Skip to Main Content

The Nobel Prize-winning discovery of immune checkpoint inhibitors has changed how cancer is treated. These drugs “unblock” the immune system’s normally protective pathways that prevent T cells from overreacting and potentially harming healthy cells. Immune checkpoint inhibitors work by “uninhibiting” a cancer patient’s T cells to attack his or her tumor.

While successful checkpoint therapy indicates that an individual’s immune system can control tumor growth as though the tumor is a viral infection, not everyone’s immune system seems to be up to the task: Among patients eligible for immunotherapy, only about 1 in 7 respond to it. A new tumor-scoring algorithm based on the unique genomic characteristics of an individual’s cancer, recently described in STAT, has been proposed to reduce the number of treatment failures by predicting the outcome of checkpoint therapy.

While that is a promising and welcome approach, there is more to the story that warrants a closer look.


The groundbreaking observation that women whose breast cancers had upregulated hormone receptors were likely to benefit from receptor-specific inhibitors launched a series of tremendous therapeutic successes in targeted cancer therapy. The road to developing specific and effective immunotherapies, however, has been a tougher journey than we expected. It would be ideal to have one or more biomarkers that precisely reflect an individual’s immune capacity and so predict his or her response to checkpoint inhibitor drugs, but so far, the predictive potential has been disappointing. One example is the inadequacy of the PD-L1 biomarker to clearly identify responders to PD-1 blockers, a type of checkpoint inhibitor immunotherapy currently being used.

We should be cautious about the ever-expanding effort to characterize “hot” and “inflamed” tumors, meaning those that are likely to provoke strong response by the immune system. This approach has not provided greater prediction accuracy, and the mere presence of high numbers of mutations has also failed to be an accurate predictor of treatment response. Although the recently reported predictive computer algorithm is an improvement over current markers, as we have seen before, early data often looks promising while greater experience may weaken their predictive value.


Why is an accurate predictor of the impact of a checkpoint inhibitor so hard to find? Despite intense investigations at all levels of the biomedical spectrum from basic to clinical research, we have yet to fully grasp the extraordinary complexity of the human immune system: how it recognizes tumor targets, how tumors interact with and influence patients’ immune cells, and how a patient’s treatment history may “rewrite the script” and change the rules his or her immune system follows.

In addition to focusing on the tumor and using computer algorithms to predict targets, another informed way to identify individuals for whom immunotherapy will be successful includes careful interrogation of each patient’s immune system. This task involves mapping all the mutations in an individual’s tumor and identifying precisely how well his or her T cells are equipped to recognize and kill each of those targets.

Our work and that of others to profile how patient’s T cells respond against their own tumor mutations reveals unappreciated biological complexity. In some cases, an individual’s immune system can’t see some tumor targets, or doesn’t appear to see them, and thus doesn’t know what to attack. In other cases, individuals have a greater proportion of tumor-specific “inhibigens” that suppress an individual’s immune response — effectively accelerating tumor growth rather than slowing it. Our data show that the unique ratio of an individual’s suppressive inhibigens to his or her stimulatory T cell responses may be a strong predictor of whether or not immunotherapy will work for that specific patient.

Every increment of progress is important, but our understanding of the massively complex human immune system remains incomplete. Any predictive models we use are limited to the model’s assumptions and input data — which are complicated by the fact that each patient’s tumor is unique, as is their ability to marshal an effective T cell force to attack that tumor.

There is so much more to learn about immune responses before we can recreate and predict them, and then turn that knowledge into precise and effective therapies that work as prescribers and patients expect them to.

Thomas Davis, M.D., is the chief medical officer of Genocea, a cancer-focused biotech company headquartered in Cambridge, Mass. Jessica Flechtner, Ph.D., is the company’s chief scientific officer.

Comments are closed.