Immune-related RNA-seq biomarker-based clustering reveals heterogeneous immunotherapy responses and guides subtype-specific strategies in metastatic NSCLC

Metastatic non-small cell lung cancer (mNSCLC) represents a highly heterogeneous disease with variable clinical outcomes under first-line immunotherapy plus chemotherapy. To better understand immune landscape features associated with heterogeneous response to immunotherapy, we performed biomarker-driven RNA-seq molecular clustering using known immune-related markers TIGIT, FOXP3, CD274 ( PD-L1 ), and tumor-associated macrophage (TAM) score. We analyzed a real-world cohort of 2,235 mNSCLC patients with pre-treatment tumor biopsies in the de-identified Tempus database treated with first-line PD-(L)1 plus chemotherapy. Unsupervised clustering of RNA-seq data defined four distinct immune subtypes. Real-world overall survival (rwOS) and progression-free survival (rwPFS) were assessed via Kaplan-Meier analysis with a log-rank test. Pathway enrichment using hallmark gene sets, tumor mutational burden (TMB), and immune cell composition using QuantiSeq were analyzed. Expression levels of RNA-seq biomarkers and TAM score were significantly different across identified clusters (ANOVA; p <0.001). These clusters also showed significantly differential prevalence of TMB-high and PD-L1-positive (IHC) (Chi-squared; p < 0.001, respectively), as well as characteristic pathway enrichment and immune profiles. Non-squamous/Never smoker were more frequent in Cluster 2, whereas Squamous/Current smoker were predominant in Cluster 1 (Chi-squared; Histology/Smoking, p <0.05, respectively). Survival differed significantly, being poorest in Cluster 1 and best in Cluster 3 (rwOS/rwPFS, p <0.001) (Table 1). This biomarker-driven RNA-seq analysis identified four immune clusters of mNSCLC with differential survival outcomes. This study provides a foundation for understanding tumor heterogeneity and supports the use of immune biomarkers to enable patient stratification for therapeutic combinations. Table 1. Cluster 1 (Immune-desert) N=713 Cluster 2 (TAM-enriched) N=402 Cluster 3 (Immune-hot) N=813 Cluster 4 (Myeloid-inflamed, PD-L1-high) N=302 p -value Median Survival Time (Months) rwOS/rwPFS 11.5/5.95 mo 14.8/7.33 mo 18.1/8.15 mo 16.7/6.84 mo Log-rank test; p <0.001 RNA-Seq Biomarkers (TIGIT, FOXP3, CD274 (PD-L1)) and TAM Score Uniformly low expression of all markers High TAM but low TIGIT/FOXP3/PD-L1 High TIGIT/FOXP3/PD-L1 with elevated TAM High PD-L1 with low TIGIT/FOXP3 ANOVA test; p <0.001 Pathway Enrichment TME: tumor microenvironment ↑ Oncogenic signalings and proliferation ↓ Immune -related pathways ↑ TME remodeling pathways ↑ Immune/inflammatory signaling (e.g., IFNgamma) and TME remodeling pathways ↑ Proliferation and DNA-repair pathways Welch ANOVA + Games-Howell or Kruskal-Wallis and Dunn (BH) test; adjusted p <0.05 Immune Cell Composition ↓ Lymphoid and myeloid cell infiltration ↑ M1/M2 macrophage Broad infiltration (↑ CD8, CD4, Treg, B, NK) ↑ Myeloid cell infiltration Kruskal-Wallis and Dunn (BH) test; adjusted p <0.05 TMB-High (TMB ≥10 mut/Mb) 283 (34%) 92 (20%) 254 (26%) 124 (35%) Chi-squared test; p <0.001 PD-L1-Positive (IHC; TPS ≥ 1%) 203 (36%) 174 (54%) 421 (68%) 213 (94%) Chi-squared test; p <0.001 Tumor Histology Chi-squared test; p <0.001 Squamous 232 (33%) 73 (18%) 221 (27%) 84 (28%) Non-Squamous 444 (62%) 315 (78%) 555 (68%) 201 (67%) NOS 37 (5.2%) 14 (3.5%) 37 (4.6%) 17 (5.6%) Smoking Status Chi-squared test; p <0.05 Current Smoker 118 (62%) 44 (46%) 113 (55%) 48 (59%) Never Smoker 16 (8.4%) 23 (24%) 32 (15%) 12 (15%) Ex-Smoker 56 (29%) 29 (30%) 62 (30%) 21 (26%) Unknown 523 306 606 221