (31%) initially presented with metastatic disease, whereas 40 suffered from metastatic and/or local recurrence at a median of 20.2 months (range, 1.2-111.8)after initial surgery.Most of these patients had high-risk disease at presentation based on the NIH consensus classification and the National Comprehensive Cancer Network-Armed Forces Institute of Pathology risk classification (Table 1).Information on primary tumor size and mitosis count [per 50 high-power field (HPF)] was available for 37 and 50 patients respectively; median tumor size was 145 mm (range, 25-500) and median mitosis count was 10 (range, 0-126).The primary tumor location was unknown for 4 patients, and the stomach was the primary tumor location for the majority of the remaining patients (69%; Table 1).Interestingly, while PDGFRA mutations were described more often in patients with KIT-negative GISTs by immunohistochemistry, most of the patients in our series (49 of 58, 85%) were KITpositive; however, several cases were noted to be "faintly positive" or "positive staining on few cells" (n ¼ 6).Median time from diagnosis to initiation of imatinib was 12 months (range, 0-202) for the whole cohort and 23 months (range, 3-202) for patients with initially localized disease who later developed a recurrence (n ¼ 40).Forty-four patients (76%) were treated with imatinib 400 mg daily and 14 (24%) with imatinib 800 mg daily.Of the 14 patients treated with 800 mg upfront, 11 were assigned this dose level in a clinical trial (EORTC studies).Fifteen different types of PDGFRA mutations were found and grouped into 3 different groups for analysis (1 patient with a GIST harboring a PDGFRA exon 4 mutation was described separately; Supplementary Table S1); the most common single type of mutation was the D842V substitution (Table 1).

Efficacy of first-line imatinib in patients with advanced PDGFRA-mutant GIST
Fifty-seven patients (98%) were evaluable for response, and 1 patient died of hemorrhage before the first assessment.There were 2 CRs (4%) and 8 partial responses (PR; 14%) for an overall RR of 18% (Table 2).The median PFS in the whole group was 6.4 months (95% CI, 3.1-9.7).There was no significant difference in the distribution of responses (CR þ PR) or stable disease (SD) between patients treated with imatinib 400 or 800 mg daily (P ¼ 1.0, Fisher exact test, data not shown).

Efficacy of first-line imatinib by mutation group
Patients with D842V substitution (N ¼ 32, 55.2%) had the poorest outcome, with a median PFS of 2.8 months (95% CI, 2.4-3.2).The RR in this group was 0%, and the majority of patients (21 of 31, 68%) had progressive disease (PD) as their best response (Table 2).Dose was not statistically correlated with response in patients with PDGFRA-D842V-mutant GISTs (Supplementary Table S2).Eight patients with D842V mutations had PFS longer than 6 months (range, 6.4-50.8).Four of these 8 patients had metastatic disease at time of diagnosis, whereas 4 relapsed 15, 19, 27, and 31 months after surgery.Most of these patients had mitosis counts lower than the median for the entire group of patients with D842V mutation, although this was not significant (Mann-Whitney U test, P ¼ 0.110).
Patients with non-D842V-exon 18 mutation (N ¼ 17, 29.3%) and patients with exon 12 mutation (N ¼ 8, 13.8%) had similar outcome in terms of RRs (ORR for both groups combined was 36%), and despite a difference in the median PFS (28.5 months for exon 18 vs.12.6 months for exon 12), the curves were overlapping (log-rank test, P ¼ 0.5571; Supplementary Fig. S1).The median PFS of all patients with non-PDGFRA-D842V-mutant GISTs was 28.5 months (95% CI, 5.4-51.6)and was significantly different from that of patients with D842V-mutant GIST (P ¼ 0.0001; Fig. 1).The response distribution (CR, PR, SD, and PD) between patients in the exon 18 non-D842V-mutant group and the exon 12-mutant group were not statistically different (P ¼ 0.426), whereas they differed significantly from that of the D842V group (P < 0.0001; Table 2).The only patient with an exon 4 mutation had a PR lasting 57þ months.
Overall, the PFS of patients with PDGFRA mutations other than D842V treated with first-line imatinib appeared comparable with those previously published for patients with KIT exon 11 mutations, although RRs in the former seemed lower.

OS of patients with PDGFRA-mutant GIST
The median follow-up for surviving patients was 45.3 months.Median OS for the whole cohort was 23.7 months.Here again, patients with D842V substitutions had poorer outcome than other molecular groups (median 14.7 a One patient with a D842V-mutant GIST died of gastrointestinal hemorrhage before his first assessment and was therefore not evaluable for response.months vs. not reached for D842V mutation and non-D842V mutations, respectively; Fig. 3).

Discussion
This study is the largest to report specifically on the outcome of patients with advanced PDGFRA-mutant GISTs treated with imatinib.Most previously reported trials investigating the activity of tyrosine kinase inhibitors enrolled patients with PDGFRA-mutant GISTs, these studies provided evidence that D842V substitution of exon 18 of PDGFRA is resistant to imatinib (7)(8)(9)11), but we sought to provide a more comprehensive analysis of imatinib sensitivity in PDGFRA-mutated GISTs.Our study therefore provides additional information, albeit biased by the retrospective nature of our study.Our data confirm that although PDGFRA-mutant GISTs are probably more frequent than initially reported (11% of GIST patients with genotyping information had PDGFRA mutation), metastasis in this subgroup is rare: only 60 patients identified over a 10-year period across 12 European referral centers, and only 11% (44 of 389) of patients with PDGFRA-mutant GISTs had metastasis in our study.This was already observed in clinical trials of imatinib in patients with advanced GISTs, where PDGFRA-mutant GISTs usually represented only 2% to 5% of the whole study population (7)(8)(9)11).Most of the patients in our study had CD117/KIT-positive GISTs.However, we and others (2,15) have shown that PDGFRA mutations can be found in up to 15% of patients with GIST whereas only 4% to 5% are CD117/KIT negative by immunohistochemistry.Likewise, in a gene expression study by Subramanian and colleagues, although KIT was significantly less expressed in PDGFRA-mutant GISTs than in KITmutant GISTs, only one PDGFRA-mutant GIST sample was negative for KIT using immunohistochemistry (19).These data suggest that in fact only a minority of PDGFRA-mutant GISTs are CD117/KIT negative by immunohistochemistry.
As expected from in vitro data (5,20,21), patients with D842V mutation failed to respond to first-line imatinib: most of them progressed at their first assessment (2-3 months of treatment, depending on individual institutions' standards).Some patients in this group had long lasting SD on first-line imatinib, which was already observed by others (13).In our series patients with D842V and PFS longer than 6 months tended to have lower mitotic count suggesting that this may be related to more indolent disease.Median OS in this group was close to the historical controls for patients with GISTs treated with chemotherapy (ref.16; which is considered ineffective in this disease).Taken together, these data suggest that imatinib is ineffective in this subgroup of patients and that alternative therapies should be evaluated for D842V-mutant GISTs.This finding also has implications in the adjuvant setting as PDGFRA-D842V-mutant GISTs made up approximately 10% of patients in the SSGXVIII/AIO randomized phase III study comparing 12 with 36 months for high-risk GISTs (22).Imatinib was active in patients with GISTs harboring a non-D842V PDGFRA exon 18 mutation or a PDGFRA exon 12 mutation, and median PFS in this group was close to that reported for patients with GISTs with KIT exon 11 mutations with a median PFS over 2 years (23).These patient subgroups also had longer OS (median not reached) than in patients with PDGFRA-D842V mutations.Of note, response and PFS appeared comparable in patients with non-D842V PDGFRA exon 18 mutations or an exon 12 mutation.These findings had already been suggested by subgroup analyses done on clinical trial data (7,8,13,24) as well as in vitro work (5).Interestingly, of 2 patients with D842del, one had very long PR whereas the other had PD as best response.These discrepancies suggest that both the location and the nature of the mutation are important for imatinib sensitivity (D842V vs. D842del) but is not the sole factor affecting response.Dewaele and colleagues previously reported the activity of tyrosine kinase inhibitor against DIM842-844 del of exon 18 PDGFRA and the resistance of D842V substitution, as previously observed (5,7,20).In a recently published article, Dileo and colleagues, using in silico modeling, show that the DIMH842-845del of PDGFRA exon 18 (3 patients in our study) and the V561D substitution in PDGFRA exon 12 (7 patients in our study) do not negatively affect the affinity or binding of imatinib to the kinase, unlike the D842V substitution (24).
We also report here a new mutation in PDGFRA exon 4. The tumor sample from this patient was analyzed twice in 2 different laboratories, which confirmed the presence of a L221F substitution.Exon 4 mutations have previously been reported in lung carcinomas (Catalogue of Somatic Mutations in Cancer) but never in GISTs.Although no protein modeling was done to investigate the conformational changes associated with this mutation, our clinical data (long lasting PR) indicate that this mutation is imatinibsensitive in vivo.
Overall our study provides evidence that imatinib has activity in patients with advanced non-D842V PDGFRAmutant GISTs and suggests that other approaches should be explored for patients with advanced GISTs harboring a PDGFRA-D842V substitution.Heinrich and colleagues recently showed that crenolanib, a new tyrosine kinase inhibitor targeting PDGFRa, had in vitro activity against D842V mutants (25).A phase II trial is currently open to recruitment for patients harboring such PDGFRA mutations (trial NCT01243346 www.clinicaltrials.gov).

Figure 1 .
Figure 1.PFS of patients with advanced PDGFRA-mutant GISTs on firstline imatinib according to the type of PDGFRA mutations.

Figure 2 .Figure 3 .
Figure 2. PFS with second-line treatment (n ¼ 32) according to the type of treatment (A) and according to the type of PDGFRA mutation (B).

Table 1 .
Main characteristic of the 58 patients included in this study

Table 2 .
RR to imatinib per group of PDGFRA mutation and overall