10.1038_modpathol.2016.20.pdf (8.78 MB)
mTOR Pathway Genes Drive Intratumor Heterogeneity and Morphological Progression in Clear Cell Renal Cell Carcinomas
journal contribution
posted on 2016-03-20, 14:19 authored by Salvador J. Diaz-CanoSalvador J. Diaz-Cano, R Sutherland, J Moorhead, A Blanes, R DobsonBackground: The meaning of morphological intratumor heterogeneity (ITH) in clear cell renal cell carcinomas (ccRCC) by Fuhrman grade (FG) has not been investigated at clinic-pathologic or genetic levels in the Cancer Genome Atlas (TCGA) set. Design: A systematic evaluation of 401 ccRCC included: microscopic satellites, heterogeneous clones, FG (1-4), con uent necrosis, spindle cell presence, primary and secondary growth pattern (tubular-nested-thick trabecular-solid), tumor in ltrating lymphocytes, stromal reaction (none/myxoid/desmoplastic), and edges (pushing/ in ltrative). Morphological heterogeneity (ITH+) was systematically assessed, and identi ed in 56 cases. Clinical data were also collected (gender, age, and stage). Whole-exome sequencing was performed on tumor and normal tissues from ccRCC available at the TCGA. We used a Random Forest machine learning approach comparing ITH- vs. ITH+ ccRCC.
Model Analysis
► Data was retrieved from the Pan-Cancer Analysis repository
► Functional somatic mutations unique to tumors were identi ed and represented as samples x genes mutation matrix (mutated=1, non-mutated=0).
► Pairwise Random Forest models were built for the low-intermediate-high CNAG subgroups
► Variable selection using Fisher’s Exact test was conducted to reduce the number of predictors with 50 fold cross-validation design.
Random Forest models were based on the training set using the caret package in R, and predictive accuracy measured in an independent test set.
Results: Complete histological-molecular data were available in 313 cases (179 low FG, 19 ITH+; 139 high FG, 37 ITH+). ITH+ ccRCC were mainly de ned by thick trabecular-solid secondary pattern (31/56 vs. 92/257 ITH-, P=0.007), and high FG (37/56 vs. 102/257 ITH-, P=0.001), with no differences on vascular invasion, satellites or multifocality. The age-gender-proteins model performed with an AUC of 0.875, being predictive features of ITH+ ccRCC: PBRM1 (OR = 9.562, P < 1.3E-17), BAP1 (OR = 9.144, P < 6.0E-16), TP53 (OR = 8.574, P < 4.7E-14), PTEN (OR = 0.661, P < 2.0E-10), PIK3CA (OR = 0.328, P < 5.38E-06). The proportion of C>T transitions at CpG sites increased from morphologically ITH- to ITH+ neoplasms and during grade progression. Conclusions: Morphological heterogeneity in ccRCC represents a stepwise tumor progression de ned by thick trabecular-solid growth and high nuclear grade, which is mainly driven by PI3K-mTOR pathway genes.
Model Analysis
► Data was retrieved from the Pan-Cancer Analysis repository
► Functional somatic mutations unique to tumors were identi ed and represented as samples x genes mutation matrix (mutated=1, non-mutated=0).
► Pairwise Random Forest models were built for the low-intermediate-high CNAG subgroups
► Variable selection using Fisher’s Exact test was conducted to reduce the number of predictors with 50 fold cross-validation design.
Random Forest models were based on the training set using the caret package in R, and predictive accuracy measured in an independent test set.
Results: Complete histological-molecular data were available in 313 cases (179 low FG, 19 ITH+; 139 high FG, 37 ITH+). ITH+ ccRCC were mainly de ned by thick trabecular-solid secondary pattern (31/56 vs. 92/257 ITH-, P=0.007), and high FG (37/56 vs. 102/257 ITH-, P=0.001), with no differences on vascular invasion, satellites or multifocality. The age-gender-proteins model performed with an AUC of 0.875, being predictive features of ITH+ ccRCC: PBRM1 (OR = 9.562, P < 1.3E-17), BAP1 (OR = 9.144, P < 6.0E-16), TP53 (OR = 8.574, P < 4.7E-14), PTEN (OR = 0.661, P < 2.0E-10), PIK3CA (OR = 0.328, P < 5.38E-06). The proportion of C>T transitions at CpG sites increased from morphologically ITH- to ITH+ neoplasms and during grade progression. Conclusions: Morphological heterogeneity in ccRCC represents a stepwise tumor progression de ned by thick trabecular-solid growth and high nuclear grade, which is mainly driven by PI3K-mTOR pathway genes.
