Purpose: Sarcomatoid renal cell carcinoma (SRCC) ranks among the most aggressive clinicopathologic phenotypes of RCC. However, the paucity of high-quality, genome-wide molecular examinations of SRCC has hindered our understanding of this entity.

Experimental Design: We interrogated the mutational, copy number, and transcriptional characteristics of SRCC and compared these data with those of nonsarcomatoid RCC (RCC). We evaluated whole-exome sequencing, single-nucleotide polymorphism, and RNA sequencing data from patients with SRCC (n = 65) and RCC (n = 598) across different parent RCC subtypes, including clear-cell RCC, papillary RCC, and chromophobe RCC subtypes.

Results: SRCC was molecularly discrete from RCC and clustered according to its parent RCC subtype, though with upregulation of TGFβ signaling across all subtypes. The epithelioid (E-) and spindled (S-) histologic components of SRCC did not show differences in mutational load among cancer-related genes despite a higher mutational burden in S-. Notably, sarcomatoid clear-cell RCC (SccRCC) showed significantly fewer deletions at 3p21-25, a lower rate of two-hit loss for VHL and PBRM1, and more mutations in PTEN, TP53, and RELN compared with ccRCC. A two-hit loss involving VHL predicted for ccRCC and a better prognosis, whereas mutations in PTEN, TP53, or RELN predicted for SccRCC and worse prognosis.

Conclusions: SRCC segregates by parent subtype, and SccRCC has a fundamentally different early molecular pathogenesis, usually lacking the classic 3p21-25 deletion and showing distinctive mutational and transcriptional profiles. These features prompt a more precise molecular classification of RCC, with diagnostic, prognostic, and therapeutic implications. Clin Cancer Res; 23(21); 6686–96. ©2017 AACR.

See related commentary by Bergerot et al., p. 6381

The presence of sarcomatoid histology is associated with a poor prognosis in cancers arising from various organs, and sarcomatoid renal cell carcinoma (SRCC) represents the most aggressive, treatment-resistant group of RCC. The molecular features that characterize and potentially drive the aggressive behavior of SRCC are poorly understood. Our study of SRCC suggests that it is not a single disease entity but rather multiple diseases that segregate according to the underlying parent RCC subtype. Notably, sarcomatoid clear-cell RCC (SccRCC) has a distinct tumorigenesis, often lacking copy losses at 3p21-25 and showing distinct mutational and transcriptomic features. The molecular landscape of SccRCC is likely related to its traditionally poor response to current therapies and may inform future treatments. Our findings improve the molecular classification of RCC, propose biomarkers that may aid in predicting aggressive sarcomatoid tumors preoperatively, and provide opportunities for therapeutic targeting of SRCC.

Renal cell carcinoma (RCC) is an important contributor to cancer-specific mortality (1). Sarcomatoid RCC (SRCC), which is characterized by a biphasic epithelioid carcinomatous (E-) component and a spindle cell sarcomatous (S-) component, represents an important and understudied subset of RCC. SRCC comprises up to 20% of stage IV RCC cases and shows aggressive behavior irrespective of the parent subtype with which it shares histologic characteristics (2–4). No established therapies are effective against SRCC (5). Current clinical thinking conceptualizes SRCC as a single entity (6); however, a thorough molecular understanding of the disease has been limited by the dearth of genome-wide examinations, both within specific RCC subtypes and across different parent RCC subtypes.

Large-scale multi-institutional, multiplatform genomic studies of RCC have been reported, notably The Cancer Genome Atlas (TCGA) data on clear-cell RCC (ccRCC), papillary RCC (PapRCC), and chromophobe RCC (ChRCC; refs. 7–10). Genome-wide molecular characterization of SRCC has heretofore been based on formalin-fixed, paraffin-embedded (FFPE) tissues and largely confined to the sarcomatoid ccRCC (SccRCC) subtype (11–14).

The aim of this study was to comprehensively examine the mutational, copy number, and transcriptional characteristics of SRCC and to compare these data with those of nonsarcomatoid RCC (RCC). We found that SRCC is not a single entity but rather is highly influenced by the parent RCC subtype with which it shares histologic characteristics. Moreover, we show that SccRCC exhibits a different early molecular pathogenesis, driver mutation, and transcriptional profile than ccRCC. These findings will improve the molecular classification of RCC and provide a means for predicting aggressive sarcomatoid tumors preoperatively as well as suggesting therapeutic targets, thus aiding in the management of this disease.

Patient and tumor characteristics

For this retrospective study, we obtained 40 frozen SRCC samples and nonneoplastic kidney control samples from the tissue bank of The University of Texas MD Anderson Cancer Center (Houston, TX) after informed consent and using an institutional review board–approved protocol (IRB# LAB 08-670). The study was conducted in accordance with the Declaration of Helsinki. Lesional foci representing the E- or S- component of SRCC as well as nonneoplastic kidney controls were marked on hematoxylin and eosin–stained slides, with all cases reviewed by at least two genitourinary pathologists (Supplementary Fig. S1). These foci were macrodissected from the frozen samples prior to the extraction of tissue analytes. We identified a second cohort of SRCC cases (n = 32) by examining surgical pathology reports and digital pathologic images from TCGA. The clinicopathologic features of the SRCCs derived from the MD Anderson and TCGA cohorts are summarized in Supplementary Table S1.

Whole-exome sequencing and analysis

Genomic DNA was extracted using the AllPrep DNA/RNA Mini Kit according to the manufacturer's instructions (QIAGEN). DNA concentration was quantified using a NanoDrop 1000 spectrophotometer (NanoDrop Technologies) and the Quant-iT PicoGreen Kit (Life Technologies) individually. Paired-end whole-exome sequencing (WES) was performed on frozen sarcomatoid renal tumors and corresponding normal controls using the HiSeq2000 platform. DNA mutations and copy numbers were assessed from WES data using in-house pipelines. Data were analyzed within and across sample subtypes, with details provided in Supplementary Methods S1.

RNA-sequencing expression profiling and analysis

Total RNA was isolated using the miRNeasy Mini Kit according to the manufacturer's instructions (QIAGEN). RNA was quantitatively assessed using the NanoDrop 1000 spectrophotometer (NanoDrop Technologies) and qualitatively assessed using the Agilent 2100 bioanalyzer (Agilent Technologies).

RNA-sequencing (RNA-Seq) was performed on frozen and FFPE tissues with paired-end sequencing using the HiSeq2000 platform. Gene expression was assessed using an in-house pipeline that uses tophat (15) and htseq (16), among other tools. The details are provided in Supplementary Methods S2.

DNA methylation profiling and analysis

DNA methylation profiling was performed from genomic DNA extracted from frozen sarcomatoid renal tumors and corresponding normal kidney controls using the Illumina Infinium MethylationEPIC BeadChip platform. The details are provided in Supplementary Methods S3. Assessment of gene silencing by DNA methylation was carried out as previously described (17).

Fluorescence in situ hybridization

Fluorescence in situ hybridization (FISH) using a commercial probe set (VHL/CEN3q; Abnova; catalog number FG0029) was performed on FFPE-derived tissue microarray sections from two patient cohorts. The first cohort comprised patients with SccRCC: n = 88 patients, median age at surgery was 59, with 83% dead of disease and median follow-up of 11 months. The second cohort comprised patients with ccRCC lacking sarcomatoid features: n = 66 patients (Fuhrman grade 3, n = 57; Fuhrman grade 2, n = 9), median age at surgery was 58, with 23% dead of disease and median follow-up of 70 months. The detailed FISH methodology is provided in Supplementary Methods S4.

In silico analysis of TCGA data

TCGA exome sequencing and RNA-Seq data derived from SRCC and non-SRCC cases were evaluated using the Broad Institute Firehose pipeline. TCGA data downloaded from the TCGA data coordinating center site were also used to compare SccRCC with non-SccRCC cases in terms of gene-level copy number, arm-level chromosomal copy number, exome-based somatic mutations, and VHL methylation status.

Statistical analysis

For RNA-Seq data differentially expressed gene analysis, we used the “edgeR” package (18) in R to calculate the fold change and to adjust the P value. We analyzed canonical pathways and functional networks with Ingenuity Pathway Analysis (Ingenuity Systems). The Kaplan–Meier survival analysis was used with the log-rank test to assess the statistical significance of the differences between stratified survival groups using the “survival” package in R. Other standard statistical tests were used to analyze the clinical and genomics data, including the χ2 test, Fisher exact test, Mann–Whitney–Wilcoxon test, and Cox proportional hazard analysis. Significance was defined as P < 0.05. Analyses were primarily performed using R. Heat maps were plotted based on the pheatmap function in the “pheatmap” package.

SRCC segregates molecularly according to underlying parent subtype

We examined SRCC derived from three different parent subtypes (SccRCC; sarcomatoid ChRCC or SChRCC; and sarcomatoid PapRCC or SPapRCC) and their biphasic E- and S- morphologic components. The overall mutational landscape of SRCC found in our cohort is schematically represented in Fig. 1A. Supplementary Table S2 summarizes the mutations found in our cohort; detailed sequencing data are in the process of being deposited in database of Genotypes and Phenotypes (accession numbers pending). Using both local and TCGA data, we found SRCC of different subtypes to show a mutational burden similar to that of RCC (refs. 7–9; Fig. 2A). Unsupervised clustering of copy number and transcriptional data showed that SRCC segregated according to the parent subtype rather than to the E- or S- morphologic components, as illustrated in Fig. 2B and C. Potentially actionable mutations, involving pan-cancer T200 panel genes (19), were distributed across all SRCC subtypes (Supplementary Table S2). However, mutations in 3p21-25 genes (VHL, PBRM1, SETD2, and BAP1) were only seen in SccRCC (Fig. 2D).

Figure 1.

Mutational, copy number, and methylation landscape of SRCC. A, Mutation frequency sorted by parent RCC subtype (SccRCC, SPapRCC, SChRCC, and Unclassified); unique patient samples indicated by square borders, with paired patient samples indicated by rectangular borders; E- and/or S- components from each patient assayed by WES or FISH with VHL deletions assessed by WES and FISH. B, Types of somatic mutations involving specific genes, including significantly mutated genes by mutation frequency and found by MutSig*. C, Bar plot shows mutations grouped by SccRCC or sarcomatoid non–clear-cell RCC (SNccRCC) subtype. FS, frameshift; InFS, insertion frameshift; UTR, untranslated region. D, SccRCC derived from the MD Anderson (M4) and TCGA (T4) cohorts and ccRCC (grades 1–4) from TCGA in terms of the “hits” on 3p genes (VHL, PBRM1, SETD2, and BAP1) consisting of mutations, deletions (DEL), and epigenetic silencing by methylation.

Figure 1.

Mutational, copy number, and methylation landscape of SRCC. A, Mutation frequency sorted by parent RCC subtype (SccRCC, SPapRCC, SChRCC, and Unclassified); unique patient samples indicated by square borders, with paired patient samples indicated by rectangular borders; E- and/or S- components from each patient assayed by WES or FISH with VHL deletions assessed by WES and FISH. B, Types of somatic mutations involving specific genes, including significantly mutated genes by mutation frequency and found by MutSig*. C, Bar plot shows mutations grouped by SccRCC or sarcomatoid non–clear-cell RCC (SNccRCC) subtype. FS, frameshift; InFS, insertion frameshift; UTR, untranslated region. D, SccRCC derived from the MD Anderson (M4) and TCGA (T4) cohorts and ccRCC (grades 1–4) from TCGA in terms of the “hits” on 3p genes (VHL, PBRM1, SETD2, and BAP1) consisting of mutations, deletions (DEL), and epigenetic silencing by methylation.

Close modal
Figure 2.

SRCC segregates according to parent subtype. SRCC of different parent subtypes (SccRCC, SPapRCC, and SChRCC) demonstrates (A) similar mutational burdens compared with nonsarcomatoid RCC (RCC) of different parent subtypes (ccRCC, PapRCC, and ChRCC). SRCCs segregate according to parent subtype, as shown by an unsupervised clustering analysis of (B) copy number (1,500 top genes) and (C) mRNA expression (1,000 genes). D, Mutations of 3p21-25 tumor-suppressor genes are confined to SccRCC. FS, frameshift; InFS, insertion frameshift; UTR, untranslated region.

Figure 2.

SRCC segregates according to parent subtype. SRCC of different parent subtypes (SccRCC, SPapRCC, and SChRCC) demonstrates (A) similar mutational burdens compared with nonsarcomatoid RCC (RCC) of different parent subtypes (ccRCC, PapRCC, and ChRCC). SRCCs segregate according to parent subtype, as shown by an unsupervised clustering analysis of (B) copy number (1,500 top genes) and (C) mRNA expression (1,000 genes). D, Mutations of 3p21-25 tumor-suppressor genes are confined to SccRCC. FS, frameshift; InFS, insertion frameshift; UTR, untranslated region.

Close modal

Subsets of genes are mutated across different parent subtypes of SRCC

Despite the large number of nonsynonymous mutations that were found in SRCC, relatively few genes were recurrently mutated. Indeed, among the top 30 recurrently mutated genes, only 9 were significant in SRCC according to the MutSig algorithm (q < 0.1; Fig. 1B and Supplementary Table S3A). Some gene mutations were shared across different subtypes of SRCC, most notably TP53, PTEN, NF2, and RELN (Fig. 1C).

We next assessed whether mutations in the aforementioned genes were significantly enriched in sarcomatoid tumors compared with nonsarcomatoid tumors within the same parent RCC subtype using institutional and TCGA-derived data (Supplementary Table S3B). PTEN was more frequently mutated in SccRCC versus ccRCC; TP53 mutations were associated with SccRCC and SPapRCC; and NF2 mutations were associated with SPapRCC only. RELN mutations were significantly higher in SRCC across all subtypes.

SccRCC shows a different molecular pathogenesis compared with lower grade ccRCC

Biallelic loss of tumor suppressor genes at 3p21-25 is fundamental to the pathogenesis of ccRCC (20, 21). Loss of VHL is thought to be the initiating event, followed by inactivation of other 3p21-25 genes, such as PBRM1 and BAP1 (22). Numerous additional chromosomal copy-number alterations have been proposed as contributing to the progression of ccRCC (23). In terms of arm-spanning copy-number changes, fewer chromosomal losses at 3p were associated with high-grade SccRCC (Supplementary Table S4). Differences in gene-level copy number were more dramatic, with significantly less segmental 3p21-25 deletion seen in high-grade and SccRCC spanning the VHL, PBRM1, BAP1, and SETD2 genes (Supplementary Table S5).

This finding of markedly less 3p21-25 deletions in SccRCC was maintained across a broad range of bioinformatics thresholds used for calling copy loss. We next considered whether a greater burden of contaminating normal diploid cells may have confounded our results. However, the distribution of tumor purities did not differ significantly between high grade (SccRCC + grade 4) versus lower grades (grades 1–3) of ccRCC (P = 0.28). We then used the FISH platform to assay SccRCC and ccRCC for VHL copy number. We found that sarcomatoid features correlated with significantly higher VHL/3q ratios as shown in Supplementary Fig. S2. Using FISH as the basis for copy-number calls, SccRCC showed fewer VHL deletions compared with ccRCC (P = 0.001), and high-grade ccRCC showed fewer VHL deletions compared with lower grade ccRCC tumors (P = 0.000049).

Finally, we considered whether copy-neutral LOH contributed to LOH at 3p in those cases that did not show simple 3p21-25 loss. Our analysis of SNP data from TCGA showed copy-neutral LOH to be an exceedingly rare event in ccRCC (∼1%) and to involve only one case of SccRCC (Supplementary Methods S5 and Supplementary Figs. S3–S7). We assessed independent ccRCC samples assayed by SNP array (GSE12808, GSE14670, GSE11447, and GSE9469) and also found a low incidence of copy-neutral LOH (∼2%) that was not correlated to tumor grade.

SccRCC exhibited less concurrent mutation and copy loss (i.e., two-hit loss) of the VHL and PBRM1 genes than did ccRCC across different histologic grades (Table 1). Gene silencing by methylation is another mechanism of gene inactivation. Epigenetic silencing has been reported to be mutually exclusive with mutation for VHL in ccRCC (7, 24), which is consistent with our data where VHL mutations and methylation did not overlap (Fig. 1D). We only found methylation of the VHL gene and not for the PBRM1, SETD2, or BAP1 genes (Supplementary Fig. S8). Notably, after taking methylation status into account as well as mutation and LOH, SccRCC still showed fewer two-hit losses of VHL compared with ccRCC [P < 0.001; odds ratio (OR), 0.22].

Table 1.

Fewer concurrent LOH and mutations of the VHL and PBRM1 genes in SccRCC than in ccRCC by P value (OR)

3p 21-25 gene
TestVHLBAP1SETD2PBRM1
SccRCC vs. grade 1 ccRCC 3.97E−02 (0.13) 1 (1) 1 (1) 1.93E−03 (0.04) 
SccRCC vs. grade 2 ccRCC 2.7E−06 (0.14) 1 (1.03) 2.11E−01 (0.43) 1.53E−05 (0.11) 
SccRCC vs. grade 3 ccRCC 1.49E−06 (0.13) 1 (1.01) 5.72E−01 (0.63) 1.66E−04 (0.14) 
SccRCC vs. grade 4 ccRCC 5.25E−02 (0.3) 6.97E−01 (0.58) 1 (1.21) 7.99E−03 (0.17) 
SccRCC vs. all ccRCC 1.79E−06 (0.15) 1 (0.96) 4.52E−01 (0.56) 1.79E−05 (0.13) 
3p 21-25 gene
TestVHLBAP1SETD2PBRM1
SccRCC vs. grade 1 ccRCC 3.97E−02 (0.13) 1 (1) 1 (1) 1.93E−03 (0.04) 
SccRCC vs. grade 2 ccRCC 2.7E−06 (0.14) 1 (1.03) 2.11E−01 (0.43) 1.53E−05 (0.11) 
SccRCC vs. grade 3 ccRCC 1.49E−06 (0.13) 1 (1.01) 5.72E−01 (0.63) 1.66E−04 (0.14) 
SccRCC vs. grade 4 ccRCC 5.25E−02 (0.3) 6.97E−01 (0.58) 1 (1.21) 7.99E−03 (0.17) 
SccRCC vs. all ccRCC 1.79E−06 (0.15) 1 (0.96) 4.52E−01 (0.56) 1.79E−05 (0.13) 

NOTE: Contrasts with a P value < 0.05 are in boldface.

Abbreviation: OR, odds ratio.

Mutational profile of ccRCC is predictive of sarcomatoid change and poor prognosis

With respect to ccRCC, we found that mutations in TP53, PTEN, RELN, BAP1, and SETD2 were individually associated with or anticorrelated (PBRM1) with sarcomatoid change (Table 2). A mutation in TP53, PTEN, or RELN was correlated with sarcomatoid change in ccRCC (P = 1.64E–06; OR = 6.53), including grade 4 (P = 4.68E–02; OR = 3.2; Supplementary Table S6). Further, mutant allelic frequencies in TP53, PTEN, and RELN were lower than in VHL or BAP1 (P = 0.035), suggesting that they occurred later in the evolution of the tumor. When comparing the mutational landscape of SccRCC versus ccRCC, profiling of only the E- component of SccRCC revealed similar results compared with profiling of both E- and S- components (Table 2 and Supplementary Table S6). A survival analysis showed that a mutation in TP53, PTEN, or RELN was associated with reduced overall survival in ccRCC (Fig. 3A; P = 8.72E–07). In contrast, deletion of 3p21-25 genes or a two-hit loss of VHL was associated with increased overall survival (Fig. 3B and C).

Table 2.

Mutations in SccRCC versus ccRCC across different Fuhrman grades by P value (OR)

Both S- and E- components of SccRCC vs. ccRCC
GeneGrade 1Grade 2Grade 3Grade 4All ccRCCGrades 1–3a
VHL 0.691 (1.72) 0.757 (1.12) 0.877 (1.07) 0.316 (1.55) 0.667 (1.15) 0.732 (0.91) 
BAP1 0.586 (3.46) 0.033 (2.93) 0.082 (2.27) 0.464 (0.67) 0.066 (2.05) 0.0004 (3.17) 
SETD2 0.328 (4.76) 0.085 (2.08) 0.044 (2.43) 0.305 (1.96) 0.032 (2.25) 0.068 (1.76) 
PBRM1 0.022 (0.13) 0.004 (0.32) 0.01 (0.35) 0.005 (0.25) 0.002 (0.32) 0.146 (0.69) 
PTEN 1 (1.97) 0.015 (5.08) 0.115 (2.45) 0.727 (1.67) 0.034 (3.05) 0.03 (2.78) 
NF2 1 (0.41) 0.244 (9.39) 1 (1.02) 1 (0.8) 0.485 (1.78) 0.31 (2.35) 
RELN 0.462 (0.47) 0.003 (29.8) 0.014 (13.1) 0.378 (3.37) 0.006 (10.1) 0.007 (9.12) 
TP53 1 (2.32) 0.002 (8.12) 0.001 (12.1) 0.291 (3.06) 0.001 (7.96) 0.001 (6.8) 
UPK3B 0.306 (0.23) 0.059 (15.9) 0.151 (6.3) 0.501 (4.16) 0.077 (7.3) 0.216 (3.54) 
HIF1A 1 (0.7) 0.149 (6.38) 0.06 (15.75) 1 (0.79) 0.118 (4.86) 0.01 (14.48) 
PTPRD 1 (0.41) 0.571 (1.56) 0.247 (9.28) 1 (2.45) 0.328 (3.58) 0.532 (1.75) 
Only E- component of SccRCC vs. ccRCC 
Gene Grade 1 Grade 2 Grade 3 Grade 4 All ccRCC Grades 1–3a 
VHL 1 (1.4) 0.865 (0.91) 0.733 (0.87) 0.67 (1.26) 0.873 (0.93) 0.4 (0.8) 
BAP1 0.573 (3.08) 0.072 (2.56) 0.168 (1.98) 0.429 (0.58) 0.189 (1.8) 0.001 (3.07) 
SETD2 0.58 (3.59) 0.334 (1.54) 0.207 (1.79) 0.572 (1.45) 0.229 (1.66) 0.285 (1.42) 
PBRM1 0.048 (0.17) 0.045 (0.43) 0.092 (0.47) 0.034 (0.33) 0.04 (0.43) 0.525 (0.82) 
PTEN 1 (2.14) 0.018 (5.46) 0.146 (2.63) 0.484 (1.8) 0.039 (3.28) 0.04 (2.81) 
NF2 1 (0.53) 0.202 (12.0) 1 (1.31) 1 (1.02) 0.41 (2.29) 0.265 (2.68) 
RELN 0.464 (0.45) 0.008 (29.5) 0.027 (12.5) 0.36 (3.23) 0.015 (9.65) 0.017 (8.27) 
TP53 1 (1.3) 0.098 (4.17) 0.059 (6.22) 0.676 (1.57) 0.067 (4.09) 0.032 (4.15) 
UPK3B 0.14 (0.05) 1 (3.92) 1 (1.28) 1 (1.02) 1 (1.78) 1 (0.79) 
HIF1A 1 (0.9) 0.104 (8.24) 0.041 (20.3) 1 (1.02) 0.08 (6.28) 0.007 (16.6) 
PTPRD 1 (0.53) 0.493 (2) 0.204 (11.9) 0.494 (3.14) 0.27 (4.61) 0.492 (1.99) 
Both S- and E- components of SccRCC vs. ccRCC
GeneGrade 1Grade 2Grade 3Grade 4All ccRCCGrades 1–3a
VHL 0.691 (1.72) 0.757 (1.12) 0.877 (1.07) 0.316 (1.55) 0.667 (1.15) 0.732 (0.91) 
BAP1 0.586 (3.46) 0.033 (2.93) 0.082 (2.27) 0.464 (0.67) 0.066 (2.05) 0.0004 (3.17) 
SETD2 0.328 (4.76) 0.085 (2.08) 0.044 (2.43) 0.305 (1.96) 0.032 (2.25) 0.068 (1.76) 
PBRM1 0.022 (0.13) 0.004 (0.32) 0.01 (0.35) 0.005 (0.25) 0.002 (0.32) 0.146 (0.69) 
PTEN 1 (1.97) 0.015 (5.08) 0.115 (2.45) 0.727 (1.67) 0.034 (3.05) 0.03 (2.78) 
NF2 1 (0.41) 0.244 (9.39) 1 (1.02) 1 (0.8) 0.485 (1.78) 0.31 (2.35) 
RELN 0.462 (0.47) 0.003 (29.8) 0.014 (13.1) 0.378 (3.37) 0.006 (10.1) 0.007 (9.12) 
TP53 1 (2.32) 0.002 (8.12) 0.001 (12.1) 0.291 (3.06) 0.001 (7.96) 0.001 (6.8) 
UPK3B 0.306 (0.23) 0.059 (15.9) 0.151 (6.3) 0.501 (4.16) 0.077 (7.3) 0.216 (3.54) 
HIF1A 1 (0.7) 0.149 (6.38) 0.06 (15.75) 1 (0.79) 0.118 (4.86) 0.01 (14.48) 
PTPRD 1 (0.41) 0.571 (1.56) 0.247 (9.28) 1 (2.45) 0.328 (3.58) 0.532 (1.75) 
Only E- component of SccRCC vs. ccRCC 
Gene Grade 1 Grade 2 Grade 3 Grade 4 All ccRCC Grades 1–3a 
VHL 1 (1.4) 0.865 (0.91) 0.733 (0.87) 0.67 (1.26) 0.873 (0.93) 0.4 (0.8) 
BAP1 0.573 (3.08) 0.072 (2.56) 0.168 (1.98) 0.429 (0.58) 0.189 (1.8) 0.001 (3.07) 
SETD2 0.58 (3.59) 0.334 (1.54) 0.207 (1.79) 0.572 (1.45) 0.229 (1.66) 0.285 (1.42) 
PBRM1 0.048 (0.17) 0.045 (0.43) 0.092 (0.47) 0.034 (0.33) 0.04 (0.43) 0.525 (0.82) 
PTEN 1 (2.14) 0.018 (5.46) 0.146 (2.63) 0.484 (1.8) 0.039 (3.28) 0.04 (2.81) 
NF2 1 (0.53) 0.202 (12.0) 1 (1.31) 1 (1.02) 0.41 (2.29) 0.265 (2.68) 
RELN 0.464 (0.45) 0.008 (29.5) 0.027 (12.5) 0.36 (3.23) 0.015 (9.65) 0.017 (8.27) 
TP53 1 (1.3) 0.098 (4.17) 0.059 (6.22) 0.676 (1.57) 0.067 (4.09) 0.032 (4.15) 
UPK3B 0.14 (0.05) 1 (3.92) 1 (1.28) 1 (1.02) 1 (1.78) 1 (0.79) 
HIF1A 1 (0.9) 0.104 (8.24) 0.041 (20.3) 1 (1.02) 0.08 (6.28) 0.007 (16.6) 
PTPRD 1 (0.53) 0.493 (2) 0.204 (11.9) 0.494 (3.14) 0.27 (4.61) 0.492 (1.99) 

NOTE: Contrasts with a P value < 0.05 are in boldface.

Abbreviation: OR, odds ratio.

aSarcomatoid and grade 4 ccRCC versus grades 1–3 ccRCC.

Figure 3.

Mutational and copy-number profiles associated with SccRCC are prognostic. A mutation in PTEN, RELN, or TP53 is associated with significantly lower overall survival (A), whereas deletion of 3p21-25 (B) or a two-hit loss of VHL (C) is associated with increased overall survival. CI, confidence interval.

Figure 3.

Mutational and copy-number profiles associated with SccRCC are prognostic. A mutation in PTEN, RELN, or TP53 is associated with significantly lower overall survival (A), whereas deletion of 3p21-25 (B) or a two-hit loss of VHL (C) is associated with increased overall survival. CI, confidence interval.

Close modal

SRCC shows a distinct transcriptional program

Because SRCC practically always presents as advanced-stage disease, we compared the transcriptional profile of sarcomatoid tumors with that of advanced-stage nonsarcomatoid tumors across all three RCC subtypes. SRCC showed enrichment of several pathways, including TGFβ1 signaling (activated, P = 5.67E–25; Fig. 4A; Supplementary Table S7A).

Figure 4.

SRCC shows a distinct transcriptional program. A, SRCC shows enrichment of TGFβ1 signaling across different parent subtypes. B, SccRCC shows differential regulation of various signaling pathways compared with advanced-stage (stages III and IV) non-SccRCC. C, Expression differences between the E- and S- components of SccRCC translate into discrete pathway alterations. Barplots indicate number of cases analyzed across each contrast.

Figure 4.

SRCC shows a distinct transcriptional program. A, SRCC shows enrichment of TGFβ1 signaling across different parent subtypes. B, SccRCC shows differential regulation of various signaling pathways compared with advanced-stage (stages III and IV) non-SccRCC. C, Expression differences between the E- and S- components of SccRCC translate into discrete pathway alterations. Barplots indicate number of cases analyzed across each contrast.

Close modal

We next assessed only SccRCC versus advanced-stage ccRCC. Using TCGA data, we found a large number of differentially expressed genes (n = 2,362 genes; FDR < 0.1) that translated into differential regulation of numerous signaling cascades, as illustrated in Fig. 4B and Supplementary Table S7B. Notably, alteration of the VEGF (activated, P = 1.8E–20), TGFβ1 (activated, P = 2.7E–20), and TP53 (inhibited, P = 1.8E–17) pathways was seen in SccRCC. At the level of cellular function, SccRCC showed enrichment of cell growth, movement, and proliferation processes. We performed the same analysis using expression profiling data from local samples (GSE59266) derived from FFPE tissues and found approximately 600 genes that overlapped between the two independent datasets. A pathway analysis based on the overlapping genes revealed enrichment of similar cellular functions and pathways (Supplementary Fig. S9).

Biphasic E- and S- components of SRCC show regional genetic differences

Multiple regions of a given tumor were sequenced in a subset of SRCC cases encompassing different RCC subtypes and different histologic components (E- or S-), as shown in Fig. 1A and Supplementary Table S8. The S- component showed a significantly higher overall mutational load than did the E- component across all RCC subtypes (P = 0.0215) and in the ccRCC subtype (P = 0.0199; Supplementary Fig. S10A and S10B). The mutational load among cancer-related genes (25), however, did not show significant differences between E-and S-. Moreover, despite the biphasic components harboring private mutations that were unique to that region, no single-gene mutation was significantly enriched in E- or S- regions. Interestingly, mutations in cancer-related genes were more commonly shared among the E- and S- components as compared with all gene mutations: for example, 66% shared cancer-related mutations in SRCC versus 19% shared total mutations in RCC, P = 1.15E–09; and 62% shared cancer-related mutations in SccRCC versus 19% shared total mutations in ccRCC, P = 3.57E–05. An Ingenuity Pathway Analysis of the mRNA expression profile of paired E- and S- components of SccRCC revealed activation of TGFβ1 signaling (P = 2.85E–25) as the top upstream regulator (Fig. 4C and Supplementary Table S7C). Network functions that were altered included those involved in cellular assembly and organization as well as cell movement. Together, these pathways provide potential targets for future clinical trials.

SRCC is often considered and managed as a single clinical entity, regardless of the underlying parent subtype with which it is associated. For instance, the National Comprehensive Cancer Network guidelines do not distinguish between sarcomatoid tumors of clear-cell or non–clear-cell origin in terms of recommendations for management (6). Molecular characterization of SRCC has generally been limited to the clear-cell subtype using targeted single-platform studies (11, 13, 14, 26). Our multiplatform, genome-wide analysis shows subtype-specific differences and that SRCC clusters according to parent subtype. In support of our observations, Tickoo and colleagues demonstrated upregulation of hypoxia-inducible factor (HIF) pathway markers in only the clear-cell subtype of SRCC (27). These results suggest that SRCC is not a single disease entity, but rather that its molecular biology reflects its parent subtype, which should be taken into account when managing this disease.

The signal molecular alteration in ccRCC is loss of VHL on 3p (through deletion, mutation, or methylation), leading to activation of HIF pathway molecules such as VEGF and CAIX (22). This has led to the implementation of targeted antiangiogenic agents that abrogate HIF signaling with improvements in progression-free survival (28). Other 3p genes involved in chromatin remodeling (PBRM1, SETD2, and BAP1) were subsequently described as important tumor suppressors that are lost during the pathogenesis of ccRCC (22). The relationship between VHL's mutational status and tumor progression is controversial (29–32), and we did not find a mutation in VHL to correlate with sarcomatoid features or with aggressive disease (Table 2). However, sarcomatoid ccRCC differs from non-SccRCC molecularly in that there is significantly less 3p21-25 deletion spanning the VHL, PBRM1, BAP1, and SETD2 genes and less two-copy loss of VHL and PBRM1. This is consistent with historic cytogenetic studies that have shown the presence of 3p deletions to be associated with lower grade and lower stage disease, absence of sarcomatoid features (33), and improved survival outcomes in ccRCC (33–35). More recent molecular evidence also supports our findings in that LOH at the VHL locus (36) and two-hit loss of the VHL gene in ccRCC (24) were associated with lower stage, lower grade disease and with the absence of sarcomatoid features. Regulation of PBRM1 is postulated to be influenced by tumor differentiation, with fewer mutations in higher grade disease and more aggressive tumors (12, 37, 38). By contrast, BAP1 and SETD2 have been reported to show more mutations with disease progression (7, 12, 39). Taken together, it appears that high-grade ccRCC tumors—exemplified by SccRCC—are prone not to show LOH at 3p21-25 and, notably, to retain a copy of the wild-type VHL and PBRM1 genes. Our analyses challenge the dogma that all ccRCC tumors show a 3p LOH/deletion event (22, 40). Rather, it appears that SccRCC tumor clones show a distinct early molecular pathogenesis, as illustrated in Fig. 5. This suggests that although SccRCC and ccRCC may originate from a common precursor, the molecular events are different, and SccRCC does not evolve linearly from low-grade ccRCC. Consequently, diagnostic molecular assays that fail to demonstrate a 3p or VHL deletion in a suspected ccRCC case do not necessarily rule out a clear-cell subtype of RCC, particularly when associated with higher grade or sarcomatoid features.

Figure 5.

High-grade SccRCC and low-grade ccRCC show a distinct molecular pathogenesis. Schematic illustration of the proposed sequence of events in low-grade ccRCC versus high-grade SccRCC.

Figure 5.

High-grade SccRCC and low-grade ccRCC show a distinct molecular pathogenesis. Schematic illustration of the proposed sequence of events in low-grade ccRCC versus high-grade SccRCC.

Close modal

Whereas mutations in 3p genes were confined to the ccRCC subtype, a handful of other recurrent mutations were shared by SRCC of different subtypes, namely, PTEN, TP53, NF2, and RELN. Although mutations in these genes occurred in all SRCC subtypes, their specific association with sarcomatoid changes within a given subtype varied according to the gene. Among ccRCC, a mutation in PTEN, TP53, or RELN was predictive of the sarcomatoid histologic type, whereas a two-hit inactivation of VHL or PBRM1 was predictive of the non-SccRCC histologic type (Fig. 5). Knowing whether a patient with advanced disease has sarcomatoid features is relevant because of the frequent explosive growth of sarcomatoid tumors that dissuades clinicians from offering cytoreductive nephrectomy prior to initiating systemic therapy. The ability to predict SccRCC is not feasible at present, owing to the lack of radiologic criteria or tissue biomarkers. Moreover, biopsy prediction of RCC tumor grade is notoriously unreliable, including predicting for sarcomatoid change (41).

Assaying for a handful of mutations is feasible using a clinical platform in a Clinical Laboratory Improvement Amendments (CLIA)–certified environment and may be an informative test, with prognostic and management implications. However, the intratumoral mutational heterogeneity of ccRCC, resulting from a putative pattern of branched evolution (40), limits the utility of any mutational screen based on single biopsies, because the majority of mutations are subclonal and, therefore, would not be detected by sampling a single region. Nevertheless, VHL is considered as a clonal mutation, and PBRM1 is sometimes clonal and often shared among different regions (40, 42); thus, assays for these mutations or deletions are less likely to be affected by intratumoral heterogeneity than assays for the other genes that we have proposed. Sampling strategies have not yet been devised to capture subclonal mutations, although early work is promising (42). In this context, our finding, supported by another independent cohort (14), that the S- component showed a higher mutational burden suggests that the higher grade component of a tumor should be assayed as part of a multiregional sampling approach.

The mutational pattern in SccRCC is likely a key driver of its response to existing therapies and may provide insight into new treatment approaches. TP53 mutations have been associated with decreased therapeutic response to antiangiogenic therapies and to PI3K pathway–targeting agents (43, 44). Our recent publication testing the efficacy of PI3K pathway–targeted agents, including MK-2206 and everolimus, demonstrated that mutations in TP53 were associated with reduced response to either agent (44). Duration of response to antiangiogenic agents also appears to be shorter in SccRCC than in RCC (45). Interestingly, SccRCC is associated with increased tissue expression of PD-L1 (46), and emerging data suggest that SccRCC may be associated with heightened response to checkpoint antibodies (E. Jonasch; unpublished observations). The mechanistic link between the unique genomic features of SccRCC and response to checkpoint antibodies is being actively investigated. For PTEN-deficient tumors, several drugs are currently under clinical development in this setting, including PI3K inhibitors, AKT inhibitors, and PARP inhibitors (47). Circulating biomarkers for SccRCC have not yet been formally developed, but a recent study by Pal and colleagues detected a high rate of TP53 mutations (64%) in the circulating tumor DNA of patients with ccRCC treated with later lines of VEGF inhibitors versus first-line treatment (31%; ref. 48). Whether this finding is related to sarcomatoid dedifferentiation remains to be determined.

Transcriptional differences between SccRCC and ccRCC may also be exploited to develop biomarkers of sarcomatoid change, particularly as current technologies enable high-throughput mRNA assays using FFPE material. The major advantage of a transcriptional analysis would appear to be the relatively stable gene expression signature among different regions (12, 13, 40). Our transcriptional analyses suggested that different pathways are enriched in SRCC, particularly TGFβ signaling. These may provide opportunities for therapeutic targeting in addition to biomarkers of tissue of origin.

In summary, we have shown that SRCC is not a single entity but rather multiple diseases that segregate according to the underlying parent subtype. With respect to the most common ccRCC subtype, our findings challenge the dogma that LOH at 3p21-25 is an event common to all ccRCC (20, 22, 40). Rather it appears that clinically aggressive, SccRCC has a distinct tumorigenesis, generally lacking copy losses at 3p21-25. Together with its distinct mutational and transcriptomic landscape, this line of investigation will herald a more precise molecular classification of RCC, with diagnostic, prognostic, and therapeutic implications.

N. Tannir reports receiving speakers bureau honoraria from and is a consultant/advisory board member for Bristol-Myers Squibb, Exelixis, GlaxoSmithKline, Nektar, Novartis, and Pfizer, and reports receiving commercial research grants from Bristol-Myers Squibb, Epizyme, Exelexis, and Novartis. F. Monzon has ownership interest (including patents) in Castle Biosciences. B. Czerniak is a consultant/advisory board member for Abott Molecular and Merck Sharp & Dohme Cor. G. Mills has ownership interest (including patents) in Catena Pharmaceuticals, ImmunoMet, Myriad Genetics, PTV Ventures, and Spindletop Ventures, reports receiving speakers bureau honoraria from Allostery, AstraZeneca, ImmunoMet, ISIS Pharmaceuticals, Lilly, MedImmune, Novartis, and Symphogen, is a consultant/advisory board member for Adventist Health, Allostery, AstraZeneca, Catena Pharmaceuticals, Critical Outcome Technologies, ISIS Pharmaceuticals, ImmunoMet, Lilly, MedImmune, Novartis, Precision Medicine, Provista Diagnostics, Signalchem Lifesciences, Symphogen, Takeda/Millennium Pharmaceuticals, Tarveda, and Tau Therapeutics, and reports receiving commercial research grants from Adelson Medical Research Foundation, AstraZeneca, Breast Cancer Research Foundation, Clinical Outcome Technologies, Illumina, Karus, Komen Research Foundation, Nanostring, and Takeda/Millennium Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Z. Wang, J. Karam, E. Jonasch, C. Wood, B. Czerniak, G. Mills, K. Shaw, K. Sircar

Development of methodology: Z. Wang, J.R. Canales, C. Wood, I. Wistuba, K. Chen, K. Sircar

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Karam, F. Kawakami, P. Trevisan, C.-W. Chow, P. Tamboli, N. Tannir, C. Wood, M. Varella-Garcia, B. Czerniak, G. Mills, K. Chen, K. Sircar

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Z. Wang, T.B. Kim, B. Peng, J. Karam, C. Creighton, A. Joon, N. Tannir, C. Wood, F. Monzon, K. Baggerly, B. Czerniak, G. Mills, K. Chen, K. Sircar

Writing, review, and/or revision of the manuscript: Z. Wang, T.B. Kim, B. Peng, J. Karam, A. Joon, F. Kawakami, E. Jonasch, N. Tannir, C. Wood, F. Monzon, I. Wistuba, G. Mills, K. Shaw, K. Chen, K. Sircar

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.-W. Chow, P. Tamboli, K. Shaw, K. Chen, K. Sircar

Study supervision: C. Wood, K. Chen, K. Sircar

Other (FISH analyses): M. Varella-Garcia

The authors would like to acknowledge Kim-Anh Vu, Camille Sanchez, and Ann Sutton for technical and secretarial assistance.

Z. Wang and K. Chen are partially supported by NIH grant R01CA172652. C. Creighton is supported by NIH grant CA125123. P. Trevisan is supported partially from the CAPES Foundation. M. Varella-Garcia was supported by NCI-P30CA046934 to the Cancer Center Molecular Pathology Shared Resource. K. Sircar received funding from the University of Texas MD Anderson Cancer Center (Robert J. Kleberg, Jr. and Helen C. Kleberg Foundation and CCSG grant CA016672, the Khalifa Bin Zayed Al Nahyan Foundation) and the Kidney Cancer Research Program (Monteleone Foundation). C.J. Creighton is supported by NIH grant CA125123.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Siegel
R
,
Ma
J
,
Zou
Z
,
Jemal
A
. 
Cancer statistics, 2014
.
CA Cancer J Clin
2014
;
64
:
9
29
.
2.
Shuch
B
,
Said
J
,
La Rochelle
JC
,
Zhou
Y
,
Li
G
,
Klatte
T
, et al
Cytoreductive nephrectomy for kidney cancer with sarcomatoid histology–is up-front resection indicated and, if not, is it avoidable?
J Urol
2009
;
182
:
2164
71
.
3.
Shuch
B
,
Bratslavsky
G
,
Shih
J
,
Vourganti
S
,
Finley
D
,
Castor
B
, et al
Impact of pathological tumour characteristics in patients with sarcomatoid renal cell carcinoma
.
BJU Int
2012
;
109
:
1600
6
.
4.
Cheville
JC
,
Lohse
CM
,
Zincke
H
,
Weaver
AL
,
Leibovich
BC
,
Frank
I
, et al
Sarcomatoid renal cell carcinoma: an examination of underlying histologic subtype and an analysis of associations with patient outcome
.
Am J Surg Pathol
2004
;
28
:
435
41
.
5.
Shuch
B
,
Bratslavsky
G
,
Linehan
WM
,
Srinivasan
R
. 
Sarcomatoid renal cell carcinoma: a comprehensive review of the biology and current treatment strategies
.
Oncologist
2012
;
17
:
46
54
.
6.
Motzer
RJ
,
Jonasch
E
,
Agarwal
N
,
Beard
C
,
Bhayani
S
,
Bolger
GB
, et al
Kidney cancer, version 3.2015
.
J Natl Compr Canc Netw
2015
;
13
:
151
9
.
7.
Cancer Genome Atlas Research Network
. 
Comprehensive molecular characterization of clear cell renal cell carcinoma
.
Nature
2013
;
499
:
43
9
.
8.
Linehan
WM
,
Spellman
PT
,
Ricketts
CJ
,
Creighton
CJ
,
Fei
SS
,
Davis
C
, et al
Comprehensive molecular characterization of papillary renal-cell carcinoma
.
N Engl J Med
2016
;
374
:
135
45
.
9.
Davis
CF
,
Ricketts
CJ
,
Wang
M
,
Yang
L
,
Cherniack
AD
,
Shen
H
, et al
The somatic genomic landscape of chromophobe renal cell carcinoma
.
Cancer Cell
2014
;
26
:
319
30
.
10.
Durinck
S
,
Stawiski
EW
,
Pavia-Jimenez
A
,
Modrusan
Z
,
Kapur
P
,
Jaiswal
BS
, et al
Spectrum of diverse genomic alterations define non-clear cell renal carcinoma subtypes
.
Nat Genet
2015
;
47
:
13
21
.
11.
Pal
SK
,
He
ML
,
Tong
T
,
Wu
HQ
,
Liu
XL
,
Lau
C
, et al
RNA-seq reveals aurora kinase-driven mTOR pathway activation in patients with sarcomatoid metastatic renal cell carcinoma
.
Mol Cancer Res
2015
;
13
:
130
7
.
12.
Singh
RR
,
Murugan
P
,
Patel
LR
,
Voicu
H
,
Yoo
SY
,
Majewski
T
, et al
Intratumoral morphologic and molecular heterogeneity of rhabdoid renal cell carcinoma: challenges for personalized therapy
.
Mod Pathol
2015
;
28
:
1225
35
.
13.
Sircar
K
,
Yoo
S-Y
,
Majewski
T
,
Wani
K
,
Patel
LR
,
Voicu
H
, et al
Biphasic components of sarcomatoid clear cell renal cell carcinomas are molecularly similar but distinct from non-sarcomatoid renal carcinomas
.
J Pathol Clin Res
2015
;
1
:
212
24
.
14.
Bi
M
,
Zhao
S
,
Said
JW
,
Merino
MJ
,
Adeniran
AJ
,
Xie
Z
, et al
Genomic characterization of sarcomatoid transformation in clear cell renal cell carcinoma
.
Proc Natl Acad Sci U S A
2016
;
113
:
2170
5
.
15.
Trapnell
C
,
Pachter
L
,
Salzberg
SL
. 
TopHat: discovering splice junctions with RNA-Seq
.
Bioinformatics
2009
;
25
:
1105
11
.
16.
Anders
S
,
Pyl
PT
,
Huber
W
. 
HTSeq–a Python framework to work with high-throughput sequencing data
.
Bioinformatics
2015
;
31
:
166
9
.
17.
Chen
F
,
Zhang
Y
,
Senbabaoglu
Y
,
Ciriello
G
,
Yang
L
,
Reznik
E
, et al
Multilevel genomics-based taxonomy of renal cell carcinoma
.
Cell Rep
2016
;
14
:
2476
89
.
18.
Robinson
MD
,
McCarthy
DJ
,
Smyth
GK
. 
edgeR: a Bioconductor package for differential expression analysis of digital gene expression data
.
Bioinformatics
2010
;
26
:
139
40
.
19.
Chen
K
,
Meric-Bernstam
F
,
Zhao
H
,
Zhang
Q
,
Ezzeddine
N
,
Tang
LY
, et al
Clinical actionability enhanced through deep targeted sequencing of solid tumors
.
Clin Chem
2015
;
61
:
544
53
.
20.
Sato
Y
,
Yoshizato
T
,
Shiraishi
Y
,
Maekawa
S
,
Okuno
Y
,
Kamura
T
, et al
Integrated molecular analysis of clear-cell renal cell carcinoma
.
Nat Genet
2013
;
45
:
860
7
.
21.
Nickerson
ML
,
Jaeger
E
,
Shi
Y
,
Durocher
JA
,
Mahurkar
S
,
Zaridze
D
, et al
Improved identification of von Hippel-Lindau gene alterations in clear cell renal tumors
.
Clin Cancer Res
2008
;
14
:
4726
34
.
22.
Brugarolas
J
. 
Molecular genetics of clear-cell renal cell carcinoma
.
J Clin Oncol
2014
;
32
:
1968
76
.
23.
Hagenkord
JM
,
Gatalica
Z
,
Jonasch
E
,
Monzon
FA
. 
Clinical genomics of renal epithelial tumors
.
Cancer Genet
2011
;
204
:
285
97
.
24.
Dagher
J
,
Kammerer-Jacquet
SF
,
Brunot
A
,
Pladys
A
,
Patard
J
,
Bensalah
K
, et al
Wild-type VHL clear cell renal carcinomas are a distinct clinical and histologic entity: a 10 year follow-up
.
Eur Urol Focus
2016
;
1
:
284
90
.
25.
Kandoth
C
,
McLellan
MD
,
Vandin
F
,
Ye
K
,
Niu
B
,
Lu
C
, et al
Mutational landscape and significance across 12 major cancer types
.
Nature
2013
;
502
:
333
9
.
26.
Malouf
GG
,
Ali
SM
,
Wang
K
,
Balasubramanian
S
,
Ross
JS
,
Miller
VA
, et al
Genomic characterization of renal cell carcinoma with sarcomatoid dedifferentiation pinpoints recurrent genomic alterations
.
Eur Urol
2016
;
70
:
348
57
.
27.
Tickoo
SK
,
Alden
D
,
Olgac
S
,
Fine
SW
,
Russo
P
,
Kondagunta
GV
, et al
Immunohistochemical expression of hypoxia inducible factor-1alpha and its downstream molecules in sarcomatoid renal cell carcinoma
.
J Urol
2007
;
177
:
1258
63
.
28.
Motzer
RJ
,
Hutson
TE
,
Tomczak
P
,
Michaelson
MD
,
Bukowski
RM
,
Rixe
O
, et al
Sunitinib versus interferon alfa in metastatic renal-cell carcinoma
.
N Engl J Med
2007
;
356
:
115
24
.
29.
Patard
JJ
,
Fergelot
P
,
Karakiewicz
PI
,
Klatte
T
,
Trinh
QD
,
Rioux-Leclercq
N
, et al
Low CAIX expression and absence of VHL gene mutation are associated with tumor aggressiveness and poor survival of clear cell renal cell carcinoma
.
Int J Cancer
2008
;
123
:
395
400
.
30.
Yao
M
,
Yoshida
M
,
Kishida
T
,
Nakaigawa
N
,
Baba
M
,
Kobayashi
K
, et al
VHL tumor suppressor gene alterations associated with good prognosis in sporadic clear-cell renal carcinoma
.
J Natl Cancer Inst
2002
;
94
:
1569
75
.
31.
Brauch
H
,
Weirich
G
,
Brieger
J
,
Glavac
D
,
Rodl
H
,
Eichinger
M
, et al
VHL alterations in human clear cell renal cell carcinoma: association with advanced tumor stage and a novel hot spot mutation
.
Cancer Res
2000
;
60
:
1942
8
.
32.
Kim
JH
,
Jung
CW
,
Cho
YH
,
Lee
J
,
Lee
SH
,
Kim
HY
, et al
Somatic VHL alteration and its impact on prognosis in patients with clear cell renal cell carcinoma
.
Oncol Rep
2005
;
13
:
859
64
.
33.
Klatte
T
,
Rao
PN
,
de Martino
M
,
LaRochelle
J
,
Shuch
B
,
Zomorodian
N
, et al
Cytogenetic profile predicts prognosis of patients with clear cell renal cell carcinoma
.
J Clin Oncol
2009
;
27
:
746
53
.
34.
Kroeger
N
,
Klatte
T
,
Chamie
K
,
Rao
PN
,
Birkhauser
FD
,
Sonn
GA
, et al
Deletions of chromosomes 3p and 14q molecularly subclassify clear cell renal cell carcinoma
.
Cancer
2013
;
119
:
1547
54
.
35.
Dagher
J
,
Dugay
F
,
Verhoest
G
,
Cabillic
F
,
Jaillard
S
,
Henry
C
, et al
Histologic prognostic factors associated with chromosomal imbalances in a contemporary series of 89 clear cell renal cell carcinomas
.
Hum Pathol
2013
;
44
:
2106
15
.
36.
Young
AC
,
Craven
RA
,
Cohen
D
,
Taylor
C
,
Booth
C
,
Harnden
P
, et al
Analysis of VHL gene alterations and their relationship to clinical parameters in sporadic conventional renal cell carcinoma
.
Clin Cancer Res
2009
;
15
:
7582
92
.
37.
Kapur
P
,
Pena-Llopis
S
,
Christie
A
,
Zhrebker
L
,
Pavia-Jimenez
A
,
Rathmell
WK
, et al
Effects on survival of BAP1 and PBRM1 mutations in sporadic clear-cell renal-cell carcinoma: a retrospective analysis with independent validation
.
Lancet Oncol
2013
;
14
:
159
67
.
38.
Pena-Llopis
S
,
Vega-Rubin-de-Celis
S
,
Liao
A
,
Leng
N
,
Pavia-Jimenez
A
,
Wang
S
, et al
BAP1 loss defines a new class of renal cell carcinoma
.
Nat Genet
2012
;
44
:
751
9
.
39.
Hakimi
AA
,
Ostrovnaya
I
,
Reva
B
,
Schultz
N
,
Chen
YB
,
Gonen
M
, et al
Adverse outcomes in clear cell renal cell carcinoma with mutations of 3p21 epigenetic regulators BAP1 and SETD2: a report by MSKCC and the KIRC TCGA research network
.
Clin Cancer Res
2013
;
19
:
3259
67
.
40.
Gerlinger
M
,
Horswell
S
,
Larkin
J
,
Rowan
AJ
,
Salm
MP
,
Varela
I
, et al
Genomic architecture and evolution of clear cell renal cell carcinomas defined by multiregion sequencing
.
Nat Genet
2014
;
46
:
225
33
.
41.
Abel
EJ
,
Culp
SH
,
Matin
SF
,
Tamboli
P
,
Wallace
MJ
,
Jonasch
E
, et al
Percutaneous biopsy of primary tumor in metastatic renal cell carcinoma to predict high risk pathological features: comparison with nephrectomy assessment
.
J Urol
2010
;
184
:
1877
81
.
42.
Sankin
A
,
Hakimi
AA
,
Mikkilineni
N
,
Ostrovnaya
I
,
Silk
MT
,
Liang
Y
, et al
The impact of genetic heterogeneity on biomarker development in kidney cancer assessed by multiregional sampling
.
Cancer Med
2014
;
3
:
1485
92
.
43.
Yu
JL
,
Rak
JW
,
Coomber
BL
,
Hicklin
DJ
,
Kerbel
RS
. 
Effect of p53 status on tumor response to antiangiogenic therapy
.
Science
2002
;
295
:
1526
8
.
44.
Jonasch
E
,
Hasanov
E
,
Corn
PG
,
Moss
T
,
Shaw
KR
,
Stovall
S
, et al
A randomized phase 2 study of MK-2206 versus everolimus in refractory renal cell carcinoma
.
Ann Oncol
2017
;
28
:
804
8
.
45.
Jonasch
E
,
Lal
LS
,
Atkinson
BJ
,
Byfield
SD
,
Miller
LA
,
Pagliaro
LC
, et al
Treatment of metastatic renal carcinoma patients with the combination of gemcitabine, capecitabine and bevacizumab at a tertiary cancer centre
.
BJU Int
2011
;
107
:
741
7
.
46.
Joseph
RW
,
Millis
SZ
,
Carballido
EM
,
Bryant
D
,
Gatalica
Z
,
Reddy
S
, et al
PD-1 and PD-L1 expression in renal cell carcinoma with sarcomatoid differentiation
.
Cancer Immunol Res
2015
;
3
:
1303
7
.
47.
Dillon
LM
,
Miller
TW
. 
Therapeutic targeting of cancers with loss of PTEN function
.
Curr Drug Targets
2014
;
15
:
65
79
.
48.
Pal
SK
,
Sonpavde
G
,
Agarwal
N
,
Vogelzang
NJ
,
Srinivas
S
,
Haas
NB
, et al
Evolution of circulating tumor DNA profile from first-line to subsequent therapy in metastatic renal cell carcinoma
.
Eur Urol
2017
;
72
:
557
64
.

Supplementary data