Early changes in the tumor environment may explain why immunotherapy works for some patients but not others

The efficacy of immunotherapy can vary among patients, with some experiencing long-term remission while others see no improvement. Despite efforts to predict patient response, oncologists currently lack a reliable method to distinguish these outcomes before treatment begins. A recent study led by Professor Dvir Aran from the Technion and the Henry and Marilyn Taub Faculty of Computer Science, in collaboration with Dr. Zhongyang Lin and Professor Jürgen C. Becker of the German Cancer Consortium, sheds light on this conundrum.

Their research, published in Cancer Cell, suggests that the changes occurring in the tumor microenvironment during the initial weeks of immunotherapy therapy may hold the key to understanding why some patients respond favorably while others do not. The tumor microenvironment, a complex ecosystem of immune cells, blood vessels, and structural cells surrounding the tumor, can either bolster or hinder the immune system's attack on the tumor.

While previous studies have examined this environment at a single point before treatment, the new research focuses on the early stages of therapy, when the immune system and tumor first interact. Professor Aran compares this to a chess game, where predicting the outcome requires observing how players respond to each other's initial moves.

The German Cancer Consortium emphasizes that these early interactions not only explain why patients respond or fail to respond to treatment but could also enable clinicians to anticipate the immune response's course early on. However, collecting sufficient single-cell data from patients is challenging, as current methods are expensive and labor-intensive, typically allowing for only 10 to 20 patients to be studied.

To overcome this limitation, the research team generated fresh single-cell data from melanoma patients sampled in the early days of treatment and compiled scattered datasets from various sources. By integrating 16 independent cohorts, including nearly 200 patients from multiple cancer types, the team amassed one of the largest data sets of its kind.

Using advanced computational and AI-based methods, they identified four recurring states of the tumor microenvironment that appear across different cancers. One state consists of immune-rich, inflamed environments with active immune engagement, while the other features suppressive myeloid cells or a lack of immune activity. The real discovery lies in how these states change during treatment.

In approximately half of the patients, the tumor microenvironment remains stable. In the other half, it shifts from one state to another, and the direction of this shift is crucial. Tumors that move towards inflamed, immune-rich states tend to have strong responses to therapy, while those that drift towards suppressive states are more likely to experience treatment failure.

In essence, the key lies not just in where the tumor starts but in how it can evolve. The researchers developed a "transition score" using pretreatment samples that estimates a tumor's potential to shift towards a favorable immune state once therapy begins, capturing its capacity to change rather than its initial state. This score, tested on a larger dataset of approximately 1,300 patients, accurately predicted responses, matching established biomarkers while revealing a distinct signal - the tumor's potential to evolve.

While the findings are not yet ready for clinical application, they highlight a paradigm shift in how researchers and clinicians approach cancer immunotherapy. Rather than categorizing tumors into fixed "hot" or "cold" labels, the study suggests monitoring and potentially steering the tumor's evolution. Instead of merely determining whether a tumor is favorable or unfavorable, the research emphasizes understanding and influencing how tumors respond in real-time to treatment.

The next step involves developing therapies that nudge this evolution in a favorable direction, potentially improving immunotherapy outcomes by tailoring them to individual tumor responses.