Genomic characterization of sarcomatoid transformation in clear cell renal cell carcinoma

Abstract

The presence of sarcomatoid features in clear cell renal cell carcinoma (ccRCC) confers a poor prognosis and is of unknown pathogenesis. We performed exome sequencing of matched normal-carcinomatous-sarcomatoid specimens from 21 subjects. Two tumors had hypermutation consistent with mismatch repair deficiency. In the remainder, sarcomatoid and carcinomatous elements shared 42% of somatic single-nucleotide variants (SSNVs). Sarcomatoid elements had a higher overall SSNV burden (mean 90 vs. 63 SSNVs, P = 4.0 × 10(-4)), increased frequency of nonsynonymous SSNVs in Pan-Cancer genes (mean 1.4 vs. 0.26, P = 0.002), and increased frequency of loss of heterozygosity (LOH) across the genome (median 913 vs. 460 Mb in LOH, P < 0.05), with significant recurrent LOH on chromosomes 1p, 9, 10, 14, 17p, 18, and 22. The most frequent SSNVs shared by carcinomatous and sarcomatoid elements were in known ccRCC genes including von Hippel-Lindau tumor suppressor (VHL), polybromo 1 (PBRM1), SET domain containing 2 (SETD2), phosphatase and tensin homolog (PTEN). Most interestingly, sarcomatoid elements acquired biallelic tumor protein p53 (TP53) mutations in 32% of tumors (P = 5.47 × 10(-17)); TP53 mutations were absent in carcinomatous elements in nonhypermutated tumors and rare in previously studied ccRCCs. Mutations in known cancer drivers AT-rich interaction domain 1A (ARID1A) and BRCA1 associated protein 1 (BAP1) were significantly mutated in sarcomatoid elements and were mutually exclusive with TP53 and each other. These findings provide evidence that sarcomatoid elements arise from dedifferentiation of carcinomatous ccRCCs and implicate specific genes in this process. These findings have implications for the treatment of patients with these poor-prognosis cancers.

Keywords: dedifferentiation; kidney cancer; p53; sarcomatoid; transformation.

Fig. S1.

Patient disease-specific survival.

Fig. 1.

Somatic mutations in 21 renal tumors with sarcomatoid features. (A) Somatic mutation counts in 21 tumors by tumor component. Sample IDs labeled on bottom axis. (B) Somatic mutation pattern by single nucleotide change. (C) Presence of somatic mutations and LOH for significantly mutated and genes of interest. (D) Frequency of LOH events by chromosome region in the carcinomatous (green) and sarcomatoid (red) tumor components for 14 nonhypermutated tumors with complete genome-wide LOH data. (E) Presence of LOH in chromosomal segments with significant sarcomatoid-specific LOH.

Fig. S2.

Number of somatic mutations (log scale) in each cohort (TCGA Clear Cell/KIRC; ref. (27) and clear cell and sarcomatoid component of our tumor cohort. Note hypermutated outliers (samples 1 and 2), one of which had a shared and a sarcomatoid-specific MSH2 and POLE mutation, respectively. The median number of mutations was greater for both the carcinomatous and sarcomatoid components compared with the TCGA clear cell cohort (Mann–Whitney u test, *P = 0.005 and **P < 0.001).

Fig. S3.

Phylogenetic trees of 21 ccRCC tumors. Branch and trunk lengths correspond to the number of somatic mutations in each tumor component, including shared, carcinomatous-specific, and sarcomatoid-specific mutations. Mutations in previously described ccRCC genes, Pan-Cancer genes, other recurrently mutated genes (TSG101, RQCD1, LRIF1, PTK7, and FAT family), and MMR genes in hypermutated samples are shown. Sample IDs are labeled at top left of each phylogenetic tree.

Fig. 2.

Comparison of somatic mutations in carcinomatous and sarcomatoid elements. (A) Mean number of somatic mutations by tumor component for the 19 nonhypermutated tumors. Among all mutations, 41.7% were shared between tumor components. Sarcomatoid regions had a significantly higher number of component-specific mutations (mean 45 vs. 18, P = 6.2 × 10⁻⁴ by Wilcoxon signed-rank text). (B) Mean number of nonsynonymous somatic mutations in known Pan-Cancer genes by tumor component. Sarcomatoid regions had a significantly higher number of component-specific mutations (1.42 vs. 0.26, P = 0.002 by Wilcoxon signed-rank test). (C) Ratio of nonsynonymous to synonymous mutations in known Pan-Cancer genes by tumor component.

Fig. 3.

Phylogenetic trees of six TP53-mutant, nonhypermutated ccRCC tumors. Branch and trunk lengths correspond to the number of somatic mutations in each tumor component, including shared, carcinomatous-specific, and sarcomatoid-specific mutations. Mutations in previously described ccRCC genes, Pan-Cancer genes, and other recurrently mutated genes (TSG101, RQCD1, LRIF1, PTK7, and FAT family) are shown. Sample IDs are labeled at top left of each phylogenetic tree.

Fig. S4.

Comparison of frequency of somatic mutation of most frequently mutated in sarcomatoid regions (red outline) (n = 19) and that observed in nonsarcomatoid clear cell tumors in the TCGA (blue outline) (n = 395). Heterozygous alterations are shaded gray, whereas hemizygous alterations are in white. Fisher’s exact test was used for comparisons. *P < 0.05. TP53, ARID1A, RQCD1, LRIF1, TSG101, PTK7, and FAT1/2/3 were all significantly increased in the sarcomatoid components.