Summary
Cancer is the second most common cause of premature death worldwide, and clear cell renal cell carcinoma (ccRCC) is the seventh most frequent cancer type. Although advances in imaging, diagnostics, prognostics, and treatment have improved patient outcomes, ccRCC remains associated with high rates of recurrence and progression. Unlike breast cancer and other malignancies, molecular prognostic tools are still not employed in the clinical management of ccRCC. This gap is often attributed to poor reproducibility across cohorts, limited improvements over established clinical models, and a lack of biological interpretability of candidate biomarkers.

The present work addresses these challenges by investigating the prognostic potential of transcriptomic profiling in both tumor and peritumoral compartments of ccRCC, with a particular emphasis on robust statistical modeling and biological interpretability. The thesis is presented in an article-based format, comprising three main research articles and one perspectives article that outlines recommendations for future work related to peritumoral tissue (PTT). The first study focuses on COL7A1, identified using Cox model as one of the most prognostic genes in ccRCC. We demonstrated that COL7A1 expression significantly improves survival prediction, even when considered alongside established clinical variables such as tumor stage, nuclear grade, and patient age.

The second study originated from the unexpected finding that gene expression in PTT was even more prognostic of outcome than tumor-derived gene expression. Detailed literature review revealed that PTT is transcriptionally distinct from both tumor and healthy kidney tissue, challenging the long-standing misconception of its “normalcy.” As a result, this study highlighted the need for standardized terminology, need for increased awareness that PTT is not “normal,” and a need for precise metadata annotation in future PTT research.

The third study further explored the prognostic and biological significance of PTT in ccRCC. We confirmed that PTT-derived transcriptomic signals stratify patient outcomes more effectively than tumoral signals, and identified inflammatory processes within PTT as key drivers of poor prognosis. These findings position PTT as an underappreciated but clinically relevant compartment for biomarker discovery. Finally, the fourth study investigated the immune landscape within the ccRCC tumor microenvironment. We characterized CD8⁺ILT2⁺ T cells as a distinct bystander cytotoxic population, largely antigen-independent and shaped by past viral infections rather than tumor-specific reactivity. This population may represent a reservoir of effector function that could be leveraged in future immunotherapy strategies, while also serving as a potential predictive biomarker of treatment response.

Together, these studies demonstrate that transcriptomic profiling of both tumor and peritumoral compartments can provide novel and clinically meaningful prognostic insights in ccRCC. By identifying COL7A1 as a strong single-gene biomarker, revealing the prognostic value of PTT, and characterizing immune bystander populations, this thesis advances the field of biomarker discovery and emphasizes the importance of looking beyond the tumor core when studying cancer prognosis.

Supervision
Laurent GUYON