Scientist Stories: Carolyn Bertozzi, Therapeutic Opportunities in Glycoscience

So first of all, it's a tremendous honor for me to be joining you virtually, and especially to honor the life and the career of Max Perutz, who was such an important figure in science as we know it today, and in both in biology and chemistry, and particularly in structural biology. Such a storied history at the LMB, so I'm deeply honored to be with you today, sharing in this celebration. Let me share my screen and get my PowerPoint deck fired up. So I thought I would take the time this afternoon for you to tell you about a project in the lab that started back when I was at UC Berkeley, just about 11 or 12 years ago now. And it's basically a project that started out with some some basic science and some curiosity around an interesting glycosylation phenotype that had a longstanding association with cancer, and how our learnings in the early days of the project have led quite recently to the development of a new immune therapy candidate, which has just started human clinical testing about a month ago. So it's a therapeutic and opportunity— therapeutic opportunities talk, but focusing on cancer immune therapy. So the backstory is that over half a century ago, people who study molecular changes in cancer made the observation that among those changes, almost ubiquitously, are changes in the pattern of protein and lipid glycosylation on the surface of cancer cells, and one of the most prominent glyco-phenotypes as we refer to them, is the observation of increased density or overproduction of glycans on cell surface proteins and lipids that terminate with a particular monosaccharide called sialic acid. So we call this class of glycans the sialoglycans, and the way I like to think of these observations is that a healthy cell will have a certain pattern of sialoglycans on the surface of that cell, but in the setting of a cancer, and quite prominently in solid tumors, those sialoglycans become more abundant, and sometimes there's new kinds of sialoglycan structures that you might not observe on normal healthy tissues. And so the metaphor I often use is that the healthy cell has kind of a well-manicured garden, of these sialoglycans, whereas those cancer cells in the tumor setting have like a tropical rainforest, like an unkempt jungle of sialoglycans. And the earliest paper that I have found which points to this phenotype was from 1959, but then there's dozens of these observations that have been reported in the literature in the 70 years or so since that time. And throughout those decades, although the phenotype was quite prominent, no one really had a good understanding as to why this phenotype might arise. So back in the years of 2010 or '11 or '12, there was a convergence of understanding around this phenotype that came from the field of cancer immunology. So it was over the last 20 years or so that we have come to understand how important our immune system is in protecting us from cancer most of the time. So one of the jobs that your immune cells have is to patrol your body for signs of damage in our own cells. And when a damaged cell is encountered, your immune system's job is to get rid of it before it can cause more harm. And so cancer cells are damaged cells. And they present evidence of this damage in various cell surface molecules, and that's a sign to your immune cells that it's time to get rid of this cell. And so most of the time, these damaged cells are cleared out of the body before we develop a disease. But as you might imagine, just through the process of evolution and natural selection, if a cancer cell gains the ability to basically trick the immune cells into thinking that it's healthy, that's a cancer cell that will live another day and will have the opportunity to proliferate and eventually to form a tumor and then to disseminate, and that's when you have a disease. So at a very high level, the way that the immune cell can sense whether that target cell is healthy or diseased is through engagement of a collection of receptors on the immune cell with ligands on the target cell. And those receptors can signal to the immune cell in two different ways. Some of those receptors are activating receptors, like this green shape. And those activating receptors are looking for ligands on the target cell that present evidence of damage. And if those activating receptors cluster at this synapse between the immune cell and the target cell, at high enough density, they will deliver this inhibitory signal and cause that immune cell to activate, which then leads to a response to kill the target. And the nature of the response depends on the specific type of immune cell, and there's lots of different kinds of immune cells, right? There's T cells and B cells, which are cells from the adaptive immune response, and then there's macrophages, neutrophils, NK cells, these are cells of the innate immune response. At the end of the day, all of those immune cells have the same job, which is to provoke an immune reaction against this damaged cell. But then at the same time, your immune cell has receptors that are inhibitory, and if those receptors engage ligands on the target cell, the inhibitory signal can quench the activating signal that might be occurring in parallel and block the immune cell from activating. And this is important because your immune cells have a difficult job of having to distinguish between the damaged cells and healthy cells, and there's a risk of reaction against healthy cells, which would then lead to autoimmune disease. So these inhibitory receptors play the role of protecting us from autoimmunity, and it's the balance between the inhibitory signals and the activating signals that allows the immune cell to try to make a good decision. So as you might imagine, a cancer cell might, just through natural selection, arise at a phenotype where it is presenting ligands to these inhibitory signals, and that would be an advantageous phenotype for the immune cell. So once this biology began to be understood, it became clear that one could develop immune therapies that modulate the balance between the activating and the inhibitory signal. And the big breakthrough immune therapies of the previous decade were immune therapies that block the inhibitory signaling receptor on the immune cell, or the ligands that deliver the inhibitory signal on the cancer cell. And the most prominent of these have been what we call the T-cell checkpoint inhibitors that block inhibitory receptors on a T-cell that go by the name PD-1 or CTLA-4. And the approval of monoclonal antibodies that do this was one of the great breakthroughs in oncology over the last two decades. So let me blow up that synapse between the T cell and the cancer cell in a little bit more detail, because this is basically the axis of immune modulation that's best understood in the field, and which has led to some of the most important immunotherapy drugs. And two of the big players on the T cell are receptors PD-1 and CTLA-4, and these are both inhibitory receptors, and they bind ligands on target cells, such as PD-L1 for PD-1 and a B7 family of receptors for CTLA-4. And if these receptors engage their ligands at high enough density, the inhibitory signal can quench an activating signal, which might otherwise come from the T cell receptor, interacting with MHC molecules that are presenting neoantigenic peptides. They're the products of translation of mutated DNA, and that's how they basically tell the immune system that there's been mutations and therefore damage in this cell. Now, the activating signal through the T-cell receptor involves a kinase cascade, and there are activating signatures in the cytosolic domain of the T-cell receptors. They're called ITAM domains. The immunologist will recognize these as domains that allow the association of kinases. Involved in this signaling. And the way that PD-1 quenches the signal through the T-cell receptor is through its own signaling domain in its own cytosol. So PD-1 has amino acid motifs that we call the ITIM and the ITSM domains, and they're simply docking sites for phosphatases that are called the SHP phosphatases, SHP-1 and SHP-2. And so there's this battle in the cytosol between the phosphatases and the kinases, and the SHP phosphatases will dephosphorylate these signaling intermediates in the TCR signaling cascade. So that's how PD-1 is, is able to deliver its inhibitory signal, through the recruitment of these so-called SHP phosphatases. And this important biology, first of all, was recognized with a Nobel Prize back in 2018, and second of all, has led to some very impactful immunotherapy drugs. And so they're all monoclonal antibodies, And the probably most recognizable among them is Merck's number one selling drug, which is Keytruda, and it's an antibody that binds PD-1 on the T cell and blocks the interaction with PD-L1. And Keytruda is not alone. There are now several other drugs from other companies that compete in the same space, such as Bristol Myers Squibb Opdivo. And on the other side of the synapse, there are drugs now approved that target the ligand for PD-1. Such as Tecentriq, which is Roche and Genentech's drug that binds to PD-L1. So these drugs have been transformative in the sense that there are patients with otherwise lethal forms of, of aggressive and advanced cancers that have actually been cured with these drugs. For example, they're approved in treatment of metastatic melanoma, bladder cancers, certain forms of lung cancer, And there are patients who have a dramatic response to these medicines. Those patients, however, tend to be in a minority. So for example, of patients with metastatic melanoma that meet the criteria for treatment with a blocker of PD-1 or PD-L1, maybe 20 or 30% of those patients have a durable response where the tumors shrink, and in some cases, those patients are cured. But the other majority of patients, right, the 70% or so, that don't respond, their tumors don't shrink and they don't have a durable response. And what's been missing in the field is an understanding of why so many patients who look molecularly like they should respond, they're indistinguishable from the responders in terms of their PD-L1 expression on the cancers, why don't they respond? How come blocking this pathway doesn't potentiate the immune response against those cancers? And this is where we started thinking about the glycoscience and that phenotype of having overproduction of sialic acids, which is so prominent on so many different cancers. Because it turns out there's a family of immune modulatory receptors that bind sialic acid, and they're called the sialic acid-binding immunoglobulin-like lectin, or Siglec family. And this is a family of receptors that is quite large in humans. There's 14 members of this family, and they're shown here in their cartoon form as transmembrane domain proteins, where there's a— the yellow line is the plasma membrane, and outside the cell, each member of the Siglec family has a variable number of immunoglobulin domains, Ig domains, capped with a domain that binds sialoglycans.. And what's interesting from the standpoint of their potential role in immune modulation and cancer is that inside the cell, in their cytosolic domains, many members of the Siglec family have those amino acid motifs, the so-called ITIM and ITSM domain motifs, that map precisely onto the signaling domains of the T-cell checkpoint receptor PD-1. So members of this family with the ITIM and ITIM-ITSM motifs are able to recruit SHP phosphatases to the synapse and deliver an inhibitory signal to the immune cells that they populate. But unlike PD-1, which is quite restricted to the T-cell, the Siglec family of receptors are distributed across virtually every immune cell type in the body, both adaptive immune cells like B-cells and T-cells,, as well as innate immune cells like macrophages, NK cells, neutrophils. And if you look at this cluster of Siglecs right in the middle here, which is Siglecs 7, 8, 9, and 10, as we call them, they are found on virtually every immune cell type which is known to be important in protecting us from cancer. So activated T cells express Siglec-7 and Siglec-9. Macrophages, including those that infiltrate tumors, express Siglec-7 and 9 and 10, as well as other Siglecs. So understanding the importance of checkpoint receptors like PD-1 in protecting us from cancer, and understanding how similar the signaling biology is for these inhibitory Siglecs compared to PD-1, we couldn't help but speculate that maybe the reason so many cancers seem to evolve this overly sialylated phenotype is because that allows them to engage members of the Siglec family on various immune cells, and to suppress the activity of those immune cells, just as engagement of PD-1 with PD-L1 between the T cell and the cancer cell can suppress the reactivity of the T cell. And then furthermore, maybe the reason a majority of patients who get prescribed a checkpoint inhibitor immune therapy don't respond is because those patients might have such strong inhibitory signaling coming through the Siglecs that PD-1 is not an important player. And of course, the corollary of that is that blocking the binding of Siglecs to their receptors might be an alternative and maybe even more broadly efficacious way to potentiate immunity against cancer as a therapeutic modality. So we started working on this problem, and where we started— and this takes me back now to around the year 2011 or 2012— was by looking at the role that Siglec engagement and signaling might play in modulating the activity of NK cells and macrophages. So both of these cell types have an activating receptor called Fc gamma RIII. And one of the important ways that these cell types can contribute to anti-cancer immunity is by performing a process called antibody-dependent cell cytotoxicity. Or ADCC. And the synapse that they form to kill cells through this mechanism occurs between Fc gamma RIII and antibodies that opsonize the cancer cell by binding to neoantigens on the cancer cell surface. And this is a component of our own immune reaction against cancer, so our body can produce antibodies to contribute to the anticancer response, But it's also an important mechanism for the activity of monoclonal antibody cancer drug treatment. And so there are a number of antibodies that biopharma companies have developed against cancer antigens that are used as important treatments for cancer. For example, Genentech has a drug called Herceptin, which binds HER2 overexpressed on breast and gastric cancers. And there's the very important leukemia drug Rituxan, which binds to CD20. These are two examples from about a dozen antibodies that are used to treat cancer, and this is an important mechanism for all of those drugs. And so we wondered, normally an NK cell or a macrophage would do a good job killing a cancer cell if it was able to bind to an antibody-coated target, but what if that cancer cell was overproducing sialoglycans, as so many are found to do? Would it be possible for those cancer cells to undermine the ADCC response by engaging a Siglec on the NK cell or the macrophage that delivers an inhibitory signal? And we thought this was a reasonable hypothesis because the biochemical signaling pathway of Fc gamma RIII is very similar to the T-cell receptor. And we knew that T-cell receptor signaling could be quenched by PD-1 signaling. And so would it be possible that Siglec-7 signaling could quench Fc gamma RIII signaling. And so we set up a bunch of experiments to test this, and sure enough, it turns out that Siglec-7 is a checkpoint that undermines signaling through Fc gamma RIII. And we published this back in 2014. And this was true not just in terms of the innate reaction against endogenous antibodies that a cancer patient might produce, but it was also a resistance mechanism for drugs like Herceptin and Rituxan, which might explain why, you know, in the setting of their treatment, 50% of patients don't respond to these drugs. Either. So with that initial observation, a series of work from both our lab and other labs continued to basically expand upon our understanding of the role that Siglecs play in undermining the immune reaction against cancers. And labs like Stefan von Gunten's lab over there in Scandinavia showed that in human clinical samples, you could find upregulation of Siglec-7 and -9 on NK cells, and he showed a role for them in undermining immune surveillance in human tumor settings. Ajit Varki's lab contributed to this by understanding the role that Siglec-9 plays in macrophage activity against cancers. Joy Burchell over there in the UK did beautiful work showing that Siglec-9 engagement on macrophages by a mucin called MUC1 can polarize them from the M1 to the M2 phenotype in the tumor microenvironment. Heinz Läubli did gorgeous work with animal models of breast cancer showing how important Siglec-7 and 9 are in protecting cancers from immune surveillance. Li-Ping Cheng, who is one of the sort of founding scientists in the field of immunotherapy, showed a role for Siglec-15 on macrophages in suppressing activity and blockade of Siglec-15 as a potential immunotherapy. And then finally, my own colleague Irv Weissman discovered that Siglec-10 on macrophages can contribute to immune suppression, by binding to a cancer-associated glycoprotein called CD24. And when all of these studies sort of were synthesized and brought together by CNE News in a very nice article back in 2019, it became clear that glycoscience might provide the answer to this question that was open in the field, which is why so many patients don't respond to these first-generation immune therapies targeting PD-1 and CTLA-4, and suggesting that the Siglecs and their sialoglycan targets might be a new generation of targets that might be more successful for cancer immune therapies. So here we are today in the year 2022, and the field has now shifted over towards understanding how we might drug the interaction between the Siglecs and their ligands. And so this is now pretty much the focus of my lab, as well as a company that I formed, which is located in Massachusetts, co-founded with Paul Crocker, an immunologist from Dundee, and we're trying to do basically this very thing, was develop new therapeutic strategies. So we contemplated two different ways you could think about blocking the interaction of Siglecs with their sialoglycan ligands. On the one hand, you could take a page from the playbook of the checkpoint inhibitors like Keytruda and develop blocking antibodies that bind Siglecs and prevent their interaction with ligands. And these kinds of blocking antibodies have been described. If you just do a monoclonal antibody development campaign against a Siglec and screen enough clones, you can find some that bind to these receptors and block the interaction with their sialoglycan ligands. The problem is there's not just one Siglec. It's a large family. Many of the Siglecs are expressed often in combinations on immune cells. So your typical macrophage will have Siglec-9, Siglec-7, Siglec-10, Siglec-3, and they're regulated, you know, in different settings. So you would have to know which Siglecs would you have to block of all of the Siglecs that are potentially co-expressed on a given immune cell type. And if you don't know which Siglecs to block, or if many Siglecs are contributing to immune suppression in a particular patient, then picking one Siglec might not be all that efficacious.. And so instead of going after a Siglec, not knowing which ones you would have to block for a given patient, we thought another potentially more broadly applicable approach might be to target the ligands for the Siglecs, and in particular to go after molecular features of those ligands that are common to every member of the Siglec family. So that pointed us to the sialic acids themselves. Because different Siglecs, it turns out, they bind different kinds of sialoglycans with specificities. They're not totally redundant with each other. They actually can sense what's underneath the sialic acid, and they have preferences for different linkages, different underlying structures. But at the end of the day, every member of the Siglec family requires the sialic acid residue as an important component of its binding determinant. So we conceived of a therapeutic strategy where we would develop a molecule that functions like a molecular lawnmower. That's the metaphor that we use, and it would be a molecule that basically you can target to a cancer cell and park on a cancer cell, and that would then degrade the sialosides the same way a lawnmower can just kind of cut the grass and mow the sialic acids off the cell surface depriving every Siglec of the opportunity to engage with this cancer cell. So what kind of molecule would function as such a lawnmower? And we came up with the idea of making a bispecific biologic that was part antibody, and that would park this molecule on the cancer, and then part enzyme. And the enzymes that we use are called sialidases. And the sialidase catalyzes the hydrolysis of the glycosidic linkage between the sialic acid and what lies beneath. So the first prototype molecule we made that was an antibody sialidase conjugate used the antibody trastuzumab, which is in the drug Herceptin, and it's Genentech's antibody that binds to HER2, already an approved drug for treatment of certain breast cancers and gastric cancers, and we fused to trastuzumab the sialidase from a bacterium, Salmonella. And this was a molecule that we used as a proof-of-concept molecule, not as a human therapeutic candidate, because of course a bacterial enzyme wouldn't be ideal for that purpose. It would be immunogenic in people. But this is where we started, and this was the work of a former postdoc named Han Xiao, who's now a professor at Rice University, and a former graduate student, Melissa Gray, who is now a postdoctoral fellow in Stephen Bannock's lab. At Stanford. And in thinking about how they might join the enzyme and the antibody together, we realized that at least for our proof-of-concept molecules, it would be quite simple to do this in a modular fashion using bio-orthogonal chemistries. So Liz, when you introduced me, you mentioned bio-orthogonal chemistry as something that we've been interested in for a very long time, a decade before we even started thinking about immune therapies, and bioorthogonal chemistries are simply chemistries that have selectivity such that you can perform them on complex biological molecules, and even in cells or living animals. But here we use these chemistries to do a site-specific conjugation between two highly functionalized proteins, the antibody and the sialidase. And the way we, the way we did this was we started by functionalizing our antibodies site-specifically with aldehyde groups. And we had already developed a technology to do that, which we had applied to antibody drug conjugates. And so here we just borrowed a page from our own playbook to use these chemistries to make antibody enzyme conjugates. So we install the aldehydes site-specifically using a technology called the aldehyde tag. That's all published work, and it's even been commercialized, so I'm not going to go through the details there. It suffices to say that we can position aldehydes wherever we want them, genetically, and then do a chemistry which is basically an adaptation of the Pictet-Spengler reaction, which is a named reaction from the organic literature that's over 100 years old at this point. And that Pictet-Spengler reaction occurs when you have an indole functionalized with this dimethylhydrazine. And the hydrazine reacts with the aldehyde to form a hydrazone, which then undergoes an intramolecular cyclization, which is like an electrophilic aromatic substitution on the indole, to form this carbon-carbon bonded product product, and that allowed us to introduce this little spacer molecule capped with an azide. Now, the azide is positioned to undergo a copper-free click chemistry reaction, which we enable by functionalizing the sialidase on one of its cysteine residues with this little spacer that has a dibenzocyclooctyne, and the cyclooctyne reacts with the azide in a very selective way to form this triazole product And through this series of two different reactions, we're able to position the sialidase right here, covalently linked near the C-terminus of the heavy chain of trastuzumab. So we call this molecule T-SIA for trastuzumab sialidase, and we had engineered this molecule with the hope that it would selectively remove sialic acids from HER2-positive cells, which would be breast cancer cells, without removing the sialic acids from HER2-negative cells, which would be other healthy cells in the body. And it's important that we we're not stripping sialic acids off of healthy cells, because that would potentiate autoimmunity against our healthy tissues. We don't want that. We want to strip the sialic acids off of just the cancer cells to render them susceptible to killing by immune cells. So to test the selectivity of this reagent, what Melissa did is she set up reactions where she had mixed together cells that were HER2 negative and HER2 positive in the same tube, both cell types having roughly the same amount of cell surface sialic acid. And we can quantify the sialic acids by staining our cells with a lectin called SNA and then analyzing them by flow cytometry. And so you can see this two-color plot where the HER2 is on the x-axis and the SNA staining is on the y-axis. You can see the two populations that are HER2 positive or HER2 negative and roughly having the same amount of sialic acid. Then when we add our T-SIA reagent to this tube in various doses, we can monitor its efficacy in taking off the sialic acids by just looking at changes in SNA staining. And what Melissa observed is that even at treatment with quite low doses of T-SIA, sub-nanomolar doses, she could quantitatively strip the sialic acids off the HER2-positive cells without any removal of sialic acids from the HER2-negative cells. So you can see this HER2 population has dropped from the upper quadrant to the lower quadrant, and it was only way out here at triple-digit nanomolar doses of this conjugate that we started to see a little bit of loss of sialic acid from the off-target cells, but this is a really nice therapeutic window, so to speak, in which we have good selectivity for the target cell type. So we joined forces with Heinz Laubli, who is an oncologist at University Hospital Basel, and who had was working in a breast cancer model in the mouse, a syngeneic orthotopic model in which the EMT6 breast cancer cell line is engineered to express human HER2, and then can be engrafted in the mammary fat pad. And this is a model that's notoriously resistant to the naked antibody trastuzumab. And Hines and his grad student, Michael Stanczak, showed in their own lab that when you grow tumors in these mice, and then treat them at various points with trastuzumab or saline, there's no difference in the growth of these tumors over time. But by contrast, when he administered Melissa's T-SIA conjugates at various doses, those tumors had a much harder time engrafting in these animals because they were infiltrated with immune cells that were preventing their expansion. So showing that T-SIA could be efficacious in a breast cancer model that was normally resistant to trastuzumab gave us confidence that it was time to think about how to translate this concept of the antibody sialidase conjugate into a human therapeutic. And of course, that's difficult to do in an academic lab, and when you're in the Bay Area like I am, it's sometimes easier to translate work from the lab into human therapeutics by starting a company. So as I mentioned, I've joined forces with Paul Crocker and with an entrepreneur named Jim Broderick, and the three of us founded a company called Pallion Pharmaceuticals. And Pallion is actually located in Waltham, Massachusetts, so it's outside of the Boston area. That's where Jim is located, and he's the CEO of our company, but it's also kind of the midpoint between where I live in California and where Paul Crocker lives in Scotland. Boston is about, you know, 6 to 8 hour flight for both of us. So the Pallion scientists did some really important optimization of the structure of this antibody enzyme hybrid molecule. Importantly, they were able to figure out how to make a conjugate between Trastuzumab and a human sialidase called NU-2, and as well, they've kind of re-engineered the molecules, so instead of being a chemical conjugate between the enzyme and the antibody as we had made, they have now made a genetic fusion between the human sialidase and the Fc region of the heavy chain of an antibody, so that their molecule looks like this, where the sialidase just replaces one of the Fab domains, so the molecule has a size that's more like a natural antibody, and it's not really big like the molecule we had made. And also, it's made entirely through recombinant expression without any need for chemical conjugation, and that makes the manufacturing much simpler as well. So their lead candidate has progressed to the point where we started our first Phase 1 study about a month ago, and so patients are now being dosed with Palaeon's lead antibody sialidase fusion molecule, and we're really hoping that patients will benefit from this, but we are going to test this hypothesis in humans over the next year, so it's a very exciting time for us. Meanwhile, we felt that we were flying a little bit blind because there was so much we didn't know about the natural ligands for the Siglecs in the actual setting of a tumor. We knew that different Siglecs have a preference for different kinds of sialoglycans, and that knowledge has come from many labs who have done screening of recombinant soluble Siglecs in glycan microarray binding formats, but the actual macromolecules that are supporting Siglec binding on tumors remained, for the most part, mysterious. And we thought the more we understood about them, the better job we could do in identifying patients who might respond to drugs that block Siglec ligand binding. So a former postdoc named Simon Wisnowski, who's now a new faculty member at University of British Columbia, decided to start de-orphanizing Siglecs and identifying their biological ligands using unbiased CRISPR screening platforms. And he started by taking a cell line called K562. It's a chronic myelogenous leukemia cell line where we knew, from just doing flow cytometry binding studies in our lab, that these cells were able to bind different members of the Siglec family. So they expressed ligands for Siglec-7 and Siglec-9, for example, and we thought through CRISPR screens we could identify what those actual molecules were. On the surface of the K562 cell. So we got a cell line that had been transduced with a library of single guide RNAs targeting every open reading frame in the human genome, and from that pooled CRISPR guide RNA transduced cell population, we did a flow cytometry sort using soluble Siglec-Fc reagents that had fluorescent probes on them, and then we could sort out from this population those those transduced cells that had lost their ability to bind various Siglecs, do the DNA extraction and the deep sequencing, and find what guide RNAs corresponded to Siglec ligand expression. And so here's an example of CRISPR screens that Simon did where he compared the hit list for Siglec-7 and Siglec-9. So on the x-axis, each dot here that has a significant score down on the x-axis, that's a gene that's important for the ligands of Siglec-7. By contrast, those dots that have a high score on the y-axis, those are genes that are important for the ligands of Siglec-9, and any dots that are on the diagonal are genes that contribute to the ligands for both Siglec-7 and Siglec-9. And you can see there's some genes out here on that diagonal that are the genes that encode enzymes for sialic acid biosynthesis. You would expect that those genes would be important for any Siglec, including Siglec-7 and 9. And similarly, we found lots of genes that are involved in Golgi trafficking, and that also is important for any ligand for a Siglec, because those genes bring molecules to the plasma membrane. But what was interesting is that there were genes that we saw that were very important for Siglec-7, ligands, but totally irrelevant for Siglec-9 ligands, testifying to the fact that these two Siglecs, which are actually quite related to each other as a family, they bind different things on the surface of the K562 cell. And there was one glycoprotein that really stood out as a Siglec-7-specific gene, which was the gene encoding a mucin glycoprotein called CD43., as well as genes that are involved in the biosynthesis of O-linked glycans, which are the glycans that, you know, define the mucin structure. So a mucin is a glycoprotein that has lots of serines and threonines that are clustered over extended regions of the protein that are all modified with glycans, and a consequence of having these dense clusters of O-glycans is that mucins have an extended rigid structure, and they kind of stand tall on the surface of cells, like, kind of like the giant redwood trees of the cell surface. And CD43 is an example of a mucin that actually has a long-standing association with cancer. It has been observed to be upregulated on a number of different cancer types, both leukemias and solid tumors, and in a number of clinical studies, its expression levels have been inversely correlated with clinical outcome. So here's an example of a study in which CD43 expression was found to be an adverse prognostic factor for diffuse large cell— large B-cell lymphoma. So patients that had a lot of CD43 did more poorly, right? If you looked at the Kaplan-Meier curves, they had poor survival, poor progression-free survival on chemotherapy treatments. And these kinds of studies populate the literature, but again, no one has really had a proposal as to why overexpression of CD43 might, you know, be promoting cancer progression. And so we did a deep dive on CD43's structure and glycosylation pattern on this cell type to understand why it was so important for Siglec-7 binding. And it turns out that Simon basically— I won't go through the details because this is also all published work now, but there are hundreds of glycosites in the extracellular mucin domain of CD43, but near the N-terminus, right near the tip of that protein, there's a cluster of serines and threonines right here, which are all glycosylated with the same tetrasaccharide, which is a tetrasaccharide called disialyl-T, and the T part is this black disaccharide core, which is galactose linked to N-acetyl galactosamine, on serine and threonine residues, and then the disialyl part are the two sialic acids, one on the galactose residue and one at the 6th position of this GalNAc core residue. And when CD43 is glycosylated with this structure in this cluster, this is the part that binds with high affinity to Siglec-7. It's very interesting, and no other glycosylation on this protein seems to matter, and no other sialoglycan on the surface of this of this K562 cell seems to matter either. So all of the binding between Siglec-7 and the K562 cell is mediated by this one glycoprotein through this one little motif, and that was quite stunning because it points to, I think, what is probably more the rule than the exception in glycobiology, that when carbohydrate-binding receptors bind to their ligands, they're not just seeing a carbohydrate in isolation, they're often seeing that glycan in a context, and this is a very rich context for a Siglec to be seen. And I point this out because we had another project running in parallel at the time that Simon made this observation that actually converged on this dicyl-T structure as being very important for cancer immune suppression. And The project came basically through the hands of an MD-PhD student named Benji Smith, who wanted to understand not just why this sialic acid phenotype might be of benefit to the cancer, but also to understand how this phenotype came to arise through changes inside the cancer cell during the process of tumor evolution. And we were inspired by a study that had just been published around the time that I moved to Stanford from my colleague Dean Felsher. So Dean Felsher is an oncologist whose lab focuses on understanding oncogenic programs and particularly in understanding how the Myc oncogene contributes to malignancies. And it was well known that Myc, which is a transcription factor, can drive a transcriptional program involved in cell proliferation. That's well understood, and that's why MYC is such a holy grail target in oncology, because it's overexpressed in over half of human cancers and it drives proliferation. What was not known until Dean's work was that MYC can also drive an immune suppressive transcriptional program. So he found that in certain types of leukemia, MYC can drive the expression of PD-L1, that ligand for PD-1, and that allows suppression of the T cell through checkpoint receptor engagement, as well as the expression of another so-called "don't eat me" signal called CD47, which interacts with an immune suppressive receptor on macrophages. So when we saw this paper published in Science, and we saw that MYC was driving the expression of ligands the cancer cell was presenting to immune suppressive receptors, we couldn't help but wonder whether MYC could also be involved in upregulating these sialoglycan ligands for the Siglec receptor family, because, you know, we understood this analogy. So we wanted to study this, and toward that end, we joined forces with Dean and with a postdoc in his lab named Anja Deutzmann. And Anja was working with isogenic cell lines that Dean's lab had engineered to express MYC under the control of the tetracycline repressor. So these were cell lines in which one could add doxycycline to the cell culture media to shut down the expression of MYC, or withdraw doxycycline from the cell culture media to induce the expression of MYC. And with these MYC-on and MYC-off cells, one could then do comparisons of their transcriptomes, their glycomes, and also their ability to engage Siglecs through their own ligands. And so this is exactly the kind of comparative study that Benjie did. Again, he was an MD-PhD student in my lab, a graduate of Pomona College before that, and now he's finished his PhD and is just finishing the final 2 years of his clinical work to earn his MD. And Benjie basically took cell lines that had been either MEK on or MEK off state, and then did RNA-seq, looked at their glycan profiles by mass spectrometry, and then tested their ability to bind to different Siglec-Fc reagents as a probe of their Siglec ligand expression, and what he found was quite stunning. So first of all, when he did RNA-seq analysis of the glycogenes of a certain T-ALL cell line, so acute lymphoblastic leukemia, in the MYC on versus MYC off state, he found that when MYC was on, there was a high expression of one particular gene that encodes a sialyl transferase called ST6GalNAc4, and we were really intrigued by that observation because that's a sialyl transferase that makes the Disialyl-T structure starting from the precursor trisaccharide called Sialyl-T. So this structure where there's a sialic acid, we denote that with this pink diamond, and an alpha-2,3 linkage to galactose, linked to N-acetyl galactosamine, that's a quite common structure found on healthy cells as well, but this disialyl structure is much more rare, and you would only be able to make this structure if this trisaccharide can be sialylated at the 6 position by an enzyme like ST6GalNAc4. And sure enough, when we did the mass spec glycomic analysis to look at the O-glycans in the MYC on versus MYC off state, we found that when MYC is on, the disialyl-T structure is actually the predominant O-glycan structure on this cell type, and it is on mucins like CD43. And when MYC is off and MYC levels are lower, that the prevalence of this disialyl-T structure goes down, drops in half, and correspondingly precursors to disialyl-T are enhanced in their abundance. Importantly, when we looked at Siglec-Fc binding by flow cytometry, which is now represented with a heat map, we found that the ligands for Siglec-7 and Siglec-9, the human Siglecs, as well as relatives in the mouse, Siglec-F and Siglec-E, that those ligands, or the binding of those Siglecs to their ligands, is regulated by MYC. So for example, when MYC is on, these cells bind avidly to human Siglec-9 and human Siglec-7, as well as mouse Siglec-F and mouse Siglec-E. But when MYC is off, that binding is reduced to all of those Siglecs, and when we knocked out that sialyl transferase that is regulated by MYC, the ST6GalNAc4, we entirely lost all binding activity to human Siglec-7, and there was a reduction of binding to human Siglec-9. So the effect on Siglec-7 was so dramatic that we basically positioned this disialyl-T structure as a critical determinant of Siglec-7 binding, much more so than any other Siglec, and dependent on ST6GalNAc4. So we now have this working model where we understand that ST6GalNAc4 is regulated by MYC, so when MYC is overexpressed, levels of this transcript go up, and that leads to increased expression of the sialyl transferase, which is in the Golgi compartment, and what that enzyme can then do is increase the flux in the biosynthesis of the Disialyl-T structure from its Sialyl-T precursor, which then increases the density of these Disialyl-T epitopes on the surface of the T-ALL cell, and when this high affinity ligand is present at high density, Siglec-7 can be engaged on its immune cells,, which includes activated T cells, macrophages, as well as NK cells, and that can then suppress the activity of these immune cells as they form synapses with these Disialyl-T-rich cancer cells. Further confirmation of this model came when we looked at how cancer cells grow in a mouse model when they're missing ST6GalNAc4, and therefore unable to make high levels of disialyl-T. These ST6GalNAc4 knockout cells have a very difficult time engrafting in an animal because they're riddled with immune cells, whereas cells that are able to make ST6GalNAc4, the wild-type cells, those cells are able to suppress immune infiltration and they can grow much more clearly. So you can see, this is work from Anya in Dean's lab, she inoculated mice with either safe-targeted, or ST6GalNAc4 knockout CRISPR cell lines of the ATAL host, and as a function of time, the tumors are growing much more aggressively in the wild-type or safe-targeted cells versus ST6GalNAc4 knockouts. So this is kind of genetic validation for the idea that if you can shut down the activity of ST6GalNAc4, you could have an immunotherapy consequence, and so we're right now thinking about how to design inhibitors of this enzyme, small molecule inhibitors, that could be a different type of immune therapy, and ideally one that is someday, you know, a small molecule orally available pill rather than a biologic that you would administer through infusion. So that's where we're going in the future, and we're excited about this as a kind of new target class. So to summarize what I've talked about today, I started by giving you the backstory of how we came to understand in my lab and as a field, that tumor sialoglycans can engage the Siglec receptors and drive immune suppression in the tumor microenvironment. And that's what stimulated our idea that a molecule that can basically do targeted degradation of those Siglec ligands, these antibody sialidase conjugates, that this could be a new approach to cancer immune therapy, and we are going to find out over the next few years through the clinical work at Pallion Pharmaceuticals. Then I showed you how probing MYC oncogene consequences on cell surface glycosylation led us to identify the disialyl-T glycoform of mucins like CD43 as important players, particularly in Siglec-7 biology, and that led us to identify ST6GalNAc4 as a potential target for cancer immunotherapy, in the form of a small molecule inhibitor, and that's what the future holds. So with that, I'll conclude finally by thanking all of the people that presently occupy my lab. I also have many former lab members who I called out during the presentation on slides, pictures of their faces, but they have handed the baton to this current generation, and of course everyone in the lab makes a contribution to everybody's project, either directly or indirectly. And none of this would be possible without the generous funding that I've received from the NIH, and in particular, the NIGMS, which is General Medical Sciences Institute, as well as the National Cancer Institute. HHMI has been with me now for 22 years. I'm deeply grateful to their ongoing support. And finally, to Stanford University for giving me the opportunity to do this work in a really exciting research environment. So thank you.

And now I'm very happy to entertain questions.

Fabulous. Thank you.

I might ask you to stop sharing your screen so that we can now also see other participants. So hopefully you can see people as well. That was really fantastic. So I'm going to ask participants to either raise your hand to ask a question directly, at which point Kelly will unmute you, or to type your questions in the Q&A for me to read out. So while those are coming in, I actually wanted to just start with a really naive question. It was beautifully convincing that this disialyl-T signature is a way of a cancer cell triggering the inhibitory immune response. I was sort of wondering why it hasn't evolved as an eat me signal that, you know, if it's a hallmark of a cancer cell, it's certainly, you know, a massive flashing neon light.

To—

I'm just sort of curious your thoughts about that.

Yeah, so this is a great question, uh, and evolution is a funny thing, isn't it? Um, I didn't go through the details here, but, um, if you'll indulge me that I screen my sheet— I'm going to share my screen one more time here quickly and go back to show you the family of this, of siglecs. And I mentioned that many members of the family have a signaling domains that deliver inhibitory signals, like, you know, this ITIM/ITSM motif, but you can see that I've remained silent on other members of the family, so Siglec-4, 14, 15, 16. These Siglecs actually have activating motifs in their cytosolic domains, so that they actually can deliver an activating signal to the immune cell, and so it's entirely possible that There are some activating signals that can be delivered from some of these sialoglycans, but people have not studied those Siglecs in any great detail, so we just don't know much about them. The entire family kind of fell out from the Human Genome Sequencing Project back in the early 2000s, and prior to that time, only a few of the Siglecs had been described just by having been purified from tissues and cloned the old-fashioned way. Remember, those of us of a certain age remember when you had to clone a gene because we didn't have the whole genome sequenced. But once we discovered all the members of the Siglec family, it became clear that their immune modulation is quite nuanced, and they are the product of gene duplications and expansions. They have rapidly evolved in vertebrates, so if you look at the mouse Siglecs and the human Siglecs, there's a lot of differences. Mice have fewer Siglecs than humans have. There's no direct murine ortholog of Siglec-7 or 9. They have kind of an in-betweener called Siglec-E. So the mouse is not a great model even for human Siglec biology. And it could be that in, over human evolution, we have figured out ways to recognize tumors through Siglecs, as well as to be tricked by tumors through Siglecs, and hopefully all that will play out. But let me just make a point here also, which is glycoscience is a relatively small area in biology, and the number of labs that focus on Siglec biology entirely, you can count on one hand. There's not that many people who are working on these receptors, but I would like to think that if we can show that they have a prominent role in tumor immune evasion, that will change.

Yeah, absolutely.

And yeah, okay, so I'll stop asking questions and let Kelly unmute our first questioner.

So I unmute Sean first.

Okay.

Hi, I had a question. So do you think that the not engraftment that you saw was specific to ST6GalNAc4, or would you see it with knockout of other sialic acid-involved genes?

That's a really good question. We don't know, because we didn't knock out any other sialotransferases. But we kind of took our cue from those genes that were regulated by MYC. And there were other glycogenes that we saw sort of significant changes in their transcript levels, but that sialotransferase was really so striking, so much more so than any other gene. And there were no other sialic acid-related genes, actually. The other glycogenes that we found as hits were sort of fucosyltransferases, xylosyltransferases. They were sort of in other glycan categories, but nothing nearly as striking as that one sialotransferase. It really stood out. Now, this is just for the T-ALL cell line that we were working with. You know, there are lots of other types of, you know, cancers from other lineages that might be driven by MYC, and it's possible that there are other solutions to the problem of engaging Siglecs besides just ST6GalNAc4. But in that particular cell type, this seems to be a really prominent node for immune suppression, which is what's so interesting about it. Matt Cotts will be next.

Are you there, Matt? Hi, thank you for a great talk. I was wondering whether you explored whether dimensions of the sialoglycans are somehow related to the ability of these molecules to drive inhibitory signaling, given that the size of the contacts between T cells and APCs are rather constrained by the size of the TCR-MHC complex.

All right, so let me— I'm not sure I understand the question. Can you start from the beginning?

What was the first part again? I'm sorry, I was just wondering whether you explored whether dimensions of the sialoglycans are somehow related to how well those molecules can drive inhibitory signaling. Dimensions.

You mean the dimensions? Yeah. Okay, I think I understand. So let me get a graphic to explain how I think about this. Sorry, I keep sharing my slides again and again. Let's get a synapse out here. All right, so for example, can you see that? So in these cartoons, we're pretending that the cancer cell and the immune cell are two spheres. And of course they're not, right? Their cell surface is very featured and very ruffled. And we've actually done super-resolution microscopy on synapses between T cells and cancer cells where the Siglec, where Siglec-7 in particular, has engaged mucins on the cancer cell at the same time that T cell receptors have engaged their ligands. And it's quite clear that in order for these very tall mucins to engage with Siglecs, there have to be features of the membrane, ruffles, because these mucins do tend to tower above molecules like MHC and T-cell receptors, which are globular proteins and they're shorter. And just to give you a sense of the dimensions, your typical T-cell receptor MHC complex is going to be like 20 angstroms on each side, so let's say 40 angstroms total, whereas a mucin could be 100 to 200 angstroms tall on the cell surface. It's much taller.. But you can still get patches of MUC and Siglec engagement by having membrane ruffles at the same time that there is, you know, patches of MHC T-cell receptor. And these patches, of course, are where all the signaling activity is occurring. So there's going to be patches within patches at the synapse. But as long as membranes can— are fluid and featured, it's really not a problem to have things of different dimensions engaged at the synapse, but they will be in discrete patches, which is why I kind of love this cartoon so much. Because you can really see in this electron micrograph of a real target cell and synapses with real immune cells that the synapse is very intricate and very featured, and that's how you can get these patches to engage simultaneously.

Does that help?

Yeah, thank you. Next is Mark Troll. Oh, hi. I was just wondering if it's possible to put the degrading enzymes directly into the bloodstream, or do you have to target them the way you did?

You know, so we do administer our antibody sialidase conjugates intravenously, both in animals and in human patients, and we count on the antibody component to basically give them the ability to accumulate in the tumor microenvironment where the antigen is overexpressed. But they are in the periphery as well. And in fact, we can't rule out that there, you know, some desialylation of disseminating circulating tumor cells might be occurring in the periphery, which can contribute to anti-tumor immunity. So there could be some peripheral activity going on here.

But, you know, the antibody does its job. And I just give them— What I meant to ask, and I didn't put it well, is Suppose you just put in the enzyme by itself, would it be—

without the antibody targeting? Yeah, yeah. So I didn't go through the details, they are published, but we actually mutated the enzyme, the sialidase, to detune it, to make it relatively inactive. We made mutations that elevate the Km, so the enzyme on its own has very weak binding to its substrates, and it's really not able to desialylate cells that it's just floating around amongst. It actually has to be parked on the membrane so that the local concentration of its substrates is now so high it can overcome the fact that we mutated it and made it give it a really high Km. So these enzymes, we've engineered them to be extremely proximity dependent, and so they don't indiscriminately desialylate cells that they might happen to collide with in 3 dimensions. They will only take off the sialic acids that are in their face in 2 dimensions. That was a really important part of the engineering process. Thank you.

Okay, great. I think in the interest of time, I'm going to have just one more question from Magda, who doesn't have a microphone, so I'm going to read it out for you. Is it possible that the sialoglycan signature on cancer cells is something that other immunoprivileged cells utilize, like stem cells?

Say it again. Sorry, what's the last second part again?

My internet connection got warbled. Yeah, so other immunoprivileged cells that are protected from an autoimmune response, could they also use these silo signatures to hide?

Absolutely, and in fact, that is probably the main function of the Siglecs is they help, you know, protecting us. Basically, Siglecs are not there for us to get cancer, obviously. The natural selection process for Siglecs has been host-pathogen discrimination, as well as autoimmunity protection during pregnancy and fetal development. So the Siglecs are regulated during embryogenesis, and Siglec ligands are really prominent on cells at the placental interface between the mother and the fetus, and it's known that if you knock out Siglecs in animals, there's problems with gestation, right? You can have a higher prevalence of miscarriage and fetal rejection. So things that really drive evolution are going to be things related to host-pathogen or things related to pregnancy and reproduction. Cancer is something that happens to us generally later in life. It's not really— the evolutionary pressure really isn't there so much on cancer as it is on these other things. But the sialoglycans, they're really important in microbial interactions with the human immune system. And bacteria that do a good job of infecting us often have figured out how to engage immune suppressive receptors, right? And so Group B strep is a classic example of a bacterial pathogen that decorates itself in a sialic acid coating, and that allows it to engage Siglec-9 on macrophages and neutrophils and to suppress the reactivity of those innate immune cells. So that's probably the more important function of the Siglecs, whereas tumor discrimination is really something just unfortunate for us later in life.

Yeah, that makes perfect sense. All right. So in the interest of time, we're going to call a halt to the questions so that Carolyn can have some time to meet with students and other people. Thank you so much for a really fantastic talk. It's super exciting to hear the directions you're going in. And obviously this is going to have a huge impact for lots of people, I think.

So thank you very much again for joining us. I really appreciate the questions. Thank you. Thanks, Carolyn.