Santa Monica, CA, September 8, 2016“One has the feeling that, ultimately, PD-L1 will not be the biomarker people use to select patients for immunotherapy” - Vice President, Global Medicines Development, AstraZenecaI spent the latter part of last week listening to oncology leaders share their perspective at the Biomarkers for Immuno-Oncology (I/O) program at the 2016 Immuno-Oncology Summit in Boston, MA. While the talks covered a wide range of topics, certain themes arose across multiple presentations that I think give insight into the future of companion diagnostics (CDx) for cancer immunotherapies. Here are my key takeaways from the conference:

• Multi-parameter approaches to biomarker analysis are the future of I/O CDxThough PD-L1 expression in tumors correlates with response to PD-1/L1 blockade, typically, more than half of even high-expressing patients will not respond to treatment, underscoring the need for more powerful predictive markers. Conversely, many patients not overexpressing this marker respond to I/O drugs. Many speakers at the conference suggested that multi-parameter analysis (DNA, RNA, protein, cellular) of the tumor and its microenvironment, and potentially peripheral blood, could provide a more comprehensive picture of tumor immunogenicity and a patient’s likelihood of response to immunotherapies. This hypothesis is consistent with my previous analysis of ongoing I/O clinical trials, which showed that many I/O trials are pursuing a multiple-marker approach. Much of the talk at the conference focused on tumor mutation burden (TMB) and/or microsatellite instability MSI analysis as potential markers of response to immunotherapies, including the current anti-PD-1/L1 drugs. Speakers from both Merck and Genentech highlighted data demonstrating that TMB can be predictive of response to Keytruda and Tecentriq, respectively, in certain tumors. A speaker from the FDA also identified MMR/MSI status and TMB as emerging biomarkers for which the agency is currently developing regulatory frameworks. These talks come on the heels of the recent announcement that Foundation Medicine added TMB and MSI analysis capabilities to its FoundationOne and FoundationOne Heme assays to help guide the use of immunotherapies. Additionally, our internal primary research in this field reveals that approximately half of oncology stakeholders (oncologists, pathologists, and lab directors) anticipate that NGS will have a significant impact on I/O diagnostics (driven by TMB and potentially HLA analysis), with the majority of the remaining stakeholders predicting a moderate role for NGS (<10% of surveyed stakeholders predict limited / no role). With pharma and diagnostics companies, the FDA, and oncology stakeholders all aligning behind TMB as a predictive marker, it seems likely that TMB or MSI testing will eventually be incorporated into routine clinical practice for I/O therapy selection.In addition to TMB testing, both Genentech and Merck also identified gene expression profiling, particularly interferon γ (IFNγ)-related gene expression profiles, as predictive of response to immunotherapies. Data from Merck showed that the vast majority of patients with head and neck, bladder, and gastric cancers who had a durable partial or complete response to Keytruda had relatively high expression of IFNγ-related genes while a Genentech study showed that patients with high IFNγ gene expression had a hazard ratio (HR) for Atezolizumab vs. Docetaxal at 20 months that was equivalent to that of PD-L1 high expressers (>50% PD-L1 expression on tumor cells and >10% on immune cells). Another technique commonly-mentioned as a tool for I/O biomarker analysis was multiplex IHC, which can be used to measure PD-L1 expression as well as the types, density, and location of immune cells in the tumor microenvironment, leading to my next takeaway…
• Cellular localization information may be critical to understanding tumor immunogenicity and predicting response to therapyIn our research in the field of I/O, many oncologists and pathologists we have spoken to have suggested that the “grind-and-bind” approach of deconstructing a tissue sample and binding or analyzing the target(s) of interest in the digested sample would be suitable for future diagnostic modalities, dismissing the need for the spatial context of biomarkers within a tissue sample. Multiple speakers at the conference, however, presented data reiterating that the location of PD-L1 expression and immune cells within a tumor can correlate with response to therapy. For example, data presented showed that response to PD-1 blockade in some cancers, such as metastatic melanoma or colorectal cancer, varies depending on the location of PD-L1 expression and/or T-cells within a tumor sample (i.e., within the tumor core vs. at the invasive front). The speakers highlighted research showing that CD8+ cell density in the invasive margin of a tumor is more predictive of response than CD8+ cell density or PD-1/L1 expression elsewhere in the tumor (source). Additionally, research suggests that changes in the location of PD-L1 expression and/or immune cells following treatment can be used as an early indicator of response to therapy. At least one company, HalioDx, is exploring the use of T-cell localization as a predictive biomarker for response to immunotherapy. HalioDx’s assay, which measures the densities of CD3+ and CD8+ immune cells in both the tumor core and invasive margin via IHC to arrive at a proprietary “ImmunoScore”, has already been established as a prognostic tool for colorectal cancer, and is currently being explored for its predictive potential for use with immunotherapies. Ultimately, while various emerging technologies (e.g., NGS, NanoString) show significant promise as new tools for immuno-oncology diagnostics, it is apparent that IHC tissue-based testing will remain a staple in oncology CDx. However, as the complexity of tissue analysis increases to include more markers and cell types and increasing spatial context, there is likely to be a shift towards more sophisticated IHC technologies, such as mutliplex and/or digital pathology platforms.

• Combination treatments are on the rise and may be a driver of the CDx marketWhile current checkpoint inhibitors have demonstrated impressive results as monotherapy agents, the majority of ongoing clinical trials for cancer immunotherapy are for combination treatments. One of the primary goals of I/O combinations is to address the large share of patients that are unresponsive to mono-immunotherapies (e.g., those with no PD-L1 expression and/or minimal tumor-invasive immune cells). A common strategy for I/O combination treatments is to induce immunogenicity in an otherwise non-immunogenic tumor by treating it with a primary therapy, such as chemotherapy or antigen-specific T-cells, followed by administration of an immunotherapy, such as a checkpoint inhibitor, to allow the immune system to attack the tumor. One common feature of two I/O combination therapy presentations (one from Astra-Zeneca on its investigational Olaparib / Durvalumab combination, and one from Genentech on its Bevacizumab / Atezolizumab combination) is the utilization of serial biomarker analysis to interrogate biomarkers prior to therapy, after the initial priming treatment, and again after the addition of the combination therapy. Within the context of clinical trials, these analyses allow investigators to confirm that target tumors are indeed becoming immunogenic following the run-in treatment and subsequent immunotherapy. For example, data presented on Genentech’s Bev / Atezo combo showed an increase in both intra-tumoral CD8+ cell density and immune-related gene expression following treatment with Bevacizumab, and again with Atezolizumab, in RCC patients in a phase 1b trial (source). Feedback from the presenters suggested that a serial testing strategy to confirm the onset of immunogenicity could potentially become part of the clinical paradigm for managing patients on combination therapies. If so, combination therapies could be a boon for the CDx market as patients would receive multiple tests throughout the course of treatment, as opposed to a single test for therapy selection prior to treatment – as is currently the case. Such a protocol, however, would present distinct challenges to adoption, namely the need for repeat sample collection, which would be particularly challenging for tissue-based assays, and the increased costs associated with additional testing. While many clinical trials are conducting biomarker analyses for combination therapies, it remains to be determined whether drug companies will seek to include serial testing on drug labels or whether the FDA would recommend or require it for the use of certain combinations.

Overall, the talks and discussion at the conference made it clear that the field of immuno-oncology is still in its infancy and is progressing rapidly, and that the way we diagnose and treat cancer in the near-mid future may be distinctly different than the way we do today.Disclaimer: Some of the companies listed above may be DeciBio Consulting clients or customers.---

Author: Andrew Aijian, Project Leader at DeciBio Consulting, LLC, aijian@decibio.comCo-Author: Madison Jones, Analyst at DeciBio Consulting, LLC, jones@decibio.comConnect with Andrew and Madison on LinkedInhttps://www.linkedin.com/pub/andrew-aijianhttps://www.linkedin.com/in/madison-jones-563bb5b5