Immune composition and adenosine-associated transcription in non-small-cell lung cancer after neoadjuvant PD-1 blockade.
Abstract
Background
In non-small-cell lung cancer (NSCLC), high metabolic gene expression during immunotherapy can reflect either more infiltrating cells or altered cellular activity. We tested whether separating cell abundance from cell-intrinsic transcription clarifies incomplete tumour clearance after neoadjuvant programmed cell death-1 (PD-1) blockade.
Methods
Using cross-dataset triangulation, we analysed 1173,843 post-treatment immune cells from 224 patients (103 non-major pathological response [non-MPR], 40 major response, and 81 complete response) via patient-level regressions adjusted for clinical covariates, cell proportions, and library size. Findings were compared across recurrence follow-up (n = 154), spatial profiling (n = 40), three bulk cohorts (n = 78), population data (n = 3045), and 7224 genetic tests.
Results
Non-MPR tumours had higher regulatory T-cell (Treg) proportions (+3.91% FOXP3+, +3.07% CCR8+) and elevated lymphoid glycolysis, exhaustion, Treg, and adenosine scores (all false-discovery rate q ≤ 1.09 × 10-7). Notably, adjusting for cell composition and depth abolished broad glycolysis and exhaustion differences, whereas adenosine-associated transcription persisted (difference 0.044, 95% confidence interval [CI] 0.020-0.068; q = 0.0046), peaking in terminally exhausted CD8+ T cells (0.449, 95% CI 0.328-0.571; q = 1.29 × 10-10). Paired spatial biopsies localized HK2, LDHB, and SLC16A1 to tumour regions (q ≤ 0.0057) and CTLA4 to stroma (q = 5.14 × 10-5) without response interactions. A six-component score discriminated non-MPR (apparent area under the curve [AUC] 0.791, 95% CI 0.724-0.851) and recurrence (hazard ratio 1.41, 95% CI 1.04-1.93), attenuating after pathology adjustment, but failed in pre-treatment bulk cohorts (AUCs 0.091-0.517); no genetic test passed correction (q ≥ 0.9934).
Conclusions
Incomplete response involves Treg accumulation and cell-intrinsic adenosine transcription, offering exploratory, hypothesis-generating insights for prospective validation.