Spatial Distribution of Tumor Cells in Clear Cell Renal Cell Carcinoma Is Associated with Metastasis and a Matrisome Gene Expression Signature

Authors

Prahlad Bhat
Pheroze Tamboli
Kanishka Sircar
Kasthuri Kannan

Publication Date

1-14-2025

Journal

Cancers

Abstract

Background/Objectives: Predicting the behavior of clear cell renal cell carcinoma (ccRCC) is challenging using standard-of-care histopathologic examination. Indeed, pathologic RCC tumor grading, based on nuclear morphology, performs poorly in predicting outcomes of patients with International Society of Urological Pathology/World Health Organization grade 2 and 3 tumors, which account for most ccRCCs.

Methods: We applied spatial point process modeling of H&E-stained images of patients with grade 2 and grade 3 ccRCCs (n = 72) to find optimum separation into two groups.

Results: One group was associated with greater spatial randomness and clinical metastasis (p < 0.01). Notably, spatial analysis outperformed standard pathologic grading in predicting clinical metastasis. Moreover, cell-to-cell interaction distances in the metastasis-associated group were significantly greater than those in the other patient group and were also greater than expected by the random distribution of cells. Differential gene expression between the two spatially defined groups of patients revealed a matrisome signature, consistent with the extracellular matrix's crucial role in tumor invasion. The top differentially expressed genes (with a fold change > 3) stratified a larger, multi-institutional cohort of 352 ccRCC patients from The Cancer Genome Atlas into groups with significant differences in survival and TNM disease stage.

Conclusions: Our results suggest that the spatial distribution of ccRCC tumor cells can be extracted from H&E-stained images and that it is associated with metastasis and with extracellular matrix genes that are presumably driving these tumors' aggressive behavior.

Keywords

clear cell renal cell carcinoma, metastasis, matrisome, spatial, point process, digital pathology, tumor grade

DOI

10.3390/cancers17020249

PMID

39858031

PMCID

PMC11763402

PubMedCentral® Posted Date

1-14-2025

PubMedCentral® Full Text Version

Post-print

Published Open-Access

yes