2026-04-18-pm-late-macrophages-therapeutic-gatekeepers

Date: 2026-04-18 Primary topic: single-cell-spatial-omics Cross-referenced from: single-cell-spatial-omics Word count: ~16,020

Auto-generated captions

This transcript is Vimeo auto-generated English. Expect misheard names and technical terms. Use the ripgrep/search approach for exact quotes.

Well, welcome everybody. This is the Macrophages as Therapeutic Gatekeepers from Biology to Clinical Application session. I'm Jennifer Guerrero. I'm from Brigham and Women's Hospital, and I'm joined by Dr. Tyler Miller from Case Western Reserve University School of Medicine, and Dr. Tom Marron from the Tisch Cancer Institute. We designed this session to highlight emerging translational strategies to reprogram, constrain, or selectively eliminate immunosuppressive myeloid states to improve clinical outcomes. Together, these talks, we hope to demonstrate, will emphasize how disease-specific biology can inform scalable, clinically actionable myeloid targeting strategies. So I'm the first talk. Great. Thank you very much. And, I'm going to be talking about lipid-associated macrophages, their role in tumor progression, and novel strategies that we've developed to target these cells. And so a lot of this data will be unpublished, so I want to acknowledge those who contribute to it up front. My lab has spent an extensive amount of time trying to understand the different types of myeloid cells, specifically macrophages in breast tumors. But what we're finding is that a lot of these, the diversity and heterogeneity is applicable beyond just breast cancer. We have specifically focused on lipid-associated macrophages. So I'll tell you about one of the projects where we have identified lipid macrophages are responsible for therapy resistance, and we've identified strategies to target them clinically. And then in the last bit of my talk, I'll share, really for one of the first times publicly, a novel strategy that we've developed in which we're using CAR monocytes to eliminate these lipid-associated macrophages. So we know that macrophages are abundant across solid tumors. And I always like to start with this immunohistochemistry because I think that the visual representation is quite striking. So here on the left, this is normal breast tissue, on the right is a breast tumor. The dark brown staining indicates CD68 positive macrophages, of which about 50% of the tumor are composed of these macrophages. We also know that patients who have higher numbers of macrophages, shown here in the green, have worse relapse-free survival, as well as overall survival. So why is that? Well, I always think about the homeostatic function of monocytes and macrophages. Monocytes go to sites of damage, of infection, and they facilitate repair and protective programs. They do the same thing in tumors. So cancer cells secrete factors that recruit these monocytes to the tumor. And when the monocytes get to the tumor, they differentiate to a macrophage, and they sense that there's low oxygen, poor nutrients, and so they become the type of cell that helps repair the tumor or repair these damaged cells, and in doing so, helps the cancer cells survive. These macrophages also secrete factors that lead to angiogenesis, metastasis. They themselves are impaired in their ability to present antigen. They recruit suppressive T regulatory cells. And importantly, they inhibit the recruitment and function of cytotoxic T lymphocytes. So given the abundance of macrophages and their role across solid tumors, there's been a huge interest in developing strategies to either inhibit these cells, deplete these cells, or to inhibit the things that they do, the suppressive factors that they make. And so, despite that I've spent more than a decade here, many of the things on this slide were, I was responsible for putting them on this slide, or my lab was. I am the first to admit it's not a home run. We're not curing cancer yet by targeting tumor-associated macrophages. And I think it's because we don't know how these drugs are really working. We don't know what subsets of macrophages these therapies are working on, and we don't know who is best going to benefit in what setting. And so, I think that this diversity and complexity of these cells is actually, if we could better understand it, we'll have better strategies to target them. And so this is an example of some of the diversity, where this is immunofluorescence on the same breast tumor, and five commonly used macrophage markers are shown here. You can see in this Venn diagram that there's very little overlap. So if you're talking about a CD163 positive macrophage, and I'm talking about a CD68 positive macrophage, we might not be talking about the same cell. In addition, where they're located in the tumor is important. Their metabolic programming is important. Also, their tissue specificity may be important. Every tissue in your body is 5% to 20% tissue-resident macrophages, and they have really important functions. The macrophages in your brain, they're involved in synaptic pruning, whereas macrophages in your liver or spleen are involved with clearing debris and other things. Macrophages in different tissues have different transcriptional, epigenetic programming. And so potentially, as infiltrating monocytes get to solid tumors, they may have different roles among the different tissues, and it may be something we need to consider when targeting them in different tissues. Okay. So we set forward to try to understand, in breast tumors, the diversity, the heterogeneity of macrophages, like many other groups have done across other solid tumors. And we had two approaches. One was to use publicly available data sets, and the other one is, we used an in-house cohort of primary tumors that we obtained from the Dana-Farber Brigham and Women's Cancer CenterAll of these samples had single-cell RNA sequencing, and then our cohort had a matched paraffin-embedded tissue that we could use for imaging and spatial analysis. And so to make this huge effort into just two short summary slides, we performed all the RNA sequencing, or we reanalyzed the RNA sequencing, and we identified multiple subsets of macrophages. We subclustered out the macrophages. We then used a number of strategies to try to understand what each subset was doing. So we used differential expression of the genes. We correlated different subsets of macrophages with T regulatory cells and cytotoxic T cells to try to understand that relationship between the different subsets of macrophages. We used gene set variation analysis. We used spatial analysis, as well as ex vivo functional studies. And so this is the summary of our findings, where we identified, again, these multiple subsets. We have common, the highly expressed genes are shown here on the top, and then we have these predicted functional characterizations. And so we're going to focus on two of these subsets, the GPNMB TAMs and this darker black box, and these SFP1 TAMs down here. These are really the ones that emerged as being highly suppressive, both having signatures of lipid metabolism. And so they also were the two most abundant macrophage populations out of all the macrophage subsets that we could characterize across these breast tumors. And when we looked more closely, for example, at their hallmark pathway enrichment, we see that the GPNMB TAMs have evidence of lipid metabolism, fatty acid metabolism, and these SPP1 TAMs, in addition to lipid metabolism, they also were involved in hypoxia. And so you might have noticed I didn't use M1 or M2 at all. This is my favorite slide. These states don't exist. M1 is a LPS interferon gamma-stimulated macrophage ex vivo. Fine. You can say that. But in a biological setting, this is not a state that exists. And so I urge you to characterize your macrophages by the genes that they express, by the cytokines they secrete, by the predicted functional signatures or confirmed signatures. So what are these lipid-associated macrophages? Lipid macrophages, you can actually see lipid droplets inside of them. They're well described across solid tumors. They express TREM2, CD36, SFP1, APOE, both at the translational level and then also at the protein level. In normal breast tissue, I'm not sure how well this is projecting, but CD68 macrophages are here in red, which you can see throughout the normal breast tissue. Over here in the breast cancer, you can see this yellow, which indicates uptake of these green lipid droplets. So these, again, the lipid-associated macrophages, are well described across solid tumors, and they are considered highly suppressive. We had identified them in our single-cell analysis as having relationships to suppressive T cells. And so others have also characterized lipid-associated macrophages as inhibiting phagocytosis, suppressing T cells, and involved with migration and invasion. We actually had previously identified lipid-associated macrophages as a main driver of therapy resistance in BRCA-associated triple-negative breast cancer. So we had previously published that olaparib, which is a PARP inhibitor used to treat BRCA-deficient breast cancer, in mice induced the recruitment of highly suppressive lipid macrophages. We had this BRCA 53-deficient model of triple-negative breast cancer. When we treat the mice for five days, you can see that there's an increase in macrophages. These macrophages express CSF1R and PD-L1. When we use olaparib on healthy human donor CD14 positive monocytes to understand how PARP inhibitors may be inducing macrophage phenotype, we then find that as a direct effect of the PARP inhibitor, PD-L1 increases, but more significantly, CSF1R increases. Again, this is just a PARP inhibitor on a differentiating monocyte ex vivo. And what was striking is that we had performed both proteomics and transcriptomics, and overwhelmingly, the signatures that emerged to be most significantly different in these PARP inhibitor-treated monocytes was lipid metabolism-related signatures. And indeed, we identified that CD36, which is a lipid uptake receptor, is activated to be expressed after PARP inhibitor treatment. So we're using bone marrow cells from mice. We're differentiating to macrophages with or without olaparib. And the flow cytometry here, the vehicle-treated murine bone marrow is here on the left, and then the olaparib treated is on the right. You see the expression of CD36 on the x-axis, so you're getting a major increase in CD36 lipid receptor expression. They also have high levels of PD-L1 and CSF1R, so really bad macrophages. And indeed, when we went back to our mouse model, we could identify more characteristics of these lipid macrophages. So this is this P53 BRCA-deficient breast cancer model. Again, these are immunocompetent mice. You treat the mice with olaparib for five days. You have a bit of a decrease in tumor volume. When you look just at the macrophages, these macrophages after olaparib treatment express PD-L1, the lipid uptake receptor CD36, CSF1R, TREM2, and other markers suggesting suppressive phenotype.So, at the same time we were identifying these really suppressive macrophages, we were making another discovery, along with several others at that time, which are cited here, that actually PARP inhibitors were inducing DNA damage in the cancer cell, which is part of its mechanism, but it was activating the cGAS-STING pathway, which leads to secretion of type I interferons. This type I interferon results in T cell recruitment. These T cells become activated, and they're required for PARP inhibitor monotherapy efficacy. We also saw increase in PD-L1. And so, this really suggested that potentially we could enhance PARP inhibitor responses by adding a T cell activating agent, such as some sort of immune checkpoint blockade. However, there's been more than a dozen trials at this point that have tested PARP inhibitors plus immune checkpoint blockade, and so far, those trials have failed to show that the combination is better than PARP monotherapy. This is a trial from our colleagues at Yale that tested the PARP inhibitor olaparib plus atezolizumab. And you can see here that the progression-free survival and overall survival is the same. And so we asked, "How could this be possible that if you're getting interferon responses to the PARP inhibitor, you're getting more T cells. Why isn't the immune checkpoint blockade working to enhance PARP inhibitor responses?" Well, you know the punchline. We hypothesized it was the lipid-associated macrophages that we had already identified to be present, at least in the murine tumors. And then we actually have some new mechanism of why PARP plus immunotherapy might not be working. Dr. Kenichi Shimada from my lab will present that on Monday, so go see his poster. So we had identified these lipid-associated macrophages increased in T cells in the murine models. We had identified that cGAS-STING is increased after PARP inhibitor responses. But we wondered, does this happen in the human setting? And so we worked with Dr. Philippa Lynch and Claudian Isaacs, who were the PIs of TALAFE, which is talazaparib, followed by the combination talazaparib plus avelumab in advanced breast cancer. And we have pre-treatment biopsies after one month of PARP monotherapy and then after PARP plus avelumab. For the purposes of today, I just want to show you the pre-treatment biopsies in Biopsy 2 to understand what PARP inhibitors do in the human setting. This is the amazing correlative analysis team that I get to work with. And so we have worked very hard to understand the correlatives from this trial, led again by Kenichi Shimada. So indeed, PARP inhibitor increases this immune response. You see the transcriptional regulation and increase in PD-L1, NK cells, T cells. You see that activated T cells emerge after PARP inhibitor responses, but you also get these suppressive macrophages, CD163, CSF1R positive macrophages. This is what the tumors look like. So this is a baseline sample, and then this is after talazaparib. You can see the green and the red macrophages throughout the tumor, so more macrophages. We then performed digital spatial profiling to understand if there was induction of these lipid-associated macrophages, so we can find areas high in macrophages and then perform whole transcriptome sequencing. And indeed, after PARP inhibitor therapy, there was an increase in lipid signaling, specifically in the macrophages. So this was very cool, right? The clinical data was similar to what we saw in the murine setting. So we went forward to see how can we target these lipid-associated macrophages. One of the most clear phenotypes that we had observed initially was the emergence of CSF1R positivity among these macrophages treated by PARP inhibitor. So CSF1R is expressed by mature macrophages. It's a receptor tyrosine kinase inhibitor, the signaling pathway, and its ligands are CSF1 and IL34. If you follow the CSF1R field, you know it's been tested across many solid tumors without a ton of success. However, there are two FDA approvals. One is for tenosynovial giant cell tumor, in which the patients have a translocation of CSF1. So it's just they're highly infiltrated by CSF1R positive macrophages, which is shown here, and eliminated with the anti-CSF1R therapy. And these patients derive benefit from this strategy. More recently, CSF1R axetilimab was approved for chronic GVHD. And so in our mouse model, we tested would anti-CSF1R enhance PARP inhibitor responses. And so here is our BRCA1 P53 deficient model of breast cancer. The monotherapy and vehicle-treated animals succumb quite quickly to the disease. Olaparib induces a median overall survival of about 100 days, which can be doubled out to 200 days with the addition of anti-CSF1R. Both the PARP monotherapy and the combination therapy was dependent on T cells, because when you deplete T cells, you significantly reduce the efficacy. I just want to show you that we also had the PARP plus PD-L1 arm in here. And I wanted to highlight that actually, so here in brown is PARP plus PD-L1. It looks just like PARP monotherapy. So this model might have predicted the lack of clinical responses that we saw.And so when you have the combination of PARP plus anti-CSF1R, you have increased T cells, interferon signaling, STING activation, really good anti-tumor responses with less suppressive macrophages. You reduce tumor growth and metastasis. So we've moved forward to the clinic. This is the AXLAP phase 1b trial of axcitilimab plus olaparib. It's in metastatic HER2-negative breast cancer with BRCA1 or PALB2 mutations. Patients receive 14 days of olaparib, followed by the combination of PARP plus axcitilimab. It's led by the very talented Philippa Lynch and Dr. Shapiro, and I serve as a translational PI. We're more than half enrolled. It's very exciting. All the tissue and blood are in my lab, so we'll have the opportunity to understand what's going on in these tumors. So I presented this part of the story, where the PARP inhibitor induces DNA damage of the tumor. This leads to activation of cGAS/STING, secretion of type I, recruitment of CD8 T cells. But at the same time, this response activates tumor-associated macrophages that become lipid-associated, increase lipids, CD36 expression, CSF1R, and TREM2. And so we thought, "How else might we target lipid-associated macrophages more broadly and not in the context of a PARP inhibitor?" And when we go back and look at our single-cell sequencing, we see that these two populations of lipid-associated macrophages have high expression of TREM2. Now, TREM2 is a receptor, specifically on macrophages, myeloid cells, that mediates homeostasis in tissues, really by eating up all of these different things. It's involved in phagocytosis. It's a regulator of inflammation. And so we know that TREM2, from work from others, is an unfavorable prognostic biomarker across many solid tumors. We have shown in breast cancer that high expression of TREM2 is associated with worse clinical outcome. And really importantly, TREM2's actually only expressed on tumor-associated macrophages, so these are all different tumors, and the brown staining is TREM2, but not in normal tissue. So the tissue-resident macrophages have very low expression of TREM2. Preclinically, there's been work to show that antibodies against TREM2 have demonstrated some efficacy. This is a sarcoma model from Marco Colonna's lab, where the TREM2 antibody delays tumor progression compared to the control. Anti-PD-1 makes that a little better. This is a study from Miriam Merad's lab, where they showed that anti-TREM2 results in decreased lung tumor burden. And another example from breast cancer, where anti-TREM2 delayed tumor progression. Really interesting and excitingly, in the last couple of months, two big papers came out from "Cancer Cell." One was from the Merad Brown lab at Sinai, where they used CAR T cells to get rid of full R2 macrophages, and separately, another CAR T product against TREM2 macrophages. So these CAR T cells, if you just look here, these CAR T cells against anti-TREM2 could reduce tumor burden and extend survival. In the same issue of "Cancer Cell," Ito Amit's lab also demonstrated that CAR T cells against TREM2 was effective in reducing colorectal and fibrosarcoma. So this is really good proof of principle that targeting TREM2 is important. It can reduce tumor burden. But we also know that CAR T cells have not been effective across solid tumors. There are seven FDA approvals for CAR T cells, but these are all in heme malignancies. So far, there's no approvals for solid tumors. And so both antibodies and CAR T cells have a hard time penetrating solid tumors. The CAR T cells are inhibited by macrophages in the tumor, and they're generally not pan-cancer. And so here, in the last couple of minutes, I want to propose a new strategy called macrophage replacement, where we're using monocytes, which I hope I emphasized in my first couple of slides, actively as part of their homeostatic function. They go to tumors, right? That's just what they do. And so the monocytes rapidly infiltrate the tumor that express this TREM2 CAR. Once they get to the tumor, they become CAR macrophages that express TREM2. They destroy TREM2-positive macrophages and then become the type, these CAR macrophages secrete cytokines and activate T cells and kill cancer cells. So some proof of principle, we generated human CAR monocytes. This is work led by a very talented post-doc of my lab, Dr. Dan Michaud. So we generated the effector cells. These are U937 cells that express the TREM2 CAR. No expression of TREM2. And our target cells are THP1, which have some expression of TREM2. And what we find is that the THP1, these are red. The U97 are green. And so when you put them together and see the red and green on the same plot here, indicating that there has been a phagocytosis event. However, this has not happened when the effector cells have a mesothelin CAR, an irrelevant CAR. We also know that these TREM2 CAR monocytes, upon ligand engagement, they secrete TNF alpha. And we can demonstrate additional phagocytosis using Incucyte assays, which is a better reader of actual phagocytosis and not just binding.In the murine system, we've used this really cool tool. It's the HoxB8 monocyte system, which is a progenitor self-renewal monocyte system that's under the control of estrogen. So when you remove the estrogen, the cells differentiate out into macrophages. And so we use these as our effector cells in the murine system to target these Ros264.7 Trem2 expressing cells. And indeed, you can see that there's a nice phagocytic events that occur when you put these cells together. Now, in the mouse models, we have shown that we had to pick a mouse model where we at least knew there were some Trem2 positive macrophages. So you see here the E0771 breast tumor line has Trem2 macrophages. So we started here. We IV inject about five million Trem2 CAR monocytes into the mice and then look to see what happens. So at day 10, we see that the untreated mice have this many CD8 T cells. Untransduced monocytes that we infuse in have same number of T cells. But when we use the Trem2 CAR monocytes, we see an increase in CD8 T cells. That translates to about a 50% reduction in tumor burden. And we can further enhance that by adding anti-PD-1. So we think that this approach is superior from prior cell therapy because the monocytes have superior homing, penetration relative to CAR T cell. There's a superior remodeling that happens with the Trem2 CAR monocytes as opposed to the Trem2 antibody. Importantly, Trem2 is expressed by all tumor macrophages and not tissue-resident macrophages. And so as we move forward to the clinic with this, we'll probably follow the footsteps from Charisma Therapeutics, who put both CAR monocytes and macrophages into patients. And so, we'll continue to work on this strategy by enhancing the product, by changing intracellular signaling domains, adding cytokines, and then thinking about how this will go to the clinic, either with CD14 monocytes or HSC type of approach. So I hope that I convinced you that intentional analysis of macrophage subsets is important to understand in a different specific context what types of macrophages need to be targeted for anti-cancer therapy. We identified these really suppressive macrophages that were associated with lipid metabolism. And so we think this represents a new strategy to eliminate really your choice of tumor macrophages, but we think that Trem2 is a really attractive target. So I just want to thank my lab who has done this work, our funding agencies, all of our collaborators, and I'm happy to take a few questions with the three minutes I have remaining. So if you can just come to the microphone so that the online audience can hear. Very nice work. A little bit fast, but really nice. A lot of data. So I just have two curious questions. Is this lipid macrophage terminally differentiated, or if you treat with some cytokines, they can change phenotype? And do you need to treat the CAR monocytes with something to make them go to the tumor better? Or just like- Yeah. So they are terminally differentiated macrophages, but whether we can switch them to a different direction, I actually suspect you might be able to play a little bit with their plasticity there. We haven't done that exact experiment. No, I don't think we need to treat the monocytes with anything. All right. Thank you. Congratulations. This microphone. Yeah. Great talk. My first question is about the lipid-associated macrophage. I'm kind of curious, so for the primary resected tumor, so you saw this population. How about in the breast cancer metastasis- Yeah. Great question ... this drug also have similar effect? Yeah. So Ito Amit, I believe, a couple of years ago, published a really nice paper on Trem2 macrophages in lung metastases from breast tumors, and those tumors were surrounded by Trem2 macrophages. So that data exists. We are really interested in what happens in liver metastases from breast tumors, and we see these lipid-associated macrophages actually in higher numbers. This is a project led by Gabriella Antonellis, who should be here somewhere. So we see them there, and we think it'll be effective in the metastatic setting. Great. Just kind of a follow-up question about, so if you want to target the macrophage, target the lipid-associated macrophage and the normal other type of macrophage, yeah, the strategy the same or different? Sorry, just say the last part. If you want to target the lipid-associated macrophage, will the strategy be the same as the other type of macrophages? If you want to block the macrophage tumor immunosuppressor effect, something like that. Yeah. It's similar. We can chat later. Okay. Okay. Yeah, go ahead. Thank you. So you showed a subpopulation of these locally proliferating tumor macrophages, and that's been shown before, but not really been explored. Have you guys investigated the tumor-modulating roles of these and how your therapies may impact that population? I don't know why I can't hear. What's the first part of your question? Oh, that's much better. The locally proliferating tumor macrophages that you showed- Yeah ... at the beginning, there's a lot of evidence of those out there, but nobody's really explored the tumor-modulating properties of those. So have you guys? Yeah. No, I don't think so. I mean, theBone marrow-derived macrophages are not likely to be super proliferative. It'll be the tissue-resident macrophages that are proliferating. And so where those go into which subsets, I think that they can become any of the subsets. That's really interesting question and work that's underway. Yeah. Great talk. Two quick questions. Why do you think these methods would be superior compared to an antibody that would deplete by ADCC? Yeah, great question. I think that there's several reasons. One is that antibodies have a difficult time penetrating into especially these really macrophage-rich, suppressive, hypoxic areas, without blood vessels, for example. So monocytes will get there. They're already there. So that's one advantage. Got you. Yeah. And then a quick one. Do you envision that a therapy like this might be like what's happening now with CCR8, where there seems to be some promising data in combination with PD-1? Yeah. Do you think that this will differentiate and have single-agent activity or ultimately require checkpoint inhibition? Yeah. We are aiming for the absolute best monotherapy that we can get. I think it will require cytokine delivery and some sort of T cell activation, but that might be as a single product and not require immune checkpoint blockade, for example. I see. Thanks. Yeah. Thank you. Just maybe one last quick one. Hi. Yeah. Have you seen, do your CAR monocytes express TREM2 at all? Yeah, great question. So in the mouse system, we use a TREM2 knockout bone marrow, and then in the human system, for the clinical product we're working on, it would be a knockdown strategy. And have you looked at the CAR mon to macrophage after tumor killing and stuff? Do they get reprogrammed by the microenvironment? Yeah. It's a great question, too. We're still evaluating that. In the early time points in the mice, no, it looks great. Depending on the cytokine it's loaded with, we hope that that's enough to activate a durable anti-tumor T cell response, and also we have strategies to have suppressive cytokines enforce a anti-tumor phenotype in macrophages. So, there's so many things we can do here, but it's a really important consideration. Yeah. Thank you. Okay. So now we are going to turn it over to Tyler Miller for the second presentation. Perfect. All right. Thanks, Jen. Thanks for the invitation. Thanks to the organizers for putting this session together. It's clearly a very popular one. Great to go second because Jen gave a fantastic overview of tumor-associated myeloid cells. I'm going to talk about redefining these myeloid cells and spend time on that, and then targeting these in solid tumors using brain cancer as a model system. But I think everything that we do here is probably really applicable to other tumors. I don't have any relevant disclosures. So immunotherapy has really revolutionized care for certain cancers, and I'm convinced that active immunotherapy is our best shot at curing gliomas. Unfortunately, it hasn't worked so far, and we've tried. I think there's lots of reasons for this, but the most important hurdle we have to overcome is the immunosuppressive environment that's imposed by these myeloid cells. And I love showing this picture. Here in brown is microglia in the normal cortex, and then you get a higher grade glioma. You get activation of these microglia and then proliferation of them. And in glioblastoma, you get this chock-full of immunosuppressive macrophages. They can make up to 50% of all cells in the glioma. And so the question that my lab tries to answer is: How do we overcome that myeloid suppression to create effective immunotherapies for glioma patients? And we've taken this long view approach where we have to create both a platform for discovery and translation, and that's built on two pillars. One is we have to understand the problem better to create the foundational data sets to understand these myeloid cells, and we use single cell RNA sequencing, spatial transcriptomics, and increasingly spatial proteomic data sets to understand them better. And the second really important arm is we have to have a functional model system where we can test human myeloid cells in matched human cancer cells. And I was really lucky during my postdoc that Penn came out with this GBO model. And this is where you take tumor directly from the OR, you chop it into small pieces. You don't dissociate to single cell. You keep it intact piece, like one millimeter pieces. You grow it in an orbital shaker, and that allows you two to eight weeks of time where you have the whole intact tumor microenvironment, including the myeloid cells, T cells, stromal cells, and you can perturb these during that time, and you can understand how that perturbation impacts all the different cell types in that system. After a couple of months, you lose the vast majority of the tumor microenvironment cells, except for the malignant cells. But you have the malignant cells in their own extracellular matrix. You can add back on human monocytes, T cells, CAR T cells. They will infiltrate in and start to express many of the same programs they do in the primary tumor. Okay. So I'm going to talk a little bit about our mental model here. And our view is that immunosuppressive myeloid cells are a top target to make immunotherapies more effective for our brain tumor patients and solid tumor patients. That single cell genomics combined with patient tumor organoids are a really powerful platform for both therapy discovery and translation. And I'm going to spend some time on this, that we should move beyond Louvain and Lighting clustering, which is typically projected on a UMAP, to understand our single cell RNA sequencing data. Because complicated cells like myeloid cells and cancer cells that exist on a spectrum are really challenging to understand with clustering in UMAP. So Louvain and Lighting clustering are reasonably good at identifying cell types, but pretty poor at pulling out cell activities. And I would argue that most immunotherapies, and actually most of our other therapies, aim to target cell activities. And if you can isolate out the cell activity from the identity, it really enables a lot of novel insights and new targets.And so we are now thinking about we should reprogram or remove specific myeloid cell populations, as Jen talked about, and the activities of those combined with CAR T cell therapy to make more persistent responses for our patients. So I'm going to spend a little bit of time on this. So Louvain lighting clustering really groups cells based on similarities in gene expression profile. Every cell is treated as a singular unit. You put it next to its nearest neighbor on a plot. You draw a circle around it, make it a cluster, and then all of your gene sets, your signatures, whatever you call them, are derived from looking at all the genes expressed in one cluster compared to all the other clusters. Oppose that to CNMF, consensus non-negative matrix factorization, which is what we use, aims to decompose that single cell RNA sequencing data down into distinct gene expression programs. So it looks at sets of genes that go up and down together across thousands of single cells. Those highly correlated sets of genes, we call programs. Those programs can be identity program. What cell type is it? Macrophage, monocyte, dendritic cell, or it can be an activity program. What is that cell doing? And importantly, you can have more than one program per cell. So it allows you to have sort of this matrix, where some of the programs will define, hey, that cell is a microglia, and other programs say, hey, that microglia is inflammatory, or that microglia is responding to interferon. And so we did this across a large panel of myeloid cells. In glioma, we pulled out 14 programs. Five of these programs were identity programs, microglia, macrophage, dendritic cell. Nine of these were activity programs. Five of those activity programs were sort of classic response, hypoxia response, interferon response, cycling programs. But four of them looked immunomodulatory, whether they were inflammatory or immunosuppressive. And if you looked at the programs being used by all the myeloid cells in a tumor, it were these four programs that were utilized the most. In fact, 90 plus percent of all of the myeloid cells in a brain tumor used one of these four programs. And what's interesting here, this is a very simple plot. It just says, what are the percent of microglia that express these different activity programs? And so any given microglia might express one of two or three different activity programs. What was really interesting was that these activity programs were shared across the different cell types. So for example, this one systemic inflammatory program could be expressed by all five different cell types. This is very different than what we typically do, where you take microglia, for example, and then you try to subdivide the microglia down into inflammatory, immunosuppressive, or interferon responding. And then you take the macrophages, and you take the monocytes, you subdivide them. What this tells us was that all of those inflammatory programs can be used across, and you don't need to subdivide them out. You just have to pull them out independently, and then you just apply them to that. And so that allows you then to think about what is that particular program? How is it being used across the population of all myeloid cells? So this is an example of that. This is a really nice data set from Houston Methodist that we use. So this was their UMAP and their clustering, and they basically annotated all of their cells. So they annotated this cluster as inflammatory microglia, for example. So we took all of their single cells. We looked at our programs in their single cells. And it's true that all the cells that were in their microglia cluster looked like they were expressing our microglia identity program. But they also could express one of three different activity programs. So it was a combination of different sort of cell states that were in that one cluster. So what we're really pulling out here is the principal components of myeloid cell expression phenotypes. And this really brings it home, I think. Again, UMAPs are good at pulling out cell types. It's pretty good at pulling out all the microglia clusters and macrophages and monocytes, which you can do on UMAPs. But it's really poor at pulling out these cell activities because they're expressed across multiple clusters. Like this microgli inflammatory program is expressed across four or five different clusters. And so you can never pull it out if you're taking one cluster versus all others. It just gets lost in the background. Okay, so we basically looked at all of these with a bunch of clinical correlates, and it were these four programs that were associated with survival and response to immunotherapy and all the other things. And I'll show you a little bit of that. And so these four programs are now how we think how we should define these myeloid cells. And again, so you have these two inflammatory programs. A systemic inflammatory program is like the default inflammatory program for monocytes and all over the body. Everywhere we look, if there's an inflammatory lesion there in the liver, the lung, breast cancer, it's there. And this expresses TNF alpha, OSM, IL-1 beta, and it basically secretes a bunch of cytokines. It recruits additional immune cells into that microenvironment. And then the microgli inflammatory program, which is really brain specific, it's the default inflammatory program for microgli. They see an inflammatory stimulus, they express this program. Then you have two immunosuppressive programs. They were expressed in different cells within the tumor in different places spatially. One was driven by dexamethasone in a really immunosuppressive steroid that we give our patients. The other was driven by or associated with hypoxia and necrosis and a few markers that we see. And so we're thinking about, again, clinical correlates. We had 85 different single cell RNA sequencing data sets from brain tumor patients. We looked at what are some correlates of sort of response to immunotherapy. T regs are known to be associated with that, and so we looked at what programs in our myeloid cells associate with high T regs in a tumor. It were these two immunosuppressive programs that are highest. The microgli inflammatory program was the opposite. If you look at survival, the scavenger immunosuppressive program was the one program that was associated with overall survival, not any of the identity programs. We had this nice data set that was published where they treated patients with checkpoint blockade for two weeks, then they did surgery, did single cell sequencing on those samples, put the patients back on checkpoint blockade until they progressed based on imaging.And they divided these up by non-responders and responders. And so we looked at the myeloid cells in those patients that responded and didn't respond, and it was the scavenger immunosuppressive program. If it was high in their tumor, they were a lot more likely to be a non-responder. Here's this schematic again, these four programs. Here's the real data that's based on. Each one of these is a myeloid cell in a brain tumor, and it's placed on this quadrant plot based on expression of these four programs. And what we did is we took the myeloid cells from that trial, we placed them in a quadrant plot based on their expression of our four programs, and then we labeled whether that myeloid cell was from a responding tumor or non-responding tumor. And it's like black and white, right? The responding tumors had much more of this microgli inflammatory program. The non-responding tumors had more of this scavenger immunosuppressive program. Okay, so everything else from here is unpublished data. So that was all checkpoint blockade. The question, we think, CAR T-cell therapy is a more promising therapy for our brain tumor patients. And so we partnered with a group from Penn, Donna Roik, Stephen Bagley, are running a CAR T-cell trial with a bivalent CAR targeting EGFR and IL-13 receptor alpha two. There's lots of details here that you don't really need to know for this story. But basically, they put in a Mayo reservoir and catheter in where they deliver the CAR T-cells directly into the ventricle. And what they're able to do is take CSF off before they put the CAR T-cells in. They put the CAR T-cells in, and then they can take subsequent samples of CSF to monitor that patient over time. And so when they originally looked at this, they had 44 samples of CSF. These are all the myeloid cells from those 44 samples. They had annotated these as basically cell identities. None of these correlated with any response in their trial. And so they asked us to help out. And so Reagan Murphy in my lab is a grad student. She ran CNMF on all of these cells in an unbiased fashion and pulled out what are the programs in myeloid cells in the CSF of these patients. And it turns out that many of the programs that we find have proxies of what we see in the primary tumor. Right? There's a complement program, there was a scavenger program, there was these inflammatory programs. And so we want to see what those did over time in response to CAR T-cell therapy. There's this very sort of stereotypic response for all of these patients, whether they responded or did not respond. For example, the complement immunosuppressive program at baseline is very low, less than 5% in then basically every patient in their CSF. But after seven days of product, it skyrockets to 75% of the myeloid cells in that tumor are using this program. It was dose-dependent. You get a higher dose of CAR T-cells, you get higher levels of this immunosuppressive program. And there were several other myeloid cell programs that had this sort of classic response across all these patients. But the important point is what actually correlates with survival, what correlates with response for those patients. And so again, we looked at the non-responders, responders, and it was the scavenger immunosuppressive program, honestly, it's probably the lipid macrophage program that Jen was talking about in some form, that was associated with non-responders. And we had the tumor sequencing data from just three of these patients, but the one non-responder we had had much higher levels of this in their tumor as well. And then if you do a simple thing, you just take the patients, divide them in two and say, what are the patients with the highest levels of the scavenger immunosuppressive program at baseline before these CAR T-cells go in, how do those patients respond? This is overall survival. And it's a pretty striking Kaplan-Meier curve here, where if you have low scavenger at baseline, you do better. If you have high scavenger at baseline, you do worse. Okay, so what do we do about it? And I think our view is that we should push all the myeloid cells in these tumors to this microglial inflammatory program. We want to get rid of all of these and leave it with this. We think you need some myeloid cells to have a good response, and you have to have the right ones. So my lab is systematically identifying ways to reprogram or remove the myeloid cell states. Adele and Alex are working on the immunosuppressive nature of dexamethasone and how to reduce its effect on our patients to try to get rid of this one. Alex and Kaylee are working on how to remove the scavenger program, prevent it from even being created in the first place. I'll talk a little bit about this story and paradoxically how we could prevent more of these from coming in. Bryce and a collaborator, Berkeley Gretter, are thinking about how do we reprogram the immunosuppressive macrophages that are already there. And then Adam and Lindsay are using CAR T and K cells to go against the immunosuppressive myeloid cells that are already there to just get rid of the ones that we don't want. So our lab, our whole research program is built on a high volume preclinical testing pipeline. So had the lab for about two years. We've been collecting tumors for about 18 months. We've collected over 200 brain tumors from university hospitals in those 18 months. We take multiple samples to be able to do multimodal testing on them, and we try to make organoids from every patient that we have. And then we use those organoids to do acute testing for a bunch of different therapies, basically in a way to give us the data that we need to think about an IND for a particular preclinical therapy we want to translate. This is what that looks like. You get a piece of tumor in, you chop it into one to two millimeter pieces. They sort of round up over a couple of weeks. You can put them in 96 well plates and treat these with whatever perturbation you want, including cellular therapy to see how that works. These organoids are pretty faithful to the original tumor, and they are for weeks. This is a 40-year-old male IDH mutant patient that had not been treated with dexamethasone, notably. And so it has a really complex repertoire of immune populations here, which is why I could use it as an exampleHere is a single-cell sequencing data from the primary tumor in blue, and then the organoid two weeks after spending culture in red. And you can see all of these populations still exist at two weeks in these organoids. So it's a really powerful model for us to use. And because it's a complex system, all of our readouts to all of our experiments basically need to be single cell. It can either be whole tissue, where we keep everything with intact spatial information. We use the 10x Xenium platform. We have a thousand plex custom panel. We use spatial proteomics with Akoya. We have 123 antibodies and custom panels depending on the question. Sometimes it's best to dissociate single cells, where we use the 10x LX system or LX Apex now, which has really changed the game in the capacity to do lots of these samples. And then we can do flow cytometry or single-cell PCF. Okay. So I'm going to talk about one vignette here, which is a fun translational vignette, where Alex and Kayla are thinking about how do we prevent this scavenger immunosuppressive program from being created. And we think about IL-1 beta and how do we prevent that. So we had shown previously that different sort of cytokines in the media could produce different phenotypes within our myeloid cells. And we had shown that using this myeloid organoid system, where you take an organoid without any myeloid cells, you put donor monocytes in, you let them infiltrate, and you treat them with different things, how does that impact the sort of trajectory they go down from a differentiation standpoint? And IL-1 beta was the thing that drove the scavenger immunosuppressive program the most. The question was, where is IL-1 beta coming from? And so IL-1 beta is coming from the myeloid cells. This is 85 different single-cell sequencing samples from glioma patients. This is all publicly available. You can go on and play with it yourself. So these are all the myeloid cells. It's the systemic inflammatory myeloid cells themselves that are expressing IL-1 beta, and they do so in high quantities. Alex did a fun experiment where he took organoids with no myeloid cells in them at baseline, and then added back into that system either T cells or monocytes and looked at the IL-1 beta concentrations in the media. And so if you add T cells, it doesn't really do much. But if you add T cells and monocytes, you get this large increase in IL-1 beta. If you add just a bunch more monocytes, you get a much higher level of IL-1 beta. And so it is really just the monocytes coming in, expressing the systemic inflammatory program that are driving IL-1 beta secretion. So it turns out that Alex did his homework and tried to figure out how we're going to inhibit IL-1 beta release. And in data I'm not going to show, but lots of data that he has, IL-1 beta is released through pyroptosis in these cells. They go into this tumor microenvironment, they pyroptose, and they secrete a ton of cytokines, including IL-1 beta in very high quantities. And disulfiram can actually inhibit the pyroptosis and the IL-1 beta release. And so Alex did an experiment, again took the organoids, added back PBMCs. So if you have PBMCs alone, there's no IL-1 beta that could be released. It's the combination of the PBMCs, the monocytes with the organoid that drives IL-1 beta release. If you add disulfiram into the mix, it basically drops out the IL-1 beta that gets released in that system. And importantly, this is all single-cell sequencing data now, you get reductions in the scavenger immunosuppressive program that we were trying to prevent. And so now we're in a spot where we're trying to start up a Phase 0 window of opportunity trial with disulfiram in our brain tumor patients, where between three and 14 days before that patient goes to surgery, they get disulfiram. And then we would monitor over time what that looks like for their blood and also in the tumor that our surgeons then resect. And understand, does this reduce the immunosuppressive programs we see in their tumor? And interestingly, does it also prevent us from having to give them so much dexamethasone? Because we actually think that the IL-1 beta that's in that system is helping drive some of the perivascular edema that we're seeing. And if you prevent that, can you prevent giving them dexamethasone, which then prevents the other complement immunosuppressive program? So this is sort of a hit two birds with one stone type of trial. Okay. With that, I think the goal that I have for our lab at Case Western and University Hospitals is really create a translational research program for brain cancer immunotherapy. I'm so fortunate to have fantastic clinical colleagues that we work with every day who help get us specimens, who think about the trial designs, and think about how do we get better therapies for our patients. And we have this very iterative process where we're getting samples from the clinic and understanding why that therapy works or doesn't work. And with that, I want to thank the lab that has all joined me. My lab's only been open for about two years. All these people came in to join me to do this work. It's been a really fantastic journey. And thank our collaborators and especially the funding agencies that help us do all this work, and be happy to take questions. Hi, Tyler. Great talk. Thanks. Kind of a quick question. So you mentioned about in the brain tumor, right? So the myeloid cell is the dominant population, immune cell population in the brain tumor, especially in the GBM setting. So do you think just now targeting or reprogram the myeloid cells itself will be sufficient to really kind of make the immunotherapy effective? Or you think we still need to add maybe PD-1, PD-L1, Blockade in combination or other kind of immunotherapy approach to really boost the intel? Yeah. Yeah. It's a great question. Our viewpoint on this is that we're going to need to combine all of the myeloid-targeted therapies with other immunotherapies. So our vision is that we would do sort of neoadjuvant... therapy with a number of these different targeting strategies, where before the patient got CAR T cells, for example, we would do seven days of treatment with this, and then maybe we extend that treatment in the same time we're doing CAR T cells or not, depending what that therapy is and how it impacted the CAR T cells themselves. But how do we reprogram the microenvironment so it's in a state that can receive CAR T cells or something else and have that therapy be more effective? That's how we're thinking about it. I would love to get to a point where we had a monotherapy with macrophages. Mm. Maybe we'll get there, but initially, I think we have to do this in combination. And then for brain tumors, the cancer neuroscience part of this is also important, and it's probably going to end up being that we do three or four different therapies for these patients. The question is going to be, how do we do that in sequence to be best for outcomes? Okay. Makes sense. I have always have kind of crazy idea. I was thinking, if you can hijack those kind of myeloid cells in the brain, let those cells attack the brain tumor cells directly, would that be possible? Absolutely, I think possible, right? Many of these myeloid cells have the ability of phagocytose target cells. And I think, the CAR macrophage stuff, people who do that but do it to the cancer cells, they go in and eat cancer cells, you can do that. How do we get them in the state where they're phagocytosing cancer cells, not anything else? So I think we'll get there. We're studying that. We have collaborations around how do we push them to a more phagocytic state. It won't be a single agent therapy though, I think. Thank you. Yep. On the IL-1 beta side of things, so you're targeting pyroptosis specifically, but obviously IL-1 beta is just created by these cells in a normal fashion. Have you tested doing an NLRP3 inhibitor or Anakinra or Canakinumab? Are you seeing similar results, or is it something specific to the program of pyroptosis? Yeah, it's a great question. So we tried Anakinra. It was like the very first obvious thing that we should do, right? We tried Anakinra. And it works, but it doesn't work great. And I think the reason that that is, is the stoichiometry is just a bit wrong. There's so much IL-1 beta there, and I don't know that everything acts through IL-1 receptor. So it does reduce it down, but doesn't reduce it down as much of just like shutting off IL-1 beta. What I think is interesting is, you said IL-1 beta is created by these cells and it goes everywhere, but what we've seen is if you shut that off, there's not much IL-1 beta left. And it's not coming from the T cells, and it's not coming from the cancer cells. It's just coming from these myeloid cells. And the only way that it's really being secreted that we can see is through these pores that are formed during pyroptosis and these gas dermo pores, and it gets pumped out. And so if you shut off pore formation, which is what you do with disulfiram, then it just doesn't get secreted. Have you taken those macrophage-- In pyroptosis, but if they had the proper stimulus, I'm assuming ex vivo you're doing this, right? You'd need to give them an LPS ATP stimulation to get the inflammasome to generate IL-1 release in that way, ex vivo, right? So in vivo, the macrophages will be releasing IL-1 beta, obviously, pyroptosis at a much higher degree. So that's kind of more where I'm getting at. Yeah. We don't give them anything. We literally just put them in the context of the organoid. We don't put anything else in there except putting monocytes with the organoid, and there's enough signal there from the tumor cells and whatever else is in that organoid that it signals for that cell to go under pyroptosis. Okay. Yep. Thanks. Hi. Thank you for your talk again. I'm curious how your work could potentially expand into brain metastases, such as melanoma, lung, breast, that sometimes also resist the immune checkpoint inhibitors. My understanding is that in gliomas, there are more microglia than monocyte-derived macrophages, but there's comparatively more bone marrow-derived macrophages in brain mets. So I'm just curious if these should be reprogrammed in some way to that inflammatory, more like- Yeah ... full microglia that you're proposing or- Yeah. Great question. It's something we're actively studying, and we have a sort of breast cancer brain met program going on in the lab. At least in GBM, so lower grade gliomas, more microglia than bone marrow-derived macrophages. But in GBM, there's a lot of bone marrow-derived macrophages, and probably, we think the vast majority of the immunosuppressive myeloid cells that are in those tumors are coming from the bone marrow. So I think you'll have similar targeting strategies, actually, with the brain mets. I think we may have the tumor targeting part of that's going to be different. And so the combination therapies will be different. But I think the main therapies that we're going after for the myeloid cells are going to be very similar because we see very similar sort of sets of myeloid programs in the brain mets as well. Okay. Like with the IL-1 beta, that- Yeah. I don't know specifically for that story, if it's exactly the same thing that we're seeing, if we could combine that. But if we just profile them by single cell, we see similar types of programs in the brain mets. Slightly different, but similar. Thank you. Yeah. Great. Okay. Cool. Thanks, everyone. Thanks, Tyler. All right. Our third speaker is Dr. Thomas Moran. All right. Thank you to Jen and the organizers for putting this great session together. I think there's obviously lots of interest in macrophages. And I'm going to talk a little bit about some of the more clinical work that I'm doing, that others have done to try to target some of these macrophages and some of our successes and failures. Unfortunately, more failures than success so far, but we're making headwinds. So I'm an immunologist initially by training, and I lead the Phase one program at Sinai. And my side gig is taking care of lung cancer patients, which is one of my favorite days of the week, because patients these days are doing so much better with the incorporation of immunotherapy into chemotherapy. This is my current treatment algorithm. It's based on a relatively mediocre biomarker you may have heard of called PD-L1, and we're basically just giving immunotherapy with or without chemotherapy. And we're more than doubling the median overall survival of these patients with the incorporation of immunotherapy, which is amazing, and our clinics are very happy and full. But the elephant in the room here is obviously that the majority of our patients are still dying. So we need to find a better way to harness the immune system to recognize and kill cancer. There's really two approaches that I think are most promising. Number one, really inducing vaccinal effects. So on the right here, you see the incorporation of ipilimumab in PD-L1 negative patients, and there we think we are seeing induction of de novo clones. You can also just do neoantigen vaccines, and there's lots of RNA and DNA and peptide approaches that are ongoing right now and are looking promising, particularly in those patients that don't have pre-existing immunity. But one of the things that we really focus on, and obviously is the focus of this session, is harnessing the rest of the innate and adaptive immune system to see how we can improve outcomes. Because the reality is that the vast majority of the agents that we're going to be talking about here at AACR and at ASCO are really myopically focused on CD8 T cells, which you can see are a very small minority of immune cells within the tumor microenvironment. Much more common are macrophages and neutrophils, which are pretty much ignored by 99% of therapies. But we really need to look outside of the CD8 compartment to see how we might be able to improve things. And as Jen said, we have these good and bad. I asked Gemini to make this, and she didn't even come out with an M1 and M2, which was very nice. But we have the good and bad macrophages, and I think one of the number one things we have to focus on and learn from our previous failures is not throwing the baby out with the bathwater. So not depleting both good and bad macrophages. We know that macrophages are very good at killing cancer and also clearing cancer debris. And so what we really need to find ways to either deplete the bad macrophages or potentially repolarize these macrophages into a more pro-inflammatory, anti-tumor state. And that's really the goal of many of the therapies that are being developed. And what we realize is we always like things to be black and white and very dichotomous. The reality is that there is real spectrum of functional statuses, as has been described by my colleagues here. And hopefully we see lots of plasticity in the near future as we're incorporating new therapies, so we can potentially move from one functional status to another. So there's been many approaches, and Jen very nicely described an interesting approach with CSF1R blockade, and this approach has been around for decades now. It just hasn't really worked. And so this was one of the first studies using an anti-CSF1R. This was being used in relapsed refractory solid tumors, and they did skin punch biopsies to show that even from normal tissue it was very good at depleting macrophages and a variety of myeloid subsets. The reality, though, was that you didn't really see any significant clinical efficacy there. And similarly, there's lots of small molecules that have looked at inhibiting the CSF1R receptor or the CSF1 receptor. This is one that was studied early on, about 10 years ago. You saw very nice inhibition of the receptor activation, but we saw fewer than 5% of patients actually responding to this therapy. And as we move into the immunotherapy era, combining everything with PD-1 blockade, this study that came out a few years ago showed that a small molecule inhibitor combined with PD-1 blockade in patients that had already received PD-1 blockade didn't really rescue any of the benefit there. So we still have some work to go there, and I think that Jen's work really highlights the need to improve the therapeutic window. These drugs are somewhat toxic, so if you can really increase the CSF1R on the tumor monocyte-derived macrophages you're trying to deplete, we might actually finally see some benefit of these drugs. Many groups have looked at blocking recruitment. So we talked about how most of the monocyte-derived macrophages are being recruited from the bone marrow into the blood and then through a chemokine gradient into the tumor. There's many different ligands that are being secreted that many of them can bind to CCR2 and single cell analysis from our group of Murine lung cancer model. This is KRAS mutant p53 deficient, so similar to a smoking model. We see a very similar population in our smokers where we've resected tumors. There's a population that we were interested in depleting that's very high in CCR2, and this represents the Ly6C high monocytes. These are the monocytes that are trafficking into the tumor and differentiating into monocyte-derived macrophages. And indeed, when you knock out CCR2, you actually see a significant decrease in the growth potential of these lung tumors. So this was in a very intriguing data I saw at a lab meeting, and I said, "Let's plan a trial around this." We had a very nice neoadjuvant study. These studies give you very big chunks of tumor that you're cutting out, so you can really see if your drug is working. In this study, we looked at lung and liver patients that were having their tumors removed, and they just got a brief intervention of PD-1 blockade and CCR2/5 inhibitor, so it blocks both those chemokines. And the initial data just from the CBCs, if you look at the monocytes and the lymphocytes, you actually see a drop. So we wanted to see a drop in the monocytes. We want to see decreased efflux of these monocytes, these immature monocytes, into the blood and then into the tumor. But unfortunately, I think probably because there was a bit of a CCR5 inhibition, we saw a drop in the lymphocyte counts in the peripheral blood. And when we did single cell analysis on the resected tumors, these are very small numbers, so you're not going to see anything significant here. This is more hypothesis generating. But one of the things we did see was a drop in the numbers of Trem2 and SPP1, very immunosuppressive monocyte-derived macrophages. That was the goal. The only significant thing in this very small population that was depleted that were B cells. And so that's somewhat concerning because we've had some papers recently that have really demonstrated the significant importance of humoral immunity in response to checkpoint blockade. So I think this is one example of how you can do a very small study. As long as you do a deep dive, you can really see how your drug's working, and if you're using it appropriately, or you should be pursuing a more rational combinatorial approach. Unfortunately, I'll tell you the punchline of this study that we haven't reported out yet is when you add CCR2/5 inhibitor to nivolumab, it doesn't actually improve outcomes as far as pathologic responses. So we would be moving on to something else. So what are those other things? I think macrophages have lots of these eat me signals and do not eat me signals that really dictate their functional status as well as their ability to take up living or dead tumor cells. And there's been many different therapies that have approached the eat me and do not eat me side of this. The one that probably has the most shots on goal and was the most exciting is CD47. So CD47 is a do not eat me signal expressed on tumor cells, and it binds to SIRPalpha on macrophages, and it keeps those macrophages from eating them. It basically not only blocks phagocytosis of the tumor cell, it also leads to polarization of the macrophage towards a very immunosuppressive tolerogenic phenotype. And there was this beautiful study that came out in New England Journal. This was one of the first antibodies developed to bind CD47 that was put into humans. This was used in patients with relapse refractory diffuse large B-cell or follicular lymphoma. So all these patients had progressed on rituximab, and now they were being rechallenged with rituximab, but in combination with this thing that blocks a do not eat me signal. So basically, you're giving an eat me signal along with a do not eat me signal. And then we saw very nice responses here.One of the issues with this class of therapy is that these "eat me" and "do not eat me" signals are evolutionarily conserved and a very important part of your normal homeostasis. There's high levels of CD47 on red blood cells, particularly older red blood cells. And so a class effect of CD47 blocking antibodies, even those that have been optimized somewhat, is always a bit of hemolysis. Though it seems like this probably is manageable. However, unfortunately, most of the development in this space has ceased, including with this antibody. It was being tested in myelodysplastic syndrome as well as acute myelogenous leukemia, and both those studies were negative, and so this development was tabled. But there are some other groups that are more intelligently looking at CD47 blockade with a variety of "eat me" signals like FC-enabled antibodies that might be able to see some promise. So hopefully we'll see some data here at AACR, and there'll be some data at ASCO as well. Another approach that I think is an intriguing approach is a different "do not eat me" signal. That's CD24. This is also overexpressed on many solid tumors, particularly after treatment. And it basically binds to Siglec-10, which is expressed by macrophages and both disrupts the ability of those macrophages to phagocytose cancer, but also induces a very immunosuppressive functional status in those macrophages. And so potentially we could block this, and indeed there is a first-in-class antibody against CD24. I'm actually going to be opening this trial at Mount Sinai because it's an exciting approach, and they're going to be presenting some of their preclinical and early clinical data on Monday, and the poster number is here, so I encourage you to check it out. There's many of these TAM family receptors, and these are scavenger receptors that are actually an "eat me" function that really turns off the macrophages. So it encourages the macrophages to gobble up cell debris and really induces this lipid-associated macrophage phenotype that Jen was talking about earlier. One of the members that this family has, things like MerTK that have been extensively studied, as well as AXL. I think AXL is particularly interesting. This is a receptor on macrophages that binds to phosphatidylserine, which is seen on dying cancer cells that are either dying because of a therapeutic response or also just rapid turnover in the tumor immune microenvironment. And when macrophages take up these cells through this mechanism, it actually induces a very tolerogenic phenotype. And so potentially we could block this. And there's been many attempts to block this. Unfortunately, we haven't seen much success. I would say the most promising was this drug citrovatinib. So this was not AXL specific, but this bound to AXL and a few other. Many of these tyrosine kinase inhibitors are very dirty. In this second-line study in lung cancer, I was nearly certain this was going to be a positive study. We were comparing it to docetaxel, which should be a relatively low bar. A combination of PD-1 blockade and this AXL inhibitor did not reach statistical significance as far as improving median overall survival. It is a bit of a dirtier AXL inhibitor, but there are some more specific AXL inhibitors like bemsetinib, which has been tested and looked somewhat promising in very difficult-to-treat lung cancers, such as those who have SDK11 mutations. But unfortunately, those studies have also been stopped due to a lack of ability to show a significant improvement. So I do think that this is an interesting space. I think we just need to better understand the pathway and probably the most optimal combinatorial approaches. So I could talk for about 18 hours about all the drugs that are being developed to target macrophages, but thankfully, two of our trainees just published this review in "Cancer Cell" a few weeks ago, so I highly encourage you guys to check it out. They give a very nice postmortem as well as a status of many of the drugs that are in development that we're trying in the Phase 1 unit. So, our work in the lab and our work in our Phase 1 unit is really focused on understanding how we can disrupt this emergency myelopoiesis. So, as was described by my colleagues, there are a variety of tumor factors that are really hijacking your bone marrow. They're revving up the expansion of these monocyte progenitors, the granulocyte and monocyte progenitors. They then also imprint on them a very immunosuppressive phenotype. They then leave the bone marrow, and they traffic through a chemokine gradient into the tumor microenvironment, where they differentiate into these monocyte-derived macrophages. These then enable further tumor growth and metastasis, and there's a very vicious cycle here, and we need to figure out how we could potentially disrupt this. We've tried things like CCR2 inhibition to block this. That was not effective. And so one of the main focuses is understanding how, once these macrophages are already really being differentiated, how we might be able to reverse that. Or maybe this whole differentiation step here that's happening within the tumor, how could we potentially block that as well? So we've done extensive single-cell analysis on macrophages from both human lesions. These are untreated non-small cell lung cancer. We have similar data sets in colon cancer, in bladder cancer, and in liver cancer. And we see both in lung cancer and also in this preclinical model, when we're looking at their functional status, we see this very strong IL-4 signaling pathway that is induced. So this suggests that these cells, these monocyte-derived macrophages, have seen IL-4 and that IL-4 receptor-related signaling is really significantly activated in these cells and likely is doing something to drive their polarization. So we found this intriguing and IL-4 classically in medical school, we learn about this whole nice, once again, black and white thing, which is not real. But type one versus type two immunity. So type one immunity, these are the good guys, antiviral, antitumor responses. And type two immunity is thought to be this vestigial thing. So classically, we thought of it as important for antihelminth responses. We don't see a lot of helminths in New York because we use too much bleach and Purell. So we do have lots of atopy, though. So we cannot have bake sales in New York because everyone has horrible allergies, and these atopic conditions that are very strong in type two cytokines like IL-4, IL-5, IL-13, are also very pro-tumorigenic. This whole pathway is very pro-tumorigenic, and it overlaps with some of the functional phenotypes that we're seeing in these monocyte-derived macrophages. And this is not a new observation on our part. Lisa Coussens, who's a former president of AACR, her lab has been working on this for over a decade. They had this study that looked at blocking IL-4 and combining it with classic therapies for breast cancer, which is the model that she studies. And you can see here in combination with classic therapies like radiation or taxane therapy, you actually see synergy, which makes sense because the type two immune response is very much of a healing response. So if you are causing damage with radiation or chemotherapy, that's inducing this healing response, and that healing response is actually very pro-tumorigenic. So potentially by disrupting this, by taking out the IL-4, you might see some synergy.So we saw this data and we said, "Well, that's interesting. What is it going to do with PD-1?" Which is obviously how we're now treating triple negative breast cancer, lung cancer, and many other solid tumors. So using the KP model, which is exquisitely insensitive to PD-L1, it really doesn't do anything for those mice who are developing these metastatic lung cancers. If you block IL-4, here using a neutralizing IL-4 antibody, we actually see a significant decrease in the growth potential of these KP lung tumors. We have a slightly more responsive model, the KP1 model, where PD-L1 does do something. It does decrease the tumor growth potential. You can see that it's just as good as IL-4 blockade, and if you put them together, you actually appear to see some sort of immunologic synergy, really significantly reducing tumor burden. So I saw this data and I said, "That's super exciting. How can we put that into humans?" And what we're thinking is that we have this emergency myelopoiesis. It's a variety of these tumor-derived factors. And this is already published data, so you should take a look at this paper that shows that there's really a cocktail of cytokines and chemokines that are being secreted by the tumor. They're going systemically, they're hijacking your bone marrow. They're activating type II granulocytes, which are actually up to 5% of the cells in the bone marrow, which I didn't know until we did this study. You basically are activating these type II granulocytes. They're pumping out lots of IL-4. This is then imprinting the granulocyte monocyte progenitor. These then expand, and then these very immature immunosuppressive monocytes then traffic into the tumor microenvironment. And once again, we have this very vicious cycle. But nicely in preclinical models, if we use an IL-4 neutralizing antibody, you can really cut this off right here in the bone marrow. So the bone marrow is where all the IL-4 is being made, and using the systemic IL-4 blockade or IL-4 neutralizing antibody, you cut off this entire production of these monocytes and you see very nice antitumor immunity in many preclinical models. Thankfully, in humans, we have a drug that already has lots of safety that blocks IL-4. It blocks the IL-4 receptor alpha, so disrupting both IL-4 and IL-13 signaling. This has been studied in many atopic conditions, including one of my pediatric colleagues at Mount Sinai gave this to kids as young as one, so you know that it's a very safe drug when you're giving it to a one-year-old. So I said, "Why don't we try this in patients that have lung cancer who are progressing on PD-1 blockade?" So this is what I call a rescue study. So basically, I have patients that have relapsed refractory non-small cell lung cancer. These are patients that receive first-line immunotherapy with or without chemotherapy. Most of them do well initially, and then they're continued on maintenance PD-1 blockade, so typically pembrolizumab, or if they had stage III disease, durvalumab, and they're continued on that for up to two years. Unfortunately, most patients eventually start to progress. So all we do here is we just continue that immune therapy that stopped working, and we add on three injections of dupilumab, which binds to the IL-4 receptor alpha, so blocking that type II signature. We're not selecting for patients who have lots of type II immunity per se. And we did biopsies before and on treatment and collected a lot of blood to see how this drug was actually working in vivo in humans to see if it correlated with mice. Nicely, the fourth patient we put on this trial, this was a smoker who had received chemoradiation. He had a very nice response initially, and then he went on to immunotherapy consolidation. Very soon after starting immunotherapy, he had progression not only in the initial radiated tumor bed, but also in the retroperitoneum. He had a very painful lesion here in his back. All we did was continue immune therapy, and we added on three doses of dupilumab, and the patient went into remission. Basically, his pain regimen of morphine eight times a day went to zero within a few weeks, so we knew that the drug was actually working. And he then went on to receive two years of immunotherapy, and then he stopped, as is common practice. So we had some nice responses. We had these biopsies that showed that in a subset of patients, we actually were inducing a de novo antitumor response. We were seeing formation of tertiary lymphoid structures and nice changes. And we did see about a third of patients reachieve control of their disease. And so these were patients where their disease had been progressing, so they normally would just go on docetaxel, which is a mediocre and very toxic chemotherapy. And here, a third of the patients, after we gave them those three doses of dupilumab, could just continue their previous checkpoint blockade, and they achieved significant benefit. So we're in the process of really trying to figure out what is the biomarker that identifies this subset of patients that have some sort of type II signature mediated resistance, and then there's probably something else going on in the other two-thirds of patients. One way that we're doing that is we're exploring this in the neoadjuvant setting, and so we're giving a single dose of cemiplimab, which is a PD-1 antibody used in lung cancer, as well as a single dose of dupilumab. And then just two weeks later, we're taking them to the operating room so we can see exactly what's happening to those cells. We get biopsies, and we can compare the biopsies at baseline with the tissue. We're actually running this in collaboration with the SCHUM. It's supported by a CRI grant. So our working hypothesis of this emergency myelopoiesis is that there's these tumor-derived factors that are really hijacking your bone marrow, your type II granulocytes. They're pumping out IL-4. They're imprinting on your myeloid progenitors that are then effluxing from the marrow, trafficking to the tumor, and setting up this vicious cycle. You give some dupilumab, and in a subset of patients, that can disrupt this. However, like I said, two-thirds of patients are not benefiting, but we know that there's a variety of other potential targets here, and I told you there's a cocktail of about eight different cytokines and chemokines that are coming from the tumor and leading to this effect in the bone marrow.And one of the things that we saw that was part of that cocktail are interleukin-1 alpha and interleukin-1 beta, and we see a very strong interleukin-1 alpha and beta signature in monocyte-derived macrophages. So we've been studying lung cancer in old mice. They much better recapitulate the bone marrow of old humans and also the lung status of old humans. Nine-year-olds, thankfully, do not get lung cancer, so we shouldn't be studying lung cancer in seven-week-old mice. And one of the things that we did was we looked at IL-1 beta, as Tyler had said, we know that this inflammasome activation really drives an immunosuppressive phenotype. And anybody who hears about IL-1 beta lung cancer will run out of the room because we did huge Phase III trials of canakinumab once we saw a bit of a signal from a trial in cardiovascular disease called the CANTOS study, and those trials were all negative. The addition of IL-1 beta to standard immunotherapy did absolutely nothing. Here, when we use anti-IL-1 beta, the tumors grow just as poorly as they do in a control of, once again, 72-week-old mouse, which correlates somehow with a 65-year-old human. I'm not sure of the math equation there. And also we were interested in IL-1 alpha because we actually saw a stronger signature of IL-1 alpha that was specific to the monocyte-derived macrophages, but also IL-1 alpha blockade itself did nothing. But then when Matthew, who's a superstar MD PhD student in Miriam Merad's lab, she's my partner in crime, doing all these preclinical studies. If you add IL-1 alpha and IL-1 beta together, we all of a sudden saw this very nice signature. He had this data right around the time that we were grappling with the joys of COVID in New York City and basically throwing all spaghetti at the wall and seeing what stuck, and we had tried anakinra with some success in some of our patients. And so we threw anakinra in these mice, and we also saw a very nice anti-tumor effect. So we obviously didn't want to just look at one mouse model. That's always the best way not to publish something. So in, once again, our old mice, we had some orthotopic models of colorectal cancer with a cecal injection here or pancreatic cancer. And this is not a new finding that in old mice, tumors grow faster, and that's very well established. That's why old people get cancer, one of the reasons why old people get cancer. But interestingly, we see a lot more inflammatory inflammasome activation in the old mice versus the young. And indeed, when we give anakinra to these old mice in this colon cancer model and this pancreatic cancer model, we see a significant decrease in the growth potential. So we really think that we're disrupting this inflammation that occurs in the setting also of inflammaging, which is just the chronic inflammation that we have over time as we age, particularly living in cities like New York. And we really see that we're able to disrupt this inflammaging-related pro-tumorigenic milieu. So based on that, we're going back to our rescue model, and we're basically just adding IL-1 blockade to this. And so we're doing a trial right now where we give 28 days of anakinra. This is a daily injection, which is a bit of a pain for patients, but we're trying to see if we can see some synergy between IL-4 and IL-1 blockade. So finally, I just wanted to talk about some of the work that we're doing in a much more difficult setting. I'm not sure why this is blurry, but in colon cancer, microsatellite colon cancer, we know that these are patients that do not respond to immunotherapy, and this is particularly true for patients with colorectal mets. And so this is actually from one of our patients who had their primary colon cancer removed, as well as an oligo met from their liver. And you can see that in the primary tumor, there's lots of CD8 T cells. The colon cancer is where we actually developed the Immunoscore many decades ago, so we know that there's lots of T cells there. They're just not doing their job. But in liver mets, we really see a dearth of T cells, and we see a very stromal fibrotic lesion. And this is one of the many reasons why we think we really don't see any response in patients with liver mets. This is the study of botbal, this is a souped-up CTLA-4 given with a PD-1. And we've seen that in the patients with liver mets that we put on this combination, we really don't see any benefit. While we see a significant benefit in a subset of patients, particularly in some of the neoadjuvant studies that have been running, and we're going to be updating this study in the near future. So how can we overcome that very resistant milieu? So the signature, this monocyte-derived macrophage signature that I've been talking about and that Jen and Tyler have been talking about, we really see it across tumors. So we're so myopically focused oftentimes on let's develop a lung cancer therapy, and we don't realize that the immune system can only be hijacked in a finite number of ways. And we really think that this Type 2 signature is seen in every tumor type that we've ever looked at. So we initially said, "Well, let's try IL-4 blockade." And indeed, in this model of metastatic colon cancer to the liver, so this is an AKPS model where we inject cells into the spleen and they metastasize to the liver. It's extremely insensitive to PD-1 blockade. It might actually even cause some hyperprogression if you believe in that. But if you give IL-4 receptor alpha blockade to these mice, you actually see a significant decrease in the tumor growth, and you actually see synergy with PD-1 blockade, even though you didn't see any monotherapy activity here. It's an important thing for investors who are always insisting on monotherapy activity. Here you actually see the combination. And interestingly, when we look at tissue from these mice, you see that IL-4 receptor alpha blockade induces a bit of an influx of immune cells, but you really see lots of these CD8 T cells infiltrating when you give this cocktail of the two. So we have some nice preclinical rationale. And one of the interesting things was when we were looking in colon cancer in humans, we have a large single-cell data set that we're going to be publishing soon of human patients that had untreated colon cancer that was resected. And we noticed that unlike our lung cancer patients, where monocytes that we saw in the blood had already seen IL-4, had been imprinted in the bone marrow. We don't see that in colon cancer patients, and we don't see that in these mice that have colon cancer, either an orthotopic colon cancer in their cecum or metastatic to the liver. We didn't see them in the monocytes in the blood, but we see them in the monocyte-derived macrophages in the tumor. So it suggested maybe in this setting, the IL-4 is coming from somewhere else. And indeed, with an IL-4 reporter mouse, we saw that the eosinophils, there was lots of eosinophils in our colon cancer patients, but particularly in this preclinical model. And they were really making the vast majority of the IL-4, and we see an increase in the number of eosinophils in these lesions compared to a normal liver, and all those eosinophils are pumping out IL-4. And doing some receptor ligand interaction, what we realize is that that IL-4 probably is related to fibroblast production of IL-33. So IL-33 is an alarmin. There's been a few drugs recently that have shown some promise in COPD. Targeting IL-33, it's also a Type 2 cytokine of sorts. And what we see is that in tumors, particularly the fibroblasts, are making a lot of IL-33, and we know IL-33 is very good at maintaining the survival and activity of eosinophils. And when we look at the... This is, once again, the mouse model of the liver met. We see that within the tumor, there's lots of cells that are expressing IL-33, and we see that they're all these fibroblasts, these stromal cells surrounding the tumor cells. When we look at our human samples, we also see lots of IL-33 coming from the fibroblasts. And basically, our data is really showing that your tumor-derived, and particularly CAF-derived IL-33, is activating the eosinophils in the tumor immune microenvironment. They then make IL-4 that then polarizes the monocyte-derived macrophages. We can even block this whole upstream signaling by using an anti-IL-1 RAP. This docks out or blocks not only IL-1 alpha and beta, but also ST2, the IL-33 receptor.And so here, once again, in this model that doesn't respond to PD-1 blockade, if you add anti-IL-1RAP, it appears to have a very nice anti-tumor effect and synergize. And so we have these two different models where we block IL-4 downstream or upstream. Both seem very promising. So we have this trial, a platform study now ongoing, where two of the arms of the platform study are looking at these cocktails in patients with colon cancer that's metastasized to the liver. And we also have an oligometastasectomy trial looking at the use of PD-1 and IL-4 very briefly in patients that are having tumors resected from their liver. So with that, thanks first and foremost to the patients who are part of these studies, that spend way too much time with me and my colleagues. Miriam Murad is my partner in crime, and these sorts of studies literally take hundreds of people. So thank you. Happy to take questions. Hi. Great talk. So I was curious, this is the first I've seen eosinophils really brought into the mix of the tumor microenvironment. Do you expect that you would see eosinophils driving this IL-4 production in other tumors as well? Do you think this is specific to CRC, and do you think it's also specific to MSS CRC, or would you see a different immune signature in MSI where you have a lot more of the mutagenic proteins? So in lung cancer, we actually do not see local production of IL-4. We don't see a lot of eosinophils either. Eosinophilia in the blood is a very mixed bag. If you look at studies correlating response to immunotherapy with eosinophils, some papers say eosinophilia is a great marker for response, and others say it's very immunosuppressive. They're definitely a complicated cell. They're a difficult cell to study, and so that's probably why we don't know enough about them. Interestingly, when we look at different tumors in humans, but also in the models, we see IL-4 as part of all of their stories, but it's always coming from somewhere else. So in ovarian cancer, we're actually seeing IL-4 being produced by ovarian cancer cells themselves, and strangely, we actually see this not only in human samples, but also in pre-clinical models of ovarian cancer. And so I think that it's more complicated than we realize. I was just telling you, we have shared immune programs throughout. I think IL-4 is a shared immune program throughout. It's just where the IL-4 is coming from might be different from one tumor to another. It might also be different from one patient to another. Thank you. My question is, do you have quantified the result about how many myeloid cell the human body mobilized? You mean in the setting of cancer, how many are being mobilized? Yes. I think we all showed a picture that showed that sometimes the majority of immune cells in a tumor, so literally tens of billions of cells, are actually macrophages, and those are primarily monocyte-derived macrophages. So we see this emergency myelopoiesis, this expansion of granulocyte, monocyte progenitors, really across all the tumors, and you see a significant expansion compared to any sort of healthy control in the bone marrow and the blood, then obviously the tumor is chock-full of these cells. The half-life is a real big question. We don't know that in humans. We know that they're constantly being recruited and enabling tumor growth and metastasis, but we don't know how long they actually live in humans. Unfortunately, that's very difficult to track. So we don't know, are they turning over once a week, every 72 hours, or are they sticking around for a few months? So how great is that recruitment? And that's an important question if we're thinking about finding ways to block recruitment, for sure. Actually, my question, it also has a specific setting. It's the cancer patient after the treatment, after chemo or radiation treatment, and how much the bone marrow mobilize the myeloid immune cell get into the tumor. Do you have estimation? No. I don't have specific data on that. I think that we know that the tumor is releasing all these factors, so as you kill off tumor with radiation and chemotherapy, you could argue you're going to have fewer cancer cells and fewer stromal cells pumping out those cytokines and leading to expansion in the bone marrow. But also, as part of this healing response to those very toxic insults, that might actually do the opposite and really drive that expansion. So it's hard to really know. I bet it's going to be very different from one patient to another. Thank you. One more question. Yeah. Really nice data. My question is, does this IL-4 blockage change the PD-L1 expression in the tumor or in the macrophage that could also interfere with the association with PD-L1 treatment? That is a very interesting question, and I do not know. We're looking at the tissue right now, the pre- and on-treatment biopsies, and so we'll see if we see a difference. The issue is that in the responders, they're going to have more T cells pumping out interferon, and that's going to induce upregulation of PD-L1. So it's going to be hard to sort of isolate the effect of the IL-4 versus just the IL-4 is leading to better anti-tumor immunity, and then you have more PD-L1. Okay. Thank you, Tom. And before we conclude this session, I just want to acknowledge the involvement or engagement in this room. We had two full overflow rooms and hundreds of people online, so I think we won in terms of how many people attended our session. Sure. But I think it reflects the movement in the field. So, we know now that it's not if we should target macrophages, it's just that we need to target them and how to do that. So, thank you all for being part of this movement, and thank you to Dr. Marin and Dr. Miller