CHICAGO – Researchers at the American Society of Clinical Oncology annual meeting here this weekend showcased scenarios in which predictive biomarkers can identify cancer patients who are likely to experience greater tumor shrinkage and longer survival from anti-PD-1 immunotherapies.
In one study, researchers from Johns Hopkins Sidney Kimmel Comprehensive Cancer Center and elsewhere considered colorectal cancer patients with deficiencies in mismatch repair, the biological machinery that fixes the mistakes in a newly synthesized DNA strand. In patients whose mismatch repair mechanism doesn't work properly, genetic mutations build up as DNA is copied time and time again, and can become red flags for the immune system to recognize.
Mismatch repair-deficient patients can have hundreds to thousands of mutations in stretches of repetitive DNA, called microsatellites. Tanguy Seiwert of the University of Chicago explained that there is a greater chance that the immune systems of patients with this hypermutational characteristic, known as microsatellite instability, will produce antigens on cancer cells that the immune system will detect and attack. "It's like buying a ticket for the lottery. If you buy a lot of tickets, the chance of benefit[ting] is higher," Seiwert said at the meeting. "Even if you buy fewer tickets, you might still have an opportunity. It's just less likely. I think that's the mutational analogy."
In the Phase II study, researchers led by JHU's Luis Diaz hypothesized that colorectal cancer patients with mismatch repair deficiency might respond particularly well to anti-PD-1 immunotherapy, which acts by blocking the interaction between PD-1 receptors on T cells and PD-L1 and PD-L2 receptors on tumor cells, disabling the mechanism tumor cells use to avoid a immune system attack. To test this theory, researchers enrolled patients with previously-treated metastatic cancer with Merck's Keytruda (pembrolizumab) and followed them for 20 weeks.
At the conference, researchers reported data on around 50 patients who were separated into three arms: those with mismatch repair-deficient colorectal cancer, those with mismatch repair-intact colorectal cancer; and those with mismatch repair deficiencies but other cancer types, such as endometrial, stomach, small bowel, and bile duct cancers. Patients with mismatch repair-deficient colorectal cancer and other tumor types had objective response rates of 62 percent and 60 percent, respectively, after treatment with Keytruda. None of the colorectal cancer patients with intact mismatch repair capabilities experienced a partial or complete response.
Virtually all the mismatch repair-deficient tumors showed response, JHU's Dung Le said at the meeting. "These responses were durable in a treatment refractory patient population and many of these responses are ongoing for over a year," said Le, the first author on the paper published on this study in the New England Journal of Medicine this Saturday.
Having 1,700 mutations is like putting a red flag on a cancer cell and saying to the immune system, 'Here I am.'
Median progression-free survival and overall survival for patients with mismatch repair deficiencies haven't yet been reached, but was 2.3 months and 5 months, respectively, in mismatch repair-intact colorectal cancer patients. "This data suggests that genomics is more influential than histology for mismatch repair-deficient tumors treated with anti-PD1 [therapy]," Le said at the meeting.
Patients with mismatch repair-deficient proteins have a heightened risk for Lynch syndrome, a type of hereditary colorectal cancer, and this deficiency is also implicated in a proportion of sporadic, or non-inherited, colorectal cancer. In the study, this deficiency showed up in around 5 percent of cancer patients.
Researchers used a standard, commercially available PCR test to establish whether patients had microsatellite instability. Personal Genome Diagnostics (PGDx) performed exome sequencing and tumor-specific mutation detection on a subset of patients in this trial, and identified approximately 1,700 mutations in patients with mismatch repair deficiencies, while those with mismatch repair-intact cancer had around 70 mutations.
Having 1,700 mutations "is like putting a red flag on a cancer cell and saying to the immune system, 'Here I am,'" Lynn Schuchter from the University of Pennsylvania and an ASCO expert said. "It allows the immune system to recognize these cancer cells as foreign."
The current study was funded by Swim Across America, the Commonwealth Fund, the Ludwig Center at Johns Hopkins, and the National Institutes of Health.
Mismatch repair as a predictive marker
Diaz, an associate professor of oncology at JHU and cofounder of PGDx, told GenomeWeb that this marks the first time genomics has been used to guide immunotherapy.
Assessing mismatch repair status using PCR or immunohistochemistry has been around since the 1990s, and is part of the standard of care for all colorectal cancer patients. Then, about four years ago, Diaz recalled that he and his colleagues began noticing that patients with mismatch repair deficiencies appeared to be responding better to checkpoint inhibitors, such as Keytruda, than those with intact mismatch repair functions.
Testing for mismatch repair status is generally performed in order to identify patients with Lynch syndrome and for purposes of family genetic counseling. Le noted that oncologists also commonly check for these deficiencies in endometrial cancer patients.
According to her estimates, such tests cost a few hundred dollars, but Le said further study is necessary before mismatch repair status can be used as a predictive marker to gauge immunotherapeutic benefit broadly across cancer types. Researchers hope to reproduce the results from this Phase II study in a larger trial in order establish mismatch repair deficiency as a predictive marker for anti-PD1 therapy. Also, Merck plans to begin enrolling patients into the Phase II KEYNOTE-164 study, which will investigate Keytruda's safety and efficacy in advanced colorectal cancer patients based on their mismatch repair status.
Meanwhile, PGDx has developed a next-generation sequencing approach for measuring mismatch repair that can inform drugmakers' efforts in immuno-oncology. The company this week launched the ImmunoSelect-R laboratory service, through which it will perform CancerXome analysis and predict neoantigens using its proprietary bioinformatics capabilities.
In the Kimmel-led study, researchers also looked at responses according to the expression of PD-L1, the biomarker drugmakers are most commonly using to characterize responders and non-responders to anti-PD-1 drugs. However, in the colorectal cancer setting, patients' response didn't differ significantly according to PD-L1 expression.
"It all comes back to the cancer genome," said Luis, whose group is investigating whether mismatch repair deficiency can identify which patients with prostate cancer and brain tumors respond to anti-PD-1 drugs.
Moving toward standard of care?
The US Food and Drug Administration to date has approved two anti-PD-1 drugs: Keytruda for advanced melanoma patients, and Bristol-Myers Squibb's Opdivo (nivolumab) for advanced melanoma and for second-line treatment for advanced squamous non-small cell lung cancer. None of these indications specify the use of a predictive biomarker to identify best responders. But drugmakers developing anti-PD-1 drugs are all exploring diagnostic strategies to gain better insights into the responder and non-responder populations for these agents.
Herbst said that while the data on PD-L1 as a predictive biomarker in this setting may be hypothesis generating, it shouldn't yet be used to determine which patients should not get Opdivo.
In another Phase III trial presented at the ASCO meeting, researchers led by Luis Paz-Ares of Hospital Universitario Virgen Del Rocio compared advanced, non-squamous NSCLC patients' responses according to PD-L1 expression to Bristol-Myers Squibb's anti-PD-1 drug Opdivo (nivolumab) or docetaxel. Researchers determined PD-L1 expression status on tumor cells using Dako's IHC test, and patients were stratified into three groups based on whether they had PD-L1 expression in 1 percent or more, 5 percent or more, or 10 percent or more of tumor cells.
In the overall population of approximately 600 patients, those on Opdivo had a 27 percent lower risk of death than those receiving docetaxel. However, patients with a high level of PD-L1 expression and treated with Opdivo had between a 41 percent and 60 percent reduction in the risk of death compared to those on the standard chemotherapy arm.
Overall survival approximately doubled with Opdivo compared to docetaxel across the PD-L1 expression continuum, Paz-Ares said at the meeting. However, when PD-L1 was not expressed in the tumor, there was no difference in overall survival between the treatment arms. Treatment-related adverse events, including serious toxicities, were also lower for those receiving the anti-PD-1 drug compared to those treated with docetaxel.
This is the second Phase III study to demonstrate Opdivo's superiority over docetaxel in advanced NSCLC, and based on the data, Roy Herbst from the Yale School of Medicine said that the drug might become the new standard of care in previously-treated non-squamous NSCLC. However, at the meeting, Herbst said that while the data on PD-L1 as a predictive biomarker in this setting may be hypothesis generating, it shouldn't yet be used to determine which patients should not get Opdivo.
PD-L1 is “a dynamic marker, and that in itself is a very difficult problem,” Herbst observed. “There is heterogeneity in multiple tumors and multiple pathways within a tumor where you can get different measurements.” Then, there are other considerations that make PD-L1 a challenging biomarker to define, such as when to test the patient to accurately reflect the characteristics of the tumor and whether to test PD-L1 on just tumor cells or also on the immune cells, as Genentech is doing.
Paz-Ares acknowledged that while PD-L1 was a good positive predictor of Opdivo's response, some patients who lack PD-L1 expression still might benefit from treatment. Other drugmakers developing anti-PD-1 and anti-PD-L1 immunotherapies have observed similar trends in lung cancer indications. "Still, we don't have the right negative biomarkers" for predicting who should not get treatment, Paz-Ares noted.
Ultimately, the studies by Le's group and Paz-Ares' team showcase two different ways of predicting response to anti-PD-1 drugs. While Le and colleagues focused on the role of genetics to differentiate responses, Paz-Ares' work used markers related to the mechanism of anti-PD-1 treatments.
UPenn's Schuchter observed that the study by Diaz and colleagues is "not about the immune characteristics of the tumor, it's not about PD-L1 expression," but is asking "is there something about the genetic makeup? Is there something about how many red flags are on a cancer cell that really predicts response? It's a different way of thinking [about] which patients are going to respond to the immunotherapy."