Outcome of EGFR-mutated NSCLC patients with MET-driven resistance to EGFR tyrosine kinase inhibitors

Background Several mechanisms of acquired resistance to EGFR tyrosine kinase inhibitors (TKIs) in EGFR-mutated NSCLC have been described including the T790M mutation and MET amplification. Whereas T790M mutation confers prolonged survival and sensitivity to 3rd generation TKIs, data are lacking on clinical features and outcome of MET-driven resistant EGFR-mutated NSCLC patients. Methods Patients with metastatic EGFR-mutated NSCLC displaying high MET overexpression or MET amplification, detected on a biopsy performed after progression on EGFR TKI, were identified in 15 centers. Clinical and molecular data were retrospectively collected. Results Forty two patients were included. The median overall survival (OS), and the median post EGFR TKI progression overall survival (PPOS) were 36.2 months [95%CI 27.3-66.5] and 18.5 months [95%CI 10.6-27.4] respectively. Nineteen out of 36 tumors tested for MET FISH had MET amplification. A T790M mutation was found in 11/41 (26.8%) patients. T790M-positive patients had a better OS than T790M-negative patients (p=0.0224). Nineteen patients received a MET TKI. Objective response was reported in 1 out of 12 evaluable patients treated with a MET inhibitor as a single agent and in 1 of 2 patients treated with a combination of MET and EGFR TKIs. Conclusion MET-driven resistance to EGFR TKI defines a specific pattern of resistance characterized by low objective response rate to MET inhibitors given alone and overlapping with T790M mutations. Further studies are warranted to define adequate therapeutic strategies for MET-driven resistance to EGFR TKI.


INTRODUCTION
EGFR mutations are found in 10% of non small cell lung cancer (NSCLC) in Caucasians and 40% in Asians [1]. Treatment of advanced EGFR-mutated NSCLC patients relies on EGFR tyrosine kinase inhibitors (TKIs), which demonstrated superiority over chemotherapy as first-line therapy [2][3][4][5]. However, despite initial efficacy, all the patients will eventually develop resistance to EGFR TKIs resulting in tumor progression [6]. The most frequent mechanism of resistance is the T790M mutation, a second EGFR mutation that can be successfully targeted with third generation EGFR TKIs which have been specifically designed to overcome T790M-driven resistance [7,8] Besides T790M mutation, bypass activation of other tyrosine kinase receptors including MET or HER2 is the second most common mechanism of resistance to EGFR TKI. MET amplification leads to overexpression and constitutive activation of the receptor, thus activating the PI3K pathway and bypassing EGFR [9]. MET amplification has been detected in 5 to 22% of patients with an acquired resistance to EGFR TKI [9][10][11][12][13][14]. MET amplification is highly associated with high overexpression of MET in NSCLC (p<0.001) [15,16]. Moreover, high MET overexpression with a 3+ immunoscore (IHC3+) by immunohistochemistry was recently found in 27% of EGFR mutated NSCLC with acquired resistance to EGFR TKI [17].
Little data is available about clinical characteristics and outcome of EGFR-mutated NSCLC patients with MET-driven resistance to EGFR TKIs. Moreover, optimal treatment of these patients is still unknown. Preclinical data and case reports suggest that MET-amplified EGFRmutated NSCLC are addicted to both MET and EGFR and that combination of MET and EGFR TKIs is required to overcome this mechanism of resistance [9,18,19].
Whereas available data and specific treatments emerge for T790M-driven resistance in EGFR-mutated NSCLC patients, MET-driven resistance is still an unexplored field [20][21][22]. In the present study, we report clinical features, outcome and treatment in a series of EGFR-mutated NSCLC patients with MET-driven resistance to EGFR TKI.

Clinico-pathological and molecular characteristics
Forty six patients with metastatic NSCLC displaying both EGFR mutation and MET overexpression or MET gene amplification were retrospectively identified in 15 centers. Four patients were excluded : 3 had only a biopsy performed before EGFR TKI initiation, and 1 had no data available on the treatments received. Forty two patients were included. Re-biopsies of these patients had been performed from May 2011 to May 2016.
Patient clinical characteristics are summarized in Table 1. All 42 patients were diagnosed with metastatic lung adenocarcinoma. Median age was 65.1 years (range 30-82.7). The majority of patients were women (66.7%) and never smokers (70.7%). Most of the EGFR mutations detected on the initial biopsy were exon 19 deletions or exon 21 L858R point mutations.
The most frequent site of rebiopsy was the lung and re-biopsies were performed, in 90.5% of the cases, after the Response Evaluation Criteria in Solid Tumor (RECIST) progression on EGFR TKI, which was given as first or second line treatment (Supplementary Table 1). The median time between EGFR TKI initiation and rebiopsy was 15.6 months (range 2.1-61.3).
Among the 42 patients included in the study, 36 tumor re-biopsy samples were tested for MET FISH and 19 (52.8%) were found MET amplified ( Figure 1). MET IHC was performed on the re-biopsy of 36 patients and all displayed a high level of MET expression (IHC3+). Six patients had a MET FISH but no MET IHC on their re-biopsy and conversely 6 other patients had MET IHC and no interpretable MET FISH. No difference was found regarding the MET amplified status between patients with an EGFR exon 19 deletion or an EGFR exon 21 mutation on the initial biopsy (Supplementary  Table 2 ). Re-biopsies of 34 patients were tested for the BRAF mutations and none harbored a mutation of this oncogene. Otherwise no histological transformation in small cell lung cancer was reported among the 42 patients of the study. www.impactjournals.com/oncotarget

Clinical outcome and EGFR TKI treatment characteristics
Characteristics of initial EGFR TKI therapy are shown in Table 2. All patients received a first or second generation EGFR TKI. The overall response rate (ORR) was 82.1%, and the median progression free survival (PFS) was 11.1 months [95%CI 7.6-14.1]. In 73.2% of the cases, tumor progression involved a new lesion. The main site of new metastasis was the lung. 27 patients (66%) developed more than one progressive lesion at EGFR TKI resistance. The median post-progression overall survival (PPOS) and the median overall survival (OS) were respectively 18 Table 2).

Impact of the T790M mutational status
Eleven among the 41 patients (26.8%) tested for the T790M mutation were T790M positive. In 9 patients, the T790M mutation was detected in the tumor re-biopsy displaying MET overexpression or MET amplification. In the remaining 2 patients, the T790M mutation was detected in circulating free DNA in one case and in a tumor biopsy obtained before the re-biopsy displaying MET overexpression and MET amplification in another case. Three patients had both MET amplification and T790M mutation in the same sample. The T790M-positive patients were significantly younger than T790M-negative patients and the time between EGFR TKI initiation and rebiopsy was significantly longer in T790M positive patients (Table 1 and Supplementary Table 1). T790M-positive patients had a better OS (median OS 43.1 vs. 32.2 months, p=0.0224) than T790Mnegative patients ( Figure 2B). There was also a trend to

Treatment with MET inhibitors and third generation EGFR TKI therapy
Nineteen patients received a MET inhibitor, mostly crizotinib, including 13 MET-amplified patients ( Table 3). The MET inhibitor was given as a monotherapy in 15 patients and in combination with an EGFR TKI in 4 patients. The MET inhibitor was used as 2 nd or 3 rd line in most of the cases. Objective response was reported in 1 out of 12 evaluable patients treated with a MET inhibitor as single agent and in 1 out of 2 evaluable patients treated with a combination of MET and EGFR inhibitors ( Figure 3A and 3B). The MET inhibitor was stopped because of elevated liver enzymes in 2 patients and diarrhea in one patient. Two of these patients received the MET inhibitor in combination with an EGFR TKI. The median times between re-biopsy and MET inhibitor initiation and first EGFR TKI withdrawal and MET inhibitor initiation were respectively 2.3 months and 3.1 months. Oncotarget 105108 www.impactjournals.com/oncotarget Ten patients received a 3 rd generation EGFR TKI (Table 4) mostly osimertinib. Two partial responses were reported among the 5 T790M-positive evaluable patients who were treated following detection of MET overexpression or MET amplification. 3 patients were treated with 3 rd generation EGFR TKI before they received a MET inhibitor.

DISCUSSION
Bypass activation of tyrosine kinase receptors is a well described mechanism of resistance to EGFR TKIs in EGFR-mutated NSCLC. MET-driven resistance has been reported in up to 22% of patients with acquired resistance to EGFR TKIs [9]. Still, although the biological basis of this mechanism of resistance has been extensively studied, there is very little data on the clinical characteristics and outcome of patients with MET-driven acquired resistance to EGFR TKIs. In this multicenter retrospective study, we report for the first time clinical features, response to MET inhibitors and outcome of 42 metastatic EGFRmutated NSCLC patients with MET amplification or MET overexpression, as assessed on a post-progression re-biopsy. Because of the very low number of patients with MET-driven resistance to EGFR TKIs, we performed a multicentric retrospective study. Therefore, we could not perform a central analysis of RECIST responses, MET FISH, and MET IHC analysis. The absence of MET FISH and MET IHC analysis on the initial biopsy is also challenging. Indeed, we cannot exclude that some patients might have a MET overexpression or a MET amplification before EGFR TKI therapy. Most patients with EGFR exon 19 deletions in our study received a first generation EGFR TKI and were not treated with afatinib. There were no specific clinical characteristics of the patients included in our study but we observed a low rate of objective response to MET inhibitors when used as a monotherapy and a substantial rate of concomitant T790M mutations, which were still partially associated   with efficacy of 3 rd generation EGFR TKIs and favorable prognosis, even in association with MET overexpression or MET amplification. In our study, MET amplification was defined using the criteria provided by Schildhaus et al. to define highlevel MET amplification, i.e. as an average MET gene copy number (GCN) per cell ≥ 6 or a ratio MET/CEP7 ≥ 2 or the presence of MET clusters [15]. Indeed these criteria were already widely used across our pathological centers to define positive MET FISH. However, many other definitions of MET amplification based on FISH have been proposed based on various thresholds for MET/CEP7 ratio or mean MET GNC [21,[23][24][25][26][27]. Unlike the mean MET GNC, the ratio MET/CEP7 is thought to discriminate real amplification from polysomia. Recent data showed that a MET/CEP7 ratio >5 was able to discriminate lung adenocarcinoma with no other driver mutations and was associated with high objective response rate to crizotinib [27,28]. However, the right definition of a positive FISH that would allow identification of MET gene amplification remains to be determined. The ability to detect MET amplification through next generation sequencing may favor routine screening and harmonization of the definition of MET amplification.
In our study, EGFR mutated NSCLC patients with either MET amplification or MET overexpression (MET IHC3+) were included. In lung adenocarcinoma, MET amplification is significantly associated with IHC3+ MET overexpression [16]. In our study, all the patients with a MET amplification who also underwent a MET IHC were scored 3+. MET overexpression, regardless MET amplification status, has been found to induce addiction to the MET pathway, and was recently found in 27.1% of EGFR mutated NSCLC with acquired resistance to EGFR TKI [29] [17]. Moreover, an ongoing clinical trial evaluating the efficacy of combining MET and EGFR TKIs in EGFR-mutated NSCLC patients with MET-driven resistance to EGFR TKIs includes patients with both MET amplification or MET overexpression (NCT01610336). In patients without MET amplification, the cause of MET overexpression at the time of re-biopsy may involve other molecular alterations including MET exon 14 splicing sites mutations which have been reported to be associated with MET overexpression [30] and may be involved in resistance to EGFR TKI [31]. Because of the unavailability of most of the tumor samples, we could not test for other genomic alterations.
In our cohort, metastatic EGFR mutated NSCLC with a MET-driven resistance to EGFR TKI did not display specific clinical features compared to those observed in previous studies. Median PPOS (18.5 months) was also in the range of what has been reported in previous studies focusing on patients with acquired resistance to EGFR TKIs (14.3-20 months) [13,32,33]. Moreover, although MET GCN alterations and MET overexpression have been associated with poor prognosis in resected NSCLC [34,35], we did not find poor outcome for the patients included in our study. These findings suggest that the prognostic impact of MET activation might depend on the stage of the disease and on the oncogenic environment. We observed only one objective response out of 12 patients treated with MET inhibitor monotherapy. In preclinical models, a double inhibition of EGFR and MET pathways was required to overcome MET driven resistance to EGFR TKI [9,36]. Cases of EGFR mutated MET amplified NSCLC responding to combined MET and EGFR TKIs have been reported [18,19]. This dual inhibition approach is currently evaluated in several phase I/ II clinical trials in this setting using various EGFR (gefitinib, erlotinib, EGF816, osimertinib) and MET (capmatinib, volitinib, tepotinib) TKIs (NCT02468661, NCT01610336, NCT02335944, NCT02374645, NCT02143466, NCT01982955). Preliminary results from a phase II study, evaluating the combination of capmatinib and gefitinib in EGFR mutated NSCLC patients who progressed on EGFR TKI, reported an ORR of 19% in patients with MET IHC3+ or MET IHC≥2+ and MET GCN≥5 and raised to 30% in the subgroup of patients with MET GCN≥6 [37].
In our study, 11 patients were T790M positive. Three had both a MET amplification and a T790M mutation on the same sample. The coexistence of these mechanisms of resistance has been previously described [10]. Compared to the T790M negative, the T790M positive patients in our study had a better OS and a trend to a better PPOS. Several studies reported that, amongst patients with EGFR TKI acquired resistance,T790M-positive patients had a better PPOS than T790M-negative patients [13,32,38]. However, none of these studies evaluated the impact of concomitant MET amplification and overexpression. Of note, Gou et al. found that patients with a T790M mutation and MET overexpression had a worse PPOS than patients with only MET overexpression or the T790M mutation alone. This discrepancy with our results might be due to the small size of both cohorts, and variations in the characteristics of the study population (higher rate of never smokers and caucasians in our study). We cannot also exclude the influence of confounding factors such as the younger age and the use of 3 rd generation EGFR TKI in T790M positive patients. Indeed in our study, 9 T790M-positive patients received a 3 rd generation EGFR TKI and 3 partial responses were achieved. Very few data are available regarding the efficacy of 3 rd generation TKIs in the context of multiple resistance mechanisms. Our results suggest that 3 rd generation TKIs may retain activity against T790M-positive tumors in some patients, even in the presence of MET activation, which may be due to the spatial heterogeneity of resistance mechanisms rather than co-existence of 2 resistance mechanisms in the same tumor cells.
Finally, MET-driven resistance to EGFR TKI defines a specific pattern of resistance characterized by low objective response rate to MET TKIs given alone and overlapping with T790M mutations. Even when associated with MET dysregulation, the T790M mutation www.impactjournals.com/oncotarget was still associated with relative efficacy of 3 rd generation EGFR TKIs and prolonged survival. Further studies are warranted to define adequate therapeutic strategies for MET-driven resistance to EGFR TKI.

Patients
We constituted a multicenter retrospective observational cohort of patients identified among 15 French centers. Inclusion criteria were documented diagnosis of metastatic NSCLC, detection of an EGFR mutation on tumor sample at diagnosis, treatment with at least one EGFR TKI and detection of MET overexpression or MET amplification after the time of clinical or radiological progression on EGFR TKI. Post-progression re-biopsy and MET status assessment were routinely performed in participating centers during the study period. Clinical and pathological data were retrospectively collected in each center for all included patients. Best overall response, defined as the best response from the start of treatment until disease progression, was assessed by investigators from available follow-up exams in each center using Response Evaluation Criteria in Solid Tumor (RECIST) v1.1. Crizotinib, gefinitib, and osimertinib were respectively prescribed with the following doses and schedules : 250mg bid, 250mg daily, 80mg daily. The study was approved by a national ethic committee (CEPRO 2016-001).

Histological and molecular analyses
Histological and molecular analyses of formalinfixed paraffin-embedded (FFPE) samples were prospectively assessed by local pathologists in accredited and quality controlled laboratories, as part of the routine procedure. MET immunochemistry and MET FISH analyses were realized according to locally certified and nationally approved procedures. MET overexpression was defined as a 3+ MET Immunoscore (≥ 50% of tumor cells showing high-intensity staining) on MET immunochemistry using MET monoclonal antibody (clone SP44 Ventana). MET amplification was defined by Fluorescence In Situ Hybridization (FISH) as a mean MET GNC per cell ≥ 6, or a Ratio MET/CEP7 ≥ 2, or the presence of MET clusters [15]. Patients were considered "T790M positive" if an EGFR T790M mutation was detected in post-progression circulating tumor DNA, on the re-biopsy on which MET overexpression or MET amplification was found, or on a tumor sample collected before the re-biopsy.

Statistical analyses
Categorical variables are expressed as frequencies and percentages. Quantitative variables are expressed as medians (range). Normality of distribution was assessed graphically and by using the Shapiro-Wilk test. Bivariate analyses were realized to assess the sub-group comparability (MET amplified vs. MET non-amplified and T790M+ vs. T790M-). Chi-Squared tests or Fisher's exact test (when expected cell frequency <5) were used to study the association between categorical variables and different groups. A Mann-Withney U test was used to compare age, time between EGFR TKI initiation and rebiopsy, and duration of TKI EGFR therapy between groups. We estimated and compared overall survival (OS), post progression overall survival (PPOS) and progression free survival (PFS) between the study groups (MET amplified vs. MET non amplified and T790M+ vs. T790M-) using the Kaplan-Meier Method and log-rank test. OS was measured from the date of metastatic NSCLC diagnosis to the date of death from any cause or last follow-up. The PPOS was measured as the time from EGFR TKI RECIST progression to death from any cause. The PFS was defined as the time from treatment start to disease progression or death from any cause. Objective response rate (ORR) was defined as the percentage of patients with partial or complete response to the indicated treatment. Statistical testing was conducted at the 2-tailed α level of 0.05. Data were analyzed with SAS software version 9.3 (SAS Institute,Cary, NC).

Authors contributions
Simon Baldacci : study design and conception, acquisition of data, analysis and interpretation of data, drafting manuscript Julien Mazieres : acquisition of data, critical revision Pascale Tomasini : acquisition of data, critical revision Nicolas Girard : acquisition of data, critical revision Florian Guisier : acquisition of data, critical revision Clarisse Audigier-Valette : acquisition of data, critical revision Isabelle Monnet : acquisition of data, critical revision Marie Wislez : acquisition of data, critical revision Maurice Pérol : acquisition of data, critical revision Pascal Dô : acquisition of data, critical revision Eric Dansin : acquisition of data, critical revision