Since the debut of immune checkpoint inhibitors (ICIs) in 2011, the percentage of cancer patients responding to these treatments has risen from less than 1% to more than 10%. Identifying responders early would greatly benefit even more patients and improve the allocation of healthcare resources. Hence the search for predictive biomarkers.
As of this writing, the U.S. Food and Drug Administration (FDA) has approved nine ICIs, all monoclonal antibodies. One ICI drug targets the CTLA-4 pathway, one disrupts LAG-3, and seven work through the PD-L/PD-L1 checkpoint mechanism.
ICI drugs are much-hyped because when they work they really work, occasionally even “curing” metastatic cancers. The problem is that only about 10% of patients respond meaningfully, and median survival benefits are on the order of four to five months. Most cancer patients, including those with tumors considered to be sensitive to ICI agents (e.g., melanoma, non-small cell lung cancer and a few others), either do not respond or become treatment-resistant. Additionally, toxicity, primarily in the form of autoimmunity, can be significant and even fatal. Financial issues must also be considered, as the cost of these treatments is substantial (approximately $10,000/dose in the U.S., with multiple doses required).
The U.S. FDA has approved two predictive companion biomarkers for the monoclonal antibody checkpoint inhibitor pembrolizumab (KEYTRUDA®). PD-L1 (programmed death ligand 1), a 40kDa transmembrane protein, suppresses the adaptive immune system during pregnancy or as a consequence of tissue transplantation, autoimmune diseases, hepatitis, and other conditions. The other biomarker, mismatch repair deficiency (MMR), refers to the inability to repair DNA mismatches that develop during normal DNA replication, or result from exposure to chemicals or even anticancer drugs. A third, emerging biomarker, which will likely also be approved as a companion diagnostic, is tumor mutation burden. Bringing up the rear in this constantly evolving field is alteration or deficiency in DNA damage response (DDR) pathways, which is involved in a broad range of tumors.
No biopsy required
Biomarkers can only make a difference when samples are accessible and provide enough material to analyze. Since blood usually qualifies on both points, assays based on peripheral blood collection are highly desirable.
Circulating tumor-derived DNA (ctDNA), a popular quantity in predictive oncology, is now entering the mainstream of clinical trial design, translational research, and predictive care. ctDNA methods were developed in response to the poor predictive ability of standard-of-care imaging (e.g., PET scanning and MRI) and the cost and difficulty of obtaining biopsies. Yet, while ctDNA methods have advanced rapidly, they still generally suffer from limited sensitivity and utility due to their minimal detectable tumor burden requirements. Predictability falls off rapidly in the case of early-stage cancer or undetectable metastases.
Personalis, which specializes in cancer genomics, is interested in ctDNA on several fronts. The company’s recent patent, issued in April, 2022, “Methods and Systems for Genetic Analysis,” has a generic-sounding title but is narrowly focused on ctDNA quantitation of MRD. Personalis describes the method as “a partially personalized assay, combining both tumor-informed and prespecified (tumor naïve) content.”
Tumor-informed methods are highly sensitive for detecting MRD, and also support tracking of individual-specific genetic variants. Pre-specified, tumor-naïve content includes variants not initially detected and which may emerge to confer drug resistance.
The patent award comes fast on the heels of publication of a study, "A machine learning algorithm with subclonal sensitivity reveals widespread pan-cancer human leukocyte antigen loss of heterozygosity" (HLA LOH) in Nature Communications. HLA LOH allows cancer cells to escape immune recognition by deleting HLA alleles, which causes suppression of tumor neoantigens. Few methods exist for detecting and characterizing HLA LOH reliably.
The study describes development of “deletion of allele-specific HLAs” (DASH), a machine learning algorithm for detecting HLA LOH from paired tumor-normal sequencing data and demonstrates increased sensitivity of HLA LOH detection compared with existing methods, and most importantly suggests a pathway to clinical utility for DASH.
Dave Delano, Ph.D., Director of Product Management for Biomarker Discovery Products at Personalis, explains that ImmunoID NeXT®, combines exome-scale DNA and RNA sequencing with advanced analytics to provide a multidimensional view of both the tumor and the tumor microenvironment (TME).
“The platform was designed to overcome the limitations of small panels, which become obsolete when new genetic biomarkers or therapeutic targets are identified. Comprehensive coverage of all genes, DNA and RNA, tumor and normal tissue, and immune biology serves to future-proof the data and provide visibility to emerging investigational biomarkers as they emerge. It also creates a path to development of companion diagnostics in collaboration with pharmaceutical companies. The goal was to ensure that potential, high-value information wouldn’t be missed and to make the most of what is often a very limited amount of available sample.”
Two “birds,” one assay
Multi-omics strategies are another way to get the most out of limited samples, which is the goal of a comprehensive, multi-omics, blood-based assay from Freenome.
A poster presented at ASCO recently described a study that revealed signatures of ICI treatment responses in patients with melanoma, kidney cancer, and non-small cell lung cancer, identifying 13 transcription factors and 269 genes potentially implicated in treatment resistance and a possible epithelial mesenchymal transition signature in responders. Subsequent analysis on a subset of lung cancer patients also identified treatment response biomarkers. The Freenome platform combines signatures from both tumor- and non-tumor-derived sources.
Many of these biomarkers were already known, but as noted by Francesco Vallania, Ph.D., Manager and Staff Computational Biologist at Freenome, “The goal of this study was not necessarily to identify unknown genes or transcription factors, but rather to determine which ones were consistently associated with ICI resistance across multiple cancers and could be measured non-invasively in plasma.”
One huge issue with biomarker studies is they may reveal more than medical science is able to deal with, specifically that drugs may not exist to address the underlying mechanisms the biomarkers signify.
Vallania dismisses these concerns. “The biomarkers we identified are based on ICI therapies being used today, so it will not require development of new therapies. However, as this is a retrospective correlative study, validation studies would be needed to confirm these results before being used to select and stratify patients for ICI. While some of these indicators have been implicated in the underlying biology of immunotherapy resistance, for example the JAK-STAT pathway, they are not necessarily monitored during treatment.”
Additionally, Freenome uncovered additional biomarkers that have not traditionally been associated with ICI treatment.
“Non-tumor-derived signals, such as those that arise from the immune system, often result from the interplay between the tumor and the response of the immune system. While both sources of signals are considered equally important, ICI resistance is generally dependent on the immunological state of the patient, which is weighted toward the non-tumor derived biomarkers.”
The $64,000 question
Biomarkers alone have limited value without treatments targeting the specific etiologic or prognostic factors those biomarkers represent: Nobody ever got sick and died from a biomarker. Will more (and better) biomarkers lead to more effective utilization of ICI treatments, or do they represent another biotech “science project” that deflects from more fruitful pursuits?
“Many of our partners are dedicated to answering this question,” Delano tells Biocompare. “Our approach is to enable study designs that thoroughly leverage established biomarkers while accelerating the discovery and application of novel biomarkers that help redefine standard practice in drug development and patient care.”