Nature Reviews Drug Discovery, August 2014.

High-profile failures imperilled emerging anticancer MET inhibitors, but trials guided by a new biomarker are beginning to restore confidence.

 Ken Garber

 The failure of a Phase III lung cancer trial of the MET-specific antibody onartuzumab shocked the community and dented hopes for an important anticancer drug class when Genentech announced the results in March. Genentech halted the METLung trial early, after an interim analysis showed that onartuzumab in combination with the endothelial growth factor receptor (EGFR) inhibitor erlotinib did not provide any meaningful benefit over erlotinib alone. ArQule’s putative MET inhibitor tivantinib had failed 18 months earlier in a similar Phase III lung cancer trial. The value of MET as a target in its most promising major indications, lung cancer and gastric cancer (where GlaxoSmithKline’s foretinib failed in Phase II) seemed in doubt.

But this year’s annual meeting of the American Society for Clinical Oncology (ASCO) signalled the return of MET as a viable target. Critics of Genentech’s METLung trial described flaws in the patient selection strategy, while several companies presented promising early-stage results using a better-performing biomarker. “We continue to be convinced that MET is a good target,” says Charles Baum, CEO of Mirati Therapeutics, a developer of MET inhibitors.

Many are welcoming the reprieve. MET, a receptor tyrosine kinase, is a popular target, with at least 20 compounds in active clinical development (Table 1).  It has varied pro-tumour effects, activating signalling pathways that promote cell proliferation, survival, invasion and angiogenesis. But because these downstream pathways are not solely controlled by MET, MET inhibition will only be effective if MET is activated and driving tumour growth. Selecting the right patients is therefore key to Met inhibitor development.

Patient selection strategies, however, are anything but straightforward. Unlike finding activating gene mutations and translocations, there is no practical method to easily detect activated MET. When hepatocyte growth factor (HGF), MET’s ligand, binds the receptor, MET dimerizes and autophosphorylates. “MET expression by itself does not mean that it’s active; you need the phosphorylation event,” says Ravi Salgia, a thoracic oncologist at the University of Chicago, USA. But although antibodies against phosphorylated residues could theoretically reveal MET activation, the receptor phosphorylation sites quickly degrade ex vivo. Other tools must be used on clinical samples.

Catching FISH

Genentech selected patients using an immunohistochemistry (IHC) assay that measures MET protein expression. The company tested this biomarker in onartuzumab’s Phase II lung cancer trial. Although the trial failed overall, a pre-specified population of 66 patients who were MET-positive (as determined by IHC) did dramatically better on the combination therapy than did MET-negative patients (J. Clin. Oncol. 31, 4105–4114; 2013). “It was pretty compelling data,” says Bruce Johnson, a thoracic oncologist at the Dana-Farber Cancer Institute in Boston. “I thought the Phase III trial was a pretty good idea, and I put some of our patients on it.”

Onartuzumab is a single-arm antibody that effectively prevents ligand–receptor binding without triggering receptor dimerization and activation. But the trial was stopped early for futility. “If anything, [onartuzumab patients] did a little bit worse,” notes Johnson, who was surprised at the outcome.

Not everyone was taken aback. “There were many flaws in their Phase II study, and I think many of us would have very confidently predicted that they would have a negative Phase III study,” says Ross Camidge, a medical oncologist at the University of Colorado in Denver, USA. “I don’t think they had strong preclinical evidence to justify their combination.” More importantly, Federico Cappuzzo, Director of Medical Oncology at Ospedale Civile in Livorno, Italy, told the audience at the ASCO onartuzumab trial session that Genentech’s IHC assay over-identified likely responders by a full order of magnitude. The assay pulled in one-half of the patients who were screened but, based on his own and others’ studies, “the best scenario is… that these agents could be effective in 3–4% of our patients,” said Cappuzzo.

In the earlier Phase II trial, Genentech’s IHC cut-off point for MET positivity had boosted the survival curve for patients with high MET expression versus low MET expression. But owing to low numbers of patients, and because the trial was a negative study overall, “the real worry was that it was a complete statistical fluke,” says Camidge, “which seems to have been borne out by the results of their Phase III.”

None of this means that MET inhibition doesn’t work. “It just means you have to define who it works in,” says Camidge. Camidge presented new results at ASCO from an ongoing Phase I trial of Pfizer’s crizotinib, a small-molecule inhibitor of MET and anaplastic lymphoma kinase (ALK), using MET amplification measured by fluorescence in situ hybridization (FISH) as the selection biomarker. Whereas IHC uses an antibody to stain MET at the cell surface, FISH images the copies of a target gene in the nucleus. In a prospectively selected cohort of six lung cancer patients with highly MET-amplified tumours, defined by a MET to centromere 7 (MET/CEP7) copy ratio of five or more, single-agent crizotinib induced a 67% objective response rate, with durable responses.

“Gene amplification is better, we think, than IHC,” says James Christensen, Chief Scientific Officer of Mirati. Christensen worked at Pfizer before joining Mirati in 2013, and wrote the crizotinib trial’s MET amplification inclusion criteria. The crizotinib team defined amplification using the MET/CEP7 ratio rather than total copies of the MET gene because extra copies of MET should represent a true selection by the cancer cell, whereas extra copies of the entire chromosome 7, which includes MET, may be less biologically relevant.

But by this definition, very few lung cancer patients have highly MET-amplified tumours. In the crizotinib trial, such patients represented only 0.8% of the total number of patients screened. That translates to 2,000 lung cancer patients per year in the United States, barely enough patients for companies to screen for in clinical trials. Baum and Christensen think that adding Met-mutated tumours on top of amplification cases could bring the population up to 5–7% of lung cancers, which would be plenty to justify patient screening if a cost-effective multiplex diagnostic could be developed. (This estimate was partly based on recent evidence suggesting that MET-activating mutations are more common than previously thought. (Nature 9 Jul 2014.)

Data from other companies at ASCO confirmed the usefulness of MET-amplification biomarkers. Phase I data from Amgen’s small-molecule MET inhibitor AMG 337 showed that 6 out of 14 patients with MET-amplified solid tumours achieved an objective response, including 5 with gastroesophageal cancers. AbbVie’s monoclonal antibody ABT‑700 showed a 50% response rate in MET-amplified solid tumours, and AbbVie is now selecting such patients prospectively. Incyte’s INC280 also showed good activity in patients with MET-amplified solid tumours “You’ve seen that MET-amplified patients can be highly responsive to a small-molecule inhibitor,” says Baum. “That’s helped to show people that that’s the right track.”

Defining cutpoints

The best cutpoint for MET amplification has yet to be defined, says Camidge, and criteria may well become more inclusive as more MET-amplified patients enrol. “It’s the appropriate way to develop targeted drugs, “ he says. “You don’t commit to a Phase III with your best guess at a predictive biomarker.” That was Genentech’s mistake, he adds. “If you do the work early on and you really feel confident in your predictive biomarker signal, then you are much more likely to have a positive registration study.”

Genentech contends it did just that. The company found in Phase II that IHC predicted outcome better than did FISH (Clin. Cancer Res. 31 Mar 2014). Retrospective analysis of Phase III results also showed that “MET FISH did not predict for… positive outcomes in METLung,” writes See Phan, Associate Group Medical Director of Genentech, in an e-mail. But, crucially, Genentech defined FISH positivity as a MET/CEP7 ratio of two or more or a minimum of five total copies of MET per cell, much more inclusive definitions than Pfizer’s MET/CEP7 ratio of five or more. This could account for why Genentech’s FISH assay poorly predicted the response to onartuzumab. “MET amplification is a continuous variable,” notes Camidge. “Very low MET amplification… may not be relevant to the cancer; it may just be noise in the cancer cell.” In the end, Genentech rejected FISH for IHC, which casts an even wider net.

There is a lesson here, says Camidge. One strategy is to “grab the biggest slice of the pie, running the risk that you have a negative study,” he says. Another approach is to test the drug in a smaller but more secure market slice. “Get your drug licensed,” Camidge says, “and then gradually look for the other groups in which it works. The second appears to be a much more successful strategy.”

IHC MET expression biomarkers are, however, also still alive. Genentech is using IHC in at least one other onartuzumab Phase III trial. ArQule, too, is using it in a Phase III liver cancer trial of tivantinib. (Last year, two groups reported that tivantinib mostly works by targeting microtubules, not by inhibiting MET (Clin. Cancer Res. 19, 2381–2392; 2013; Cancer Res. 73, 3087–3096; 2013). ArQule strongly disputes these studies.) Amgen is also using IHC to select patients for a Phase III trial of its HGF-specific antibody in gastric cancer, and Lilly is doing the same for its Phase II MET-specific antibody in lung and

gastric cancers. “I think [these trials] do have a chance of success,” says Salgia, who stresses that IHC assays vary widely depending on which MET epitope is targeted and which antibodies are used.

And no single biomarker fully captures the complexity of the response to MET inhibitors. “It’s just not going to be a straightforward one-to-one patient selection strategy,” says Christensen. The functionally redundant RON receptor tyrosine kinase may be as important as MET, sayts Salgia.  And the effects of MET may also vary owing to autocrine versus paracrine activation, expression of certain transcription factors, the mutation status of ubiquitin ligases, and the presence or absence of phosphorylated MET in the nucleus. And different drugs may need different response biomarkers depending on their specific mechanism of action. Co‑targets like EGFR must also be considered.

The fact that MET amplification, when strictly defined, reliably picks out responders is a good beginning though. MET patient selection in lung cancer isn’t proving as easy as EGFR, ALK or ROS1, but one must grab the low-hanging fruit first, says Camidge. “There’s a lot of fruit on the tree, and you have to find a way of getting to it.”

Table 1 | Selected MET inhibitors in development

Agent                                      Company                                Phase

Onartuzumab (MetMab)*      Genentech/Roche                     III

Rilotumumab*                         Amgen                                       III

Cabozantinib (Cometriq) ‡    Exelixis                                        III

Tivantinib                                 Arqule                                        III

LY2875358*                             Eli Lilly                                       II

Ficlatuzumab*                         Aveo Pharmaceuticals                II

Foretinib                                   Exelixis                                        II

Crizotinib (Xalkori) §             Pfizer                                             II

INC280                                    Novartis/Incyte                           II

AMG337                                  Amgen                                          II

MSC2156119J                         Merck Serono                               II

ABT-700*                               AbbVie                                            I

MGCD265¶                             Mirati Therapeutics                       I

Volitinib (HMPL‑504)          Hutchison MediPharma                  I

SAR125844                             Sanofi                                             I

Altiratinib#                             Deciphera Pharmaceuticals           I

LY2801653                             Eli Lilly                                           I

*Antibody drug. ‡Multikinase inhibitor against MET, RET, and vascular endothelial growth factor 2 (VEGFR2). Approved for medullary thyroid cancer. §Multikinase inhibitor against MET and anaplastic lymphoma kinase (ALK). Approved for ALK-positive non-small-cell lung cancer. II Multikinase inhibitor against MET and VEGFR2. ¶Multikinase inhibitor against MET, AXL and, VEGFRs. #Multikinase inhibitor against MET, the angiopoietin 1 receptor TIE2 and VEGFR2.