The Arrival of Personalized Immunotherapy?

A longstanding debate in the field of tumor immunology has to do with the importance of shared vs. unique antigens in cancer immunity. Should researchers be attempting to target antigens shared among different tumor types or antigens found only in one particular patient’s tumor? A new paper by Cancer Research Institute (CRI) scientists makes a strong case for the latter.

Pramod Srivastava, MD, PhD, a professor of immunology and medicine at the University of Connecticut, and a long-time CRI-funded scientist, has been studying the idiosyncratic genetic fingerprints of tumors for more than 30 years. At a time when most researchers in the field were placing their bets on shared antigens, Srivastava was focused elsewhere—on tumor rejection antigens specific to each tumor. He found that he could immunize mice against a particular cancer by injecting them with protein fragments taken from cells of the same tumor.

In the early 1990s, Srivastava created a new vaccine based on this approach, called Oncophage. It was the first therapeutic cancer vaccine to be approved anywhere in the world. (It was approved in Russia in 2008 for the treatment of early-stage kidney cancer.)

Despite this advance, the main focus of the field continued to be on shared antigens. Beginning in 1991, when Thierry Boon and colleagues isolated the first shared cancer antigen, MAGE, the field placed its highest hopes on using such antigens in off-the-shelf vaccines that could be given to patients who might all share this same cancer-associated tag. Since then, numerous clinical trials have been conducted using such antigens as MAGE, Melan-A, and NY-ESO-1 in therapeutic cancer vaccines. Though useful for generating knowledge of immune responses, the cancer vaccine approach using shared antigens has largely failed to achieve the clinical successes that researchers had hoped for. What Srivastava calls the “denouement” of this approach was the large phase III MAGE vaccine trial conducted by GlaxoSmithKline, which was terminated in 2014 after failing to meet its therapeutic end points.

Into this somewhat cloudy climate has stepped Srivastava once again to promote his alternative approach to making cancer vaccines. In a recent paper published in The Journal of Experimental Medicine, Srivastava and colleagues present convincing evidence that not only can unique tumor antigens protect against cancer, but also that the few specific antigens that confer the most protection can be predicted ahead of time from amidst the large crop of mutations found in tumor cells.

The approach relies on the ability of new genomic technologies such as high-throughput DNA sequencing to quickly and cheaply decipher the entire genetic code of a cancer cell. From this ream of information they are able to identify DNA sequences that differ from those of normal cells, and then use knowledge of how antigens bind to a protein called MHC to predict which tumor mutations are likely to trigger the strongest immune response.

Dr. Srivastava and his team of researchers, including lead author and former CRI-funded graduate fellow Fei Duan, showed that the unique cancer antigens identified using their genomic method, when used in vaccines, could confer tumor immunity to mice in a high percentage of cases.

“Using novel tools reported in this study,” the authors write in their paper, “we are able to identify with high accuracy the small number of neoepitopes (among the vast numbers of potential neoepitopes) that truly elicit immunological protection against tumor growth.”

In other words, Srivastava’s method allows researchers to find an immunogenic needle in the cancer haystack.

Replication Mistakes

To understand how this approach works, one has to understand a bit more about how mutations arise in the first place. Mutations are mistakes in DNA replication. When a cell reproduces, it must make a copy of its DNA to distribute into daughter cells. Each time this DNA replication process occurs, there is a small chance that an error may be made—say one DNA base is swapped with another. When these mutations occur in genes that make proteins that regulate cell proliferation, cancer can result. Cancer cells with damaged proteins have an impaired ability to repair DNA damage, and cancer cells also reproduce more than other cells, which means that mutations tend to accumulate rapidly in cancer cells. Eventually these mutations will render a protein so unlike its normal counterpart that the immune system may recognize it as foreign or dangerous. It is these “foreign” looking cancer proteins researchers want to be able to identify in order to make a vaccine for cancer. But up until now, they had no way of predicting which of the many hundreds of mutated proteins in a cancer cell would be the ones a particular person’s immune system recognized as dangerous.

Asked how he thinks this approach may alter cancer treatment, Srivastava says it will “allow creation of personalized therapeutic cancer vaccines for each patient based on the genetic make-up of that patient’s cancer as it differs from that very patient’s normal genetic material.” In other words, a personalized immunotherapy.

The Remembrance of Things Past

Though Srivastava’s approach goes against the grain of recent thinking, there is good historical precedent for it. Indeed, some of the earliest experiments in tumor immunology support the approach. The first experiments to “put tumor immunology on the map,” so to speak, were those of researchers such as Richmond Prehn, Joan Main, and George Klein, who showed that mice could be immunized against specific tumors. When mouse tumor cells are killed by radiation and injected into a mouse, the animal does not develop cancer when it is subsequently injected with living cells from the same tumor. Importantly, this protection is only conferred against the specific tumor that was used in the vaccination, and not to other tumors. In other words, each tumor seemed to be immunogenetically distinct. To Srivastava, the field unwisely steered away from the tumor-specific approach in the years since these experiments. (Srivastava titled one of his published critiques, “The Necessity of Remembrance of Things Past,” an allusion to Proust.)

Part of the reason for the shift in attention is practical: it was much easier for researchers to search for what tumors have in common than to focus on what makes each tumor different from all other tumors. Until recently, the idea of creating a vaccine based on the unique antigens of every new patient seemed an impossible task to most investigators—there was, after all, no way to tell which antigens among the many hundreds in a tumor cell should be used. As Srivastava noted in a 1993 paper, “The prospect of identifying immunogenic antigens of individual tumors from cancer patients is daunting to the extent of being impractical.” Perhaps, too, the idea of being able to identify a kind of “magic bullet” for cancer was too seductive to give up.

But Srivastava believes the time for changing our thinking has finally arrived. “I do believe that we should stop immunizing patients with shared antigen vaccines,” he notes.

A Minority Opinion

Dr. Srivastava first became interested in unique tumor antigens back when he was a graduate student in India in the 1980s. Although he lacked any background in immunology or cancer immunology, he made an important discovery:  he identified a class of proteins called heat shock proteins that were able to confer immunity to cancer in rodents.

Heat shock proteins (HSPs) are common housekeeping proteins that play a role in assembling and transporting proteins between cell compartments and repairing misfolded proteins in times of stress. They are found in every cell, of nearly every organism on earth. Such ubiquitous proteins might seem an unlikely place to find a cancer treatment, but Srivastava found that these proteins, when injected into rats and mice, were effective at preventing the development of cancer—but only the specific cancer from which the HSPs were obtained.

Spurred on by his discovery, he decided to seek out the world’s foremost authority on tumor immunology, Lloyd J. Old, MD, and pick his brain about the field. Srivastava found Old in his office at Memorial Sloan Kettering Cancer Center one day in the early 1980s. “When I met Old for the first time, he spent about two hours with me, during which he painted for me this large canvas of cancer immunology,” Srivastava recalls. “The uniqueness of antigens, which my study had also shown, stood out for me on that canvas, and I was mesmerized by that idea.”

Srivastava joined Old’s lab at Sloan Kettering as a CRI postdoctoral fellow in 1984, and poured himself into the careful study of unique cancer antigens for the next 30 years. He eventually solved the mystery of how such a common type of protein could mediate tumor immunity. It turns out that HSPs naturally come pre-loaded with bits of essentially every protein in a cell; they therefore serve as very effective carriers of the full panoply of antigens in a cancer cell—including those that distinguish the cancer cell from normal cells.

Srivastava was thrilled with the implications of his finding but, over time, he found himself increasingly at odds with both Dr. Old and the rest of the cancer immunology field on the importance of unique antigens. While Dr. Old early on in his career had been interested in unique antigens, he largely shifted his attention away from these antigens to the shared family of cancer antigens—a move that Srivastava thought was ill-advised.

“I was never shy about pointing this out to him, nor [about] pointing out that in my opinion, he was making a big mistake, as was everyone else in the field.”

With his new paper, Srivastava might just finally convince his fellow cancer immunologists to give unique antigens a second look.

Jedd Wolchok, MD, PhD, a leading cancer immunologist at Memorial Sloan Kettering Cancer Center and director of CRI’s clinical program, agrees that the new research is promoting a change in thinking.  “The work of Duan and Srivastava has revealed the importance of potentially the most biologically relevant antigenic targets,” he says, referring to the unique antigens generated by mutations.

But Wolchok doesn’t go quite as far as Srivastava does in discounting a place for shared antigens in cancer immunotherapy. As evidence, he points to the use of genetically engineered T cells that target shared antigens such as NY-ESO-1, which are generating clinical successes in patients.

Despite their differences of opinion on the subject of cancer antigens, Old and Srivastava remained close friends and colleagues for more than 30 years, even having weekly phone conversations to discuss research. In recent years, before Old’s death in 2011, Srivastava says their thinking had begun to merge again.

“I think that when Jim Allison published his really nice paper in 2008 in Cancer Research doing an in silico analysis of human breast and colon cancers, and showing that such neo-antigens could actually exist, Dr. Old began to see my arguments in a new light.”

Srivastava credits his mentor with having shaped his development as a tumor immunologist and says he will always feel grateful for the influence of Dr. Old. He even dedicated the new scientific paper to him.

An unexpected but fitting end to the story: the paper came online on Lloyd Old’s birthday.

The full citation for the article is:  F. Duan, J. Duitama, S. Al Seesi, C. M. Ayres, S. A. Corcelli, A. P. Pawashe, T. Blanchard, D. McMahon, J. Sidney, A. Sette, B. M. Baker, I. I. Mandoiu, P. K. Srivastava. Genomic and bioinformatic profiling of mutational neoepitopes reveals new rules to predict anticancer immunogenicity. J Exp Med. Published September 22, 2014, doi:10.1084/jem.20141308

This research was supported by a grant from the Cancer Research Institute to Pramod Srivastava.