Nature Biotechnology, April 2017

The first cancer drug approval based on a marker, not a tumor type, appears immi­nent. The FDA has set a June 9 action date to decide whether Merck’s checkpoint inhibi­tor Keytruda (pembrolizumab) can be used with a test for microsatellite instability in previously treated cancer patients. “It will be the first pan-tumor marker in the history of oncology or FDA approval,” says Luis Diaz, head of solid tumor oncology at Memorial Sloan-Kettering Cancer Center in New York and a Keytruda trial lead investigator. Keytruda, an anti-programmed death-1 (PD- 1) antibody made by the Kenilworth, New Jersey–based pharma, is already approved for metastatic melanoma, Hodgkin lym­phoma and certain lung and head and neck cancers (Nat. Biotechnol. 33, 1217–1218, 2015). But should the US Food and Drug Administration (FDA) give the go-ahead to the drug/test combination, its blanket approval across all tumor types would mean that all advanced cancers will probably be tested for high-level microsatellite instabil­ity (MSI-H) or its underlying cause, a DNA repair pathway defect. This would be the first purely molecular indication for a cancer drug—a major advance in clinical oncology.

An estimated 1 in 25 tumors (all types combined) displays a high degree of micro­satellite instability. This hypermutation phe­notype usually results from replication errors due to a failure in the DNA mismatch repair system (MMR).

A deficient MMR generates a mutation burden typically one to two orders of mag­nitude higher than in MMR-proficient cells. Tumor cells are unable to correct errone­ous base insertions or deletions, leading to many subsequent missense mutations across the genome. Because short tandem repeats, called microsatellites, are especially vulner­able to mutations, they have proved useful as hypermutation markers.

This dramatically increased mutation bur­den in turn makes MMR-deficient tumors very sensitive to checkpoint blockade, because many of the randomly mutated genes generate mutant proteins seen as foreign by T cells. Early clinical data seemed to back this hypothesis. An investigator-sponsored phase 2 trial at Johns Hopkins University in Baltimore showed four of ten MMR-deficient advanced colorectal cancer patients had an objective response to Keytruda by standard criteria, whereas no patients with MMR-proficient tumors responded. In patients with cancers other than colorectal, who were also MMR-deficient, five of seven responded (N. Engl. J. Med. 372, 2509–2520, 2015.)

Expanded results presented at the 2016 annual meeting of the American Society of Clinical Oncology (ASCO) (updated from J. Clin. Oncol. 34, 142S, abstract 3003, 2016) were even more dramatic: of 30 patients with MMR-deficient, non-colorectal can­cers, 9 (30%) had a complete response; 7, a partial response; and 5, stable disease. These tumor types (i.e., pancreatic and prostate) have not previously shown good responses to checkpoint inhibitors. “Really impressive results,” comments Sanford Markowitz, a medical oncologist at Case Western Reserve University in Cleveland, who is not involved in the trials. Trial patients “have already been heavily pretreated with conventional thera­pies,” he notes. “Any response rate would be highly unexpected, and to see complete responses and durable responses is sim­ply remarkable.” At ASCO, Johns Hopkins oncologist Dung Le reported a 57% one-year progression-free survival and 81% overall survival for the 30 non-colorectal patients “This isn’t a therapy that extends life by a few weeks or months,” says Diaz. “This has the potential to transform a patient’s disease from a lethal one to one that is in remission.”

Johns Hopkins investigators, not Merck, first hypothesized that MMR-deficient tumors would be sensitive to immunother­apy. The enabling discovery came in 1993, when a Hopkins group showed that micro­satellite instability in Lynch syndrome, an inherited cancer predisposition syndrome, was caused by mutated genes in the MMR pathway (Cell 75, 2515–1225, 1993). None of the researchers thought this would lead to a therapy, recalls Hopkins molecular geneticist Bert Vogelstein. “A defect in repair would, if anything, make these tumors more difficult to treat,” he says, because they would more easily develop resistance mutations. But in 2008 Vogelstein and immunologist Jim Allison, then at Memorial Sloan-Kettering, found that many mutant peptides from human tumors are presented by the major histocompatibility complex molecules of immune cells (Cancer Res. 68, 889–892, 2008). They speculated that checkpoint blockade could be an effective therapy, because of the multiple antigen targets avail­able to T cells, nearby T cells could be acti­vated against them.

Ironically, it took the complete failure of PD-1 checkpoint blockade, in colorec­tal cancer patients treated with Bristol- Myers Squibb’s drug Opdivo (nivolumab; N. Engl. J. Med. 366, 2443–2454, 2012) to trig­ger an MMR-based clinical trial. In two phase 1 trials, only one of 20 patients had responded to the checkpoint inhibitor. Johns Hopkins trial investigators Suzanne Topalian and Drew Pardoll, puzzled, went to Vogelstein’s labora­tory to ask why he thought that one patient only had responded. “We said, ‘Oh, we know what must be going on. That patient must have mismatch repair deficiency’,” says Vogelstein. “We were certain.” The patient’s tumor, sure enough, tested positive. “As soon as that result came back, we immediately designed the [Keytruda] trial,” says Vogelstein.

Although many Keytruda-induced responses of MMR-deficient tumors to date are durable—up to three years in some patients, says Diaz—drug resistance is always a possibility. “That will be the question long term,” says Markowitz. “Is there resistance, and what’s the mechanism?” Because most of the neoantigens are passenger mutations, entirely inconsequential to the tumor, there should be selective pressure against them during treatment, he notes.

Groups advocating for patients are eagerly awaiting FDA approval. The Lustgarten Foundation in Bethpage, New York, which funds pancreatic cancer research, has already been paying for free MMR-deficiency testing for any pancreatic cancer patient, to encour­age enrollment in the clinical trial. It will con­tinue paying for testing after approval until insurance reimbursement is in place. With an estimated 2% of pancreatic cancers MMR-deficient, “there could be a thousand or two thousand patients who have this,” says foun­dation president Kerri Kaplan. “That would be a huge win.” At ASCO, Le reported on four pancreatic cancer patients, presenting scans for one who had a dramatic response. “When they had the proof of those four patients, that was really what got us to pay for the testing,” says Kaplan. (The foundation is also helping fund the trial.)

Diaz says treatment of MMR-deficient tumors with checkpoint inhibitors off-label is already happening, with some insurers reim­bursing for MMR-deficient colorectal cancer according to one report (Clin. Adv. Hematol. Oncol. 14, 476–479, 2016). “If I were seeing pancreatic cancer, prostate cancer, GI tumors, which generally have low rates of response to PD-1…I would certainly test even if there were only a 5% chance of finding some­body with mismatch repair deficiency and a high mutational load,” said Mario Sznol, an oncologist at Yale University in New Haven, Connecticut, at ASCO. “That patient could have substantial benefit.”

“If you’re a medical oncologist you’d be hard pressed not to want to use this drug for someone who has a mismatch repair–defi­cient tumor,” agrees Markowitz, who consid­ers FDA approval certain. “I think the clinical trials going forward will clearly want to be looking at this up front and in combinations, but I suspect the initial approval will probably be for monotherapy and for salvage [prior treatment failures].”

Merck is not including a companion diag­nostic in its application, because tumor MMR deficiency testing is already widely available and cheap. Indeed, it has long been standard practice to test colorectal cancer patients for Lynch syndrome, so that genetic counseling can be offered to affected families. (The test costs between $100 and $150, says Kaplan.) Tests can either detect MMR protein absence with immunohistochemistry (IHC), or quan­tify microsatellite instability using PCR, comparing tumor tissue to normal tissue. Neither test is completely sensitive, because an MMR gene may be qualitatively, but not quantitatively, changed by mutation and thus stain even though it’s functionally disabled, while PCR testing for microsatellites may miss the defect because the signal is absent due to intratumoral heterogeneity. “If the person has a tumor block that’s chock full of tumors, microsatellite instability [by PCR testing] is the best,” says Vogelstein. “If not, mismatch repair deficiency [by IHC analy­sis] is the best.” Because testing is so cheap relative to treatment, Vogelstein recommends doing both tests.

Such testing will likely become standard. “Clearly a patient who has failed standard of care and has no hope, or little hope, it’s hard to argue that that patient shouldn’t be tested,” says Vogelstein. “This therapy could be lifesaving.”

The FDA presumably is considering data from several ongoing Merck-sponsored tri­als, in addition to the original Johns Hopkins trial. If the FDA makes Keytruda treatment completely independent of tumor type, the impact could go beyond the small minority of patients who will actually qualify for the treatment. “It will have an effect on driving personalized medicine forward,” says Diaz, “in a way that… individual, disease-specific markers have not.”

Ken Garber,  Ann Arbor, Michigan