When translational researchers, drug developers, and clinical scientists seek immuno-oncology (IO) foresight, they often find it in biomarkers—or better still, the patterns formed by multiple biomarkers. Single biomarkers convey little of cancer’s complexity, leaving much unclear about the signaling pathways that sustain or suppress malignancies, the mechanisms that activate or deactivate specific immune defenses, or the contingencies that culminate in “hot” or “cold” tumor microenvironments (TMEs). Multiple biomarkers, unlike single biomarkers, can capture a lot of cancer’s complexity. What’s more, biomarkers in a multiplex panel can contextualize each other, resulting in models that can extrapolate from current circumstances to future outcomes.
Single, isolated biomarkers can determine matters of degree, or they can supply yes-or-no answers, like those displayed by a Magic 8-Ball. More elaborate means of divination, however, are needed to sift through disparate signals from multiple biomarkers and deliver satisfactory readings.
Various interpretive approaches were discussed at an Immuno-Oncology 360° (IO360°) event that was held recently in New York, NY. Although the event covered a range of topics, including IO-related business development and clinical deployment issues, much of its 360° sweep was devoted to IO biomarkers
Mechanisms of resistance
At the IO360° event, a panel of scientists looked at how biomarkers and companion diagnostics can be used to predict responses to immunotherapy. One of the panel’s scientists, Alessandra Cesano, MD, PhD, chief medical officer of NanoString Technologies, directly addressed the single/multiple biomarker issue. She contrasted “dichotomy markers,” which cover oncogenic mutations, with multiplex panels, which may cover various immunological dimensions.
When administration of a targeted therapy is being considered, we might ask, “Is the relevant oncogenic mutation present or not?” When developing an immuno-oncology therapy, however, “we are not looking at the tumor so much as we are looking at the immune response to the tumor,” Cesano said. The markers are more heterogenous and dynamic over time, and over different tissues, and so even biomarkers that have been repeatedly shown to enrich for responders to checkpoint inhibitors—PD-L1 immunohistochemistry and microsatellite instability—have suboptimal negative and positive predictive values.
“Looking at a single marker for such a complex phenomenon that could have so many mechanisms of immune resistance is definitely a failure to start with,” she continued. IO-based multiplex panels look in parallel at different mechanisms of immune resistance. Such panels, she suggested, may be more valuable than single analytes in isolation.
Myeloid-derived suppressor cells
NeoGenomics Laboratories looks for biomarkers in the TME, identifying the relationship among cells that contribute to immunosuppression and promote cancer cell proliferation. By serially repeating a process of immunostaining, imaging, and deactivation of the fluorescent dyes, the company’s MultiOmyx platform can visualize up to 60 biomarkers on a single 4-µM formalin-fixed paraffin-embedded slide, categorizing immune cells by type and revealing their spatial relationships, that is, their locations with respect to each other and with cancer cells.
During a presentation at the IO360° event, Josette William Ragheb, MD, PhD, a medical director at NeoGenomics, compared the immune landscape for normal bone marrow with that for bone marrow from acute myeloid leukemia (AML) patients. For example, she characterized the changes in the immune cells surrounding the leukemic cells.
Her analysis emphasized monocytic and granulocytic myeloid-derived suppressor cells (M-MDSCs, G-MDSCs) and protumoral (M2-type) tumor-associated macrophages (TAMs) in the bone marrow of AML patients. She noted that the MultiOmyx platform, which processes information about quantity, staining intensity, and spatial relationships of the different cells present, can be used to compare the effects of therapies.
Now that such platforms are becoming available, cells of the immune environment are emerging as biomarkers that could be used to predict therapeutic outcomes. Also, these cells may qualify, Ragheb suggested, “as therapeutic targets of many immunomodulating agents.”
Single-cell cytokine secretion profiles
Single-cell suspensions don’t present spatial context like needle biopsies do, but the cytokines each individual cell produces still tell a compelling story. For example, cytokine secretion profiles correlate with, and can help predict, clinical outcomes in immunotherapy, said Will Singleterry, PhD, director of business development at IsoPlexis. To back this assertion, he presented data from the company’s many collaborations.
“We’re able to couple cytokine secretion back to the individual cells that produce them,” Singleterry asserted. “We can capture up to about 40 secreted proteins per cell, and we do that for thousands of cells at a time. We’re seeing that the responders have these really polyfunctional T cells that secrete two or more cytokines … in an extremely intense manner.”
IsoPlexis has defined a metric, the Polyfunctional Strength Index (PSI), that quantifies the overall activity of a sample and serves as a surrogate marker for immune fitness. It is, essentially, the product of the number of polyfunctional responders and the intensity of response.
By employing the PSI, researchers at the University of California, San Diego, determined which subtle CRISPR edits led to the best candidate construct. The PSI has also been used at the NIH, where the metric facilitated the optimization of a manufacturing protocol. (A three-day activation resulted in a more functional CAR T-cell product than did a five-day activation.) According to the NIH scientists, the PSI of preinfusion CAR T-cell products showed significant correlation with the objective response of patients and neurotoxicity, the latter of which correlated with interleukin-17 production.
IsoPlexis is currently for research use only. “But everybody,” noted Singleterry, “is looking for a product release assay.”
Fresh histoculture tumor models
Sometimes the best way to assess the potential of a therapy is to see what happens when it’s given to a tumor. That’s the idea behind Mitra Biotech’s CANscript technology—a fully human ex vivo histoculture model system that uses fresh tumors and patient blood. “It preserves the phenotype and the molecular profile of the tumors even though we’re putting them in a dish,” said Mark Paris, PhD, the company’s director of translational applications.
Paris’ presentation at the IO360° event focused on results demonstrating CANscript’s ability to assess the impact of traditional and novel immunomodulating agents. This ability is important, Paris pointed out, because it is, ultimately, about “the prediction of clinical response.”
CANscript is underpinned by a series of phenotypic assays. In a large clinical study managed by Mitra Biotech, these assays were deemed predictive. A machine learning algorithm was trained by replicating in the assay the therapy received by about 850 patients, using about a dozen phenotypic endpoints. It was then validated on another 850 patients. “We went in, predicted their response, and then tracked their outcomes in the clinic,” recalled Paris. “Same drug, same patient… That’s the basis of our M-score.”
The company has collected statistics on around a dozen different tumor types and more than 80 clinical regimens, with the M-score being Mitra Biotech’s deliverable to oncologists as well as biopharma clients. The latter use the tissue and supernatant to probe the mechanistic and molecular events engendered by their compounds, and “layer on the M-score to understand where the phenotypic effects are predicting a clinical response,” Paris noted.
The immune system keeps a record of everything to which it has been exposed. To tap into this record, Adaptive Biotechnologies has developed a specialized antigen receptor sequencing platform. “It does just one thing,” said Lanny Kirsch, MD, the company’s senior vice president of translational medicine. “It enumerates, specifies, and quantifies each and every B and/or T cell in any sample of interest.” The platform has been used to look for dominant clones in certain lymphoid malignancies, and to subsequently monitor disease status.
“The very same principle can be applied to immunotherapy,” he continued. Inside a solid tumor, T cells may form a pattern indicating that they recognize the tumor as being foreign. If so, patients with such a tumor may be more likely to respond to an immunomodulatory agent. If the same tumor-associated clones in the blood expand when given an immunomodulatory agent, it may be likely that that those patients responded to the immunotherapeutic. “The more the repertoire is changing,” Kirsch added, “the more likely it is that clones that have been held at bay or sequestered are now being released by the intervention. And those are the patients may be more likely to have adverse events in terms of the immunotherapeutic.”
After describing how Adaptive’s platform is being explored in oncology and immunotherapy, Kirsch presented the idea behind a recently announced collaboration with Genentech. “The same platform,” he said, “can be used to link a sequence with a target, and therefore develop a therapeutic.” He concluded with the mention of a collaboration with Microsoft to “universalize the linkage between a sequence and its target by doing a kind of in silico structural biology using machine learning,” with the aspirational goal of antigen mapping multiple diseases from a single blood test.