Immune checkpoint inhibitors (ICIs) have revolutionized cancer treatment since the FDA approval of ipilimumab in 2011, with seven ICIs now approved across approximately 15 tumor types. However, these therapies demonstrate efficacy in only 20-30% of cancer patients, creating an urgent need for accurate predictive biomarkers to optimize patient selection and treatment outcomes.
Current FDA-Approved Biomarkers Face Significant Challenges
The three FDA-approved predictive biomarkers—PD-L1, microsatellite instability/deficient mismatch repair (MSI/dMMR), and tumor mutational burden (TMB)—each present unique advantages and limitations that complicate clinical implementation.
PD-L1: Widespread Use Despite Limited Accuracy
PD-L1 expression, the most widely used biomarker, suffers from poor diagnostic accuracy due to multiple testing variables. Four different FDA-approved immunohistochemistry methods utilize distinct antibodies, scoring systems, expression thresholds, and cell types, creating confusion in clinical practice. Analysis of 45 PD-L1 FDA approvals from 2011 to 2019 revealed predictive value in only 28.9% of cases.
The biomarker's limitations extend beyond technical variability. Up to 20% of patients with PD-L1-negative tumors benefit from PD-1/PD-L1 antibodies, while PD-L1 expression can change temporally and spatially, further reducing predictive reliability.
MSI/dMMR: High Efficacy but Limited Prevalence
MSI-H/dMMR status, approved for pembrolizumab treatment across all solid tumors in 2017, demonstrates impressive clinical outcomes with a 39.6% objective response rate and 78% of responders maintaining responses for six months or longer. The biomarker's rationale is compelling—dMMR tumors accumulate thousands of mutations, generating more neoantigens and enhanced immune infiltration.
However, MSI prevalence varies dramatically across cancer types. While endometrial (~30%), gastric (~20%), and colorectal (~15%) cancers show substantial rates, major cancer types like non-small-cell lung cancer, breast cancer, and prostate cancer exhibit only 1-2% prevalence, limiting broad clinical utility.
TMB: Technical Complexity Hinders Implementation
TMB, measuring mutations per megabase DNA, received FDA approval for pembrolizumab in 2020 based on the KEYNOTE-158 study. High TMB correlates with increased neoantigen production and enhanced immune responses across multiple tumor types.
Despite its biological rationale, TMB faces significant technical challenges. Different calculation methods, reporting formats, and cutoff values complicate clinical practice. Whole-exome sequencing limitations include high cost, long turnaround times, and limited availability, while targeted NGS panels require standardization of gene content, sequencing depth, and bioinformatics pipelines.
Gene Signature Biomarkers Show Superior Predictive Performance
Recognizing the limitations of single biomarkers, researchers have developed gene expression-based signatures that capture multiple aspects of immune response biology.
T Cell-Inflamed Gene Expression Profile
The 18-gene T cell-inflamed signature, validated across nine tumor types in 220 patients, represents genes related to IFN-γ signaling, cytotoxic effector molecules, antigen presentation, and T cell cytokines. ROC analysis demonstrated superior predictive value compared to PD-L1 alone, with responders showing high signature gene expression while non-responders exhibited low expression levels.
TIDE: Capturing Immune Dysfunction
The Tumor Immune Dysfunction and Exclusion (TIDE) signature focuses on unfavorable tumor environments, measuring T cell dysfunction and exclusion mechanisms. Unlike signatures predicting response, high TIDE scores indicate non-response. Validation studies showed TIDE outperformed both TMB and PD-L1 for anti-PD1 and anti-CTLA4 therapies.
Melanocytic Plasticity Signature
The 45-gene melanocytic plasticity signature (MPS), developed through mouse melanoma models, reflects tumor cell multipotency and differentiation status. Low MPS scores, indicating later-stage melanocytic differentiation, correlated with ICI response. In comparative analyses, MPS demonstrated the highest ROC area under the curve values, surpassing TIDE, TMB, and PD-L1.
B Cell-Focused Signatures
Recent discoveries highlight B cells within tertiary lymphoid structures (TLS) as critical predictors of ICI response. B cell-rich immune populations in TLS, particularly switched memory B cells, correlate with improved outcomes across multiple cancer types. TLS signatures dominated by B cell-specific genes predict both ICI response and overall survival independent of TMB.
Combinational Approaches Enhance Predictive Power
The limited overlap between biomarker-predicted responders suggests these markers capture different aspects of tumor immunobiology, supporting combinational approaches for improved prediction.
GEP+TMB Integration
Combining T cell-inflamed gene expression profiles with TMB across 22 tumor types in KEYNOTE clinical trials revealed complementary predictive value. Patients with both high GEP and high TMB showed the highest response rates, while those with low levels of both markers demonstrated no response. This joint approach offered higher sensitivity and greater predictive value than either biomarker alone.
MPS+TIDE Combination
The combination of melanocytic plasticity and immune dysfunction signatures showed improved predictive performance in melanoma patients. Patients with low MPS and low TIDE scores exhibited significantly longer progression-free and overall survival, while those with high scores in both signatures showed the poorest outcomes.
Future Direction: Integrated Nucleic Acid Signatures
The complexity of ICI response mechanisms necessitates comprehensive biomarker approaches. An optimal predictive biomarker would integrate multiple DNA and RNA markers in a single assay, capturing tumor microenvironment factors, neoantigenicity, and intrinsic tumor characteristics.
This integrated signature should include four key categories: TME-related RNA biomarkers (T cell-inflamed, dysfunction/exclusion, and B cell signatures), tumor multipotency genes, neoantigenicity-related DNA markers (TMB, mismatch repair genes, MSI panels), and high-impact genes like TGFB1, SOX10, and POLE/POLD1.
Such comprehensive panels would overcome conflicting results from individual biomarkers while maintaining clinical feasibility through next-generation sequencing platforms. By analyzing multiple contributing factors to ICI response, integrated assays would likely demonstrate enhanced predictive value, particularly for immunologically cold tumors where current biomarkers show limited utility.
The evolution from single biomarkers to integrated signatures represents a critical advancement in precision immunotherapy, promising more accurate patient selection and improved treatment outcomes across diverse cancer types.
