Institute data scientists identify who may be at risk from revolutionary cancer treatment

January 13, 2022

By Christopher Vaughan

A class of cancer drugs called immune checkpoint inhibitors have been a boon to cancer patients, often leading to remission in cancers that were previously almost universally fatal. Unfortunately, many patients undergoing this treatment will also suffer severe immune-related toxicities, and there has been no reliable way to tell in advance which patients are at risk. Institute researchers have now used data science techniques to draw a picture of which patients are likely to experience this dangerous side effect.

“Checkpoint inhibition has revolutionized oncology, but unfortunately many patients given the strongest forms of this cancer therapy will experience severe side effects,” said Aaron Newman, PhD, a member of the Stanford Institute for Stem Cell Biology and Regenerative Medicine. “We wanted to know if we could use data science to determine in advance who will be vulnerable to these toxicities.”

Newman is the senior author on a paper describing this research, published January 13, 2022 in the journal Nature Medicine. Newman is an assistant professor in the Department of Biomedical Data Sciences and a member of the Institute for Stem Cell Biology and Regenerative Medicine. He is also a member of the Stanford Ludwig Center and a Chan Zuckerberg Biohub Investigator.

When immune cells are doing their jobs correctly, they are eliminating infected, precancerous or damaged cells. But immune cells also have the capacity to attack the body’s own tissues, behavior that results in various autoimmune diseases. In order to keep such autoimmune attacks under control, immune cells use “checkpoint” genes that rein in immune cell activity. Unfortunately, cancer can activate this mechanism to suppress immune attacks on cancer cells.

Therapies that reverse that process and inhibit those checkpoints have saved or prolonged the lives of many cancer patients who previously had very low odds of survival. But about 60 percent of patients with melanoma who undergo the strongest forms of checkpoint inhibition therapy have severe side effects, forcing many of these patients to stop treatment. These side effects can affect any tissue in the body, and they can even be lethal. “In many cases, the cancer may end up advancing more rapidly not because it can’t be treated, but because patients cannot maintain regular treatment with immunotherapy,” Newman said.

Adding to the difficulty was the fact that there was no way to tell who would experience this toxic side effect and who wouldn’t. Newman and researchers from Stanford, Washington University in St. Louis, and Yale University set out to see if they could find a common pattern among patients who were treated and subsequently experienced severe toxicities. “We wanted to know, ‘How can we improve this immunotherapy so that we maximize the benefits and minimize the risks?’” Newman said.

The group, including co-first author and former Stanford PhD student Alexander Lozano, analyzed blood samples from over seventy skin cancer patients who had undergone the therapy in the past and used clinical records to understand what side effects these patients experienced. They used a wide variety of analytical techniques such as single-cell RNA sequencing, mass cytometry, bulk RNA sequencing, and CIBERSORTx, a technique developed in Newman’s lab to understand the cellular components of large numbers of bulk samples. 

The group wanted to see if these analytical techniques gave an answer that converged on a suspected culprit in these toxicities. And behold, they found two. “We found that patients who before treatment had a high number of what are called activated CD4 effector memory T cells, and also had a great deal of sequence diversity in the receptors displayed by T cells that recognize antigens, were much more likely to have these severe immune toxicities after checkpoint inhibition therapy,” Lozano said. 

“Our working hypothesis is that patients who have this profile—high numbers of CD4 memory T cells and high T cell sequence diversity—have either a tendency toward developing autoimmune diseases or have subclinical autoimmune attacks on their tissue,” Newman said. Since checkpoint genes are there precisely to put the brakes on immune cells so that they don’t attack our own tissues, it makes sense that a therapy directed at suppressing these checkpoint mechanisms would increase the likelihood of our immune cells mounting an attack on our own, healthy cells, he added. 

To further explore this idea, the researchers did the same sort of analysis on people who did not have cancer but did have different autoimmune disorders, including lupus and inflammatory bowel disease. They found that the same pattern of high CD4 memory T cells in the peripheral blood was also strongly associated with these autoimmune diseases. 

For cancer patients undergoing checkpoint inhibition therapy, the results could mean that clinicians will be better able to pick a therapy that works but does not produce severe immune toxicities. “They might decide not to use the strongest immunotherapy, or they could more closely monitor patients likely to experience severe adverse effects,” Newman says. “This is another example of how data science can lead to “precision medicine”—therapies and disease management that are customized for individual patients to provide the best possible outcomes.”