GU Cancers 2019 | TGF-β/EMT gene signature in low-responders to pembrolizumab

Petros Grivas

Identifying patients who are likely to respond to a specific therapy by biomarkers is useful when it comes to treatment selection. Petros Grivas, MD, PhD, of the Seattle Cancer Care Alliance, Seattle, WA, reports that tumor tissues from urothelial cancer patients treated on the KEYNOTE-052 trial (NCT02335424) were looked at to identify the gene signatures associated with resistance to pembrolizumab. Speaking at the 2019 Genitourinary Cancers Symposium, held in San Francisco, CA, Dr Grivas explains that a higher stromal EMT/ TGF-β signature was found in non-responders.

Transcript (edited for clarity):

It’s a very interesting question that we have in the field of urothelial cancer: Which patients respond better to checkpoint inhibition? For example, we have the KEYNOTE-052 trial that is evaluating the activity and safety of pembrolizumab, a single agent, in patients who are cisplatin unfit. Cisplatin ineligible, in the first line, chemotherapy naïve setting. Patients have not received chemotherapy for advanced disease, but they’re unfit for cisplatin, and received pembrolizumab at a flat dose, and overall this trial showed an overall response rate of 29%. That actually led to the FDA and EMA approval of pembrolizumab in this first line space of advanced urothelial cancer. Of course, the question is, who are the patients who are responding better, and how do we develop, discover, and evaluate biomarkers that can help us identify patients who have a higher chance to respond to those checkpoint inhibitors?

In that context, we went back and we evaluated the tumor tissue from patients who enrolled in the KEYNOTE-052 trial, and we tried to evaluate it’s signature that is present on the immunosuppressive environment. The microenvironment of the cancer, especially in the stroma. And we called this signature TGF-beta signature, because it was mainly mediated by TGF-beta pathway. And also, EMT, epithelial mesenchymal transition, because we think that this gene expression signature related to EMT might have immunosuppressive properties. So this TGF-beta EMT, gene signature in the tumor microenvironment, mainly in the stroma around the tumor, might suppress the immune system and might correlate with lower responses to checkpoint inhibition. We did this, that biomarker study, and we saw indeed that patients who had tumors that had over expression of this immunosuppressive signature, that high EMT TGF-beta signature, they had lower response to pembrolizumab as consistent with our hypothesis.

We did one step further, and we also combined this suppressive signature with a different signature that is more immunostimulatory. It’s an 18 gene signature that’s related to the interferon-gamma pathways. And the higher this 18 gene signature score is, the higher the chance of the immune system to be stimulated. The opposite to before. And that’s higher chance of response to checkpoint inhibition. And when we combine these two signatures, we call it the GP signature, the stimulatory. And the EMT TGF-beta signature, we’re able to separate patients to those who have extremely high response. If they have GP high, and TGF-beta EMT low, and vice versa. So had a very interesting data, trying to tease out, separate responders versus non-responders.

Of course, this data are hypothesis generating at this point. Are not conclusive. But definitely are very interesting and hopefully can be validated in other clinical trials and help us ideally define biomarkers that can help us select patients in the future for checkpoint inhibition. However, this data, despite how promising they are, are not practice changing yet. They are mostly hypothesis generating. And we will work to generate the manuscript and, so we can disseminate the knowledge in the community.

Share this video