IN WHAT APPEARS TO BE the largest blood-based biomarker study of a checkpoint inhibitor, an RNA transcript–based gene classifier was able to predict for melanoma patients’ response to tremelimumab. The study was recently published in the Journal for ImmunoTherapy of Cancer.1
Philip Friedlander, MD
“Our study, in many ways, offers proof-of-concept that we can take a blood sample before treatment with immunotherapy—in this case, a checkpoint inhibitor—and predict for efficacy,” lead investigator Philip Friedlander, MD, of the Tisch Cancer Institute at Mount Sinai, New York, told The ASCO Post.
Tremelimumab is an anti–cytotoxic T-lymphocyte–associated protein 4 (CTLA-4) antibody that is being evaluated in different tumor types. According to Dr. Friedlander, although the test’s predictive ability was shown solely for tremelimumab, this approach is likely to apply (though perhaps with a different gene signature) to other checkpoint inhibitors as well, including ipilimumab (Yervoy) and antibodies targeting the programmed cell death protein 1 (PD-1) or its ligand (PD-L1).
“The goal is to be able to identify biomarkers of response in pretreatment blood samples, which are much more easily obtained than tissue biopsy,” he said. “This will optimize treatment planning and efficacy and will help us avoid ineffective treatments and therefore toxicity and unnecessary costs as well.”
Predictive Ability of 15-Gene Classifier
DR. FRIEDLANDER AND HIS TEAM collected blood samples from 210 treatment-naive melanoma patients receiving tremelimumab in a multicenter phase III study, which formed the discovery data set. Another 150 treatment-refractory patients receiving tremelimumab in a phase II study formed the validation data set.
“Our study offers proof-of-concept that we can take a blood sample before treatment with immunotherapy—in this case, a checkpoint inhibitor—and predict for efficacy.”— Philip Friedlander, MD
Tweet this quote
Gene expression for 169 mRNA transcripts was measured using quantitative polymerase chain reaction. Among the 169 genes examined were those associated with inflammation, immunity, and the CTLA-4 pathway, as well as oncogenes and genes found to discriminate melanoma from normal tissue in previous exploratory studies.
They ultimately identified the optimal model to be a 15-gene signature that consists of 9 predictors and 6 nonpredictive enhancer variables, containing ADAM17, CDK2, CDKN2A, DPP4, ERBB2, HLA-DRA, ICOS, ITGA4, LARGE, MYC, NAB2, NRAS, RHOC, TGFB1, and TIMP1. Specific patterns of expression of these 15 genes appear to be needed for a robust anticancer response, the authors believe. “Expression of these genes may identify patients whose immune systems are already primed to have an anticancer immune response,” they wrote in the article.
They applied the model to both data sets to examine its ability to predict outcomes prior to treatment. The 15-gene pretreatment classifier model achieved an AUC of 0.86 (P < .0001) for objective response and 0.6 (P = .0066) for 1-year survival in the discovery set. In the validation data set, AUCs of 0.62 (P = .0455) and 0.68 (P = .0002), respectively, were observed.
Next Steps
DR. FRIEDLANDER COMMENTED on the need to find a biomarker for efficacy to checkpoint inhibitors, many of which are in development. “There are many components to the immune system, and there are many drugs in clinical development that can target these components. The problem is that in different patients, tumors will grow by evading the immune system in different ways, and subpopulations will respond to different agents,” he explained. “The risk in clinical trials is that we don’t know ahead of treatment exactly which immune abnormalities exist in a given patient that can effectively be targeted. Although specific immune modulatory drugs may work in a subset of patients, the trials may not be powered to detect efficacy in small subsets, and therefore overall sutdy results could be negative.”
A relatively simple blood-based biomarker in the pretreatment setting could help decipher the defects in a given patient’s immune system, and therefore it may be informative for drug selection, both in clinical trials and ultimately in clinical practice, he predicted.
A future step will be to apply the mRNA gene–classifier approach to anti–PD-1/PD-L1 antibodies and to checkpoint inhibitors, both in melanoma and in other malignancies. Although the genes may be different from those comprising the 15-gene signature for tremelimumab, the approach would be similar. A pretreatment blood-based biomarker might also predict for toxicity, which also would help guide treatment selection, he added. ■
DISCLOSURE: For full disclosures of the study authors, visit https://jitc.biomedcentral.com/.
REFERENCE
1. Friedlander P, Wassmann K, Christenfeld AM, et al: Whole-blood RNA transcript-based models can predict clinical response in two large independent clinical studies of patients with advanced melanoma treated with checkpoint inhibitor, tremelimumab. J Immunother Cancer 5:67, 2017.
