Artificial intelligence-powered spatial analysis of tumor microenvironment in non-small cell lung cancer patients who acquired resistance after EGFR tyrosine kinase inhibitors.

Abstract

8536 Background: We evaluated dynamic changes in the tumor microenvironment (TME) after EGFR tyrosine kinase inhibitor (TKI) treatment using an artificial intelligence (AI)-powered spatial TME analyzer and assessed the predictive efficacy of immune checkpoint inhibitors (ICIs) as monotherapy or in combination therapy. Methods: An AI-powered whole-slide image (WSI) analyzer (Lunit SCOPE IO, Lunit, Seoul, Korea) segmented cancer area (CA), stromal area (CS), and identified tumor-infiltrating lymphocytes (TILs), tertiary lymphoid structures (TLS), fibroblasts (Fibs), and endothelial cells (ECs) in tumor tissue. We analyzed 143 non-small cell lung cancer (NSCLC) samples post-resistance to EGFR TKIs from two cohorts: 1) patients (pts) treated with ICIs at Samsung Medical Center, Korea (October 2015–July 2022), and 2) pts from the ATTLAS phase 3 trial comparing atezolizumab plus bevacizumab, paclitaxel, and carboplatin (ABCP) versus pemetrexed plus carboplatin (PC). Among these, 89 pts received ICI monotherapy, and 54 were from the ATTLAS trial (ABCP: 36, PC: 18). Paired pre-treatment samples were available for 89 pts (62.8%), and whole transcriptome sequencing was performed on 42 samples. Results: In the combined pre- and post-TKI samples, TLS area per CA correlated with the TLS signature (ρ=0.439, P=0.003), Fibs with the cancer-associated fibroblast signature (ρ=0.581, P<0.001), TILs with the interferon-gamma signature (ρ=0.498, P<0.001), and ECs with the angiogenesis signature (ρ=0.315, P=0.042), but not VEGF signatures (ρ=0.183, P=0.71). Post-TKI samples showed reduced TILs in CA (P=0.045) and increased ECs in CA (P=0.005), with no significant changes in Fibs (P=0.819) or TLS area (P=0.884). Changes differed by EGFR mutation subtype: L858R mutations were linked to increased ECs (P=0.009), while T790M mutations and exon 19 deletions (19del) were linked to reduced TILs (P=0.033, P=0.045). Higher TILs in CA were associated with better overall response rate (ORR, 41.7% vs. 9.7%, P=0.003) and progression-free survival (PFS, 4.9 vs. 1.8 months, HR=0.41 [95% CI: 0.21–0.79]). Similarly, higher EC levels in CA correlated with improved ORR (19.3% vs. 3.7%, P<0.01) and PFS (2.0 vs. 1.4 months, HR=0.44 [95% CI: 0.28–0.71]). In the ATTLAS cohort, these factors were associated with clinical benefits from ABCP, with a significant association for TILs (HR=0.42 [95% CI: 0.19–0.91, P=0.027]) and a marginal association for ECs (HR=0.29 [95% CI: 0.07–1.15, P=0.067]). Conclusions: EGFR-TKI alters the immune landscape of NSCLC. Higher TILs or ECs in CA were significantly associated with favorable outcomes to ICI or combination treatment.