Background

MET, also called c-MET or hepatocyte growth factor (HGF) receptor, is a receptor tyrosine kinase discovered as an oncogene in the 1980s [1, 2]. Independent research found that HGF or scatter factor (SF) was a ligand for MET [3]. Upon binding of HGF to MET, the kinase domain phosphorylates growth factor receptor-bound protein 2 (GRB2) and GRB2-associated binding protein 1 (GAB1) and activates diverse downstream signaling pathways important in cancer, including the ERK/MAPK, PI3K-Akt/PKB, Crk-Rap, and Rac-Pak pathways [4]. These pathways form distinct branches that interact to regulate cell proliferation, invasion, migration, angiogenesis, development, organ regeneration, and tumorigenesis [4].

Activating MET mutations are found in diverse human cancer [5]. For instance, activating mutations in the kinase domain are a feature of both hereditary and some non-hereditary forms of papillary renal cell carcinoma [6]. MET amplification is seen in 5 to 20% of non-small cell lung cancer (NSCLC) and gastric cancer [7]. MET overexpression correlates with poor treatment outcome in some malignancies [8].

Drugs that target MET include inhibitors of the HGF/-MET pathway, MET antibodies, and MET kinase inhibitors [8]. Cabozantinib and crizotinib are both Food and Drug Administration (FDA)-approved multikinase inhibitors that are also potent suppressors of MET [9, 10]. Crizotinib induced responses in some patients with highly MET-amplified lung cancer [11]. Studies with the multi-target MET inhibitor cabozantinib have shown significant activity against a variety of solid tumors including melanoma, as well as renal and non-small cell lung, liver, medullary thyroid, breast, and ovarian cancer, but this activity is likely due to other targets of cabozantinib, such as VEGFR or RET [9]. In the phase I setting, patients with MET amplification did not respond to MET inhibitors (but the number of treated patients was small) [5].

Because of their non-invasive nature, liquid biopsies are increasingly used in the clinical setting. Indeed, numerous studies showed a relatively good correlation with tissue sequencing and the potential to detect actionable alterations [12,13,14,15,16]. In this study, we analyzed MET alterations in the plasma-derived circulating tumor DNA (ctDNA) of 438 patients with diverse malignancies and explored the relationship between MET alterations, demographics, as well as other molecular alterations and clinical outcomes.

Methods

Patients

We reviewed the clinicopathology and clinical outcomes of 438 consecutive patients with cancer for whom ctDNA testing had been performed and who were seen at the UC San Diego Moores Cancer Center from June 2014 to July 2016. Pathology was reviewed at UCSD. Data was abstracted from the electronic medical record. This study was performed and consents were obtained in accordance with the UCSD Institutional Review Board guidelines [PREDICT-UCSD (Profile Related Evidence Determining Individualized Cancer Therapy); NCT02478931].

Next-generation sequencing

Digital sequencing of ctDNA (DNA) in all patients was performed by Guardant Health, Inc. (Guardant360, Redwood City, California, http://www.guardanthealth.com/guardant360/), a Clinical Laboratory Improvement Amendment (CLIA)-certified and College of American Pathologists (CAP)-accredited clinical laboratory. The analytical and clinical validation of Guardant360 was conducted in accordance with evidentiary standards established by the Standards for Reporting of Diagnostic Accuracy (STARD), REporting of tumor MARKer Studies (REMARK), Evaluation of Genomic Applications in Practice and Prevention (EGAPP), and the recent Next-generation Sequencing: Standardization of Clinical Testing (Nex-StoCT) biomarker guidelines [17]. As described in Lanman et al. [17], 5–30 ng of ctDNA was isolated from plasma (two 10 ml Streck tubes drawn for each patient) and sequencing libraries were prepared with custom in-line barcode molecular tagging and complete sequencing at 15,000× read depth (~ 4000 unique double-stranded cfDNA fragments, each represented by 3–5 sequencing reads). The panels utilize hybrid capture followed by NGS of the critical exons in a panel of 54–70 genes (Additional file 1: Table S1 to S3) and report all four major types of genomic alterations (point mutations, indels, fusions, and copy number amplifications). Post-sequencing bioinformatics matches the complementary strands of each barcoded DNA fragment to remove false positive results [17]. The variant allele fraction (VAF) is computed as the number of mutated DNA molecules divided by the total number (mutated plus wild type) of DNA fragments at that allele; VAF is reported as a percentage. The majority of cell-free DNA is wild type (germline); thus, the median VAF of somatic alterations is < 0.5%. The analytic sensitivity reaches detection of one to two single-mutant fragments from a 10-ml blood sample (0.1% limit of detection), and the analytic specificity is greater than 99.9999% [17].

For 144 patients, a 54-gene panel was used, which identified potential tumor-related alterations in 54 cancer-related genes (Additional file 1: Table S1) including copy number amplifications in ERBB2, EGFR, and MET (indels and fusions were not detected as part of this panel). For 272 patients, a 68-gene version of the original panel (expanded to all four major alteration types) was used, and for 22 patients, the most recent 70-gene panel version (further expanded to amplifications in 18 genes and fusions in 6 genes) was applied (Additional file 1: Table S2 and S3). Only non-synonymous alterations were included in our analysis.

In addition, 263 patients (~ 60%) of the 438 patients with ctDNA test also had CLIA/CAP-accredited next-generation sequencing (NGS) performed on tumor tissue (FoundationOne™, Cambridge, Massachusetts, http://www.foundationone.com.) (N = 182 to 315 gene panels).

Statistical and outcome analysis

Patient characteristics were summarized using descriptive statistics. Medians and respective 95% confidence intervals and range were calculated, whenever possible. Associations between categorical variables were tested using a binary logistic regression model. Linear variables were tested using the Mann-Whitney U test for univariable analyses and a multiple linear regression model for multivariable analyses.

Time to metastasis/recurrence was defined as the time interval between diagnosis and first metastasis/recurrence (whichever came first) or last follow-up date (patients who had not recurred/developed metastases at last follow-up were censored on that date). Overall survival (OS) was defined as the time from diagnosis to death or last follow-up date for patients who were alive (patients still alive at the last follow-up were censored on that date). Estimations for the time to first metastasis/recurrence and OS were done using Kaplan-Meier analyses and were compared among subgroups by the log-rank test for univariable analysis or Cox regression models for multivariable analysis. All statistical analyses were performed by author MS with SPSS version 24.0.

Results

Patient demographic characteristics

The median age of patients at diagnosis was 57.5 years (CI 95%, 54.5–59.1). Women comprised 52.1% (N = 228) of the population. The majority of patients were Caucasian (69.2%, N = 303). The most represented tumor types were gastrointestinal (28.1%, N = 123), brain (24.9%, N = 109), lung (23.2%, N = 102), and breast (11.6%, N = 51) cancers. The majority of the patients had recurrent or metastatic disease at the time of blood draw used for testing (71.2%, N = 312) (Table 1).

Table 1 Demographics comparison of 438 patients with or without MET alterations
Full size table

MET alterations and associations with patient characteristics

Overall, MET alterations were observed in 31 of the 438 patients whose ctDNA was tested (7.1%). Sixteen patients had a MET amplification only; 13 had a somatic mutation only; and two had both an amplification and a somatic MET mutation (Fig. 1a).

Fig. 1
figure 1

a MET alterations representation. Pie chart representing the different types of MET alterations identified in our cohort comprising 438 patients. b Overall survival and time to first metastasis/recurrence. Kaplan-Meier curves depicting the overall survival (left panel) and the time to first metastasis/recurrence (both from diagnosis time). P values are from univariable analysis. For more details, refer to Tables 4 and 5

Full size image

Univariate analyses

MET alterations were significantly (positively) associated with lung cancers (11.8%) and the presence of metastasis/recurrence at the time of blood draw and negatively associated with brain tumors 1.8% (all P < 0.05) (Table 1). We then examined the associations with other genomic alterations and found that MET alterations correlated in univariable analysis with alterations in TP53, EGFR, PIK3CA, BRAF, ARID1A, ALK, and PTEN genes (all P < 0.05, Additional file 1: Table S4). We also investigated the potential associations between MET alterations and the location of metastatic sites and showed a significant correlation with lymph nodes and bone metastasis (both P < 0.02, Additional file 1: Table S4). Of note, patients with MET alterations had a median of two metastatic sites versus a single site for patients without MET alterations (P = 0.001). The latter conclusions remained unchanged if the patients with brain tumors, lymphoma/leukemia, thymoma, and NF were excluded, with P values of < 0.05 for association between MET alterations and lymph node or bone metastases and increased number of metastases.

Multivariate analyses

In the multiple logistic regression model (that included any parameters that were significant (P ≤ 0.05) in univariate analysis), the only variables that remained statistically associated with MET alterations were aberrations in TP53 (P = 0.001) and PTEN (P = 0.003) genes, as well as an increased incidence of bone metastasis (P = 0.007) (Table 2). The univariate association between MET alterations and lung cancer, the negative association with brain tumors, and the correlation with metastases at the time of blood draw did not remain significant in multivariate analysis. Multivariable analysis also demonstrated that MET alterations correlated with an increased number of alterations (median of 4 alterations vs 1, P = 0.001) (Table 3). These conclusions remained valid if the patients with brain tumors, lymphoma/leukemia, thymoma, and NF were excluded, with P values of < 0.01 for association between MET alterations and TP53 or PTEN alterations, bone metastases, and increased number of metastases.

Table 2 Multivariable analysis of the variables associated with MET alterations*
Full size table
Table 3 Association with the number of alterations*
Full size table

The other variables that were significantly associated with a higher number of alterations in the multivariable analysis were alterations in TP53, PIK3CA, BRAF, KRAS, or MYC genes as well as the presence of metastasis in the lymph node or adrenal tissue (Table 3).

MET alterations and survival outcomes

Overall survival

Overall, 213 patients (213/438, 48.6%) had died at the time of our analysis (20/31 with MET alterations; 193/407 without MET alterations detected in their ctDNA). A log rank test demonstrated a poorer survival (time from diagnosis until death or last follow-up date) for patients bearing MET alterations (30.6 months vs 58.4, P = 0.013). In the multivariable analysis, only lung cancer as well as the presence of ARID1A, KRAS, ALK, and MYC alterations and liver metastasis remained significant predictors of a poorer survival (all P < 0.50, Fig. 1b) (Table 4).

Table 4 Overall survival analysis (N = 438 patients)*
Full size table

Time to metastasis/recurrence

Patients with MET alterations had a significantly shorter median time to metastasis/recurrence, with a median of 1.0 months (95%CI could not be computed) versus 10.4 months (95%CI 6.9–14.0) (P = 0.003) (Table 5). MET alterations remained significantly associated with a shorter time to metastasis/recurrence in a multivariable analysis (P = 0.044), along with the presence of liver (P = 0.022) or lymph node metastases (P < 0.001). Inversely, breast cancer correlated with a longer time to metastasis/recurrence (P < 0.001) (Fig. 1b and Table 5).

Table 5 Time to metastasis/recurrence from diagnosis (N = 438 patients)*
Full size table

Comparison with tissue testing

As noted, 438 patients had ctDNA testing; 263 of these patients also had tissue NGS performed.

Of the 31 patients with MET alterations in ctDNA, 18 also had tissue testing (Foundation Medicine see the “Methods” section). The median time interval between the blood draw and the tissue biopsy for these 18 patients was 6.1 months (95% CI (2–13.7); range (0.2–32.6)). In most of the patients with both types of testing, the ctDNA test was performed after the tissue testing (16/18 cases). Only two patients who had a MET alteration identified in their ctDNA also had a MET alteration found in their tissue testing (11.1%; MET amplification and MET Y501C, one patient each). In these two patients, the time interval between the ctDNA and tissue biopsies was 1.8 and 15.3 months, and both tissues used for the testing were from the primary tumor.

In only one patient, a MET amplification was detected in the tissue and not in ctDNA (1 of 263 total patients who had NGS tissue testing). The tissue test was performed on a pancreatic tumor that was surgically removed, and the ctDNA test was done more than 1 year later (recent scans showed appearance of new pulmonary and liver nodules 2 months prior the ctDNA testing).

Overall, MET alterations were detected at a significantly lower frequency in tissue (3/263 patients, 1.14%) compared to ctDNA (31/438, 7.1%), P = 0.0002. Further, of eight patients who harbored MET alterations in ctDNA, and had ctDNA and tissue testing within 2 months of each other, only one patient showed a similar MET alteration in the tissue. Six of the seven patients with only ctDNA positive for MET alterations had bone metastases.

Discussion

This is the largest study interrogating the feasibility and utility of MET alteration detection through blood-derived ctDNA. Liquid biopsy is a non-invasive method to find genomic aberrations and is increasingly utilized in the clinical setting as reflected by the non-small cell lung cancer National Comprehensive Cancer Network (NCCN) guidelines [18, 19]. Our study demonstrated that MET ctDNA alterations were detected in 7.1% of patients with solid tumors. This detection rate is higher compared to previous tissue studies [5, 20, 21]. For instance, a study of MET tissue amplification determined by fluorescent in situ hybridization (FISH) demonstrated that 2.6% of 1115 solid tumor specimens were positive [5]. Furthermore, the MSKCC-IMPACT study showed that 3% of patients had MET tissue alterations [20, 21]. The differences in rate of MET alterations between our study and the other studies could be due to following reasons: (i) our technology detected both single nucleotide substitutions and amplifications, while the previous investigation by Jardim et al. at MD Anderson Cancer Center [5] discerned only amplifications (though the MSKCC study would have discerned both single nucleotide substitutions and amplifications [21]); (ii) our study used blood-derived ctDNA, which could capture shed tumor DNA from multiple sites, while the previous reports used tissue-based testing, which would only detect aberrations in the piece of tissue biopsied. Consistent with the above observations, our tissue NGS testing also showed significantly lower rates of MET alterations than the ctDNA NGS: 3 of 263 patients (1.1%) (who also had ctDNA tests) were positive for MET alterations in tissue versus 31 of 438 ctDNA-tested patients (7%) being positive for MET alterations (P = 0.0002). Further, of 18 patients positive for MET ctDNA alterations who also had tissue NGS, only 2 (11.1%) were also positive for a MET alteration by tissue NGS. The biologic underpinnings of discordance between ctDNA and tissue NGS results have been previously documented and include spatial (intra-tumor and inter-tumor) and temporal heterogeneity in genomic anomalies in cancers along with the fact that ctDNA is comprised of DNA that has leaked into the circulation from diverse metastatic sites while tissue NGS reflects only the tissue specimen analyzed [14].

The question that arises is whether or not biological explanations can specifically account for the higher rate of MET alterations in ctDNA. Of interest in this regard, our study showed that bone metastases were independently correlated with MET alterations (Additional file 1: Table S2 and Table S4). Indeed, 14 of 31 patients (45%) positive for MET ctDNA alterations had bone metastases (Additional file 1: Table S4). Of possible relevance in this regard, MET inhibitors are known to show efficacy in bone lesions [22, 23]. In the COMET-1 trial, 682 patients with castrate-resistant prostate cancer who progressed after docetaxel and androgen modulators (abiraterone and/or enzalutamide) were randomly assigned to either cabozantinib (MET inhibitor) or prednisone. Although there was no difference in overall survival, 42% of cabozantinib-treated patients showed bone scan response compared to 3% of prednisone-treated patients (P < 0.001), albeit without prostate surface antigen (PSA) response [24]. In the METEOR trial, 658 patients with advanced renal cell carcinoma who progressed with at least one VEGFR small molecule inhibitor were randomized to the MET inhibitor cabozantinib or the mTOR inhibitor everolimus [25]. In a sub-group analysis, patients randomized to cabozantinib arm with bone metastasis were associated with better overall survival (OS) (hazard ratio (HR) 0.54, 95% confidence interval (CI) 0.34–0.84) compared with non-bone metastasis (HR 0.71, 95% CI 0.55–0.91) [25]. These data together with our results raise the possibility that cancer clones with MET alterations preferentially localize to bone and may therefore explain bone responses after MET inhibitor therapy. Tissue biopsies are rarely performed on bones (none of our 18 patients with MET alterations in ctDNA who also had tissue NGS had a bone biopsy). Of interest in this regard, MET is prominently expressed (as determined by immunohistochemistry) at the site of bone metastases in renal cell cancer [26]. It is therefore conceivable that the high rates of ctDNA positivity for MET alterations, which strongly and independently correlated with bone metastases in our study, reflect shed MET alteration-bearing DNA from bone lesions (Additional file 1: Table S2 and Table S4).

Our study also found that MET ctDNA alterations are associated with poor prognosis, including decreased survival and shorter time to recurrence/metastasis (Tables 4 and 5, Fig. 1b). These results are consistent with those found by correlating MET alterations found in tissue NGS with outcome in specific malignancies, such as astrocytomas [27], breast cancers [28], genitourinary malignancies [29], and ovarian [30] or gastric/esophageal cancers [31]. Specific genes that were co-altered with MET in multivariate analysis of our study participants include PTEN and TP53 (Table 2); previously, another study has also shown the association between MET and PTEN abnormalities [5]. Overall, MET alterations significantly correlated with a higher number of alterations, which may explain—at least in part—the relatively limited efficacy of MET inhibitors as single agents for the treatment of MET-altered advanced malignancies observed in prior studies [5, 29,30,31]. Indeed, if patients with MET alterations generally have multiple genomic abnormalities, it is likely that combination therapy, rather than monotherapy with a MET inhibitor, may be necessary to achieve salutary effects.

Our study has several limitations. First, only 60% of our patients (N = 263) with ctDNA analyses also had tissue sequencing and the median time between the two tests in the patients with MET-altered ctDNA was about 6 months. Second, our patients did not have bone biopsies done, which would be of interest to determine if MET alterations are found in such biopsies, since ctDNA MET alterations correlate independently with the presence of bone metastases. Third, though our study with 438 patients is the largest to date, the rarity of MET alterations suggests that investigations of even greater numbers of patients may be worthwhile in order to best understand the biology and correlations of MET alterations.

Conclusions

In summary, our study demonstrated that assessment of MET genomic aberrations by liquid biopsy is feasible. We found that MET ctDNA anomalies were associated with bone metastases, multiple genomic alterations, and a poorer prognosis, including poorer overall survival and a shorter time to recurrence/metastases. Further studies are needed to better understand the biologic relationship between MET alterations and bone lesions, and next-generation trials with MET inhibitors may require combinations of drugs that address the genes such as PTEN that are frequently co-altered in these patients.