All right, our last speaker for today is Dr. Matthew Pease. His topic is Glioblastoma comprises of two molecular distinct immunological signatures. Not this one. This is the one we just did. Last one. Yes. Thank you. Hello. My name is Matthew Pease. I'm a third-year neurosurgery resident at the University of Pittsburgh Medical Center. My talk is entitled Glioblastoma Comprises of Two Molecular Distinct Immunological Signatures. I received grant support for this research project through the Research Update in Neuroscience for Neurosurgeons, or the RUN grant, from when I attended the course in 2017. We all know that GBM is the most common in lethal primary malignant brain tumor, with an average overall survival in the 12 to 16-month range. Over the past 10 to 15 years, there's been very few advances in terms of traditional chemotherapy and radiation treatments aimed at prolonging survival. By and large, single therapy targeted have failed in GBMs due to the large heterogeneity of this tumor. This heterogeneity has prompted many groups to try to divide GBMs into different subgroups to try to predict response to various therapies. One group, the VERSHAC group, divided GBMs based on gene expression data into proneural, neural, classical, and mesenchymal subgroups. Immunotherapy is an emerging therapeutic intervention that's been very successful in other cancer types, producing prolonged, increasing prolonged survival. However, the response to GBMs from immunotherapy has largely been unexplored. No group has systematically looked at immune gene expression in GBMs, and this is where we're coming in. This is a roadmap to my study. We performed, we downloaded gene expression data for 509 GBMs in the cancer genome atlas and looked at the most variable expressed immune-related genes. Based on this data, we then performed clustering analysis and found two clusters of GBMs that had differential expression of immune-related genes. We looked at differentially expressed genes and performed pathway analysis to find potential targets for immune therapies. And then, based on the different groupings of GBMs, we're hoping to guide potential clinical trials. As mentioned earlier, we downloaded the gene expression data from 509 GBMs that's publicly available in the cancer genome atlas. All the data was appropriately normalized. We then looked at the 200 most variably expressed immune-related genes. We performed an unsupervised, unbiased k-means clustering and found two distinct subgroups of GBMs with differential expression of immune-related genes. We then used what's called a proportion of appropriate clustering to identify the optimal number of clusters of these tumors. We found that two is the optimal number, meaning that they're, that looking at the data, there is two different groups is the optimal number. Then, after using k-means clustering, we used hierarchical clustering and found similar results. We found two distinct groups of GBMs. When we changed the number of immune gene signatures that we're looking at from 200 to 300 to 500, we found that two still remains the optimal number of clusters. Now, we had two different groupings of GBMs using our different clustering techniques, one with the k-means and the other with the hierarchical clustering. We wanted to see if there was a lot of concordance between the two groups. We found that when varying the number of genes in clustering method, between 90 and 100 percent concordance between the two groups of GBMs that we found. We found that kappa-K coefficient was .91 when comparing the 200 most commonly or most largely variated immune-related genes. So, what does this all mean? When we're looking at immune gene expression data in a large subset of GBMs, almost 510 GBMs, that they cluster into two groups when independently analyzed by two different clustering techniques. And we're hoping to use these two groups to guide potential interventions and find new therapies. We compared our clustering and classification system to the VariHEAT classification system. We found that group one tumors, which were very immune-poor, which I'll talk about in a minute, had a significant increase in the pro-neural and classical subtypes of tumors. We found that group two tumors, which were much more immune-rich, had a higher percentage of mesenchymal tumors. 96% of the mesenchymal tumors ended up in the group two of GBMs. One interesting thing to note is that only 60% of the group two tumors were mesenchymal. Mesenchymal tumors are known for being very immune-rich, and the immune-rich mesenchymal GBMs may even have a shorter overall survival. But there's still about 40% of the patients in our group two tumors that are very immune-rich that were not identified with other classification systems, suggesting that our classification system may be superior in identifying immune-rich GBMs. Despite some data that shows that mesenchymal immune-rich GBMs may have decreased in survival, we found no changes in MGMT-promoted methylation, age, gender, overall survival, or progression-free survival between our two groupings of tumor. Then, with our two tumors defined, we looked at differentially expressed genes across the entire genome. We found 170 genes that were highly differentially expressed with a Bonferri correction factor of 1.5. Here's a list of some of the most differentially expressed genes, and it gives us insight into the tumor genesis mechanisms of these two groups of GBMs. Group one tumors were very immune-poor. They had a lot of alterations in cell cycle pathways and cell signaling pathways and some tyrosine kinase pathways. For example, EGFR was significantly different, as was PI3K-AKT pathway. This is a pathway that, in several mouse models of cancer, has been known to suppress the immune system through decreasing the anti-cancer T cells. Our group two tumors were very immune-rich. In fact, they had a 10 to 20 fold increase in expression of immune-related genes compared to group one. You can see several immune-related proteins that can be found here, such as CD14, which is found on macrophages and monocytes, and CD37, which is a cell surface glycoprotein expressed on mature B cells. We then performed a pathway analysis to try to find which pathways were differentially altered in the group one and group two tumors. As seen in the previous slide, group one tumors predominantly had alterations in the cell cycle pathway, whereas group two tumors showed differences in immune system pathways. Although no tumor provides an overall survival benefit, or no grouping, this does give some insight into the tumorigenesis and potential ways that we can attack these tumors through alterations in different pathways. Then we looked through alterations in the genetic pathways and genes to try to find potential actionable targets. Nuvulamab is an FDA-approved immune monotherapy for melanoma that's increased overall and progression-free survival. It works by blocking the binding of PD-1 to its ligands in activated T cells. Nuvulamab has a very limited experience in glioblastoma. For recurrent GBMs in patients with biologic mismatch repair, the Nuvulamab is able to provide durable cancer responses, but this is a relatively rare childhood cancer syndrome. What we found is our group two tumors statistically significantly at the RNA level produced a lot more PD-L1. We think that this may be able to guide potential future clinical trials by identifying a subset of GBMs that would be amenable to Nuvulamab therapy. In conclusion, we attempted to classify a large set of GBMs based on immune gene expression. We looked at 509 GBMs from the cancer genome atlas. We performed a blinded clustering analysis on the immune-related genes and found two different groups. We found that one group was probably driven by alterations in the cell cycle, and two are synkinesis, and another subset of GBMs that were altered in the immune pathway. We hope that this can potentially guide future clinical trials or find potential immune-related targets in GBMs. I wanted to thank everyone for staying all the way to the end and give a special thanks to my mentors, Samir Agnirati and Nduka Mungkulur. Thank you.

glioblastoma

immunotherapy

gene expression data

immune-related genes

Nivolumab