Whole-exome sequencing may improve prediction of cancer immunotherapy response

Researchers at New York University, Weill Cornell Medicine and New York Genome Center have developed and tested a two-step approach that uses whole-exome sequencing to find genes and pathways that predict response to immunotherapy.

“Before prescribing an expensive therapy that may have side effects, we want to be able to predict who is going to benefit from it,” Neville E. Sanjana, PhD, assistant professor of biology at NYU, assistant professor of neuroscience and physiology at NYU Grossman School of Medicine, a core faculty member at New York Genome Center and co-author of the study, published in Nature Communications, said in an interview with Healio. “That’s the first-order question. The deeper question — which is not answered here, but could be addressed in future work — would be, can we then predict which specific drug would be the right regimen?”

Focusing on more genes

Several biomarkers — such as age, tumor type and tumor mutational burden — are known to be associated with responses to immunotherapy. In particular, tumor mutational burden, which is calculated by assessing the number of mutations across the genome but without attention to the specific locations of the mutations, is a recognized predictor frequently used to determine if a patient is a suitable candidate for immune checkpoint inhibitors.

The researchers sought to determine whether evaluating mutations in specific genes or collections of genes known as pathways could improve the ability to predict which patients will respond to immunotherapy. Whole-exome sequencing analyzes the part of the genome that codes for proteins to look for mutations that may be involved in the disease. This approach sequences approximately 20,000 genes, or 2% of the entire human genome.

Sanjana and colleagues combined data from six prior immunotherapy studies that included patients with melanoma, lung cancer, bladder cancer, and head and neck cancer. All participants underwent whole-exome sequencing of their tumors and received treatment with an immune checkpoint inhibitor (either anti-PD-1 or anti-CTLA-4).

However, even after bringing together all of these study populations, the investigators yielded only 319 patients.

“These small studies had difficulty identifying genes that were predictive — one reason is that there are 20,000 genes in the genome,” Sanjana said. “So, even in tumors where there are mutations, that’s a lot more genes than mutations when you only have a few hundred patients. It’s very hard to achieve robustness and statistical significance when you have 20,000 competing genetic hypotheses and so few patients to look at.”

The researchers circumvented this issue by repurposing a model called fishHook, which differentiates mutations that drive cancer from background mutations that occur randomly and are not involved in cancer. The fishHook model was developed by Marcin Imieliski, MD, PhD, associate professor at Weill Cornell Medicine and New York Genome Center and co-author on the current study. The model adjusts for a range of factors that may impact the rates of background mutations. For example, larger genes tend to have an increased likelihood of accumulating mutations because of their size. In addition to gene size, fishHook uses thousands of other covariates, such as sequence context, epigenetic modifications and genome replication. Each of these can impact the number of mutations that occur in a specific gene.

Responders vs. nonresponders

With the help of this model, the researchers deployed a two-step approach. The first step involved looking at the sequencing from all patients to identify genes with a higher mutational burden than expected. From this step, they identified six genes with unusually high mutational burdens — regardless of whether they were a responder or nonresponder.

“In step two, we took the genes that were mutated above background levels in all patient tumors and then separated the patients and tumors into responders and nonresponders,” Sanjana said. “Now, instead of testing 20,000 genes — the number of genes in the genome — we tested just this small group of genes. We wanted to see which of these genes enrich for mutations either in responders or in nonresponders.”

Sanjana and colleagues found two of the genes — KRAS, which is often mutated in lung cancer, and BRAF, the most commonly mutated gene in melanoma —enriched in patients who responded to immunotherapy. Conversely, they found two other genes — TP53 and BCLAF1 —enriched in those who didn’t respond to immunotherapy.

‘A decades-long promise’

The researchers applied the same two-step approach on collections of genes called pathways. Through this, they determined that MAPK signaling, p53-associated and immunomodulatory pathways also predicted immune checkpoint inhibitor response.

Next, the researchers combined the four genes and three pathways with other predictive variables such as age, tumor type and tumor mutational burden to create a tool called the Cancer Immunotherapy Response CLassifiEr (CIRCLE). CIRCLE demonstrated an ability to predict immunotherapy response 11% better than tumor mutational burden alone. CIRCLE also effectively predicted cancer survival after immunotherapy.

“This is less about the specific genes and pathways, although those are indeed interesting,” Sanjana said. “It’s more about the approach. What I think is going to be exciting is taking the same approach and applying it not to 500 patients, but 5,000 patients or 50,000 patients. Discoveries from larger cohorts are going to be powerful and will help us really pinpoint which patients are most likely to respond to immunotherapy.”

Sanjana and colleagues validated the CIRCLE approach using data from 165 additional patients with cancer who underwent treatment with immunotherapy. They found that CIRCLE yielded predictive data beyond that acquired from tumor mutational burden alone.

Research in the future will involve testing CIRCLE on larger cohorts of patient data, with the anticipation that the model will improve with data from thousands rather than hundreds of patients. With these larger cohorts, the researchers also hope to be able to analyze which patients will likely respond to different specific immunotherapies.

“This two-step approach is powerful and has not been applied in the context of immunotherapy previously,” Sanjana said. “Although precision medicine has been a decades-long promise in cancer genetics, I think the field is now starting to deliver on it, albeit slowly. CIRCLE is one piece within the much larger puzzle to find the right therapy for each tumor.”

For more information:

Neville E. Sanjana, PhD, can be reached at nsanjana@nygenome.org.