Among the various MET aberrations, MET amplification and METex14 have emerged as valid predictive biomarkers for MET inhibition. Despite challenges that have limited the development of MET inhibitors, we can learn from the latest efforts in biomarker-based therapy to better identify the patients who will benefit from treatment with these agents.
In this issue of Clinical Cancer Research, Hong and colleagues (1) and Van Cutsem and colleagues (2) report the results of early-phase clinical trial studies of AMG 337, a highly selective small-molecule MET inhibitor, in patients, respectively, with advanced solid tumors (phase I), as well as those with MET-amplified gastric (G)/gastroesophageal junction (GEJ)/esophageal (E) adenocarcinoma and other biomarker-selected solid tumors (phase II). This new clinical study results reignite the important lingering discussion of issues regarding how MET acts as a cancer “driver” and how to optimize the therapeutic MET (c-MET)-targeting strategies. In light of these new study data at this time juncture of the long journey of targeting MET in cancer therapy, it is only fair to ask ourselves, “Are we there yet?”
The effort behind the MET signaling research and targeting in cancer therapy has spanned over 30 years. The MET receptor tyrosine kinase (RTK) was discovered in 1984. The natural ligand for MET is hepatocyte growth factor (HGF; also known as scatter factor, SF), which can activate MET-expressing epithelial cells' intracellular signaling through autocrine, paracrine, or endocrine fashion. MET is a multifunctional RTK, long known to play key roles in developmental and cancer signaling. MET/HGF signaling is a “hallmark of cancer,” driving cancer “invasion and metastasis” via multiple mechanisms of activation (Fig. 1). Despite extensive preclinical studies in the past decades that unraveled MET signaling intricacies, and more recently clinical MET-targeting drug development studies, the optimal inhibitory strategies remain elusive. The MET clinical therapeutics development since 2003 has turned into an unexpected marathon journey.
The articles by Hong (1) and Van Cutsem (2) and their colleagues again demonstrated the clinical activity and feasibility of targeting MET in cancer therapy using a small-molecule MET inhibitor, AMG 337. These studies reflect the ongoing renewed enthusiasm in MET-targeted therapy clinical development, despite previous setbacks and challenges. However, these early-phase studies results apparently continue to raise more questions than answers. In the first-in-human phase I study of AMG 337 by Hong and colleagues (1), AMG 337 was found to be tolerable with manageable toxicities, with an MTD and recommended phase II dose determined to be 300 mg orally every day (MTD for twice a day dosing was not reached). The study comprised a dose-escalation phase cohort in a modified 3 + 3 + 3 design using 25 to 400 mg every day and 100 to 250 mg twice a day regimens in advanced solid tumors, and a dose expansion phase cohort with the MTD 300 mg every day dosing conducted in patients with “MET-amplified” tumors. Of note, grade ≥3 treatment-related adverse events (AE) occurred in 21% of patients, with the most common being headache (n = 6) and fatigue (n = 5). Interestingly, headache was a dose-dependent and dose-limiting toxicity (DLT) and considered to be the result of AMG 337 being a potent inhibitor of adenosine transporters, which can be alleviated with caffeine. Overall, this study yielded an objective response rate (ORR) of 10% (11/111) in all patients, regardless of the MET status, whereas the dose expansion cohort (MTD: 300 mg every day) with MET-amplified tumors resulted in a higher response rate of 30% (8/27). Most responders were found to have either high MET amplification or MET protein overexpression. Interestingly, the median (range) duration of response in all patients was 202 (51–1,430+) days, rivaling with that of the MET-amplified patients at 197 (64–1,430+) days. In a phase I study with heavily pretreated and heterogeneous population of diverse tumor types, the reported response rates ought to be considered reasonable. Nonetheless, the dose expansion cohort represents a molecularly enriched oncogene-addicted population and one would expect a higher response rate. To draw comparative reference, the phase I study of crizotinib in the ALK fusion–positive advanced non–small cell lung cancer (NSCLC) molecularly enriched population resulted in an ORR of 61% (87/143), including three complete responses and 84 partial responses. Of note, the response rate appeared to be independent of lines of treatment, and the median duration of response was 344 days. Ironically, crizotinib was originally intended as a MET inhibitor in the early drug development, and was soon “repurposed” as an ALK inhibitor after the landmark discovery of EML4-ALK as a lung adenocarcinoma–transforming oncogenic fusion in 2007.
In examining the responses observed in this phase I AMG 337 study, it is worth pointing out that the one patient who experienced complete response (CR) had stage IV MET-amplified distal esophageal adenocarcinoma with MET FISH ratio 25 (on the 200 mg every day dose schedule). Seventy percent (70%) of the partial responders (PR) were patients with MET-amplified tumors. The remaining 30% PR were seen with tumors of unknown MET amplification status, including one (mesothelioma) with MET IHC 50%, intensity 2+. Among all patients, the median MET FISH ratio for responders versus nonresponders was 5.3 versus 1.1, suggestive of a strong association.
Because AMG 337 was noted to yield tumor responses in a subset of MET-amplified patients with GEJ/G/E tumors, a multicenter phase II single-arm study of AMG 337 was conducted in MET-amplified GEJ/G/E tumors and NSCLC (2). In this phase II AMG 337 study by Van Cutsem and colleagues (2), “MET amplification” was selected as the putative predictive biomarker and tested using a MET IQFISH assay with the defining threshold being MET/CEN-7 ratio ≥2. The biomarker analysis in the study was conducted through a single central laboratory by IQFISH. This is the same cutoff as adopted in the phase I study by Hong and colleagues (1), where MET amplification was defined by FISH also as MET/CEN-7 ratio ≥2.0, or next-generation sequencing (NGS) in local testing. However, with considerable disappointment, the ORR in the phase II study cohort 1, measurable MET-amplified GEJ/G/E tumors (mean MET:CEN-7 ratio 6.2; range 2.0–20.4), was merely 18%, with median duration of response being 6 months, whereas no responses were observed in cohort 2. No tumor response was evident in the three patients with MET-amplified NSCLC (mean MET:CEN-7 ratio 4.7; range 2.7–8.6) who received AMG 337 in cohort 2C. Also, with the protocol-permitted review of this study revealing a lower than expected ORR based on preliminary data from the first-in-human phase I AMG 337 study, enrollment in all AMG 337 studies was terminated early, potentially impacting the final study results. Notably, there was no analysis of the study data with different MET:CEN-7 ratio cutoff levels in either of the AMG 337 studies here. Such biomarker cutoff level data could be very useful for overall insight into the response results, although it might not readily resolve all the relevant questions. One challenge in measuring MET amplification is that it is a continuous variable and the optimal cutpoint as a predictive biomarker can be elusive. Another challenge in genomic amplification assay is that it can be conducted using different methods (FISH, qPCR, NGS) with different user-defined cutoffs. Previous studies highlighted the correlation between MET gene amplification levels and response rates under MET inhibitor crizotinib treatment in NSCLC. ORRs observed were 0%, 17%, and 67% in the low-MET (MET:CEN7 ratio, ≥1.8–≤2.2), intermediate-MET (ratio, >2.2–<5.0), and high-MET (ratio, ≥5.0) groups, respectively. It is necessary to also differentiate between true amplification versus aneuploidy or high polysomy of the MET genomic region on chromosome 7. Moreover, the cutoff level defining MET positivity is often set at different levels in different studies, confounding the interpretation of data and proper comparison among various studies. Although MET gene copy number gain can be a result of either true gene amplification versus chromosomal aneuploidy/polysomy, there could be underlying biological and signaling differences as both the HGF and EGFR genes are also located on chromosome 7, as is the MET gene. Studies have defined high level of MET amplification as mean MET/CEN-7 ratio of ≥5 as optimal definition for MET genomic copy number gain–driven lung cancer, which apparently did not harbor other concurrent genomic driver aberrations as observed in other tumors with lower ratio levels. Finally, with the emerging insights of MET exon 14 (METex14) skipping variants as predictive biomarker of MET-targeted therapy, we now appreciate that there can also be some overlap between MET amplification and METex14, although either of them alone have been reported to be independently predictive of MET inhibitor response.
These above confounding factors could account for the overall modest response efficacy in the phase II AMG 337 study (2), which adopted a relatively low cutpoint of MET:CEN-7 ratio ≥2. On the other hand, regarding the question of whether MET protein overexpression acts as a predictive biomarker in MET therapy, lessons can be drawn from prior failed MET inhibitor development using onartuzumab and tivantinib, for example, METLung, METIV-HCC, and MARQUEE. Without dwelling on study details here, these conflicting and less-than-conclusive late-phase MET-targeted therapy study results unavoidably dampened the enthusiasm for further development of the entire class of anticancer agents. Certainly, the methodologic challenges of measuring MET expression level using IHC are also well recognized as in recent cancer immune checkpoint therapy biomarker development using PD-L1 IHC. Another potential confounding factor in assessing MET amplification as AMG 337 biomarker is that other concurrent genomic aberrations could impact the treatment response, for example, HER2 amplification (not excluded in the phase II study). In fact, as the authors pointed out, HER2 pathway activation has indeed been identified as an alternative resistance mechanism against MET inhibition. However, the mean (range) MET/CEN-7 ratio among the 8 responders was 7.7 (2.4–12.0), whereas the mean (range) ratio among the 39 nonresponders was 7.1 (2.0–20.4), thus appearing to be quite comparable between the two response groups.
So, what can we learn here and how can we reconcile the results from the two early-phase MET inhibitor studies that devoted emphasis in evaluating the MET amplification as a potential predictive biomarker for AMG 337? The response rates from MET-amplified patients in the phase I study was 30%, whereas it dropped to only 18% for the cohort 1 patients with advanced MET-amplified G/GEJ/E adenocarcinoma and none in the NSCLC cohort. Although reasons for the differences between the two studies' response data are not immediately clear, some lessons can be drawn from these important studies with potentially useful corrective measures to be taken in future. In the past few years, we witnessed a resurgence of enthusiasm in MET inhibitor development with a renewed focus of biomarker(s)-guidance basket clinical trial studies (3). Emphasis has been invested in the “actionable” MET “driver” events of MET amplification, METex14 (4), and the most recently described MET fusions. Some early-phase MET-targeting clinical studies (e.g., CBT-101) continue to explore MET overexpression (by IHC) as potential biomarkers in addition. Where should we go next at this cross-road of MET inhibitor clinical development marathon journey? First, more attention is needed to better define and refine the cutpoint of MET copy number gain and true amplification level to associate with treatment response. Second, we need to raise the bar and ensure that a more comprehensive MET genomic status analysis is included as biomarker evaluation in MET-targeting studies (i.e., amplification and mutations, METex14, MET fusions, and even protein expression; ref. 5). A broader tumor profiling strategy also could potentially help account for concurrent genomic alterations that might negatively impact MET-targeting inhibitor response in otherwise MET biomarker(s)–positive tumors; however, its requirement in clinical trial design might also pose practical challenges and impact patient accrual. Third, we should advocate for newly biopsied tumor tissues immediately prior to the investigative study treatment whenever possible and deemed safe to do so. Certainly, one needs also a balanced consideration in the study design due to this biopsy requirement's possible negative impact on the patient accrual rate. Also, posttreatment tumor tissues would be quite helpful in evaluating pharmacodynamics, which were not available from the AMG 337 studies here. Finally, one should not underestimate the effort needed in targeted therapy drug development. The phase I study lasted 5 years total. The phase II study screened 2,101 patients to uncover only 132 MET-amplified patients, with eventually 60 patients enrolled. More time-efficient and cost-effective screening methods and strategies would certainly be helpful in future. Although we are not there yet; in a marathon, the loudest applause is not only for the first athlete that crosses the finish-line. For some, lasting memories and impact on the race itself come from later finishers, who after enduring trials and tribulations, solidly complete the long journey.
Disclosure of Potential Conflicts of Interest
P.C. Ma reports receiving speakers bureau honoraria from Merck, AstraZeneca, Takeda, Bayer, and Bristol-Myers Squibb, and is a consultant/advisory board member for AstraZeneca, Caris Life Sciences, Takeda, Cymeta, and Apollomics. No other potential conflicts of interest were disclosed.
This work is supported by IDeA-CTR support NIH/National Institute of General Medical Sciences (NIGMS) grant U54GM104942 (to P.C. Ma).