Spatial transcriptomic profiling to identify stroma-based prognostic groups in oncogene-addicted lung cancer.

Abstract

8611 Background: Driver genomic aberrations determine prognosis and therapy in non-small cell lung cancer (NSCLC). In this context, some other genomic features such as TP53 mutations confer poor outcomes. However, other mechanisms leading to worse prognosis in patients (pts) sharing the same genomic profile remain incompletely characterized. Methods: NSCLC pts from three cohorts based on their molecular profile ( EGFR -mutated [EGFRm], KRAS -mutated [KRASm] and ALK/ROS1/RET fusions) were included. Baseline formalin-fixed paraffin-embedded (FFPE) tumor samples were analyzed using NanoString GeoMx Digital Spatial Profiling. Regions of interest (ROIs) were selected based on histopathological features and fluorescently labeled antibodies for tumor (Tm) and stromal (St) areas. RNA expression of >1,800 genes from selected ROIs was assessed using GeoMx Cancer Transcriptome Atlas. Leiden algorithm was used for clustering, and differential gene expression and other statistical analyses were performed using R software. Results: A total of 189 pts (100 EGFRm, 56 KRASm, 33 fusions [17 ALK , 7 ROS1 and 9 RET rearrangements]) were identified. Separate analysis of Tm and St compartments revealed an association between genotype and transcriptome in the first, whereas the St transcriptome failed to correlate with the molecular profile. Unsupervised clustering of the Tm compartment identified four subgroups with distinct gene expression. Tm Cluster 4 (30 pts) was mainly composed of KRASm pts (86.7%) and presented the poorest prognosis, with a median overall survival (mOS) of 7.5 months (mo) ( p <0.0001). Within St compartment, four prognostic clusters were also identified. St Cluster 1 (76 pts: 43.4% EGFRm, 40.8% KRASm, 15.8% fusions) and St Cluster 4 (23 pts: 21.7% EGFRm, 52.2% KRASm, 26.1% fusions) included pts from all three cohorts and showed significantly decreased OS (17.4 mo and 15.6 mo, respectively; p <0.0001). Genes involved in matrix remodeling and epithelial-mesenchymal transition signaling were overexpressed in St Cluster 1, while St Cluster 4 exhibited a highly immunogenic profile; no correlation with TP53 mutation status was observed. Conclusions: Spatially resolved transcriptomic profiling suggests that stromal composition could identify oncogene-addicted NSCLC pts with poor prognosis regardless of molecular subtype, potentially guiding the development of novel therapeutic strategies. Further validation studies are warranted.