Pathologists have made much progress in understanding the diversity — and to some extent, the aggressiveness — of different renal cell carcinoma (RCC) tumors. But a key gap remains the ability to reliably and objectively predict patients’ responses to treatment and their survival.

In an effort towards filling that knowledge gap, a collaboration of researchers recently published an extensive transcriptomic and genomic analysis of clear-cell or sarcomatoid RCC tumors that had been collected prior to the phase 3 IMmotion151 trial ( Identifier: NCT02420821). Results reported last year demonstrated prolonged progression-free survival (PFS) in patients with metastatic disease who received a combination of the vascular endothelial growth factor (VEGF)-targeting monoclonal antibody (mAb) bevacizumab and the programmed death ligand 1 (PD-L1)–targeting checkpoint inhibitor atezolizumab vs patients who only received the antiangiogenesis treatment sunitinib, a VEGF-targeting tyrosine kinase inhibitor (TKI).1

Now, the new research, published in November 2020 in Cancer Cell, identified 7 molecular subsets of clear cell RCC with distinct angiogenesis, cell-cycle, immune, metabolism, and stromal programs, which are often also associated with somatic gene mutations. These subsets exhibited different clinical responses to the mAb alone or with the checkpoint inhibitor, providing a molecular explanation of patients’ responses to treatment and aiding the development of prognostic and predictive biomarkers in RCC.

Continue Reading

“It’s an incredibly valuable dataset that will have important implications for the development of predictive biomarkers for the 2 therapies that they examined,” said Marcin Cieslik, PhD, assistant professor of computational medicine & bioinformatics and assistant professor of pathology in the department of pathology at Michigan Medicine, who wasn’t involved in the new research.

The study focused on tumor samples — mostly from nephrectomies — collected from 823 of the 915 patients who participated in the trial no longer than 2 years before enrollment. The tumors were either of primary (625) or metastatic (198) origin at collection (although the trial focused on metastatic disease). And, 688 were of clear cell histology, though the remainder had varying degrees of sarcomatoid components.

After generating whole-transcriptome profiles of the samples through RNA sequencing, the researchers used a machine learning algorithm to cluster the samples into discrete groups based on the top 10% most variable genes.

This revealed 7 biologically distinct tumor groups. Clusters 1 and 2 (representing 12% and 30% of the tumors, respectively) were enriched for angiogenesis-related genes, including genes involved in the VEGF pathway. Tumors assigned to cluster 3 (19%) tended to have low expression of angiogenesis and immune-related genes but moderate expression of ones involved in cell cycle regulation. Clusters 4, 5, and 6 (14%, 9%, and 13%) had notably low expression of angiogenesis-related genes but pronounced cell cycle- and anabolic metabolism-related transcriptional signatures. Cluster 4 also exhibited a notably immunogenic, inflammatory kind of signature. Tumors in cluster 7 (3% of samples) were enriched with small nucleolar RNAs — a class of RNAs with unclear biological significance in this context.

Interestingly, tumors in the first 2 angiogenesis-related clusters tended to fall into favorable prognostic risk categories according to the Memorial Sloan Kettering Cancer Center (MSKCC) classification. Patients with these tumors had longer PFS in both treatment arms of the trial, irrespective of treatment.

Notably, the more “proliferative,” cell cycle–focused phenotypes in clusters 4, 5, and 6 occurred more frequently in poor-risk MSKCC groups.

For those in the low-angiogenesis clusters — such as the inflammatory-type cluster 4 and the proliferative cluster 5 — the atezolizumab-bevacizumab combination significantly improved PFS as well as overall response (OR) compared with sunitinib. This underscored the importance of PD-L1 blockade in low-angiogenesis subgroups, as well as the potential of using immune and angiogenesis elements as biomarkers of response to the checkpoint and angiogenesis blockade in advanced RCC patients, the researchers wrote.

These transcriptomic signatures also correlated with somatic gene mutations. For instance, tumors with a pronounced angiogenesis phenotype such as cluster 2, often had mutations in PBRM1 and KDM5C. Meanwhile, clusters 4, 5, and 6 — characterized by increased cell cycle and anabolic metabolism — tended to have more frequent functional depletion of tumor suppressor genes such as CDKN2A/B and TP53. Those mutations were associated with overall worse prognoses but improved PFS with the combination treatment. Also, patients with loss-of-function mutations in ARID1A and/or KMT2C tended to have improved PFS on the combination treatment. Profiling this group of genes could be used to guide therapy selection for patients, the authors wrote.

The findings complement a smaller transcriptome analysis of tumors from patients enrolled in the previous phase 2 trial, but this study provides “another order of magnitude of insight into the biology,” said Brian Rini, MD, Ingram professor of cancer research at Vanderbilt University and chief of clinical trials at the Vanderbilt-Ingram Cancer Center.3 To him, the transcript-based machine learning classification tool won’t be  immediately practice changing — for one, because it would be difficult to use in the clinic — but it does represent a step toward tailoring patients’ treatment closely to the underlying biology of their tumors, he said.

In addition to angiogenesis and inflammation-based features that drive kidney cancer, he said the findings also underscore the importance of metabolic pathways that appear in some of the clusters, and the need to develop drugs that target metabolic pathways to expand the currently limited repertoire of treatments clinicians have to treat advanced RCC. “I think that these metabolic pathways are probably more important than we maybe realized,” he said.

Samuel Peña-Llopis, PhD, junior group leader and coordinator of the School of Oncology in the German Cancer Consortium (DKTK) and German Cancer Research Center (DKFZ) at the University Hospital Essen noted that the team’s transcriptome-based classification tool is insightful for basic and translational research. However, he noted it’s unlikely to be routinely used in clinical settings, in part because it requires

analysis of some 3000 genes, making it much more challenging to validate across hospitals than the simpler PAM50 classification used in breast cancer, which is based on the expression of 50 genes.

Where the study has most clinical value is its characterization of associations of mutations in PBRM1, CDKN2A/B, TP53, ARID1A, and KMT2C withimproved PFS when treated with atezolizumab plus bevacizumab than sunitinib, Dr Peña-Llopis said. “Just sequencing those genes could be informative. And then it’s not going to require this molecular classification.”

The team’s observation that inactivating mutations in PBRM1, BAP1, and CDKN2A/B tend not to overlap with one another is broadly consistent with other research. Dr Peña-Llopis and his colleagues, for instance, have previously proposed a molecular classification for clear cell RCC based on BAP1 and PBRM1 mutations, which are typically mutually exclusive. 4 Inactivating mutations in BAP1, which is currently determined by immunohistochemistry in the clinic, is associated with greater tumor aggressiveness and poor survival, whereas aberrations in PBRM1 are associated with better patient prognoses.

Dr Cieslik added that Dr Rini’s classification isn’t the first time researchers have tried to stratify RCC patients based on their gene-expression profiles, but it’s one of the best powered and is most comprehensive.5 Although he agreed that the clustering tool will likely be more useful in the research realm than in the clinical realm, it’s provided invaluable insight into the role of genetic mutations; this information could be used to augment current clinical risk stratifications. “If we see that CDKM2A loss is [significantly more] common in cluster number 5 than cluster number 1, it clearly tells us something about the role of CDKM2A in driving those hyperproliferative phenotypes in the tumor.”

Dr Cieslik also praised a separate analysis the team undertook of sarcomatoid tumors, which typically correlate with poor prognosis and don’t respond to VEGF inhibition alone. Interestingly, the tumors’ transcriptomes corresponded with a proliferative, nonangiogenesis phenotype, and tended to be enriched for clusters 4, 5, and 6, helping to explain their aggressiveness. But these tumors also exhibited increased PD-L1 expression, shining a light on why they’re often responsive to immunotherapy. That observation is extremely valuable, Dr Cieslik said, in that it strengthens the rationale for using checkpoint inhibitor-based therapy in these patients. “This is, I think, tremendous.”

In the meantime, Dr Rini is working on designing trials based on the clusters to identify the intensity and number of treatments that will be most effective in treating their specific tumor phenotypes. For instance, one could randomize patients whose tumors display an angiogenesis phenotype to investigate whether angiogenesis inhibitors are sufficient, or whether they do better with the addition of immune checkpoint inhibitors. “I think really in the big picture, it’s using the knowledge gained to design the next wave of trials and find the next new group of drug targets.”

Disclosure: Some of the authors of the discussed study disclosed financial ties to the pharmaceutical industry. For a full list of disclosures, please refer to the original article.


  1. Rini B, Powles T, Atkins MB et al. Atezolizumab plus bevacizumab versus sunitinib in patients with previously untreated metastatic renal cell carcinoma (IMmotion151): a multicenter, open-label, phase 3, randomized controlled trial. Lancet. 2019;393(10189). doi:10.1016/S0140-6736(19)30723-8
  2. Motzer RJ, Banchereau R, Hamidi H et al. Molecular subsets in renal cancer determine outcome to checkpoint and angiogenesis blockade. Cancer Cell. 2020;38. doi:10.1016/j.ccell.2020.10.011
  3. McDermott DF, Huseni MA, Atkins MB et al. Clinical activity and molecular correlates of response to atezolizumab alone or in combination with bevacizumab versus sunitinib in renal cell carcinoma. Nat Med. 2018;24. doi:10.1038/s41591-018-0053-3
  4. Peña-Llopis S, Vega-Rubín-de-Celis S, Liao A et al. BAP1 loss defines a new class of renal cell carcinoma. Nat Genet. 2012;44(7). doi:10.1038/ng.2323
  5. D’Costa NM, Cina D, Shrestha R et al. Identification of gene signature for treatment purpose to guide precision oncology in clear-cell renal cell carcinoma. Sci Rep. 2020;10. doi:10.1038/s41598-020-58804-y

This article originally appeared on Cancer Therapy Advisor