Tissue and Peripheral T Cell Receptor Repertoire Predicts Immunotherapy Response and Progression-Free Survival in NSCLC Patients

Sci Rep 2025 AI 7 Explanations View Original
Original Paper (PDF)

Unable to display PDF. Download it here or view on PMC.

Plain-English Explanations
Pages 1-2
Using the Immune System's Fingerprint to Predict Immunotherapy Success

The challenge: Immunotherapy with pembrolizumab has transformed treatment for advanced non-small cell lung cancer (NSCLC), yet only 20-40% of patients benefit. Identifying who will respond before starting treatment is critically important but remains elusive.

The approach: This study examined T cell receptor (TCR) repertoire patterns in both tumor tissue and blood samples collected before treatment from 12 advanced NSCLC patients. TCRs are unique molecular fingerprints on immune T cells, and their distribution may reveal whether a patient's immune system is primed to fight cancer.

Why it matters: Current biomarkers like PD-L1 expression are unreliable in predicting response. The TCR repertoire offers a window into the functional state of anti-tumor immunity, potentially guiding treatment decisions and sparing non-responders from unnecessary side effects and costs.

TL;DR: Researchers analyzed immune cell receptor patterns in tumor tissue and blood to see if they could predict which NSCLC patients would benefit from pembrolizumab before treatment begins.
Pages 2-3
Next-Generation Sequencing of the TCR Repertoire from Tissue and Blood

Patient cohort: Twelve patients with advanced NSCLC (stages IIIA-IVB) treated with pembrolizumab at first-line were enrolled. Patients were classified as responders (complete/partial response or stable disease) or non-responders (tumor progression) based on CT scans at 3 months.

Sequencing approach: Formalin-fixed paraffin-embedded (FFPE) tumor tissue and peripheral blood mononuclear cells (PBMCs) were collected before treatment. Using the Oncomine TCR Pan-Clonality Assay, researchers sequenced the TCR beta (TCRbeta) and gamma (TCRgamma) chain CDR3 regions with next-generation sequencing on 22 total samples.

Key metrics analyzed: The analysis measured TCR richness (number of unique clones), convergence (shared sequences), diversity (Shannon diversity index), and evenness (how uniformly clones are distributed). Clonal space was segmented into top 1%, 3%, and 5% categories to assess dominant clone behavior.

TL;DR: TCR beta and gamma chains were sequenced from tumor tissue and blood collected before pembrolizumab treatment, measuring clone diversity and evenness across 22 samples from 12 patients.
Pages 4-5
Low Tumor-Infiltrating TCR Evenness Predicts Response and Longer Survival

Key finding: Responders showed significantly lower tumor-infiltrating TCRbeta evenness than non-responders (p=0.044). Evenness close to zero means a few T cell clones dominate, while evenness near 1 means all clones are equally distributed.

Predictive power: Tumor-infiltrating TCRbeta evenness predicted immunotherapy response with an AUC of 0.86. An evenness threshold below 0.795 achieved 60% sensitivity and 100% specificity for identifying responders.

Survival link: Patients with low tumor-infiltrating TCRbeta evenness (below 0.8441) had significantly longer progression-free survival (PFS) than high-evenness patients (p=0.013). This suggests that pre-existing clonal expansion of specific anti-tumor T cells enables better response to checkpoint blockade.

TL;DR: Responders had more lopsided T cell receptor distributions in tumors (lower evenness), meaning dominant T cell clones were already targeting the cancer before pembrolizumab was started.
Pages 5-7
Specific TRBV and TRBJ Gene Frequencies Also Predict Immunotherapy Outcome

Tissue-based gene markers: Tumor-infiltrating TCRbeta genes TRBV6.5, TRBV11.3, and TRBJ2.1 were significantly lower in responders than non-responders. Notably, low TRBV11.3 frequency predicted longer PFS (p=0.002), and TRBJ2.1 predicted response with AUC=0.94 (100% sensitivity, 85.7% specificity).

Blood-based gene markers: Circulating TRBV5.3 was significantly higher in responders, while TRBV27, TRBV28, TRBJ2.1, and TRBJ2.6 were lower. Circulating TRBJ2.6 was particularly powerful, predicting response with a perfect AUC of 1.00 at a frequency below 0.0055.

Interpretation: These specific gene usage patterns may reflect T cells with particular antigen-binding preferences that are relevant to tumor recognition, though the exact neoantigens involved remain unknown. The pattern differs from findings in other cancers, suggesting tumor-type specificity.

TL;DR: Certain T cell receptor gene segments in both tumor tissue and blood were significantly different between responders and non-responders, with circulating TRBJ2.6 achieving perfect discrimination.
Pages 5, 7
Random Forest Signatures Combine Multiple TCR Variables for Superior Prediction

Tissue signature: A Random Forest model combining tumor-infiltrating evenness, TRBV6.5, TRBV11.3, and TRBJ2.1 frequencies achieved AUC=0.83. A positive result (probability above 0.230) predicted response with 100% sensitivity and 71.4% specificity, ensuring no potential responders are missed.

Blood signature: The circulating model combining TRBV5.3, TRBV27, TRBV28, TRBJ2.1, and TRBJ2.6 achieved AUC=0.92, with 75% sensitivity and 100% specificity. This blood-based model would not expose any non-responder to unnecessary treatment.

Clinical tradeoff: The tissue model is more appropriate if the goal is to treat all potential responders (maximizing benefit), while the blood model is preferable for avoiding unnecessary treatment in non-responders. Both have clinical utility, and blood-based testing offers an advantage for tumors too small for adequate biopsy.

TL;DR: Machine learning models combining multiple TCR features achieved AUC of 0.83 (tissue) and 0.92 (blood), offering clinically meaningful tools for treatment selection.
Page 7
TCR Gamma Chain Shows Similar Trends but Does Not Reach Statistical Significance

Gamma chain findings: Responders also tended to have lower tumor-infiltrating and circulating TCRgamma evenness, mirroring the TCRbeta findings. However, none of these differences reached statistical significance (p=0.149 tissue, p=0.114 blood).

Biological context: TCRgamma-delta T cells represent only 5% of circulating T cells, but play important roles in both systemic and local immune surveillance of tumors. The limited diversity of gamma-delta TCR rearrangements may reduce sensitivity for detecting repertoire differences.

Knowledge gap: Very little is known about the TCRgamma repertoire as an immunotherapy biomarker in solid tumors. This study is among the first to examine it in first-line pembrolizumab-treated NSCLC, opening a new research avenue that requires larger cohort validation.

TL;DR: Gamma chain T cell receptor patterns showed the same directional trends as beta chain findings but were not statistically significant, pointing to future research directions.
Pages 5, 7, 8
Promising Pilot Findings Require Validation in Larger Diverse Cohorts

Primary limitation: The study enrolled only 12 patients, making it an exploratory investigation. The small sample size limits statistical power, increases the risk of overfitting in the Random Forest models, and prevents definitive conclusions about clinical applicability.

Retrospective design: As a retrospective observational study, confounding factors such as age differences between responders (older) and non-responders (younger) could influence TCR repertoire patterns. Age was associated with lower TCRbeta/gamma evenness in tissue, complicating causal interpretation.

Future directions: Prospective multicenter trials enrolling larger, more diverse patient populations are needed to validate these TCR biomarkers. Future work should also explore whether TCR dynamics during treatment provide additional predictive value, and whether these signatures apply across NSCLC subtypes and treatment combinations.

TL;DR: Small cohort size is the primary limitation; validation in larger prospective studies is essential before TCR repertoire analysis can enter routine clinical practice.
Citation: Open Access, 2025. Available at: PMC12635383.