Addressing Research Into the Association Between Tumor Mutations and Responses in RCC

Eliezer Van Allen, MD

In recent years, investigators have aimed to best understand how tumor and immune microenvironmental cells interact in the context of immune checkpoint blockade, how somatic and germline interactions contribute to cancer development, how tumors and immune systems interact during disease progression, and how therapy resistance occurs, as well as how to best bring these insights into the clinic, according to Eliezer Van Allen, MD. He added that this question sparked from ongoing research aiming to understand and predict selective responses, acquired resistance, and intrinsic resistance.¹

Van Allen, who is the chief of the Division of Population Sciences, as well as a physician at Dana-Farber Cancer Institute; and an associate professor in medicine at Harvard Medical School, both located in Boston, Massachusetts, discussed immuno-oncology in relation to precision cancer medicine at The New York Academy of Sciences Frontiers in Cancer Immunotherapy Symposium 2024.

Early Research into Mutations and Tumor Responses

Van Allen explained that continued investigations contributed to early research identifying factors such as tumor mutational burden and other intrinsic molecular features that affect tumor responses to immune checkpoint blockade, such as neoantigens. He added that ongoing work continues to evaluate all the molecular features that contribute to selective response and resistance to immune checkpoint blockade in tumor cells and their microenvironments.

To best answer ongoing research questions, investigators specifically evaluated patients with kidney cancer, aiming to understand why some patients respond to immunotherapy and others do not. For example, investigators used results from the phase 2 IMmotion150 study (NCT01984242) to evaluate the clinical activity and molecular correlates of response to the PD-L1 inhibitor atezolizumab (Tecentriq) alone or in combination with the VEGF inhibitor bevacizumab (Avastin) vs sunitinib (Sutent) in 305 patients with renal cell carcinoma (RCC).²

Findings derived from this exploratory biomarker analyses revealed that tumor mutation and neoantigen burden were not associated with progression-free survival (PFS). Angiogenesis, T-effector/IFN-γ response, and myeloid inflammatory gene expression signatures were also strongly and differentially associated with PFS within and across treatment arms. Furthermore, these molecular profiles indicate that prediction of outcomes with anti-VEGF and immunotherapy agents may overcome resistance to immune checkpoint blockade.

“Loss of function in the somatic mutations in a gene called PBRM1 were associated with response to immunotherapy,” Van Allen explained when highlighting additional research in this patient population. “The response was not subtle... We noticed in a post-hoc analysis that we did as part of this initial wave of research that the signals seemed to be coming in aggregate when we combined all the datasets that we had at the time from patients who had already received VEGF TKIs…and then subsequently went on to receive checkpoint blockade, so they were getting it in a very specific clinical context. That’s where the signals seemed to be coming from.”

Seeking to validate this research, investigators led a confirmatory analysis of the phase 3 CheckMate 025 trial (NCT01668784) and examined whether mutations in PBRM1 correlated with selected responses in an independent, randomized study. Notably, CheckMate 025 evaluated treatment with nivolumab (Opdivo) vs everolimus (Afinitor) in patients with pretreated advanced or metastatic clear cell RCC.³

It was found that PBRM1-mutated tumors were more likely to respond to checkpoint blockade, a signal that was observed consistently; this signal was not present in the control group of patients who received everolimus. PBRM1 mutations were associated with prolonged PFS (HR, 0.67; 95% CI, 0.47-0.96; P = .03) and overall survival (HR, 0.65; 95% CI, 0.44-0.96; P = .03).^1,3

Van Allen said that this observation was made within the context of patients initially treated with VEGF TKIs who later received checkpoint blockade.¹ However, he noted that it’s now widely acknowledged that this signal doesn’t replicate when the clinical context varies, such as in patients treated upfront with checkpoint blockade alone or in combination with VEGF TKIs.

This discrepancy has been confusing, Van Allen explained, but subsequent research indicates that the signal is linked to HIF-dependent endogenous retroviruses (ERVs). Therefore, mutations in PBRM1 only contribute to a similar response when the HIF dependency and VEGF TKIs are removed, revealing the involvement of ERVs.

Understanding Available Precision Medicine Treatment Options

The many options for multimodal data for enhanced interference create an abundant precision medicine information landscape, which is made up of spatial multiplexed assays, single-cell omics, bulk RNA sequencing, bulk whole-exome/whole-genome sequencing, immunohistochemistry-stained slides, panel/targeted sequencing, and hematoxylin and eosin–stained slides. When considering how to use each unique imaging or sequencing approach, it may be possible to simultaneously measure prevalent features regarding tumor immune interactions in the context of immune checkpoint blockade, properties of tumor cells, agents’ interactions with each other, and emerging features that may be relevant for predicting or understanding which tumors are more likely to respond.

Van Allen explained that this here is where it is possible broaden perspectives in cancer care. This serves as just one instance of artificial intelligence (AI) technology that could prove beneficial in addressing questions within immuno-oncology. This specific disease area still has room for enhancement, as evidenced by continued research. Ongoing focuses will lie heavily on biological interpretability and models, emphasizing the overarching goal to use these tools to aid in the understanding of underlying reasons behind tumor responses. Investigators have employed similar approaches successfully in other contexts, including genomics and imaging.

To further evaluate the use of AI technology in relation to immuno-oncology for precision cancer medicine, investigators evaluated the integration of biology and machine learning for molecular learning. The findings from this study revealed unmet needs regarding the future use of AI in precision medicine.

“We published [a study] a few years ago [investigating] the biologically informed neural network; it takes the concept of these networks…and actually forces biological signaling into the network, which helps to inform interpretability after postdoc interpretability... We’ve also applied it to immunotherapy... It did not work. It doesn’t stratify very well,” Van Allen reported.

The Future of Precision Medicine in Relation to Ongoing Research

Taking into consideration both the strides and setbacks that this investigational landscape has undergone in recent years, Van Allen asked: “What can we be doing as a community to answer these [remaining] questions more effectively?”

One approach, he noted, is to incorporate more features that reflect the breadth of interests in cancer biology and immunology. This can involve both coding and noncoding mutations, as demonstrated in previous research focusing on somatic noncoding discovery. Additionally, it may be beneficial to leverage single-cell transcriptomes of tumor cells and the surrounding microenvironment to enhance oncologists’ abilities to predict immune responses and outcomes more accurately.

However, it’s critical to view these efforts through a current and forward-looking lens and to consider more advanced techniques that can be integrated into existing systems to reassess these questions with new perspectives. This may entail exploring and adopting different and new approaches, as well as incorporating graph-based approaches, similar to those used in single-cell imaging, into genomic analyses.

Furthermore, recent discussions have highlighted the potential utility of foundational models. The use of these models in image-based work has been somewhat limited; however, incorporating both unique foundational models and single-cell foundations may be indicative of noteworthy results.

“Although this is all fun and interesting for biological discovery and biomarker developments, I am a clinician first. The benchmark is to impact patient care. Therefore, we want to take these kinds of models and algorithms and bring them into clinical decision-making. That’s what we’re trying to do now,” Van Allen concluded.

Disclosures: Van Allen discloses the following: Advisory/consulting roles with Tango Therapeutics, Genome Medical, Genomic Life, Enara Bio, Janssen, Manifold Bio, Monte Rosa, Novartis Institute for Biomedical Research, Riva Therapeutics, and Serinus Bio; receipt of research support from Novartis, BMS, and Sanofi; equity with Tango Therapeutics, Genome Medical, Genomic Life, Syapse, Enara Bio, Manifold Bio, Microsoft, Monte Rosa, Riva Therapeutics, and Serinus Bio; institutional patents filed on chromatin mutations and immunotherapy response, and methods for clinical interpretation; intermittent legal consulting on patents for Foaley & Hoag; and Editorial Board participation with Science Advances.

References

1. Van Allen E. Computational immuno-oncology for precision cancer medicine. Presented at: Frontiers in Cancer Immunotherapy 2024; May 21-21, 2024; New York, NY.
2. 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(6):749-757. doi:10.1038/s41591-018-0053-3
3. Braun DA, Ishii Y, Walsh AM, et al. Clinical validation of PBRM1 alterations as a marker of immune checkpoint inhibitor response in renal cell carcinoma. JAMA Oncol. 2019;5(11):1631-1633. doi:10.1001/jamaoncol.2019.3158