Unraveling Immune Biomarkers in Endometrial Cancer Therapy

Guest expert Sacha Gnjatic, PhD, discusses groundbreaking research from a national immunotherapy network aimed at identifying biomarkers that predict response to cancer treatments, with a focus on endometrial cancer and novel drug combinations.

Sacha Gnjatic, PhD: My name is Sacha Gnjatic. [I’ve been] a professor at the Icahn School of Medicine at Mount Sinai since 2013. I'm in the Department of Immunology there, but I'm also the co-director of the Human Immune Monitoring Center, which is a type of facility that uses samples from human blood or tumors to try to understand immune correlates. Before joining Mount Sinai, where I've been for over 10 years, I was at Memorial Sloan Kettering and the Ludwig Institute for Cancer Research. My mentor there was Dr Lloyd Old who was one of the big names in tumor immunology. Prior to that, I did my PhD in France at the Cauchy Institute, focusing on tumor immunology. I moved as a postdoc and stayed in the US.

Please share an overview of this research study about immunological biomarkers and what you were hoping to find. 

Dr Gnjatic: I’d like to first give context about why we came to analyze this study. I'm a tumor immunologist by training. My PhD is in immunology, but I've always worked with cancer patient samples and human samples. The goal of most of my research has been to understand how the immune system interacts with tumors in an antigen-specific manner. I spent a lot of the previous years trying to understand how T cells and antibodies recognize proteins that are specifically expressed in tumors, specifically a family of antigens called cancer testis antigens. This brought me to have some expertise in trying to understand how the immune system is modulated either spontaneously by the presence of a tumor or because of interventions such as immunotherapy. 

One thing that became clear is that we need markers that can predict which patients will respond to these immunotherapies, versus [those] who may not respond at all or who may respond temporarily and then progress through the treatment. Many labs are interested in these questions because it would be much easier to not treat patients if we knew the treatment may not work. But the problem is that all these efforts have been in individual labs and not coordinated. 

What is happening now is that the National Cancer Institute (NCI), through the Cancer Moonshot Program, awarded a grant for U24 Cancer Immune Monitoring and Analysis Centers (CMAC). The goal was to take 4 institutions within the US and work together to identify biomarkers of response in a way that can be compared across the different sites. That way when we have a finding in one study, it could potentially be applicable more broadly. 

My colleague, Dr Kim-Schulze, principal investigator (PI) at the Mount Sinai Center, and I are part of this network coordinated by the NCI. We also collaborate with MD Anderson, Dana-Farber Cancer Institute, and Stanford University. As a network, we first compared what type of assays we can perform on these samples and then harmonize them. We spend many years sharing specimens across the different labs and making sure that we get similar results with the same specimens. That way, we can then say that wherever the assay is done, the results are comparable. 

With the NCI, we then selected a series of different immunotherapy trials that are innovative because they have new drugs and combinations. Having a robust specimen collection plan at baseline, during treatment, and after treatment is important so that we can understand how treatments work. We test tumors and blood before and after treatments. We look for different, high-dimensional methodologies for plasma, blood, and tumors that range from genomic and proteomics to even microbiome analysis. The idea is to then identify biomarkers that can be associated with response to treatment. 

Thus far in the field, there are not that many biomarkers that have been defined for response to immunotherapy. Mostly, it's PD-L1 expression for treatments that target the PD-1 pathway, as well as tumor mutation burden. But these are insufficient. Alone, they are not good enough to select which patients should undergo immunotherapy or not, so we are trying not to define novel markers that are based on the immune system's response. 

To go back to your question, one of those trials was studying endometrial cancer, which is one of the most common and lethal gynecological cancers in the US. In a study conducted by PI Stéphanie Leureux at the Princess Margaret Cancer Centre, they studied a novel combination of PD-1 agent nivolumab and cabozantinib—which is a well-known anti-angiogenic small molecule. We chose this study is because it had a really interesting design. The study was randomized to test nivolumab alone versus nivolumab plus cabozantinib. The study also had a third arm that addressed patients previously treated with immunotherapy to see if this new combination could potentially be better. Right now, the new standard of care for endometrial cancer is immunotherapy. This allows us to look at potential things we can do when a patient has already received immunotherapy without results. 

The clinical data for this study showed an improvement in progression-free survival for the patients who received the combination treatment versus the patients who received nivolumab alone. It is quite interesting to understand the mechanisms that are specific to the combination compared to the single agent and if it is possible that this could become the new standard of care in the future. We selected this study because we could get specimens from blood, as I mentioned before, during and after treatment, as well as the tumor at baseline. In some cases, the patients who progressed with single-agent nivolumab were offered the possibility to cross over into the combination and, at that time, get another biopsy. 

This particular paper focuses on blood samples. We're currently working on the tumor samples as well for another paper based on these tumor samples, but here we looked at blood biomarkers because it is much easier to detect something from blood because it’s easier to sample than tumors. We still considered the genomic information of the tumors and did whole-exome sequencing to understand the genomic subtypes.  

What are some of the key findings? How do these findings enhance our current understanding of treatment options? 

Dr Gnjatic: As I mentioned, some of the patients who received the combination did better than patients who received nivolumab alone. Our interest was to try to define other markers of this combination that could further categorize patients by those who responded versus those who still progressed even through the combination treatment. As part of the network I described earlier, we have defined a series of assays that we perform on all the samples. This includes proteomics with the Olink platform, which examines cytokines and other inflammatory markers from blood, as well as immuno-oncology-related proteins. This is circulating plasma, as well as the composition of the blood lymphocytes, and by time-of-flight (CyTOF) or flow-type mass cytometry. 

Finally, we looked at antibodies to known tumor antigens also in circulation in an assay we call grand serology. We're looking at about 20 known tumor antigens to see if they are spontaneously immunogenic in these patients—meaning we can detect antibodies against them. All this was done in the context of the knowledge of the tumor, which has a different genomic background as well. 

When we applied these assays longitudinally across samples at baseline, during and after treatment, we saw markers that were identified and associated with specific treatment arms. For example, some markers only come up in the cabozantinib arm, and others go down. We saw that VGFR2 was reduced for the cabozantinib arm, which is interesting because cabozantinib targets VGFR2. This was like a control that it had an on-target effect here. Interestingly, the directionality was down. IL-12 also reduced while other markers, like H0-1, went up with cabozantinib treatment. Some markers increased with either treatment and were probably more attributed to the nivolumab effect, which is in both arms. This includes PD-1 increases, as well as things like interferon-gamma and CXCL9. That allows us first to understand the differential mechanisms between these 2 treatments and see that some markers are specifically associated with certain treatment arms. But the overarching question is what's associated with these responses? Response is defined as complete or partial and stable disease versus progressive disease. 

We conducted a standard assessment in these patients and also investigated what is associated with overall survival and progression-free survival. Overall survival was newly reported data for this paper, it wasn't reported in the previous study. We found that we had some strong markers that were associated with both patients who progressed more quickly, as well as those patients who had shorter progression-free survival and overall survival. These 2 markers were CCL23 and CSF1. Both are macrophage-related cytokines. They're macrophage chemoattractants. If a patient had high markers for those, they had a lower chance of survival. If they had low markers for this, they would do better. This was specifically found in the combination arm only, not in the single-agent arm. There is something in the combination that works better if the macrophage-related markers are low. 

This opens up potential for either selecting patients upfront based on these types of markers to see if they may have a bit more benefit. In future studies, maybe this is something that should be considered prospectively to be tested. This also potentially opens avenues of novel treatments that could be combined and could target macrophages. 

At this point, we don't yet have confirmation that the macrophages in the tumor are related to what we detected in circulation with these cytokines, but we assume that maybe they're representing some of these markers. We're currently working on the tissue samples that we obtained from this study to also confirm the hypothesis that maybe having macrophages could be circumventing the effect of therapy, and having a lower overall circulating level helps with the combination treatment. 

The other marker that we found associated with survival was an antibody to a known tumor antigen called NY-ESO-1, which I've studied for a long time in the past because it's one of the most spontaneously immunogenic proteins in tumors. When it's expressed, about half of patients will eventually develop an antibody response, which is usually also associated with a T cell response. We saw that patients at baseline who already had these elevated antibodies ended up doing better than those who didn't have these antibodies. 

One more that I didn't mention yet are markers associated with T cell activation in blood, such as ICOS ligand and some other T cell markers that are elevated in patients who did better. It draws a picture that patients who benefited from the combination treatment probably already had some type of pre-existing immune signature that shows tumor recognition in some ways. These may be the patients who can derive the most benefit from these therapies compared to patients who don't have those. I think this is potentially an interesting way to select future trials. 

Is there anything else you’d like to share with our audience about the future of this research? 

Dr Gnjatic: As I mentioned, we are doing a follow-up study using tumor samples because, in the end, that is where the proof is. I'd love to be able to define patients who are predisposed to respond at baseline from only from the blood peripheral markers, but I think we need to further justify what is happening at the tumor site to confirm this hypothesis. 

I also want to highlight the people who worked the most on this study: Vladimir Roudko, Diane Marie Del Valle, Seunghee Kim-Schulze, and Edgar Gonzalez-Kozlova. They are the ones who helped coordinate the transfer of the samples and the cleaning of the data. They especially helped with the generation and analysis of the primary data. 

We're using some innovative ways to analyze this association with response using these mixed-effect models. Even though this is a randomized trial, we can still potentially adjust for some of the variables that we may not have anticipated affecting our biomarkers. Whenever we do these studies, we first look to see if there is anything in our Olink panel or antibodies that may be associated with something we didn't predict, such as sex, age, or tumor subtype. If there is, we can use that and put it in the model so that we can adjust for it and have more robust interpretation of the data. 

We are hoping to now pool this study with a bunch of other studies that we're doing through the CMAC network. Right now, we have selected 37 clinical trials, and most of them examine nivolumab or pembrolizumab—the PD1 agents—as part of their backbone in different combinations. It will be interesting to see if there are commonalities across these other studies in the future. One of our goals is to do more cross-trial analyses and see whether any biomarkers could work beyond just endometrial cancer. These are some of our long-term goals.