Stroma and vessel characteristics in cancer : impact on prognosis and response to treatment

Abstract

A series of pre-clinical and clinical studies imply vessel and pericyte status as determinants of tumor growth, metastasis and response to treatment. These studies thus imply biomarkerpotential of these features. Earlier studies of vessels and pericytes have largely applied semiquantitative approaches. In the studies of this thesis, novel tools for quantification of vesseland tumor stroma-related features were developed and applied to different sets of clinically well-annotated tumor collections.

Analyses of perivascular status in ovarian cancer and colorectal cancer identified independently expressed marker-defined subsets of perivascular cells with differential associations with survival. These studies also identified novel significant associations between specific oncogenic mutations and vascular phenotypes.

Studies analyzing stage II/III colon cancer samples, derived from a randomized adjuvant study, identified two novel stroma-related “metrics” that acted as markers for benefit of adjuvant chemotherapy. Firstly, high vessel density in the invasive region, but not tumor center, was identified as a marker that characterized patients benefiting from adjuvant treatment. Secondly, a digital-image-analyses-derived “metric”, related to high complexity of the tumor stroma interface, also defined a group showing benefit of adjuvant treatment. Notably, both novel markers showed statistically significant interactions with treatment supporting their relevance as response-predictive markers.

Furthermore, cell-type-specific analyses of claudin-2 expression in colorectal cancer indicated that CAF-expression of this marker was specifically associated with benefit of oxaliplatin-based treatment for metastatic colorectal cancer.

Taken together, our findings suggest continued exploration and validation of stroma-derived features, for development of clinically meaningful prognostic and response-predictive tissuebased biomarkers.