Sarcomatoid Renal Cell Carcinoma: The Apple Doesn't Fall Far from the Tree

In this issue of Clinical Cancer Research, Wang and colleagues make the most ambitious sequencing effort to date in the realm of sarcomatoid renal cell carcinoma (sRCC; ref. 1). sRCC can arise in any histologic subtype of RCC and is characterized histologically by the appearance of spindle-shaped mesenchymal cells; most troublesome is the observation that sRCC is associated with a worse outcome and poor response to targeted therapies (2). Several other sequencing studies have been performed in this field, but what sets this work apart from the others is the large number of samples, the use of frozen tissues instead of formalin-fixed, parafin-embedded (FFPE) tissues, and the inclusion of sRCC samples from papillary RCC (pRCC) and chromophobe RCC (chRCC) backgrounds in addition to clear cell RCC (ccRCC).

The authors first observe, through unsupervised clustering of copy number and transcriptome data, that sRCC samples segregated according to the parent subtype rather than the epithelial (E) or sarcomatoid (S) morphologic components. Several previous studies have demonstrated the clonal and related nature of S and E components of the same tumor, including initial experiments using X-chromosome inactivation analysis (3) and more recent experiments using RNA-sequencing (RNA-Seq) analysis (4). Thus, it is to be expected that sRCC samples cluster with the parent subtype.

Despite the segregation by parental subtype, several important similarities were revealed that suggest common pathways to sarcomatoid differentiation. Despite a large number of nonsynonymous mutations being found in sRCC tissues, only nine genes were found to have significant, recurrent mutation across all sRCC samples: VHL, c10orf113, PTEN, TP53, BAP1, NF2, TMEM97, CALML3, and IL15. Within each histologic subtype, only a few mutations were found to be significant, and of these, only RELN mutations were significantly higher in sRCC across all subtypes (Fig. 1A). RELN encodes the protein Reelin, which is an extracellular matrix serine protease best known for regulating microtubule function in neurons and neuronal migration (5). However, RELN appears to be an underappreciated, frequently mutated gene in many cancers, as it is altered in more than 10% of samples from over 30 of The Cancer Genome Atlas (TCGA) datasets—the vast majority of which are truncation and missense mutations (6). Although these mutations are listed as “putative passengers,” the high frequency across specific cancer types and emerging molecular studies suggest otherwise. It was recently found that Reelin is an essential negative regulator in the TGFβ1-induced cell migration process in an esophageal cancer model, as knockdown increased migration and overexpression prevented TGFβ-induced migration (7). Furthermore, Reelin expression is controlled by Snail, a master regulator of epithelial–mesenchymal transition (EMT), and knocking down Reelin leads to increased expression of mesenchymal markers. The sarcomatoid differentiation process is thought to be related to EMT (2); thus, loss of Reelin function through mutation may be an important player in this process.

Interestingly, the transcriptomic/pathway analysis performed by Wang and colleagues identified increased TGFβ signaling as the most significantly altered pathway in S versus E components and in sarcomatoid ccRCC versus nonsarcomatoid ccRCC samples (1). The loss of RELN function would be predicted to activate the pathways downstream of TGFβ and thus might contribute to this signal. We previously performed a transcriptomic analysis on a much smaller set of E and S components from sarcomatoid ccRCC samples (4). A comparison of our pathway analysis to that of Wang and colleagues shows striking similarities (Fig. 1C). Thirteen of the top 20 affected “upstream regulators” are shared between the datasets, and nearly all of the others are simply further down on the list of the respective datasets. We also found that TGFβ signaling was increased in S versus E components, but it was 19th on our list. Thus, TGFβ signaling along with activated pathways such as FOXM1, CCND1, PTGER2, and ERBB2 and inhibited pathways such as TP53, CDKN1A, and NUPR1 are likely involved in the sarcomatoid phenotype and deserve further study. We also found that the Aurora kinases contributed to the signature of many of the most significantly altered upstream regulators in our pathway analysis and went on to show that overexpression of Aurora kinase A caused increased proliferation of RCC cells. The Aurora kinases were also prevalent in the lists of genes contributing to the top-20 regulated pathways in the study by Wang and colleagues, thus supporting our findings that the Aurora kinases play an important role in sarcomatoid differentiation and may provide a valuable drug target to combat this disease.

In addition to specific mutations, Wang and colleagues found that the S component of tumors had a greater mutational burden than the E component of sarcomatoid samples and that sarcomatoid ccRCC samples had a higher mutational burden than nonsarcomatoid ccRCC samples (1). As high mutational burden has been associated with response to checkpoint inhibitors (8), it is possible that sRCC might be sensitive to these agents. Indeed, several case reports have recently been published that demonstrate partial and complete responses to nivolumab in patients with sRCC (9, 10). For a cancer that has been so resistant to targeted therapies, these results hold great promise.

One other significant observation from the study by Wang and colleagues was that there was more VHL loss in low-grade ccRCC than in high-grade or sarcomatoid ccRCC. Furthermore, VHL loss was associated with increased survival and combined VHL/PBRM1 alterations negatively predicted for sarcomatoid development. These data support the accumulating evidence that suggests that loss of heterozygosity at the VHL locus (11) and two-hit loss of the VHL gene in ccRCC (12) are associated with lower stage, lower grade disease. Furthermore, these data suggest that sRCC does not evolve linearly from low-grade RCC. As these data mature, they will likely impact treatment decisions—how aggressively to treat and with which agent—for RCC patients.

From a broader view, sarcomatoid differentiation is a fascinating phenomenon that occurs in many solid tumors. Although clustering data in Wang and colleagues' work clearly suggest that sarcomatoid sections are more closely related to epithelial sections from the same specimen than to other sarcomatoid tissues, direct comparisons of mutations and transcriptome changes in the E and S components were performed only in the ccRCC background. It would be very interesting to see which, if any, E versus S differences are shared among ccRCC, RCC, and chRCC, and possibly other solid tumors that show sarcomatoid differentiation (bladder, lung, etc.). Such “shared differences” may demonstrate common causes and pathways of sarcomatoid differentiation across tumor types, or they may support the hypothesis that sarcomatoid differentiation is truly parental tumor type specific. If commonalities are found to contribute to sarcomatoid differentiation across multiple tumor types, it could lead to novel pathways that could be targeted by drugs in “basket” clinical trials, which could speed development of drugs for these relatively rare cancers.

N. Agarwal is a consultant/advisory board member for Clovis, Eisai, Exelixis, EMD Serono, Genentech, Merck, Novartis, and Pfizer. S.K. Pal reports receiving speakers bureau honoraria from Genentech, and is a consultant/advisory board member for Astellas, Bristol-Myers Squibb, Exelixis, Genentech, Ipsen, and Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: P. Bergerot, N. Agarwal, S.K. Pal

Development of methodology: P. Bergerot, S.K. Pal

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.K. Pal

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Agarwal, S.K. Pal, J. Jones

Writing, review, and/or revision of the manuscript: P. Bergerot, N. Agarwal, S.K. Pal, J. Jones

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.K. Pal

Study supervision: N. Agarwal