Genomic characterization of baseline and post-progression tumors in IMmotion010, a randomized, phase 3 study of adjuvant (adj) atezolizumab (atezo) vs placebo (pbo) in patients (pts) with high-risk localized renal cell carcinoma (RCC).

4510 Background: Although adj atezo did not prolong disease-free survival (DFS) in the IMmotion010 trial, accompanying studies identified post-nephrectomy serum KIM-1 levels to be potentially predictive of benefit with atezo. We investigated whether tissue genomics could complement these findings. Methods: Tumors were obtained from patients pre-treatment and (if additional consent obtained) at disease recurrence. Whole transcriptome profiles were generated using TruSeq RNA Access Technology (Illumina). Previously, non-negative matrix factorization (NMF) was performed in a separate phase 3 study in advanced RCC (IMmotion151), establishing 7 molecular subgroups (NMF 1-7) (Motzer et al Cancer Cell 2020). For each IMmotion010 tumor, we derived signature scores dichotomized at the median as well as NMF1-7 subtype using random forest. Clinical outcome was assessed within these groups alone and with the addition of baseline serum KIM-1 levels, dichotomized into KIM-1 high (KIM-1 H ) and KIM-1 low (KIM-1 L ) using the previously established cutoff of 86 pg/mL. Where possible, we sought to characterize the evolution of molecular profiles at disease recurrence. Results: Baseline tissue was obtained from 754 pts, reflecting 97% of the intention-to-treat population. Tumors from KIM-1 H patients were enriched in myeloid, granulocyte and proliferation gene signatures at baseline. Among 722 pts for whom both serum KIM-1 and NMF subtype could be characterized, pts in cluster 6 (stromal/proliferative) appeared to derive benefit from atezo (n=50) (Table). Across all patients, no difference in outcome was observed among baseline Teff H and Teff L subsets. Within the KIM-1 H population, Teff H tumors were associated with longer DFS with atezo vs pbo. Paired baseline/recurrence tissue was obtained from 80 pts (atezo: 49; pbo: 31). At recurrence, tumors exhibited increased stromal and proliferation gene signatures, regardless of treatment, reflected in an increased proportion in NMF6 (baseline: 6%; progression: 22%). Exploratory analyses also revealed a decreased MHC-I signature after treatment with atezo. Conclusions: This is the first report of tissue genomic profiling in a phase 3 adjuvant immune checkpoint inhibitor study in RCC. While certain molecularly defined subsets may carry predictive value, serum KIM-1 remains the most robust predictor of outcome with atezo. Analyses of progression biopsies highlight an evolution in genomic profile and offer insights into mechanisms of relapse. Clinical trial information: NCT03024996 . Subgroup n DFS HR Serum KIM-1 KIM-1 H 290 0.7* KIM-1 L 443 1.13 T-effector signature Teff H 367 0.87 Teff L 366 0.97 NMF subtype NMF1 82 1.14 NMF2 271 1.04 NMF3 131 0.96 NMF4 62 0.88 NMF5 60 0.77 NMF6 50 0.25* NMF7 66 1.62 T-effector signature in KIM-1 H pts KIM-1 H Teff H 147 0.59* KIM-1 H Teff L 143 0.93 *P<0.05.