Genomic Meta-analysis of Clear-cell Renal Cell Carcinoma (ccRCC): Aggregating Tumors to Resolve the Complexity of ccRCC

While VHL inactivation and subsequent upregulation of pathways controlling cellular oxygen sensing have long been recognized as a genetic alteration present in most clear-cell renal cell carcinomas (ccRCCs), recent studies using high-throughput sequencing have identified dysregulation of chromatin remodeling pathways as another central feature of ccRCC. In one of the largest studies to date, The Cancer Genome Atlas (TCGA) combined DNA sequencing, RNA sequencing, and DNA methylation across 417 ccRCC tumors and identified 19 significantly mutated genes with recurrent mutations in the PI3K/AKT pathway, widespread DNA hypomethylation associated with mutations of SETD2, and aberrations in the SWI/SNF chromatin remodeling complex with possible downstream effects on multiple other pathways [1]. However, there is substantial intratumoral heterogeneity within ccRCC, with multiregion sampling demonstrating that up to 69% of mutations are not present across every sequenced site within a tumor [2]. It has not been fully elucidated how this intratumoral heterogeneity affects tumor response to targeted therapies, but presumably unidentified subclonal mutations could impact classification of responders versus nonresponders in clinical trials. As targeted systemic therapies are transforming the treatment of advanced RCC, accurate characterization of the genomic alterations present in individual tumors is essential for choosing the optimal therapy for patients.

In this issue of European Urology, Bui and colleagues [3] present the first meta-analysis combining previously published genomic data from both primary and metastatic ccRCC tumors to assess the prevalence of gene mutations and copy number alterations. Overall, this comprehensive analysis highlights the contemporary challenges associated with genomic profiling for clinical use, which include intratumoral heterogeneity within ccRCC [2], variability in both tumor processing (frozen vs formalin-fixed, paraffin-embedded [FFPE] tissue), and sequencing techniques (Sanger vs next-generation sequencing [NGS]). Following the PRISMA methodology for systemic reviews, the meta-analysis merges genomic data from an impressive number of samples, including 14 696 patients with 14 299 primary tumors and 969 metastatic samples, to characterize mutation prevalence and copy number changes present in ccRCC tumors. To put this in perspective, the largest genomic reports for ccRCC before this study included just hundreds of patients. Using the combined genomic data set with a total of 10 874 primary RCCs, Bui and colleagues found that the four most commonly mutated genes in ccRCC are VHL, PBRM1, SETD2, and BAP1, with a mutation prevalence of 64%, 36%, 20%, and 13%, respectively. While these four genes were previously identified in the TCGA KIRC data set [1], this meta-analysis allows for evaluation of factors impacting the mutation rate and demonstrated that NGS identified a significantly higher mutation rate for VHL and SETD2 than Sanger sequencing did. On the contrary, and important for the development of future studies and technologies, there was no difference in the mutation detection rate by nucleic acid source (FFPE or frozen tissue).

However, perhaps the most interesting finding of this meta-analysis is the differences in the genomic profile between primary ccRCC tumors and metastatic tumors. First, the metastatic sites demonstrated significantly higher mutation prevalence for ten genes (VHL, PBRM1, SETD2, BAP1, KDM5C, TP53, PTEN, TSC1, TET2, and ARID1A) compared to primary human tumors. Copy number alterations also exhibited differential prevalence between primary and metastatic tumors, with significantly higher prevalence of chromosome 1p36.11, 9p21.3, and 18 losses, and of chromosome 1q21.3, 7q36.3, 8q, and 20q11.21 gains. While these findings could be because of the greater tumor cellularity sometimes seen in metastases, they could also be explained by metastatic subclones within the primary tumor. In this vein, Bailey and colleagues [4] previously showed that inactivation of the CDKN2A tumor suppressor gene (located at 9p21.3) results in spontaneous metastases in a genetically engineered murine model of ccRCC. Regardless of the cause, given the possible association of mutations with response to systemic treatment agents [57], this meta-analysis highlights the importance of obtaining tissue from metastatic sites, as opposed to primary tumors, for identification of somatic mutations.

While this study describes genomic characterization of the largest cohort of ccRCC tumors to date, there are limitations associated with the observational nature of systematic reviews. First, there was substantial heterogeneity in mutation prevalence across different studies, possibly reflecting the etiologic heterogeneity of ccRCC in different populations. For example, the ABCB1 mutation frequency was 50% for a Chinese cohort, but a mere 2% in the USA-based TCGA study. In this meta-analysis, it is unclear whether the heterogeneity observed is due to study differences or intratumoral heterogeneity. Differences in NGS sequencing depth could impact mutation frequencies and the impact of sequencing depth on variations in mutation frequency between studies was not evaluated. In addition, while the meta-analysis focused on ccRCC, other histologic subtypes of RCC were included, comprising 1.7% of samples, and could be an alternative source of heterogeneity. Further-more, while such a large tumor cohort could provide substantial power for evaluating the clinical relevance of the genomic alterations identified, the lack of clinical annotation for this combined data set prohibits any analyses of downstream survival or therapeutic responses.

Moving forward, future studies should continue to address the challenges imposed by intratumoral heterogeneity in the therapeutic landscape for ccRCC. This study demonstrates that in samples with multiregion sampling (ie, matched primary tumor and metastases), additional mutations are identified, as evidenced by an increase in the prevalence of the most frequently mutated genes. Similarly, in TRACERx Renal, extensive sampling (>20 biopsies) in a subset of 15 tumors demonstrated that seven biopsies are required to detect >75% of all variants [8]. However, the number of biopsies required was lower for homogeneous than for heterogeneous tumors and future studies should evaluate characteristics delineating the two groups, such as tumor size. The repertoire sequencing (Rep-Seq) method, in which bulk tumor tissue left over after pathologic analysis is homogenized for pooled sequencing, may bypass the heterogeneity issue and improve tumor mutation detection [9]. Alternatively, sequencing of cell-free tumor DNA derived from plasma is both sensitive and useful in detecting targetable mutations in several other cancers [10]. This developing technology might help in overcoming barriers related to sequencing of tumor tissue from both primary and metastatic lesions.